{
 "schema_version": "0.44",
 "paper": {
  "title": "What does sMAPE reward? An analytical and simulation study",
  "contributions": [
   "C1",
   "C12",
   "C4",
   "C15",
   "C6",
   "C3",
   "C9",
   "C10",
   "C7",
   "C17",
   "C18",
   "C19",
   "C20"
  ],
  "autonomy": "human_supervised",
  "contribution_record": "complete",
  "origin": "native",
  "exploration": {
   "lab_record": "A7",
   "questions": 3,
   "hypotheses": {
    "total": 20,
    "supported": 16,
    "not_supported": 0,
    "inconclusive": 4,
    "revised": 0,
    "abandoned": 0,
    "open": 0,
    "from_exploration": 1
   },
   "tests": {
    "total": 39,
    "supported": 30,
    "not_supported": 3,
    "inconclusive": 6
   },
   "runs": {
    "total": 18,
    "confirmatory": 14,
    "exploratory": 4,
    "failed": 0
   },
   "not_claimed": [
    {
     "lab_event": "H1",
     "lab_plan": "PL5",
     "outcome": "supported",
     "because": "reworded",
     "into": "C1",
     "why": "The statement was reworded after this verdict and decided again under PL8.R2, whose verdict is the claim's. This verdict: PL5's rule for supported is met: the proof (paper H1 in proofs.md, AR1) is complete; the independent checker AC4 confirmed it (CK1, full); and the deciding run RN5's numerical checks of its formulas are within the plan's tolerances."
    },
    {
     "lab_event": "H2",
     "lab_plan": "PL5",
     "outcome": "supported",
     "because": "reworded",
     "into": "C2",
     "why": "The statement was reworded after this verdict and decided again under PL8.R2, whose verdict is the claim's. This verdict: PL5's rule for supported is met: the proof (paper H2 in proofs.md, AR1) is complete; the independent checker AC4 confirmed it (CK2, full); and the deciding run RN5's numerical checks of its formulas are within the plan's tolerances."
    },
    {
     "lab_event": "H3",
     "lab_plan": "PL5",
     "outcome": "supported",
     "because": "reworded",
     "into": "C3",
     "why": "The statement was reworded after this verdict and decided again under PL12, whose verdict is the claim's. This verdict: PL5's rule for supported is met: the proof (paper H3 in proofs.md, AR1) is complete; the independent checker AC4 confirmed it (CK3, full); and the deciding run RN5's numerical checks of its formulas are within the plan's tolerances."
    },
    {
     "lab_event": "H23",
     "lab_plan": "PL9",
     "outcome": "supported",
     "because": "reworded",
     "into": "C3",
     "why": "The statement was reworded after this verdict and decided again under PL12, whose verdict is the claim's. This verdict: PL8's rule for supported is met for the current wording: the proof (paper H3 in proofs.md, AR1.R2) is complete; the independent checker AC4 confirmed the current wording (CK3, full), as did the supervisor (CK16); and the deciding run RN8's numerical checks of its formulas are within the plan's tolerances (RN8 repeats RN5 byte for byte). Part (b), the small-error approximation, is checked by RN8's H9 lines (0 violations; the fifth-order coefficient 0.833, as the proof gives 5/6)."
    },
    {
     "lab_event": "H4",
     "lab_plan": "PL5",
     "outcome": "supported",
     "because": "reworded",
     "into": "C4",
     "why": "The statement was reworded after this verdict and decided again under PL12, whose verdict is the claim's. This verdict: Verdict renewed after H4 was corrected twice: H4.R3 (the median with equal weights is a flat point, not necessarily the target), then H4.R4 (a sufficient condition for it to be the target). PL5's rule for supported is met for the current wording: the proof (paper H5, steps 1 to 10) is complete; AC4 confirmed it (CK13, full); and RN5's numerical checks of its formulas are within the plan's tolerances. The earlier checks CK5 (disputed) and CK12 (confirmed) were of earlier wordings."
    },
    {
     "lab_event": "H33",
     "lab_plan": "PL9",
     "outcome": "supported",
     "because": "reworded",
     "into": "C4",
     "why": "The statement was reworded after this verdict and decided again under PL12, whose verdict is the claim's. This verdict: PL8's rule for supported is met for the current wording: the proof (paper H5, steps 1 to 10 in proofs.md, AR1.R2) is complete; the independent checker AC4 confirmed the current wording (CK13, full), as did the supervisor (CK18); and the deciding run RN8's numerical checks of its formulas are within the plan's tolerances (RN8 repeats RN5 byte for byte). Its direction result is checked by RN8's H6 lines (balance residual 8.6e-9; 0 settings misplaced)."
    },
    {
     "lab_event": "H5",
     "lab_plan": "PL5",
     "outcome": "supported",
     "because": "reworded",
     "into": "C5",
     "why": "The statement was reworded after this verdict and decided again under PL8.R2, whose verdict is the claim's. This verdict: PL5's rule for supported is met: the proof (paper H6 in proofs.md, AR1) is complete; the independent checker AC4 confirmed it (CK6, full); and the deciding run RN5's numerical checks of its formulas are within the plan's tolerances."
    },
    {
     "lab_event": "H16",
     "lab_plan": "PL5",
     "outcome": "inconclusive",
     "because": "revised",
     "into": "C4",
     "why": "Merged into H4 by the supervisor's decision DC5: its direction result (the target lies on the side of the median that carries more weight) is now a corollary of H4; its balance in quantile terms and its Newton step are dropped as adding little."
    },
    {
     "lab_event": "H18",
     "lab_plan": "PL5",
     "outcome": "supported",
     "because": "reworded",
     "into": "C6",
     "why": "The statement was reworded after this verdict and decided again under PL8.R2, whose verdict is the claim's. This verdict: PL5's rule for supported is met: the proof (paper H7 in proofs.md, AR1) is complete; the independent checker AC4 confirmed it (CK7, full); and the deciding run RN5's numerical checks of its formulas are within the plan's tolerances. RN5's check that the global minimiser lies on the heavier side (unequal weights) gave True; the lab could not record a boolean as a result, so it is in RN5's output file results.json."
    },
    {
     "lab_event": "H8",
     "lab_plan": "PL5",
     "outcome": "supported",
     "because": "reworded",
     "into": "C7",
     "why": "The statement was reworded after this verdict and decided again under PL8.R2, whose verdict is the claim's. This verdict: PL5's rule for supported is met: the proof (paper H8 in proofs.md, AR1) is complete; the independent checker AC4 confirmed it (CK8, full); and the deciding run RN5's numerical checks of its formulas are within the plan's tolerances."
    },
    {
     "lab_event": "H9",
     "lab_plan": "PL5",
     "outcome": "inconclusive",
     "because": "revised",
     "into": "C3",
     "why": "Merged into H3 by decision DC5: the small-error approximation is part (b) of H3."
    },
    {
     "lab_event": "H10",
     "lab_plan": "PL5",
     "outcome": "inconclusive",
     "because": "revised",
     "into": "C2",
     "why": "Merged into H2 by decision DC5: the comparison of high and low forecasts with equal absolute errors is part (b) of H2."
    },
    {
     "lab_event": "H11",
     "lab_plan": "PL5",
     "outcome": "inconclusive",
     "because": "abandoned",
     "why": "Dropped by decision DC5: it applied a standard large-sample theorem; the efficiency of the sMAPE minimiser against the sample median is reported from the simulation PL2 instead, and part (iii) of H14 is dropped with it."
    },
    {
     "lab_event": "H12",
     "lab_plan": "PL5",
     "outcome": "supported",
     "because": "reworded",
     "into": "C8",
     "why": "The statement was reworded after this verdict and decided again under PL8.R2, whose verdict is the claim's. This verdict: PL5's rule for supported is met: the proof (paper H10 in proofs.md, AR1) is complete; the independent checker AC4 confirmed it (CK10, full); and the deciding run RN5's numerical checks of its formulas are within the plan's tolerances."
    },
    {
     "lab_event": "H13",
     "lab_plan": "PL5",
     "outcome": "supported",
     "because": "reworded",
     "into": "C9",
     "why": "The statement was reworded after this verdict and decided again under PL8.R2, whose verdict is the claim's. This verdict: PL5's rule for supported is met: the proof (paper H4 in proofs.md, AR1) is complete; the independent checker AC4 confirmed it (CK4, full); and the deciding run RN5's numerical checks of its formulas are within the plan's tolerances."
    },
    {
     "lab_event": "H19",
     "lab_plan": "PL5",
     "outcome": "supported",
     "because": "reworded",
     "into": "C10",
     "why": "The statement was reworded after this verdict and decided again under PL8.R2, whose verdict is the claim's. This verdict: PL5's rule for supported is met: the proof (paper H11 in proofs.md, AR1) is complete; the independent checker AC4 confirmed it (CK11, full); and the deciding run RN5's numerical checks of its formulas are within the plan's tolerances. Part (iii), the efficiency rate, was dropped by DC5; RN5's checks of it no longer decide anything."
    },
    {
     "lab_event": "H34",
     "lab_plan": "PL9",
     "outcome": "supported",
     "because": "reworded",
     "into": "C12",
     "why": "The statement was reworded after this verdict and decided again under PL12, whose verdict is the claim's. This verdict: PL8's rule for supported is met: the proof (paper H6 for part (a); paper H5, steps 3, 6, 8 and 9, for part (b), in proofs.md, AR1.R2) is complete; the independent checker AC4 confirmed H34 as worded (CK25, full; recorded 20 seconds after PL8 was first written, as PL8.R2 states); and the deciding run RN8's numerical checks of the formulas it rests on (paper H5 and H6: the minimiser, the median, the balance) are within the plan's tolerances. The supervisor's checks of H4 and H5 (CK18, CK19) cover both parts; no separate supervisor check of H34 was recorded."
    },
    {
     "lab_event": "H35",
     "lab_plan": "PL10",
     "outcome": "supported",
     "because": "reworded",
     "into": "C17",
     "why": "The statement was reworded after this verdict and decided again under PL12, whose verdict is the claim's. This verdict: PL10's rule for supported is met: the proof of H35 (proof-msmape.md, AR8) is complete, every step naming what it rests on; the independent checker AC4 derived H35 from its statement alone, then compared it with the proof, and recorded its check as confirmed, full (CK26); and the deciding run RN9's checks meet the tolerances: the identities of (a) and (b) within 1.2e-16, no violation of the bounds or of 0 <= m < 1, and a zero forecast of a zero actual scoring 0. Steps 4 and 7 of the proof, which the checker suggested and had derived itself, were added after its check."
    },
    {
     "lab_event": "H39",
     "lab_plan": "PL12",
     "outcome": "not_supported",
     "because": "reworded",
     "into": "C18",
     "why": "The statement was reworded after this verdict and decided again under PL17, whose verdict is the claim's. This verdict: As worded, H39 does not hold in every case it allows. The independent checker (CK30) gave a counterexample to part (c): a density with bounded support, for which H(c_2) can be 0 or 1 and the stated limit is 1/0 or the ratio undefined; the proof's step 8 adds the condition 0 < H(c_2) < 1 but the statement did not. Part (b) also needs sd(D) > 0, and the closing sentence claimed more than the parts give. In addition one of RN13's checks of part (c) missed its tolerance (3.5% against 3% at sigma = 200, a slow convergence, not an error). The mathematics of (a), (b) and (d) was confirmed. The statement is corrected as H43."
    },
    {
     "lab_event": "H39",
     "lab_plan": "PL13",
     "outcome": "not_supported",
     "because": "reworded",
     "into": "C18",
     "why": "The statement was reworded after this verdict and decided again under PL17, whose verdict is the claim's. This verdict: PL13's numerical tolerances are all met by RN14 (the absolute error's ratio is within 0.0035 of its limit at sigma = 2000 and closer than at sigma = 200 in all six cases), but the checker's counterexample to the wording of part (c) (CK30) stands whatever the run shows. The statement is corrected as H43, which PL15 decides."
    },
    {
     "lab_event": "H43",
     "lab_plan": "PL15",
     "outcome": "inconclusive",
     "because": "reworded",
     "into": "C18",
     "why": "The statement was reworded after this verdict and decided again under PL17, whose verdict is the claim's. This verdict: Neither rule is met for H43. The run RN16 meets every tolerance and the checker found parts (a) to (d) right as now worded, but disputed the closing sentence again (CK34): 'sMAPE loses it, at the rate sigma^(-1/2) when h(c_1) != h(c_2)' says sMAPE loses its power in every case, which is shown when h(c_1) + h(c_2) > 0 (the ratio is then of order sigma^(-5/2) when the two densities are equal) but not when both are 0, as a density with bounded support allows; no counterexample was found. The sentence is to be narrowed in a further correction."
    },
    {
     "lab_event": "H45",
     "lab_plan": "PL16",
     "outcome": "inconclusive",
     "because": "reworded",
     "into": "C18",
     "why": "The statement was reworded after this verdict and decided again under PL17, whose verdict is the claim's. This verdict: Neither rule is met for H45. The run RN17 meets every tolerance, and the checker found parts (a) to (d) and four of the five closing sentences right, but disputed one (CK36): 'with a ratio of order sigma^(-5/2)' reads as the exact order, and only an upper bound is shown (the ratio is 0 for a symmetric density with c_1 = -c_2). It is to read 'at most of order'."
    },
    {
     "lab_event": "H40",
     "lab_plan": "PL12",
     "outcome": "not_supported",
     "because": "reworded",
     "into": "C19",
     "why": "The statement was reworded after this verdict and decided again under PL15, whose verdict is the claim's. This verdict: As worded, H40 calls the ratio of the two limiting variances the large-sample value of the relative efficiency, which is defined from the variances of finite test sets. The independent checker (CK31) confirmed all of the mathematics (the strict inequality, consistency, the conditions of the theorem, the variance, the constants and both rates) and gave a counterexample to that phrase: log-actuals with very heavy tails, under H40's conditions, for which the sample median has infinite variance at every n. RN13's numerical checks of the formula and of the two rates met their tolerances. The statement is corrected as H44."
    }
   ]
  },
  "study": "smape-target",
  "licence": "CC-BY-4.0",
  "highlights": [
   "C1",
   "C12",
   "C4",
   "C15",
   "C6"
  ],
  "summary": "Which forecast does sMAPE reward? sMAPE has been criticised for its asymmetry and its behaviour near zero, but the forecast that minimises it had been found only in special cases. We show that for positive values sMAPE depends only on the ratio of actual to forecast, not on their difference: it is the hyperbolic tangent of half the absolute log error. So it scores a forecast k times too high and one k times too low the same, and is asymmetric only for equal absolute errors. Being bounded, it has a finite expected value for every distribution. Its target is a median of the actuals, each weighted by how close it is to the forecast by ratio. This is the median when the log of the actual is symmetric and unimodal; otherwise it can move to the side of the median that carries more weight. With two well-separated groups of actuals, the median becomes a local maximum. On intermittent demand the target is 0 once the share of zeros passes a threshold below one half. Otherwise sMAPE aims at a weighted median of the non-zero actuals. As the actuals spread out, expected sMAPE approaches its bound for every forecast, so forecasts become hard to tell apart, and the target moves from the median to the mode of the log. Where sMAPE and MAE both target the median, a test set's sMAPE minimiser estimates it less precisely than the sample median, increasingly so as the actuals spread out.",
  "classification": {
   "subfield": "openalex:subfields/2613",
   "keywords": [
    "forecast evaluation",
    "error measures",
    "sMAPE",
    "point forecasts",
    "consistent scoring functions",
    "proof",
    "simulation study",
    "msMAPE",
    "Suilin loss",
    "intermittent demand"
   ],
   "topics": [
    "openalex:T11918",
    "openalex:T11652",
    "openalex:T10136"
   ]
  },
  "declarations": {
   "funding": [],
   "competing_interests": "Ivan Svetunkov runs EntryLib and wrote the review (entrylib:2026.000008) whose claims about sMAPE this study tests."
  },
  "sections": [
   {
    "title": "What sMAPE measures",
    "claims": [
     "C1",
     "C2",
     "C3"
    ]
   },
   {
    "title": "What sMAPE rewards for positive actuals",
    "claims": [
     "C9",
     "C4",
     "C5",
     "C12",
     "C6"
    ]
   },
   {
    "title": "Zeros and negative actuals",
    "claims": [
     "C7",
     "C11"
    ]
   },
   {
    "title": "Widely spread actuals",
    "claims": [
     "C10",
     "C18"
    ]
   },
   {
    "title": "Other measures",
    "claims": [
     "C8"
    ]
   },
   {
    "title": "sMAPE modifications",
    "claims": [
     "C17",
     "C20",
     "C16",
     "C21"
    ]
   },
   {
    "title": "Precision on a test set",
    "claims": [
     "C15",
     "C19"
    ]
   },
   {
    "title": "Numerical checks",
    "claims": [
     "C13",
     "C14"
    ]
   }
  ],
  "id": "entrylib:2026.000013",
  "correction_reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6.",
  "version": 3,
  "date": "2026-11-18",
  "status": "corrected",
  "previous_version": "entrylib:2026.000013@v2",
  "status_notice": {
   "date": "2026-11-18",
   "reason_code": "correction",
   "statement": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6.",
   "issued_by": "Ivan Svetunkov (ORCID 0000-0001-7826-0281)"
  },
  "changes": [
   {
    "op": "replace",
    "path": "/artefacts/0/description",
    "value": "Proofs of the formal claims, version 3, in Markdown with LaTeX: section H3 with part (c); H5 shortened to the weighted median; new sections H12 (the ordinary median, with the checker's example), H13 (signal-to-noise ratios) and H14 (large-sample efficiency). Checked independently by AC6 (CK27 to CK31, CK34 to CK37) and corrected as it proposed; after its checks the notation was made consistent (no second meaning for k, a, G), which AC6 read and found to change nothing. Sections H1 to H14 prove C1, C2, C3, C9, C4, C5, C6, C7, C11, C8, C10, C12, C18 and C19, in that order. Sections H1 to H11 were also checked independently by lab actor AC4 (proof-check-report.md) and by the supervisor.",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/artefacts/0/uri",
    "value": "https://entrylib.org/p/2026.000013/v3/files/files/proofs.md",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/artefacts/0/version",
    "value": "sha256:90c89f3395e068a7a2762e2839e86894e4efe488dd09ca3fa988e7ce0cc1df51",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/artefacts/1/description",
    "value": "The same proofs, version 3, rendered as PDF.",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/artefacts/1/uri",
    "value": "https://entrylib.org/p/2026.000013/v3/files/files/proofs.pdf",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/artefacts/1/version",
    "value": "sha256:767603774315d5d392a7fed0c7fb35531c52a4f68f836b0d944aa5c83d5cdc35",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/artefacts/2/uri",
    "value": "https://entrylib.org/p/2026.000013/v3/files/files/smape_weights.png",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/artefacts/3/uri",
    "value": "https://entrylib.org/p/2026.000013/v3/files/files/proof-check-report.md",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/artefacts/4/description",
    "value": "Proofs of C17 (record hypothesis H41: the Suilin loss beside sMAPE; the measure renamed, content as before) and of C20 (record hypothesis H42: the msMAPE of Smyl beside sMAPE), in Markdown with LaTeX; checked independently by AC6 (CK32, CK33). The msMAPE term is written s_nu here, as in this claimset; the recorded statement writes p. In version 2 the file was proof-msmape.md.",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/artefacts/4/uri",
    "value": "https://entrylib.org/p/2026.000013/v3/files/files/proof-suilin.md",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/artefacts/4/version",
    "value": "sha256:cc1edb21c3cdff150d3bdd5c455a96849e3c04617fad200aad53f32cf63f03ac",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/artefacts/5/description",
    "value": "Every result of the lab record (17182 results of 17 runs) at full precision: the lab result's ID, its run, the quantity as the run's script named it, its conditions and its value (both as JSON). The evidence's tables give the values to four significant digits.",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/artefacts/5/uri",
    "value": "https://entrylib.org/p/2026.000013/v3/files/files/smape-target.results.csv",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/artefacts/5/version",
    "value": "sha256:dbcda16de8d13b8449906fa1142665fc1efc4b472ebe082233f3e457c2eb65eb",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/artefacts/6/description",
    "value": "Lab record of the study: questions, hypotheses, plans, runs, results and decisions, as an append-only, hash-chained event log (Claimset lab record specification v0.22, begun 2026-11-16; exported with claimset 0.44.6).",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/artefacts/6/uri",
    "value": "https://github.com/entrylib/study-planning/blob/032a092a1879211ed0c8e383b5895ce81ab83e7b/lab/lab.jsonl",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/artefacts/6/version",
    "value": "sha256:26886d9cdd338f2df8c3c9f78b6458897789cf4236b28c36a34946ece8aa2de8",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/artefacts/7/uri",
    "value": "https://entrylib.org/p/2026.000013/v3/files/files/smape-target.RN1.results.csv",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/artefacts/8/uri",
    "value": "https://entrylib.org/p/2026.000013/v3/files/files/figures/RN1-inverse_Gaussian_shape_1000.png",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/artefacts/9/uri",
    "value": "https://entrylib.org/p/2026.000013/v3/files/files/figures/RN1-inverse_Gaussian_shape_200.png",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/artefacts/10/uri",
    "value": "https://entrylib.org/p/2026.000013/v3/files/files/figures/RN1-inverse_Gaussian_shape_100.png",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/artefacts/11/uri",
    "value": "https://entrylib.org/p/2026.000013/v3/files/files/figures/RN1-inverse_Gaussian_shape_50.png",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/artefacts/12/uri",
    "value": "https://entrylib.org/p/2026.000013/v3/files/files/figures/RN1-inverse_Gaussian_shape_20.png",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/artefacts/13/uri",
    "value": "https://entrylib.org/p/2026.000013/v3/files/files/figures/RN1-inverse_Gaussian_shape_10.png",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/artefacts/14/uri",
    "value": "https://entrylib.org/p/2026.000013/v3/files/files/figures/RN1-log-skewed_left_sd_0.5_skew_-2.png",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/artefacts/15/uri",
    "value": "https://entrylib.org/p/2026.000013/v3/files/files/figures/RN1-log-skewed_left_sd_0.5_skew_-1.png",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/artefacts/16/uri",
    "value": "https://entrylib.org/p/2026.000013/v3/files/files/figures/RN1-log-skewed_left_sd_0.5_skew_-0.5.png",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/artefacts/17/uri",
    "value": "https://entrylib.org/p/2026.000013/v3/files/files/figures/RN1-log-skewed_right_sd_0.5_skew__2.png",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/artefacts/18/uri",
    "value": "https://entrylib.org/p/2026.000013/v3/files/files/figures/RN1-log-skewed_right_sd_0.5_skew__1.png",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/artefacts/19/uri",
    "value": "https://entrylib.org/p/2026.000013/v3/files/files/figures/RN1-log-skewed_right_sd_0.5_skew__0.5.png",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/artefacts/20/uri",
    "value": "https://entrylib.org/p/2026.000013/v3/files/files/figures/RN1-lognormal_sigma_0.5.png",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/artefacts/21/uri",
    "value": "https://entrylib.org/p/2026.000013/v3/files/files/figures/RN1-log-skewed_left_sd_1_skew_-2.png",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/artefacts/22/uri",
    "value": "https://entrylib.org/p/2026.000013/v3/files/files/figures/RN1-log-skewed_left_sd_1_skew_-1.png",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/artefacts/23/uri",
    "value": "https://entrylib.org/p/2026.000013/v3/files/files/figures/RN1-log-skewed_left_sd_1_skew_-0.5.png",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/artefacts/24/uri",
    "value": "https://entrylib.org/p/2026.000013/v3/files/files/figures/RN1-log-skewed_right_sd_1_skew__2.png",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/artefacts/25/uri",
    "value": "https://entrylib.org/p/2026.000013/v3/files/files/figures/RN1-log-skewed_right_sd_1_skew__1.png",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/artefacts/26/uri",
    "value": "https://entrylib.org/p/2026.000013/v3/files/files/figures/RN1-log-skewed_right_sd_1_skew__0.5.png",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/artefacts/27/uri",
    "value": "https://entrylib.org/p/2026.000013/v3/files/files/figures/RN1-lognormal_sigma_1.png",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/artefacts/28/uri",
    "value": "https://entrylib.org/p/2026.000013/v3/files/files/figures/RN1-bimodal_equal_modes.png",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/artefacts/29/uri",
    "value": "https://entrylib.org/p/2026.000013/v3/files/files/figures/RN1-bimodal_left_mode_higher.png",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/artefacts/30/uri",
    "value": "https://entrylib.org/p/2026.000013/v3/files/files/figures/RN1-bimodal_right_mode_higher.png",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/artefacts/31/uri",
    "value": "https://entrylib.org/p/2026.000013/v3/files/files/figures/RN1-normal_straddling_zero_mean_0.png",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/artefacts/32/uri",
    "value": "https://entrylib.org/p/2026.000013/v3/files/files/figures/RN1-normal_straddling_zero_mean_0.5.png",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/artefacts/33/uri",
    "value": "https://entrylib.org/p/2026.000013/v3/files/files/figures/RN1-normal_straddling_zero_mean_1.png",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/artefacts/34/uri",
    "value": "https://entrylib.org/p/2026.000013/v3/files/files/figures/RN1-lognormal_sigma_0.5_with_zeros_p0_0.25.png",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/artefacts/35/uri",
    "value": "https://entrylib.org/p/2026.000013/v3/files/files/figures/RN1-lognormal_sigma_0.5_with_zeros_p0_0.42.png",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/artefacts/36/uri",
    "value": "https://entrylib.org/p/2026.000013/v3/files/files/figures/RN1-lognormal_sigma_0.5_with_zeros_p0_0.47.png",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/artefacts/37/uri",
    "value": "https://entrylib.org/p/2026.000013/v3/files/files/figures/RN1-lognormal_sigma_0.5_with_zeros_p0_0.6.png",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/artefacts/38/uri",
    "value": "https://entrylib.org/p/2026.000013/v3/files/files/figures/RN1-lognormal_sigma_1_with_zeros_p0_0.25.png",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/artefacts/39/uri",
    "value": "https://entrylib.org/p/2026.000013/v3/files/files/figures/RN1-lognormal_sigma_1_with_zeros_p0_0.37.png",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/artefacts/40/uri",
    "value": "https://entrylib.org/p/2026.000013/v3/files/files/figures/RN1-lognormal_sigma_1_with_zeros_p0_0.42.png",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/artefacts/41/uri",
    "value": "https://entrylib.org/p/2026.000013/v3/files/files/figures/RN1-lognormal_sigma_1_with_zeros_p0_0.6.png",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/artefacts/42/uri",
    "value": "https://entrylib.org/p/2026.000013/v3/files/files/smape-target.RN2.results.csv",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/artefacts/43/uri",
    "value": "https://entrylib.org/p/2026.000013/v3/files/files/figures/RN2-flatness_target.png",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/artefacts/44/uri",
    "value": "https://entrylib.org/p/2026.000013/v3/files/files/figures/RN2-flatness_expected_sMAPE.png",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/artefacts/45/uri",
    "value": "https://entrylib.org/p/2026.000013/v3/files/files/figures/RN2-flatness_sensitivity.png",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/artefacts/46/uri",
    "value": "https://entrylib.org/p/2026.000013/v3/files/files/figures/RN2-flatness_agreement.png",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/artefacts/47/uri",
    "value": "https://entrylib.org/p/2026.000013/v3/files/files/figures/RN2-flatness_minimiser.png",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/artefacts/48/uri",
    "value": "https://entrylib.org/p/2026.000013/v3/files/files/smape-target.RN7.results.csv",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/artefacts/49/uri",
    "value": "https://entrylib.org/p/2026.000013/v3/files/files/figures/RN7-efficiency_lognormal.png",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/artefacts/50/uri",
    "value": "https://entrylib.org/p/2026.000013/v3/files/files/figures/RN7-efficiency_log-Laplace.png",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/artefacts/51/description",
    "value": "Every result of lab run RN10 (PL11: the Suilin and sMAPE targets) at full precision, one row per result, sorted by setting: the lab result's ID, its run, the quantity, its conditions and its value (both as JSON). The evidence's tables give selected values to four significant digits. The run named the Suilin loss msMAPE, as version 2 of the paper did: in this file the measure \"msMAPE\" and the quantity \"relative difference of the msMAPE target from the sMAPE target\" are the Suilin loss's.",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/artefacts/51/uri",
    "value": "https://entrylib.org/p/2026.000013/v3/files/files/smape-target.RN10.results.csv",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/artefacts/52/description",
    "value": "The Suilin loss beside sMAPE: near zero, the two targets as a share of the median against the share of zeros, for lognormal outcomes at medians 1 and 100, with the zero thresholds at median 1; away from zero, the relative difference of the Suilin target from the sMAPE target for each distribution; on Poisson counts, the two targets against the mean. The figure was drawn when the paper called the Suilin loss msMAPE, and labels it so. Drawn by lab run RN11 from the recorded results of RN10 (lab/runs/RN11/figures/msmape_targets.png).",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/artefacts/52/uri",
    "value": "https://entrylib.org/p/2026.000013/v3/files/files/figures/RN11-msmape_targets.png",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/artefacts/53/description",
    "value": "The study's code: the distributions, the exact quadrature of expected sMAPE, the formula checks, the three simulations (targets.py, efficiency.py, flatness.py), and the targets of the Suilin loss (msmape_targets.py) and of msMAPE on intermittent demand (msmape_intermittent.py).",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/artefacts/53/uri",
    "value": "https://github.com/entrylib/study-planning/tree/6aa7a7ab87bf5c4e40d704a7492cd9af9ef481db/claimset-code/smape-target",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/artefacts/53/version",
    "value": "6aa7a7ab87bf5c4e40d704a7492cd9af9ef481db",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "add",
    "path": "/artefacts/54",
    "value": {
     "kind": "data",
     "uri": "https://entrylib.org/p/2026.000013/v3/files/files/smape-target.RN15.results.csv",
     "version": "sha256:c9c7808e3d002533c95d0f102ad903c1a5d7799ff6d23866507fff9eda260023",
     "licence": "CC-BY-4.0",
     "availability": {
      "level": "public"
     },
     "columns": [
      {
       "name": "lab_result",
       "dimension": "the result event of the lab record"
      },
      {
       "name": "run",
       "dimension": "the lab run that produced it"
      },
      {
       "name": "quantity",
       "dimension": "the quantity, as the run's script named it"
      },
      {
       "name": "conditions",
       "dimension": "the result's conditions, as JSON"
      },
      {
       "name": "value",
       "variable": "the value of the quantity, as JSON, at full precision"
      }
     ],
     "description": "Every result of lab run RN15 (PL14: the targets of sMAPE, msMAPE and the Suilin loss on intermittent demand) at full precision, one row per result, sorted by setting: the lab result's ID, its run, the quantity, its conditions and its value (both as JSON). The evidence's tables give selected values to four significant digits.",
     "id": "A55"
    },
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "add",
    "path": "/artefacts/55",
    "value": {
     "kind": "figure",
     "function": "data_plot",
     "uri": "https://entrylib.org/p/2026.000013/v3/files/files/figures/RN15-msmape_intermittent.png",
     "version": "sha256:591a7ac9e2f7b9c1561fc3dfb53c9127c53609687cc9bcac9d33f07540a32390",
     "licence": "CC-BY-4.0",
     "availability": {
      "level": "public"
     },
     "shows": [
      "A55",
      "E34.T1"
     ],
     "description": "The forecast with the lowest expected score (the target) against the share of zero actuals, on intermittent demand: one panel per family of settings (Poisson counts, negative binomial counts, lognormal demand sizes with zeros), one line per measure (sMAPE, msMAPE with nu = 0.1, 0.5 and 1, the Suilin loss). Drawn by lab run RN15 from its own results (lab/runs/RN15/figures/msmape_intermittent.png).",
     "id": "A56"
    },
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/claims/1/summary",
    "value": "A forecast twice too high and one half too low score the same. For equal absolute errors the higher one scores better. This compares forecasts for one known actual. Which forecast is best when the actual is uncertain is claim C4: sMAPE does not in general reward forecasting too high.",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/claims/2/checks",
    "value": [
     {
      "by": "P3",
      "outcome": "confirmed",
      "extent": "full",
      "notes": "Derived from the statement alone (earlier wording: old H3 and H9), then compared with the final proofs.md (paper H3), which agrees; the bound in (b) follows from x - x^3/3 <= tanh x <= x. Concavity is in the absolute log error on [0, inf); 'minimised by the median' is the small-error approximation, as stated."
     },
     {
      "by": "P2",
      "outcome": "confirmed",
      "extent": "full",
      "notes": "The supervisor read the proof of paper H3 in proofs.md (AR1.R2; rendered as AR2.R2) and confirmed it: first \"Confirmed\" (DC6), then, after the last corrections of paper H4, H5 and H9, \"All looks fine now. NExt\"."
     },
     {
      "by": "P5",
      "outcome": "confirmed",
      "extent": "full",
      "notes": "Derived independently. With $z=\\ln y-\\ln f$, $s=|e^z-1|/(e^z+1)=\\tanh(|z|/2)$. (a) $\\tanh(u/2)$ on $u\\ge0$ is 0 at 0, strictly increasing, concave (derivative $\\tfrac12\\operatorname{sech}^2(u/2)$ positive and decreasing), below 1 with limit 1. (b) $x-x^3/3\\le\\tanh x\\le x$ for $x\\ge0$ (the lower bound: $\\frac{d}{dx}(\\tanh x-x+x^3/3)=x^2-\\tanh^2x\\ge0$); with $x=|z|/2$ this gives $0\\le|z|/2-s\\le|z|^3/24$, so $200\\cdot\\mathrm{mean}(s)\\approx100\\cdot\\mathrm{mean}|z|$. (c) Averaging (b): $\\bar s\\le a/2\\le\\bar s+k/24$; with $\\bar s(f^*)\\le\\bar s(m)$, $a(f^*)/2\\le\\bar s(f^*)+k(f^*)/24\\le\\bar s(m)+k(f^*)/24\\le a(m)/2+k(f^*)/24$, which is the stated bound; the log-MAE is minimised at any median of the $\\ln y_t$, i.e. at any median $m$. The proof (proofs.md H3) does exactly this; its only unproved step is the elementary inequality $x-x^3/3\\le\\tanh x$, closed by the derivative argument above. I recomputed the counterexample of step 8 (actuals $1,1,e,e^4,e^4$: sums 2.734 at the median $e$ against 2.390 at 1), which supports \"not in general at it\". Every part holds as worded."
     }
    ],
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/claims/2/hypothesis/lab_event",
    "value": "H36",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/claims/2/planned_at",
    "value": "2026-11-18T12:02:12Z",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/claims/2/statement",
    "value": "For $y > 0$ and $f > 0$, the [[D1:sMAPE term]] is $s = \\tanh(|\\ln y - \\ln f|/2)$ ([[D4:log error]] $\\ln y - \\ln f$). (a) It is 0 when $y = f$, strictly increasing and concave in $|\\ln y - \\ln f|$, and bounded by 1, which it approaches as $|\\ln y - \\ln f|$ grows. (b) $0 \\le \\tfrac12|\\ln y - \\ln f| - s \\le \\tfrac1{24}|\\ln y - \\ln f|^3$; so when all log errors are small, sMAPE is about 100 times the mean absolute error of $\\ln f$ as a forecast of $\\ln y$ ([[D19:log-MAE]]). (c) The log-MAE of a test set is minimised by its median; sMAPE is then minimised near the median, not in general at it. For positive actuals $y_1, \\dots, y_n$ with a median $m$, a forecast $f^*$ that minimises their sMAPE satisfies\n$$\\frac1n\\sum_t|\\ln y_t - \\ln f^*| - \\frac1n\\sum_t|\\ln y_t - \\ln m| \\le \\frac1{12}\\cdot\\frac1n\\sum_t|\\ln y_t - \\ln f^*|^3:$$\nits mean absolute log error exceeds the smallest possible by at most a twelfth of its mean cubed absolute log error. The minimiser is the median itself only under further conditions on the distribution of the actuals.",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/claims/2/summary",
    "value": "sMAPE measures the error on the log scale. For small errors it is half the absolute log error; for large errors it levels off. For small errors its best forecast is close to the median, but not in general the median.",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/claims/2/support",
    "value": [
     {
      "kind": "formal_proof",
      "evidence": "E20"
     },
     {
      "kind": "empirical_quant",
      "evidence": "E31"
     }
    ],
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/claims/2/test/lab_plan",
    "value": "PL12",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/claims/2/test/not_supported",
    "value": "A counterexample to the hypothesis as worded, found by the proof, the checker, the supervisor or the deciding run, that a correction of the proof cannot remove.",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/claims/2/test/supported",
    "value": "The proof of the hypothesis as worded here, in the study's proof documents (proofs.md, proof-suilin.md), is complete, every step naming what it rests on; the independent checker (an agent of another model than the author's: Claude Opus 5.5, registered as an actor before it checks) records a check of the hypothesis as worded with outcome confirmed; and the deciding run's numerical checks of its formulas meet the tolerances of the design.",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/claims/3/checks",
    "value": [
     {
      "by": "P3",
      "outcome": "disputed",
      "extent": "full",
      "notes": "Derived from the statement alone, then compared with the final proofs.md (paper H5). The balance, the weighted median, the properties of the weight and the 'above'/'below' cases hold. The last clause fails: equal weights at the median make it a stationary point, not necessarily the minimum, and proof step 7 does not show it. Counterexample (numerical): ln Y ~ 0.41408 N(-1.48848, 0.2^2) + 0.03 N(0, 0.3^2) + 0.55592 N(1.13795, 1) has median 1 and equal weights there, a flat inflection at the median, and a single local minimum at ln f = -1.36. Fix: drop 'at the median when the two weights are equal', or add 'if the median is a local minimum'.",
      "resolution": "The statement was corrected, and the checker confirmed the corrected wording in a later check listed here."
     },
     {
      "by": "P3",
      "outcome": "confirmed",
      "extent": "full",
      "notes": "Checked the corrected wording against my own derivation and the revised proofs.md (paper H5, new step 8). The clause disputed in CK5 now says that with equal weights the median is a flat point, the target only if it is a local minimum, and possibly a flat inflection; that is what holds, and my counterexample (single local minimum at ln f = -1.36, median at 1) is an instance. Needs Y positive with no atoms, as stated."
     },
     {
      "by": "P3",
      "outcome": "confirmed",
      "extent": "full",
      "notes": "Checked the added sufficient condition against my own derivation and the revised proofs.md (paper H5, steps 8 to 10). The second derivative of the expected term in ln f is g(c) - E[sech^2 tanh]/2, as I derived independently; zero first derivative plus the stated strict inequality gives a strict local minimum at the median, which is the target when there is a single local minimum. The clause needs ln Y to have a density continuous at ln m, as stated; my counterexample sits at equality, consistent with it."
     },
     {
      "by": "P2",
      "outcome": "confirmed",
      "extent": "full",
      "notes": "The supervisor read the proof of paper H5 in proofs.md (AR1.R2; rendered as AR2.R2) and confirmed it: first \"Confirmed\" (DC6), then, after the last corrections of paper H4, H5 and H9, \"All looks fine now. NExt\"."
     },
     {
      "by": "P5",
      "outcome": "confirmed",
      "extent": "full",
      "notes": "Derived independently. For $y\\ne f$, $\\partial_f s(y,f)=2w(y,f)(\\mathbf 1\\{y<f\\}-\\mathbf 1\\{y>f\\})$ with $w=y/(y+f)^2\\le1/(4f)$; since $P(Y=f)=0$ and the difference quotients are bounded by $1/(2(f-\\epsilon))$ near $f$, dominated convergence gives $\\frac{d}{df}\\mathrm E[s(Y,f)]=2\\mathrm E[w(\\mathbf 1\\{Y<f\\}-\\mathbf 1\\{Y>f\\})]$ at every $f>0$, so a minimiser (necessarily $f>0$: $f\\le0$ scores 1 on every positive actual while $\\mathrm E[s]<1$ for $f>0$) satisfies the stated balance. With $r=y/f$, $w=\\frac1f\\,r/(1+r)^2=\\frac1f/(r+2+1/r)$: symmetric under $r\\to1/r$, largest at $r=1$, decreasing as $r$ moves away from 1 on either side. The proof (proofs.md H5) matches step by step; the local uniformity of the dominating bound in $f$ is left implicit but is immediate as above. Its use of H4 (existence) is not needed for the \"only if\" statement but is harmless. Every part holds as worded."
     }
    ],
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/claims/3/depends_on",
    "value": [
     "C9"
    ],
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/claims/3/hypothesis/lab_event",
    "value": "H37",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/claims/3/planned_at",
    "value": "2026-11-18T12:02:12Z",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/claims/3/relations",
    "value": [
     {
      "type": "explains",
      "note": "C5 found sMAPE's preferred candidate close to the median on other distributions; C4 shows why: the target is a weighted median, on the side of the median that carries more weight (C12).",
      "reference": "B4"
     },
     {
      "type": "addresses",
      "reference": "B12",
      "note": "Tofallis (2015) states that minimising sMAPE does not produce an established measure of location; C4 identifies the target as a median weighted by y/(y+f)^2."
     },
     {
      "type": "qualifies",
      "reference": "B10",
      "note": "Hewamalage et al. (2023) say sMAPE may tend towards selecting a slightly overestimating model; C4 shows the target is a weighted median, which lies on the side of the median that carries more weight, above or below it (C12)."
     },
     {
      "type": "qualifies",
      "target": "C2",
      "id_type": "local",
      "note": "C2 compares two forecasts of one known actual. For an uncertain actual, the forecast sMAPE rewards is the weighted median of C4, not a forecast that is too high."
     }
    ],
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/claims/3/statement",
    "value": "Let the actual $y$ be positive and take no single value with positive probability. A forecast $f$ has the lowest expected sMAPE only if the actuals below $f$ and those above it carry equal total [[D5:weight]] $w(y,f) = y/(y+f)^2$:\n$$\\mathrm E[w(y,f)\\,\\mathbf 1\\{y<f\\}] = \\mathrm E[w(y,f)\\,\\mathbf 1\\{y>f\\}].$$\n So the [[D6:sMAPE target]] is a median of the actuals weighted by $w$, where the ordinary median gives every actual the same weight. An actual's weight is largest when it equals the forecast, is the same for $y/f = r$ and $1/r$, and falls as $y/f$ moves away from 1 in either direction.",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/claims/3/summary",
    "value": "For any distribution, the forecast sMAPE rewards balances the actuals below it and above it. Each actual is weighted by how close it is to the forecast by ratio.",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/claims/3/support",
    "value": [
     {
      "kind": "formal_proof",
      "evidence": "E22"
     },
     {
      "kind": "empirical_quant",
      "evidence": "E31"
     }
    ],
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/claims/3/test/lab_plan",
    "value": "PL12",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/claims/3/test/not_supported",
    "value": "A counterexample to the hypothesis as worded, found by the proof, the checker, the supervisor or the deciding run, that a correction of the proof cannot remove.",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/claims/3/test/supported",
    "value": "The proof of the hypothesis as worded here, in the study's proof documents (proofs.md, proof-suilin.md), is complete, every step naming what it rests on; the independent checker (an agent of another model than the author's: Claude Opus 5.5, registered as an actor before it checks) records a check of the hypothesis as worded with outcome confirmed; and the deciding run's numerical checks of its formulas meet the tolerances of the design.",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/claims/4/summary",
    "value": "When the logarithm of the actual is symmetric with a single peak, the weights balance at the median, and the median is the forecast sMAPE rewards. Lognormal and log-Laplace actuals are examples.",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/claims/5/depends_on",
    "value": [
     "C3",
     "C9",
     "C12"
    ],
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/claims/11/about/object",
    "value": "the ordinary median of the actuals",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/claims/11/about/relation",
    "value": "equals, or lies on a stated side of,",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/claims/11/checks",
    "value": [
     {
      "by": "P3",
      "outcome": "confirmed",
      "extent": "full",
      "notes": "Checked against my own derivations of H5 and H4 and the proofs in proofs.md (paper H6; paper H5 steps 3, 6, 8, 9). (a) holds even with unimodality in the weak sense, since an integrable density cannot be symmetric about two points; (b) holds because a zero first derivative and a positive second derivative in ln f make the median a strict local minimum, which is the target when there is a single local minimum. Caveat: (b) is 'more general' than (a) only where the density in (a) is continuous at ln m."
     },
     {
      "by": "P5",
      "outcome": "confirmed",
      "extent": "full",
      "notes": "Derived independently, on the log scale $\\phi(t)=\\mathrm E\\tanh(|\\ln Y-t|/2)$. (a) $\\phi'(\\ln m)$ has the sign of (weight below) minus (weight above). With a single local minimum, $\\phi$ is decreasing before it and increasing after it: if it were not monotone on one side, a continuous function that is below 1 and tends to 1 at both ends would have an interior local maximum and then a second local minimum. So a negative derivative at $m$ puts the target above $m$, a positive one below. The proof's step 2 says this tersely but correctly. (b) Writing the kernel $-\\operatorname{sign}(v)\\operatorname{sech}^2(v/2)=-\\operatorname{sign}(v)+\\operatorname{sign}(v)\\tanh^2(v/2)$, the first part gives $\\tfrac12(2G(t)-1)$, which is differentiable at $t$ with derivative $g(t)$ when $g$ is continuous at $t$, and the second has a bounded continuous derivative. So $\\phi''(\\ln m)=g(\\ln m)-\\tfrac12\\mathrm E[\\operatorname{sech}^2(x/2)\\tanh(|x|/2)]$ exists under continuity of $g$ at $\\ln m$ alone. With $\\phi'(\\ln m)=0$ and $\\phi''(\\ln m)>0$, $m$ is a strict local minimum, hence the target. This agrees with H12 step 4, which interchanges derivative and expectation without justification; the decomposition above closes that gap. The existence claim \"Without it, $m$ can be a flat point of inflection, with the target elsewhere\" is supported in the proof only by a reference to an example built by an earlier checker that is not given. I built one myself: $\\ln Y-\\ln m\\sim0.4061\\,N(-1.5,0.3^2)+0.0548\\,N(0,0.5^2)+0.5390\\,N(1.1587,1)$ has median 0, $\\phi'(0)=\\phi''(0)=0$, $\\phi'\\ge0$ on both sides of 0 (a flat inflection) and a single local minimum near $t=-1.25$ (numerical grid). Including such an explicit example in the proof would make step 6 self-contained. (Under the single-minimum hypothesis a failing inequality can only mean equality, $\\phi''=0$; a strict local maximum is impossible, so \"flat point of inflection\" is the right description.) (c) holds via H6. Note that H6 assumes unimodality \"strictly on a set of positive probability\", which H38 (c) does not state; the conclusion holds for every symmetric, weakly unimodal density anyway: $P(|X-t|\\le c)<P(|X|\\le c)$ for $t\\ne0$ on a set of $c$ of positive measure (otherwise $g$ would be constant on $(0,\\infty)$), and $\\tanh$ is strictly increasing, so no correction to the statement is needed. Lognormal and log-Laplace qualify. Every part holds as worded."
     }
    ],
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/claims/11/depends_on",
    "value": [
     "C3",
     "C4",
     "C5",
     "C9"
    ],
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/claims/11/hypothesis/lab_event",
    "value": "H38",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/claims/11/name",
    "value": "When sMAPE targets the ordinary median",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/claims/11/planned_at",
    "value": "2026-11-18T12:02:12Z",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/claims/11/statement",
    "value": "Let the actual $y$ be positive and take no single value with positive probability, let $m$ be its ordinary (unweighted) median, and let its expected sMAPE have a single local minimum. With the forecast at $m$, compare the total [[D5:weight]] $w(y,m) = y/(y+m)^2$ of the actuals above $m$ with that of the actuals below it. (a) If the actuals above $m$ carry more weight, the [[D6:sMAPE target]] lies above $m$; if those below carry more, it lies below $m$. (b) If the two weights are equal, $m$ is a point where expected sMAPE is flat. It is then the target if, in addition,\n$$2g(\\ln m) > \\mathrm E\\Big[\\operatorname{sech}^2\\Big(\\frac{\\ln y - \\ln m}{2}\\Big)\\tanh\\Big(\\frac{|\\ln y - \\ln m|}{2}\\Big)\\Big],$$\nwhere $g$ is the density of $\\ln y$, continuous at $\\ln m$: this makes $m$ a strict local minimum. Without it, $m$ can be a flat point of inflection, with the target elsewhere. (c) When $\\ln y$ has a density symmetric about $\\ln m$ and unimodal, $m$ is the only target; lognormal and log-Laplace actuals are examples.",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/claims/11/summary",
    "value": "The forecast sMAPE rewards lies above the ordinary median when the actuals above it carry more weight, and below it when those below do. When the weights balance and enough actuals lie near the median, the median is rewarded.",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/claims/11/support",
    "value": [
     {
      "kind": "formal_proof",
      "evidence": "E29"
     },
     {
      "kind": "empirical_quant",
      "evidence": "E31"
     }
    ],
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/claims/11/test/lab_plan",
    "value": "PL12",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/claims/11/test/not_supported",
    "value": "A counterexample to the hypothesis as worded, found by the proof, the checker, the supervisor or the deciding run, that a correction of the proof cannot remove.",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/claims/11/test/supported",
    "value": "The proof of the hypothesis as worded here, in the study's proof documents (proofs.md, proof-suilin.md), is complete, every step naming what it rests on; the independent checker (an agent of another model than the author's: Claude Opus 5.5, registered as an actor before it checks) records a check of the hypothesis as worded with outcome confirmed; and the deciding run's numerical checks of its formulas meet the tolerances of the design.",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/claims/12/relations",
    "value": [
     {
      "type": "supports",
      "target": "C4",
      "id_type": "local",
      "note": "Computed on 34 distributions, the target lies on the side of the median that carries more weight, as C4 and C12 prove.",
      "results": [
       "E13.T1.r6",
       "E13.T1.r15",
       "E13.T1.r18",
       "E13.T3.r6",
       "E13.T3.r15",
       "E13.T3.r18"
      ]
     },
     {
      "type": "supports",
      "target": "C5",
      "id_type": "local",
      "note": "Computed on lognormal outcomes, the target is the median, as C5 proves.",
      "results": [
       "E13.T1.r13",
       "E13.T1.r20",
       "E13.T3.r13",
       "E13.T3.r20"
      ]
     },
     {
      "type": "supports",
      "target": "C6",
      "id_type": "local",
      "note": "Computed on bimodal outcomes, the target is two forecasts, one near each mode, when the modes are equal, and the larger mode's when they are not, as C6 proves.",
      "results": [
       "E13.T1.r21",
       "E13.T1.r22",
       "E13.T1.r23",
       "E13.T2.r1",
       "E13.T3.r21",
       "E13.T3.r22",
       "E13.T3.r23"
      ]
     },
     {
      "type": "supports",
      "target": "C7",
      "id_type": "local",
      "note": "Computed on lognormal outcomes with zeros, the target switches to 0 at a share of zeros below one half, as C7 proves.",
      "results": [
       "E13.T1.r27",
       "E13.T1.r28",
       "E13.T1.r29",
       "E13.T1.r30",
       "E13.T1.r31",
       "E13.T1.r32",
       "E13.T1.r33",
       "E13.T1.r34",
       "E13.T3.r27",
       "E13.T3.r28",
       "E13.T3.r29",
       "E13.T3.r30",
       "E13.T3.r31",
       "E13.T3.r32",
       "E13.T3.r33",
       "E13.T3.r34"
      ]
     },
     {
      "type": "supports",
      "target": "C11",
      "id_type": "local",
      "note": "Computed on normal outcomes straddling zero, the target lies away from zero, on the side of the mean, as C11 proves.",
      "results": [
       "E13.T1.r24",
       "E13.T1.r25",
       "E13.T1.r26",
       "E13.T2.r2",
       "E13.T2.r3",
       "E13.T2.r4",
       "E13.T3.r24",
       "E13.T3.r25",
       "E13.T3.r26"
      ]
     }
    ],
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/claims/15/estimate/estimands/difference/variable",
    "value": "relative difference of the Suilin target from the sMAPE target",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/claims/15/name",
    "value": "The Suilin target beside the sMAPE target",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/claims/15/relations",
    "value": [
     {
      "type": "supports",
      "target": "C17",
      "id_type": "local",
      "note": "Computed on actual distributions, the Suilin target stays at the sMAPE target away from zero and moves near zero, as C17's two cases imply.",
      "results": [
       "E16.T1",
       "E16.T2"
      ]
     }
    ],
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/claims/15/statement",
    "value": "Computed from the distributions themselves, the [[D29:Suilin target]] differs little from the [[D6:sMAPE target]] away from zero and on counts, and differs near zero: it is within 0.3% of the sMAPE target on the six distributions with median about 100, and the same as the sMAPE target on every [[D30:Poisson counts]] distribution studied; on lognormal actuals with median 1 and $\\sigma = 1$ it is 0.985 instead of 1, and the share of zeros at which it switches to the forecast 0 is 0.470 instead of sMAPE's 0.394, while at median 100 the two shares are 0.3937 and 0.3940.",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/claims/15/summary",
    "value": "Computed on actual distributions, the forecast the Suilin loss rewards is almost the same as sMAPE's away from zero and on counts. It differs near zero, where the Suilin loss needs a larger share of zeros before it rewards a zero forecast.",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/claims/16/checks",
    "value": [
     {
      "by": "P3",
      "outcome": "confirmed",
      "extent": "full",
      "notes": "Derived from the statement alone, then compared with proof-msmape.md. Write t = |y|+|f|. The denominator max(t+eps, 0.5+eps) is t+eps exactly when t >= 0.5, and the two branches agree at t = 0.5 (both 0.6), so m is continuous there. (a) For t >= 0.5, t > 0, so |y-f| = s t and m = s t/(t+eps); the factor t/(t+eps) = 1 - eps/(t+eps) is increasing, equals 5/6 at t = 0.5, is strictly below 1 and tends to 1, so it lies in [5/6, 1): the bound 5/6 is attained, 1 is not. s - m = s eps/(t+eps) <= eps/(t+eps), so m is close to s when t is large compared with eps. m is not a function of y/f: (1,2) and (10,20) have the same ratio and the same s = 1/3, but m = 1/3.1 and 10/30.1. On counts, y >= 1 gives t >= 1, and y = 0 gives t = |f|, so only a zero actual with |f| < 0.5 falls outside this case; this holds for forecasts of any sign and any real value. (b) For t < 0.5, m = |y-f|/0.6; m(0,0) = 0; m(0,f) = |f|/0.6 for |f| < 0.5; s(0,f) = |f|/|f| = 1 for every f != 0, and 1 is the largest value of s because |y-f| <= |y|+|f|. (c) m >= 0, and |y-f| <= t < t+eps <= denominator gives m < 1 strictly in both branches, with supremum 1 not attained (opposite signs, t large); a bounded term has a finite expectation under every joint distribution of actuals and forecasts, so expected msMAPE lies in [0, 200). Checked with exact rational arithmetic on 400000 random pairs of mixed signs and scales, including zeros and the boundary t = 0.5: no violation. Every step of proof-msmape.md (1 to 6) holds. Three clauses of the recorded statement are true but have no step in the proof, whose restated Statement omits them: that m is close to s for values large compared with eps (needs the bound s - m <= eps/(t+eps)), that m is no longer exactly a function of y/f (needs one counterexample such as the one above), and that 1 is the largest value of the sMAPE term (follows from the triangle inequality used in step 6 but is not said). The opening sentence, that m has the properties of the sMAPE term except near zero, is a summary and holds only as parts (a) to (c) qualify it: exact dependence on the ratio y/f alone is lost at every scale, not only near zero, and m never reaches the value 1 that s reaches for opposite signs or a zero actual."
     },
     {
      "by": "P5",
      "outcome": "confirmed",
      "extent": "full",
      "notes": "Derived independently. With $\\ell=|y|+|f|$, the denominator $\\max(\\ell+\\delta,0.5+\\delta)$ is $\\ell+\\delta$ exactly when $\\ell\\ge0.5$. (a) Then $\\ell>0$, $|y-f|=s\\ell$, $\\tilde s=s\\,\\ell/(\\ell+\\delta)$, and $\\ell/(\\ell+\\delta)$ is increasing from $0.5/0.6=5/6$ towards 1, never reaching 1. It depends on scale: $(1,2)$ and $(10,20)$ have equal $s$ but different $\\tilde s$. On counts, $y\\ge1$ gives $\\ell\\ge1$, and $y=0$ gives $\\ell=|f|$, so only a zero actual with $|f|<0.5$ falls outside. (b) For $\\ell<0.5$, $\\tilde s=|y-f|/0.6$; $(0,0)$ scores 0, $(0,f)$ scores $|f|/0.6$, and $s(0,f)=1=\\max s$ for $f\\ne0$. (c) In case (a), $\\tilde s\\le s\\cdot\\ell/(\\ell+\\delta)<1$ since $s\\le1$. In case (b), $\\tilde s\\le\\ell/0.6<5/6$. And $\\tilde s\\ge0$, so every expectation is finite. The proof (proof-suilin.md, first proof) matches each step. Every part holds as worded."
     }
    ],
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/claims/16/hypothesis/lab_event",
    "value": "H41",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/claims/16/name",
    "value": "The Suilin loss behaves like sMAPE except near zero",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/claims/16/planned_at",
    "value": "2026-11-18T12:02:12Z",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/claims/16/relations",
    "value": [
     {
      "type": "extends",
      "reference": "B14",
      "note": "Hewamalage, Bergmeir and Bandara (2021) evaluate forecasts with the Suilin loss, Suilin's modification of sMAPE for series with values near zero; C17 shows that it keeps sMAPE's behaviour away from zero and becomes an absolute error near zero."
     }
    ],
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/claims/16/statement",
    "value": "The [[D28:Suilin term]] has the properties of the [[D1:sMAPE term]] except near zero. (a) Where $|y|+|f| \\ge 0.5$,\n$$\\tilde s(y,f) = s(y,f)\\,\\frac{|y|+|f|}{|y|+|f|+\\delta}:$$\nthe sMAPE term shrunk by a factor between $5/6$ and 1 that tends to 1 as $|y|+|f|$ grows. So for actuals and forecasts large compared with $\\delta$, $\\tilde s$ is close to $s$, though no longer exactly a function of the ratio $y/f$; on counts, where every actual is 0 or at least 1, this case covers every pair except a zero actual with $|f| < 0.5$. (b) Where $|y|+|f| < 0.5$, $\\tilde s(y,f) = |y-f|/(0.5+\\delta)$, an absolute error: a zero forecast of a zero actual scores 0, and a forecast $f$ with $|f| < 0.5$ of a zero actual scores $|f|/(0.5+\\delta)$, where the sMAPE term gives every non-zero forecast of a zero actual its largest value, 1. (c) $0 \\le \\tilde s < 1$ everywhere, so the expected Suilin loss of every forecast is finite.",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/claims/16/summary",
    "value": "The Suilin loss behaves like sMAPE except near zero, where it becomes an absolute error.",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/claims/16/support",
    "value": [
     {
      "kind": "formal_proof",
      "evidence": "E30"
     },
     {
      "kind": "empirical_quant",
      "evidence": "E31"
     }
    ],
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/claims/16/test/lab_plan",
    "value": "PL12",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/claims/16/test/not_supported",
    "value": "A counterexample to the hypothesis as worded, found by the proof, the checker, the supervisor or the deciding run, that a correction of the proof cannot remove.",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/claims/16/test/supported",
    "value": "The proof of the hypothesis as worded here, in the study's proof documents (proofs.md, proof-suilin.md), is complete, every step naming what it rests on; the independent checker (an agent of another model than the author's: Claude Opus 5.5, registered as an actor before it checks) records a check of the hypothesis as worded with outcome confirmed; and the deciding run's numerical checks of its formulas meet the tolerances of the design.",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "add",
    "path": "/claims/17",
    "value": {
     "id": "C18",
     "name": "sMAPE loses its power to tell forecasts apart as the actuals spread out",
     "summary": "For widely spread actuals a test set tells two forecasts apart less and less well under sMAPE: the signal-to-noise ratio of their score difference falls like one over the square root of the spread. The absolute log error and MAE keep their power.",
     "statement": "Let $\\ln y = \\mu + \\sigma\\varepsilon$, where $\\varepsilon$ has a fixed density $h$ with bounded continuous first and second derivatives and distribution function $H$, and compare two forecasts at fixed quantile levels, $f_1 = e^{\\mu+\\sigma c_1}$ and $f_2 = e^{\\mu+\\sigma c_2}$ with $c_1 < c_2$, by their [[D32:signal-to-noise ratio of a comparison]]. (a) For the [[D1:sMAPE term]], as $\\sigma$ grows,\n$$\\mathrm E[D] = \\frac{4\\ln 2\\,(h(c_1) - h(c_2))}{\\sigma} + O(\\sigma^{-3}), \\qquad \\operatorname{Var}(D) = \\frac{4(2\\ln 2 - 1)(h(c_1)+h(c_2))}{\\sigma} + O(\\sigma^{-2}).$$\nSo when $h(c_1) \\ne h(c_2)$ the ratio is\n$$\\frac{2\\ln 2}{\\sqrt{2\\ln 2 - 1}}\\,\\frac{|h(c_1)-h(c_2)|}{\\sqrt{h(c_1)+h(c_2)}}\\,\\sigma^{-1/2}\\,(1 + O(1/\\sigma)):$$\nit falls to zero like $\\sigma^{-1/2}$, and the size of the test set that keeps the mean difference a given number of standard errors from zero grows in proportion to $\\sigma$. (b) For the absolute log error $|\\ln y - \\ln f|$ ([[D19:log-MAE]]), $D = \\sigma(|\\varepsilon - c_2| - |\\varepsilon - c_1|)$, so the ratio, wherever $\\mathrm{sd}(D) > 0$, does not depend on $\\sigma$. (c) For the absolute error $|y - f|$ ([[D11:MAE]]), as $\\sigma$ grows the ratio tends to\n$$\\frac{|2H(c_2) - 1|}{2\\sqrt{H(c_2)(1 - H(c_2))}}$$\nwhen $0 < H(c_2) < 1$; the limit is positive unless the higher forecast is the median. (d) For the squared error $(y-f)^2$ ([[D10:RMSE]]), when $\\mathrm E[y^2]$ is finite the ratio is $|2\\,\\mathrm E[y] - f_1 - f_2|/(2\\,\\mathrm{sd}(y))$; for lognormal actuals, $e^{\\sigma^2/2}$ times the ratio tends to 1 as $\\sigma$ grows. For widely spread actuals, then: the absolute log error keeps its power to tell two forecasts apart wherever its ratio is defined. The absolute error keeps it when $0 < H(c_2) < 1$ and the higher forecast is not the median. sMAPE loses it at the rate $\\sigma^{-1/2}$ when $h(c_1) \\ne h(c_2)$. It loses it faster, with a ratio of at most order $\\sigma^{-5/2}$, when $h(c_1) = h(c_2) > 0$. The squared error on lognormal actuals loses it faster than any power of $\\sigma$.",
     "about": {
      "subject": "D32",
      "relation": "falls like the inverse square root of the spread of log actuals for",
      "object": "D1",
      "comparator": "D19",
      "conditions": []
     },
     "type": "formal",
     "role": "principal",
     "novelty": "no_prior_found",
     "support": [
      {
       "kind": "formal_proof",
       "evidence": "E35"
      },
      {
       "kind": "empirical_quant",
       "evidence": "E32"
      }
     ],
     "provenance": "author_stated",
     "hypothesis": {
      "lab_event": "H46",
      "outcome": "supported"
     },
     "planning": "confirmatory",
     "planned_at": "2026-11-18T12:29:29Z",
     "planned_at_anchor": "git_remote",
     "test": {
      "framework": "formal",
      "supported": "The proof of the hypothesis as worded here (proofs.md, section H13, revised to it) is complete; the independent checker AC6 (Claude Opus 5.5) records a check of this wording with outcome confirmed; and the deciding run's checks meet the tolerances of PL13.",
      "not_supported": "A counterexample to the hypothesis as worded here, found by the proof, the checker, the supervisor or the deciding run, that a correction of the proof cannot remove.",
      "verdict": "judged",
      "runs": 1,
      "lab_plan": "PL17"
     },
     "checks": [
      {
       "by": "P5",
       "outcome": "disputed",
       "extent": "full",
       "notes": "Derived independently. (a) With $q(v)=1-\\tanh(|v|/2)=2/(1+e^{|v|})$, $\\mathrm E[q(\\sigma(\\varepsilon-c))]=\\frac1\\sigma\\int q(u)h(c+u/\\sigma)du=4\\ln2\\,h(c)/\\sigma+O(\\sigma^{-3})$ (the $h'$ term vanishes by symmetry; the remainder is bounded by $\\tfrac12\\sup|h''|\\int u^2q/\\sigma^3$). Likewise $\\mathrm E[q^2]=4(2\\ln2-1)h(c)/\\sigma+O(\\sigma^{-3})$ with $\\int q^2=8(\\ln2-\\tfrac12)$; the cross term is at most $4e^{-\\sigma(c_2-c_1)}$; $(\\mathrm E D)^2=O(\\sigma^{-2})$. So both expansions and their remainder orders hold, and the ratio is as stated with relative error $O(1/\\sigma)$ (from the variance), with $h(c_1)+h(c_2)>0$ when $h(c_1)\\ne h(c_2)$. A quadrature check (normal $h$, $c_1=-0.3$, $c_2=0.8$, $\\sigma$ up to 160) agrees. (b) is immediate; $D$ is bounded by $\\sigma(c_2-c_1)$. (c) $D/f_2\\to\\mathbf 1\\{\\varepsilon<c_2\\}-\\mathbf 1\\{\\varepsilon>c_2\\}$ a.s. with $|D/f_2|\\le1$, so $\\mathrm E[D/f_2]\\to2H(c_2)-1$ and $\\operatorname{Var}(D/f_2)\\to4H(c_2)(1-H(c_2))$. The stated limit therefore holds only when $0<H(c_2)<1$. The statement does not give this condition, and the proof (H13 step 8) adds it. The hypotheses allow it to fail: $h$ may have bounded support, e.g. $h\\propto(1-x^2)^3$ on $[-1,1]$, which has bounded continuous $h'$ and $h''$. If $H(c_2)=1$ the formula is $1/0$ (when $c_1$ is inside the support the ratio tends to infinity; when it is not, $D$ is constant and the ratio is undefined for every $\\sigma$). If $H(c_2)=0$, $D$ is constant and the ratio is undefined for every $\\sigma$. (d) $D=(f_2-f_1)(f_1+f_2-2Y)$ gives the ratio exactly; for lognormal actuals $e^{\\sigma^2/2}$ times the ratio is $|1-(f_1+f_2)/(2\\mathrm EY)|\\,e^{\\sigma^2/2}/\\sqrt{e^{\\sigma^2}-1}\\to1$. Every computation in the proof is correct. The dispute concerns only the missing condition in (c). Suggested wording: \"(c) ... as $\\sigma$ grows the ratio tends to $|2H(c_2)-1|/(2\\sqrt{H(c_2)(1-H(c_2))})$ when $0<H(c_2)<1$, which is positive unless the higher forecast is the median.\" Optionally, also in (b): \"wherever $\\mathrm{sd}(D)>0$\" (it is 0 when $\\varepsilon\\le c_1$ a.s. or $\\varepsilon\\ge c_2$ a.s.). The closing sentence is also loose: sMAPE loses power at the rate $\\sigma^{-1/2}$ only when $h(c_1)\\ne h(c_2)$ (otherwise faster), and the absolute error keeps it only when $H(c_2)\\ne1/2$. It would be safer as \"for widely spread actuals, comparisons of forecasts with $h(c_1)\\ne h(c_2)$ ...\".",
       "resolution": "The statement was corrected, and the checker confirmed the corrected wording in a later check listed here."
      },
      {
       "by": "P5",
       "outcome": "disputed",
       "extent": "full",
       "notes": "Re-checked every part as now worded, against the revised H13. (a) is unchanged from H39 and correct: $\\mathrm E[q]=4\\ln2\\,h(c)/\\sigma+O(\\sigma^{-3})$ and $\\mathrm E[q^2]=4(2\\ln2-1)h(c)/\\sigma+O(\\sigma^{-3})$; the cross term is at most $4e^{-\\sigma(c_2-c_1)}$; the ratio is as stated with relative error $O(1/\\sigma)$. (b), \"wherever $\\mathrm{sd}(D)>0$\", is correct. (c), now with \"when $0<H(c_2)<1$\", is correct: $D/f_2\\to\\mathbf 1\\{\\varepsilon<c_2\\}-\\mathbf 1\\{\\varepsilon>c_2\\}$ boundedly, and quadrature agrees. (d) is correct. One point remains, in the closing sentence: \"sMAPE loses it, at the rate $\\sigma^{-1/2}$ when $h(c_1)\\ne h(c_2)$\". This asserts that sMAPE loses its power in every case, but neither the statement nor the proof (whose closing paragraph repeats it) shows this when $h(c_1)=h(c_2)$. If $h(c_1)=h(c_2)>0$, it follows from (a) in one line: $|\\mathrm E D|=O(\\sigma^{-3})$ and $\\operatorname{Var}D\\sim8(2\\ln2-1)h(c_1)/\\sigma$, so the ratio is $O(\\sigma^{-5/2})$. The proof should say this. If $h(c_1)=h(c_2)=0$ (possible when $h$ vanishes at a point or has bounded support), (a) gives only $\\mathrm E D=O(\\sigma^{-3})$ and $\\operatorname{Var}D=O(\\sigma^{-2})$, which do not bound the ratio. I could not derive the claim in that case. It is true in the cases I computed: $h\\propto(1-x^2)^3$ with $(c_1,c_2)=(1.5,2)$ or $(-1.5,2)$ gives a ratio falling like $\\sigma^{-2}$; $(-1.3,1.3)$ gives 0. The gap is my own doing: my note on H39 suggested \"otherwise faster\" without having proved the zero-density case. Suggested wording: \"sMAPE loses it when $h(c_1)+h(c_2)>0$, at the rate $\\sigma^{-1/2}$ when $h(c_1)\\ne h(c_2)$ and faster when $h(c_1)=h(c_2)$\", with the one-line argument above added to the proof. Alternatively, keep only \"sMAPE loses it at the rate $\\sigma^{-1/2}$ when $h(c_1)\\ne h(c_2)$\". Smaller and optional: \"the absolute error keeps it unless the higher forecast is the median\" omits the condition $0<H(c_2)<1$ of (c) (with $H(c_2)=0$ the ratio is undefined); adding \"when $0<H(c_2)<1$\" would make it exact. With these changes I would confirm.",
       "resolution": "The statement was corrected, and the checker confirmed the corrected wording in a later check listed here."
      },
      {
       "by": "P5",
       "outcome": "disputed",
       "extent": "full",
       "notes": "Parts (a)-(d) are unchanged from H43 and correct (checked in CK30 and CK34). I judged each sentence of the new closing against them. (1) The absolute log error keeps its power wherever its ratio is defined: this follows from (b). (2) The absolute error keeps it when $0<H(c_2)<1$ and the higher forecast is not the median: this follows from (c), whose limit is then positive. (3) sMAPE loses it at the rate $\\sigma^{-1/2}$ when $h(c_1)\\ne h(c_2)$: this follows from (a). (5) The squared error on lognormal actuals: (d) gives $e^{\\sigma^2/2}\\cdot\\text{ratio}\\to1$, so the ratio is $e^{-\\sigma^2/2}(1+o(1))=o(\\sigma^{-k})$ for every $k$, for every $c_1<c_2$; this is fully covered by (d), and lognormal actuals (normal $h$) meet the hypotheses. (4), \"It loses it faster, with a ratio of order $\\sigma^{-5/2}$, when $h(c_1)=h(c_2)>0$\", is not established as worded. (a) gives $\\mathrm E[D]=O(\\sigma^{-3})$, an upper bound only, and $\\operatorname{Var}(D)=8(2\\ln2-1)h(c_1)/\\sigma+O(\\sigma^{-2})$, which is exactly of order $1/\\sigma$. So the ratio is $O(\\sigma^{-5/2})$: at most of order $\\sigma^{-5/2}$, which does establish \"loses it faster\". But \"a ratio of order $\\sigma^{-5/2}$\" naturally reads as the exact order, and that is false in general. The $\\sigma^{-3}$ coefficient of $\\mathrm E[D]$ is proportional to $h''(c_1)-h''(c_2)$, which can vanish: for a symmetric $h$ with $c_1=-c_2$, $\\mathrm E[D]=0$ exactly and the ratio is 0 for every $\\sigma$. The proof's new sentence in step 6 has the same slip: \"the mean is of order $\\sigma^{-3}$ (step 2)\", whereas step 2 gives only $O(\\sigma^{-3})$. Suggested wording: \"It loses it faster, with a ratio of at most order $\\sigma^{-5/2}$ ($O(\\sigma^{-5/2})$), when $h(c_1)=h(c_2)>0$\", and in proof step 6 \"the mean is $O(\\sigma^{-3})$ ... so the ratio is $O(\\sigma^{-5/2})$\". With that change I would confirm. Separately, I checked the revised H12 step 6 (behind H38): the weights 0.406125, 0.054825 and 0.539050 now sum to 1; the parameters are correctly marked as rounded; and \"rises through $m$, where it is flat\" matches my computation ($\\phi'\\ge0$ on both sides of 0). It is now right.",
       "resolution": "The statement was corrected, and the checker confirmed the corrected wording in a later check listed here."
      },
      {
       "by": "P5",
       "outcome": "confirmed",
       "extent": "full",
       "notes": "A word diff against H45 shows that the only change is \"at most\" in the sentence I disputed in CK36. Parts (a)-(d) are correct, as checked in CK30, CK34 and CK36. I judged each closing sentence against them. The absolute log error keeps its power wherever its ratio is defined: this follows from (b). The absolute error keeps it when $0<H(c_2)<1$ and the higher forecast is not the median: this follows from (c). sMAPE loses it at the rate $\\sigma^{-1/2}$ when $h(c_1)\\ne h(c_2)$: this follows from (a). \"It loses it faster, with a ratio of at most order $\\sigma^{-5/2}$, when $h(c_1)=h(c_2)>0$\" is now exactly what (a) gives. The mean is $O(\\sigma^{-3})$ and the variance is $8(2\\ln2-1)h(c_1)/\\sigma+O(\\sigma^{-2})$, exactly of order $1/\\sigma$, so the ratio is $O(\\sigma^{-5/2})$; the bound allows the ratio to be smaller, even 0 (symmetric $h$, $c_1=-c_2$). The squared error on lognormal actuals loses its power faster than any power of $\\sigma$: by (d) the ratio is $e^{-\\sigma^2/2}(1+o(1))$, for every $c_1<c_2$. The revised proof step 6 writes $O(\\sigma^{-3})$ and $O(\\sigma^{-5/2})$ (\"at most of that order\"), and the closing paragraph matches the statement. Every part and every closing sentence holds as worded, and the proof establishes them."
      }
     ],
     "verification": "agent_checked",
     "relations": [
      {
       "type": "supports",
       "note": "C6 recommends against sMAPE for point forecasts; C18 shows that on widely spread actuals a test set tells two forecasts apart less and less well under sMAPE.",
       "reference": "B6"
      },
      {
       "type": "supports",
       "reference": "B13",
       "note": "Kolassa (2026) concludes that sMAPE should not be used; C18 shows that on widely spread actuals a test set tells two forecasts apart less and less well under sMAPE."
      },
      {
       "type": "refines",
       "target": "C10",
       "id_type": "local",
       "note": "C10 shows that the scores of forecasts at fixed quantile levels move together like 1/σ. C18 adds the noise and shows that the signal-to-noise ratio falls like σ^(-1/2), while that of the absolute log error does not change."
      }
     ],
     "depends_on": [
      "C3",
      "C10"
     ]
    },
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "add",
    "path": "/claims/18",
    "value": {
     "id": "C19",
     "name": "The sMAPE minimiser in large samples",
     "summary": "Where the log of the actual is symmetric with one peak, the forecast that minimises sMAPE on a test set is asymptotically normal about the median. It is asymptotically as precise as the sample median for narrow distributions and loses its precision as the actuals spread out.",
     "statement": "Let $\\ln y$ have a density $g$ that is symmetric about $\\ln m$, unimodal ($g(\\ln m + v_1) \\ge g(\\ln m + v_2)$ whenever $|v_1| \\le |v_2|$, strictly on a set of positive probability) and continuous at $\\ln m$. Write $x = \\ln y - \\ln m$, $A = \\mathrm E[\\operatorname{sech}^2(x/2)\\tanh(|x|/2)]$ and $B = \\mathrm E[\\operatorname{sech}^4(x/2)]$. (a) $A < 2g(\\ln m)$. (b) Over test sets of $n$ independent actuals, the [[D8:sMAPE minimiser of a test set]] $\\hat f_n$ (a forecast with the lowest sMAPE of the test set) tends to $m$ in probability, and $\\sqrt n(\\ln\\hat f_n - \\ln m)$ is asymptotically normal with mean 0 and variance $B/(2g(\\ln m) - A)^2$; for the sample median the variance is $1/(4g(\\ln m)^2)$. The ratio of these two limiting variances, the asymptotic relative efficiency of the sMAPE minimiser to the sample median, is\n$$\\mathrm{ARE} = \\frac{(2g(\\ln m) - A)^2}{4\\,g(\\ln m)^2\\,B};$$\nthat the [[D9:relative efficiency]] of finite test sets tends to it is not shown here, since the variances of finite test sets need not tend to those of the limits. (c) For lognormal actuals with standard deviation $\\sigma$ of $\\ln y$: $\\mathrm{ARE} \\to 1$ as $\\sigma \\to 0$ and $\\mathrm{ARE}\\,\\sigma^3 \\to 6(\\ln 2)^2\\sqrt{2\\pi} \\approx 7.23$ as $\\sigma \\to \\infty$. For log-Laplace actuals with scale $b$ of $\\ln y$: $\\mathrm{ARE} \\to 1$ as $b \\to 0$ and $\\mathrm{ARE}\\,b \\to 3$ as $b \\to \\infty$. So on both families the sMAPE minimiser is asymptotically as precise as the sample median for narrow distributions and asymptotically loses all of its precision relative to it as the actuals spread out.",
     "about": {
      "subject": "D8",
      "relation": "is asymptotically normal about the median, with a stated variance and",
      "object": "D9",
      "comparator": "the sample median",
      "conditions": []
     },
     "type": "formal",
     "role": "principal",
     "novelty": "no_prior_found",
     "support": [
      {
       "kind": "formal_proof",
       "evidence": "E36"
      },
      {
       "kind": "empirical_quant",
       "evidence": "E33"
      },
      {
       "kind": "cited",
       "reference": "B15",
       "locator": "Theorems 5.7 and 5.23"
      }
     ],
     "provenance": "author_stated",
     "hypothesis": {
      "lab_event": "H44",
      "outcome": "supported"
     },
     "planning": "confirmatory",
     "planned_at": "2026-11-18T12:19:07Z",
     "planned_at_anchor": "git_remote",
     "test": {
      "framework": "formal",
      "supported": "The proof of the hypothesis as worded here (proofs.md, section H13 for the signal-to-noise ratios, H14 for the large-sample efficiency, revised to the corrected wording) is complete; the independent checker AC6 (Claude Opus 5.5) records a check of the corrected wording with outcome confirmed; and the deciding run's checks meet the tolerances of the design.",
      "not_supported": "A counterexample to the hypothesis as worded here, found by the proof, the checker, the supervisor or the deciding run, that a correction of the proof cannot remove.",
      "verdict": "judged",
      "runs": 1,
      "lab_plan": "PL15"
     },
     "checks": [
      {
       "by": "P5",
       "outcome": "disputed",
       "extent": "full",
       "notes": "Derived independently, with $x=\\ln Y-\\ln m$, $\\theta=\\ln f-\\ln m$, $M(\\theta)=\\mathrm E\\tanh(|x-\\theta|/2)$. (a) $k(u)=\\operatorname{sech}^2(u/2)\\tanh(|u|/2)$ is positive for $u\\ne0$ with $\\int k=2$, and $g(\\ln m+u)\\le g(\\ln m)$ with strict inequality on a set of positive Lebesgue measure (a density cannot be constant on the whole line), so $A<2g(\\ln m)$ strictly. This holds even without the \"positive probability\" clause (e.g. for a uniform $g$), so the hypothesis is stronger than needed. (b) Consistency: the proof's direct argument (step 3: a grid on $[-R,R]$ plus the $\\tfrac12$-Lipschitz bound; outside, $M_n\\ge\\tanh(R/4)P_n(|x|\\le R/2)$) is valid. A minimiser of $M_n$ exists because $M_n$ is continuous, tends to 1 at $\\pm\\infty$ and is below 1 at any data point; this is not stated. van der Vaart Thm 5.23 conditions: $|m_{\\theta_1}-m_{\\theta_2}|\\le\\tfrac12|\\theta_1-\\theta_2|$ (constant, square integrable); differentiable at 0 for every $x\\ne0$; second-order expansion $M(\\theta)=M(0)+\\tfrac12V\\theta^2+o(\\theta^2)$ with $V=g(\\ln m)-A/2>0$. I re-derived H12 step 4: writing the kernel as $-\\operatorname{sign}+\\operatorname{sign}\\cdot\\tanh^2$, $M'$ exists everywhere and is differentiable at 0 with $M''(0)=g(\\ln m)-A/2$ whenever $g$ is continuous at $\\ln m$, and $M(\\theta)-M(0)=\\int_0^\\theta M'$ gives the expansion. So continuity at $\\ln m$ suffices. Variance $(B/4)/(V^2)=B/(2g-A)^2$; sample median $1/(4g^2)$; ratio $(2g-A)^2/(4g^2B)$: all correct. (c) Recomputed: $\\int u^2k=16\\ln2$, $\\int\\operatorname{sech}^4(u/2)=8/3$, $\\int|u|k=4$; lognormal $\\mathrm{ARE}\\,\\sigma^3\\to6(\\ln2)^2\\sqrt{2\\pi}\\approx7.226$ with relative error $O(\\sigma^{-2})$; log-Laplace $2g-A=2/b^2+O(b^{-3})$, $B=4/(3b)+O(b^{-2})$, so $\\mathrm{ARE}\\,b=3+O(1/b)$ (the proof's orders are right; quadrature gives $b(\\mathrm{ARE}\\,b-3)\\approx-5.6$). Both narrow limits are 1. Quadrature agrees with all four limits. The one statement I cannot confirm as worded is the appositive in (b): \"The ratio of the two, the large-sample value of the [[TM9:relative efficiency]], is ARE\". TM9 defines $\\mathrm{RE}(n)$ as a ratio of finite-sample variances, so its \"large-sample value\" reads as $\\lim_n\\mathrm{RE}(n)$. The proof establishes only the ratio of the variances of the limiting normal distributions. Convergence in distribution does not give convergence of $n\\operatorname{Var}$, and under the hypotheses of H40 the limit can fail. For example, a symmetric, decreasing, continuous $g$ with $P(|x|>t)\\sim1/\\ln t$ makes the sample median's variance infinite for every $n$, so $\\mathrm{RE}(n)$ never approaches the finite ARE. For lognormal and log-Laplace actuals the identification is very likely true, but it would need a uniform-integrability argument that is not given. Suggested wording: \"The ratio of the two asymptotic variances, the asymptotic relative efficiency (the ratio of the variances of the limiting distributions; it is the limit of the relative efficiency only where $n$ times each estimator's variance converges to its asymptotic variance, which is not shown here), is ARE = ...\". The closing \"as precise as ... loses all of its precision\" should then say \"asymptotically\". With that rewording I would confirm every part.",
       "resolution": "The statement was corrected, and the checker confirmed the corrected wording in a later check listed here."
      },
      {
       "by": "P5",
       "outcome": "confirmed",
       "extent": "full",
       "notes": "Re-checked every part as now worded, against the revised H14. (a): $\\int k=2$, $k>0$ off 0, and $g(\\ln m+u)<g(\\ln m)$ on a set of positive measure, so $A<2g(\\ln m)$ strictly. (b): a minimiser of $M_n$ exists (now stated in the proof: $M_n$ is continuous, tends to 1 at both ends, and is below 1 at each $x_t$). The consistency argument (grid on $[-R,R]$, Lipschitz $\\tfrac12$, lower bound outside) is valid. The conditions of van der Vaart Thm 5.23 hold: constant Lipschitz bound $\\tfrac12$; differentiability at 0 for $x\\ne0$; second-order expansion with $M''(0)=g(\\ln m)-A/2>0$, which exists under continuity of $g$ at $\\ln m$ by the decomposition now in H12 step 4. The asymptotic variances are $B/(2g-A)^2$ and $1/(4g^2)$, and their ratio is the stated ARE. The corrected sentence now calls this the asymptotic relative efficiency and says explicitly that convergence of the finite-sample relative efficiency (TM9) to it is not shown; that is exactly what the proof establishes. (c): recomputed $\\int u^2k=16\\ln2$, $\\int\\operatorname{sech}^4(u/2)=8/3$, $\\int|u|k=4$. The limits $\\mathrm{ARE}\\to1$ (both narrow cases), $\\mathrm{ARE}\\,\\sigma^3\\to6(\\ln2)^2\\sqrt{2\\pi}\\approx7.226$ and $\\mathrm{ARE}\\,b\\to3$ check by hand and by quadrature. The remainder orders in steps 11–12 are right (the log-Laplace one is a relative $O(1/b)$). The closing sentence, now with \"asymptotically\", follows from (c). Every part holds as worded and the proof establishes it."
      }
     ],
     "verification": "agent_checked",
     "depends_on": [
      "C3",
      "C5",
      "C12"
     ],
     "relations": [
      {
       "type": "explains",
       "target": "C15",
       "id_type": "local",
       "note": "C15 found by simulation that the sMAPE minimiser of a test set varies more than the sample median, the more so the wider the spread. C19 gives the limiting variances of both and the rates at which their ratio falls.",
       "results": [
        "E33.T4",
        "E33.T6"
       ]
      }
     ]
    },
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "add",
    "path": "/claims/19",
    "value": {
     "id": "C20",
     "name": "msMAPE is sMAPE except for small actuals",
     "summary": "Smyl's msMAPE equals sMAPE for actuals of at least its constant ν. Below ν its denominator stops shrinking. A zero actual then no longer gives every non-zero forecast the worst score, and the expected score has no jump at the zero forecast.",
     "statement": "The [[D33:msMAPE term]] $s_\\nu$ is the [[D1:sMAPE term]] $s$ except for actuals below $\\nu$. For $y \\ge 0$, $f \\ge 0$ and $\\nu > 0$: (a) where $y \\ge \\nu$, $s_\\nu(y,f) = s(y,f)$; so when every actual is at least $\\nu$, msMAPE is sMAPE and has the same target. (b) Where $y < \\nu$, $s_\\nu(y,f) = |f-y|/(f+\\nu)$: for a zero actual it is $f/(f+\\nu)$, which is 0 at $f = 0$ and rises towards 1 as $f$ grows, where the sMAPE term is 1 for every $f > 0$. (c) $0 \\le s_\\nu \\le 1$, with $s_\\nu < 1$ where $y < \\nu$; and for every $y$, $s_\\nu(y,f)$ is continuous in $f$ at $f = 0$, so the expected msMAPE term has no jump at the zero forecast, where the expected sMAPE term jumps when zero actuals have positive probability. (d) On counts with $\\nu \\le 1$, $s_\\nu$ differs from $s$ only for zero actuals.",
     "type": "formal",
     "role": "principal",
     "novelty": "extension",
     "support": [
      {
       "kind": "formal_proof",
       "evidence": "E37"
      },
      {
       "kind": "empirical_quant",
       "evidence": "E31"
      }
     ],
     "provenance": "author_stated",
     "hypothesis": {
      "lab_event": "H42",
      "outcome": "supported"
     },
     "planning": "confirmatory",
     "planned_at": "2026-11-18T12:02:12Z",
     "planned_at_anchor": "git_remote",
     "test": {
      "framework": "formal",
      "supported": "The proof of the hypothesis as worded here, in the study's proof documents (proofs.md, proof-suilin.md), is complete, every step naming what it rests on; the independent checker (an agent of another model than the author's: Claude Opus 5.5, registered as an actor before it checks) records a check of the hypothesis as worded with outcome confirmed; and the deciding run's numerical checks of its formulas meet the tolerances of the design.",
      "not_supported": "A counterexample to the hypothesis as worded, found by the proof, the checker, the supervisor or the deciding run, that a correction of the proof cannot remove.",
      "verdict": "judged",
      "runs": 1,
      "lab_plan": "PL12"
     },
     "checks": [
      {
       "by": "P5",
       "outcome": "confirmed",
       "extent": "full",
       "notes": "Derived independently from the definition of TM38 (I did not check Kolassa's text; the check is of the statement about the term as TM38 defines it). (a) For $y\\ge\\nu>0$, $\\max\\{y,\\nu\\}=y$ and $p=|f-y|/(f+y)=s$ for $y,f\\ge0$. So with every actual at least $\\nu$ the two expected losses coincide for $f\\ge0$; negative forecasts score 1 under sMAPE on positive actuals and cannot be better, so the targets agree. (b) For $y<\\nu$, $p=|f-y|/(f+\\nu)$; at $y=0$ it is $f/(f+\\nu)=1-\\nu/(f+\\nu)$: 0 at $f=0$, increasing, tending to 1, whereas $s(0,f)=1$ for $f>0$. (c) $p\\ge0$; for $y\\ge\\nu$, $p=s\\le1$; for $y<\\nu$, $f\\ge y$ gives $f-y\\le f<f+\\nu$ and $f<y$ gives $y-f<\\nu\\le f+\\nu$, so $p<1$. The denominator is at least $\\nu>0$, so $p$ is continuous in $f\\ge0$, and bounded convergence makes the expectation continuous. For sMAPE, $\\mathrm E[s(Y,0)]=1-P(Y=0)$ while $\\lim_{f\\downarrow0}\\mathrm E[s(Y,f)]=1$: a jump of $P(Y=0)$. (d) A count below $\\nu\\le1$ is 0. The proof (proof-suilin.md, second proof) matches step by step. Every part holds as worded."
      }
     ],
     "verification": "agent_checked",
     "about": {
      "subject": "D33",
      "relation": "equals, except for actuals below its constant,",
      "object": "D1",
      "comparator": null,
      "conditions": []
     },
     "relations": [
      {
       "type": "extends",
       "reference": "B13",
       "note": "Kolassa (2026, Proposition 3) shows that the minimisers of expected msMAPE on counts are integers; C20 shows where the msMAPE term equals the sMAPE term and that it removes the jump at the zero forecast."
      },
      {
       "type": "extends",
       "reference": "B16",
       "note": "Smyl (2025) proposed msMAPE, the sparse-proof sMAPE, as Kolassa (2026) reports; Smyl's article was not read. C20 states properties of its term as Kolassa (2026, Definition 4) defines it."
      }
     ]
    },
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "add",
    "path": "/claims/20",
    "value": {
     "id": "C21",
     "name": "What msMAPE rewards on intermittent demand",
     "summary": "On counts, msMAPE with ν = 0.1 and the Suilin loss reward the same forecast as sMAPE. With ν = 0.5 or 1 it is 0 less often. On non-integer demand the zero forecast needs a larger share of zeros: sMAPE 0.394; msMAPE 0.419, 0.519, 0.565; Suilin 0.470.",
     "statement": "The forecasts that sMAPE, msMAPE with $\\nu = 0.1$, 0.5 and 1, and the Suilin loss reward were computed from the distributions themselves on the 14 [[D35:intermittent demand settings]]. (a) On each of the 10 count distributions, the [[D34:msMAPE target]] with $\\nu = 0.1$ and the [[D29:Suilin target]] equal the [[D6:sMAPE target]]. With $\\nu = 0.5$ the msMAPE target is 1 instead of 0 on Poisson counts with mean 0.7 and on negative binomial counts with mean 2, and 1 instead of 2 on Poisson counts with mean 1.5. With $\\nu = 1$ it is also 1 instead of 0 on Poisson counts with mean 0.5 and on negative binomial counts with mean 1. (b) For lognormal demand sizes with zeros, the share of zeros at which the target switches to the forecast 0 is 0.394 for sMAPE, 0.419 for msMAPE with $\\nu = 0.1$, 0.470 for the Suilin loss, 0.519 for msMAPE with $\\nu = 0.5$ and 0.565 with $\\nu = 1$. (c) In test sets of 12, 52 and 365 actuals, the forecast with the lowest score is 0 no more often under msMAPE than under sMAPE, and no more often under a larger $\\nu$. With 52 actuals from Poisson counts with mean 0.7, it is 0 in 0.6575 of the test sets under sMAPE, 0.523 with $\\nu = 0.1$, 0.1525 with $\\nu = 0.5$ and 0.0265 with $\\nu = 1$.",
     "type": "descriptive",
     "role": "secondary",
     "novelty": "no_prior_found",
     "support": [
      {
       "kind": "empirical_quant",
       "evidence": "E34",
       "results": [
        "E34.T1.r1",
        "E34.T1.r2",
        "E34.T1.r3",
        "E34.T1.r4",
        "E34.T1.r5",
        "E34.T1.r6",
        "E34.T1.r7",
        "E34.T1.r8",
        "E34.T1.r9",
        "E34.T1.r10",
        "E34.T2.r1",
        "E34.T4",
        "E34.T4.r11"
       ]
      }
     ],
     "provenance": "author_stated",
     "planning": "planned_estimate",
     "planned_at": "2026-11-18T12:14:53Z",
     "planned_at_anchor": "git_remote",
     "estimate": {
      "estimands": {
       "target": {
        "variable": "target",
        "table": true
       },
       "level": {
        "variable": "quantile level of the target",
        "table": true
       },
       "score": {
        "variable": "expected score at the target",
        "table": true
       },
       "zero_threshold": {
        "variable": "zero threshold",
        "table": true
       },
       "share_zero": {
        "variable": "share of test sets whose minimiser is 0",
        "table": true
       },
       "share_target": {
        "variable": "share of test sets whose minimiser is the target",
        "table": true
       },
       "median_minimiser": {
        "variable": "median of the test-set minimisers",
        "table": true
       }
      },
      "runs": 1,
      "lab_plan": "PL14"
     },
     "about": {
      "subject": "D34",
      "relation": "differs from",
      "object": "D6",
      "comparator": "D29",
      "conditions": [
       "D35"
      ]
     },
     "scope": {
      "data_type": "outcome distributions given exactly, and test sets drawn from them",
      "conditions": [
       "14 settings: 10 count distributions (Poisson with means 0.1 to 3, negative binomial with size 0.5 and means 0.5 to 2) and lognormal demand sizes (median 1, sigma 1) with shares of zeros 0.2 to 0.8",
       "msMAPE with nu = 0.1, 0.5 and 1",
       "test sets of 12, 52 and 365 independent actuals, 2000 test sets per setting and size"
      ]
     },
     "relations": [
      {
       "type": "supports",
       "target": "C20",
       "id_type": "local",
       "note": "Computed on intermittent demand, msMAPE needs a larger share of zeros than sMAPE before its target is 0, in line with C20 (b): a zero actual no longer gives every non-zero forecast the worst score.",
       "results": [
        "E34.T1",
        "E34.T2"
       ]
      }
     ]
    },
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/contributors/0/run_period/end",
    "value": "2026-11-18",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "add",
    "path": "/contributors/3",
    "value": {
     "id": "P4",
     "type": "agent",
     "name": "Claude Code",
     "positions": [
      "author"
     ],
     "models": [
      {
       "model": "M2",
       "role": "sole"
      }
     ],
     "run_period": {
      "start": "2026-11-18",
      "end": "2026-11-18"
     },
     "tools": [
      "code_execution",
      "file_access"
     ],
     "system": {
      "name": "Claude Code"
     },
     "label": "version 3: corrections and new derivations",
     "record_source": "harness",
     "credit_roles": [
      "methodology",
      "formal_analysis",
      "software",
      "investigation",
      "visualization",
      "writing_original_draft"
     ],
     "ai_tasks": [
      {
       "task": "analysis",
       "extent": "full"
      },
      {
       "task": "code",
       "extent": "full"
      },
      {
       "task": "writing",
       "extent": "full"
      },
      {
       "task": "json_preparation",
       "extent": "full"
      }
     ]
    },
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "add",
    "path": "/contributors/4",
    "value": {
     "id": "P5",
     "type": "agent",
     "name": "Claude Code",
     "positions": [
      "contributor"
     ],
     "models": [
      {
       "model": "M1",
       "role": "sole"
      }
     ],
     "run_period": {
      "start": "2026-11-18",
      "end": "2026-11-18"
     },
     "ai_tasks": [
      {
       "task": "review",
       "extent": "full"
      }
     ],
     "system": {
      "name": "Claude Code"
     },
     "label": "proof checking, version 3",
     "record_source": "self_reported"
    },
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/definitions",
    "value": [
     {
      "id": "D2",
      "term": "sMAPE",
      "kind": "variable",
      "definition": "Symmetric mean absolute percentage error of a point forecast $f$ over a test set of actuals $y_1, \\dots, y_n$: $\\frac{200}{n}\\sum_t |y_t - f|/(|y_t|+|f|)$, where a term with $y_t = f = 0$ counts as 0. Introduced as the symmetric MAPE by Makridakis (1993), whose denominator had no absolute values; for positive actuals and forecasts the two agree.",
      "source": "B8"
     },
     {
      "id": "D1",
      "term": "sMAPE term",
      "kind": "variable",
      "definition": "The contribution of one actual $y$ and one point forecast $f$ to [[D2:sMAPE]]: $s(y,f) = |y-f|/(|y|+|f|)$, between 0 and 1, with $s = 0$ when $y = f$ (including $y = f = 0$, where the fraction is $0/0$).",
      "symbol": "$s(y,f)$"
     },
     {
      "id": "D3",
      "term": "expected sMAPE",
      "kind": "variable",
      "definition": "The expected sMAPE of a fixed forecast $f$ under the distribution of the actual $y$: $200\\,\\mathrm{E}[s(y, f)]$, where $\\mathrm E$ is the expectation over the random actual $y$. Computed from the distribution (by integration or exact sums), never from a sample."
     },
     {
      "id": "D6",
      "term": "sMAPE target",
      "kind": "concept",
      "definition": "The value or values of $f$ that minimise expected sMAPE for a given distribution of the actual $y$, $\\arg\\min_f \\mathrm E[s(y,f)]$: the functional sMAPE rewards. It may be a single value or several."
     },
     {
      "id": "D4",
      "term": "log error",
      "kind": "variable",
      "definition": "For a positive actual $y$ and a positive forecast $f$, $z = \\ln y - \\ln f$.",
      "symbol": "$z$"
     },
     {
      "id": "D5",
      "term": "weight",
      "kind": "variable",
      "definition": "For a positive actual $y$ and a positive forecast $f$, $w(y,f) = y/(y+f)^2$: how much the actual pulls the forecast that minimises expected sMAPE, since the derivative of $\\mathrm E[s(y,f)]$ in $f$ is $2\\,\\mathrm E[w(y,f)(\\mathbf 1\\{y<f\\} - \\mathbf 1\\{y>f\\})]$. Largest at $y = f$, the same for $y/f = r$ and $1/r$.",
      "symbol": "$w(y,f)$"
     },
     {
      "id": "D7",
      "term": "quantile profile",
      "kind": "concept",
      "definition": "For an outcome distribution with quantile function $Q$ and a forecast $f$, the pull of the outcome at each quantile level $u$, $u \\mapsto w(Q(u), f)$ on $(0,1)$; the minimiser's quantile level balances its integrals below and above."
     },
     {
      "id": "D10",
      "term": "RMSE",
      "kind": "variable",
      "definition": "Root mean squared error of a point forecast $f$ over a test set of actuals $y_1, \\dots, y_n$: $\\sqrt{\\frac1n\\sum_t (y_t-f)^2}$."
     },
     {
      "id": "D11",
      "term": "MAE",
      "kind": "variable",
      "definition": "Mean absolute error of a point forecast $f$ over a test set of actuals $y_1, \\dots, y_n$: $\\frac1n\\sum_t |y_t-f|$."
     },
     {
      "id": "D12",
      "term": "MAPE",
      "kind": "variable",
      "definition": "Mean absolute percentage error of a point forecast $f$ over a test set of actuals $y_1, \\dots, y_n$: $\\frac{100}{n}\\sum_t |y_t-f|/|y_t|$; infinite when an actual is zero."
     },
     {
      "id": "D19",
      "term": "log-MAE",
      "kind": "variable",
      "definition": "100 times the mean absolute error of $\\ln f$ as a forecast of $\\ln y$: $\\frac{100}{n}\\sum_t |\\ln y_t - \\ln f|$. Unbounded; sMAPE's small-error limit (C3); minimised by the median."
     },
     {
      "id": "D28",
      "term": "Suilin term",
      "kind": "variable",
      "definition": "The contribution of one actual $y$ and one forecast $f$ to the Suilin loss:\n$$\\tilde s(y,f) = \\frac{|y-f|}{\\max(|y|+|f|+\\delta,\\ 0.5+\\delta)}, \\qquad \\delta = 0.1;$$\nthe Suilin loss is 200 times its mean over a test set. It replaces the denominator $|y|+|f|$ of the [[D1:sMAPE term]]. Proposed by Suilin (2017) in code for a Kaggle competition, as a smoothed variant of sMAPE; defined in print by Hewamalage, Bergmeir and Bandara (2021, Eq. 15), who write the constant $\\delta$ as $\\epsilon$ and the factor 2 into the measure differently (a constant factor changes no property here). Hewamalage, Ackermann and Bergmeir (2023) call it msMAPE; Smyl (2025) gives that name to a different measure, so this study follows Kolassa (2026) and calls it the Suilin loss. Version 2 of this paper called it msMAPE.",
      "source": "B14",
      "symbol": "$\\tilde s(y,f)$"
     },
     {
      "id": "D29",
      "term": "Suilin target",
      "kind": "concept",
      "definition": "The value or values of $f$ that minimise the expected [[D28:Suilin term]] for a given distribution of the actual: the functional the Suilin loss rewards, as the [[D6:sMAPE target]] is the one sMAPE rewards."
     },
     {
      "id": "D33",
      "term": "msMAPE term",
      "kind": "variable",
      "definition": "The contribution of one actual $y \\ge 0$ and one forecast $f \\ge 0$ to the sparse-proof sMAPE (msMAPE) of Smyl (2025), as Kolassa (2026, Definition 4) states it, for a small constant $\\nu > 0$ (Kolassa writes it $m$, which is the median here):\n$$s_\\nu(y,f) = \\frac{|f-y|}{f+\\max\\{y, \\nu\\}},$$\nwhich is 0 when $y = f = 0$; msMAPE is 200 times its mean over a test set (Kolassa writes the factor 2 into the term). It replaces the actual in the denominator of the [[D1:sMAPE term]] by $\\nu$ when the actual is below $\\nu$.",
      "source": "B13",
      "symbol": "$s_\\nu(y,f)$"
     },
     {
      "id": "D34",
      "term": "msMAPE target",
      "kind": "concept",
      "definition": "The value or values of $f$ that minimise the expected [[D33:msMAPE term]] for a given distribution of the actual and a given constant $\\nu$: the functional msMAPE rewards, as the [[D6:sMAPE target]] is the one sMAPE rewards."
     },
     {
      "id": "D8",
      "term": "sMAPE minimiser of a test set",
      "kind": "variable",
      "definition": "The forecast that minimises the sMAPE of one test set $y_1, \\dots, y_n$, $\\hat f_n = \\arg\\min_f \\frac1n\\sum_t s(y_t, f)$, found numerically (the sample counterpart of the [[D6:sMAPE target]]).",
      "symbol": "$\\hat f_n$"
     },
     {
      "id": "D9",
      "term": "relative efficiency",
      "kind": "variable",
      "definition": "For a test-set size $n$ and an outcome distribution where sMAPE and MAE both target the median: $\\mathrm{RE}(n) = \\operatorname{Var}(\\ln \\hat m_n) / \\operatorname{Var}(\\ln \\hat f_n)$ over test sets, where $\\hat m_n$ is the sample median (the MAE minimiser) and $\\hat f_n$ the sMAPE minimiser of the same test set. Above 1, sMAPE's minimiser varies less; below 1, MAE's does.",
      "symbol": "$\\mathrm{RE}(n)$"
     },
     {
      "id": "D20",
      "term": "ranking agreement",
      "kind": "variable",
      "definition": "For one distribution, one measure, a test-set size $n$, the measure's target and a forecast at another quantile level: the probability that a test set of $n$ outcomes gives the target the strictly lower score (ties count as disagreement); 0.5 is a coin toss."
     },
     {
      "id": "D32",
      "term": "signal-to-noise ratio of a comparison",
      "kind": "variable",
      "definition": "For a measure with term $L(y,f)$, two forecasts $f_1$ and $f_2$ and a distribution of the actual $y$: $|\\mathrm E[D]|/\\mathrm{sd}(D)$, where $D = L(y,f_2) - L(y,f_1)$ is the difference of the two forecasts' scores on one actual. Over a test set of $n$ independent actuals the mean of $D$ is $\\sqrt n$ times this many of its standard errors from zero: the larger the ratio, the smaller the test set that tells the two forecasts apart."
     },
     {
      "id": "D13",
      "term": "candidate forecast",
      "kind": "concept",
      "definition": "A point forecast fixed from the outcome distribution before any test set is drawn and the same for every test set: a quantile at a level from 0.01 to 0.99, the mean, the forecast 0, the log-mode, the minimiser of expected MAPE, or a local minimiser of expected sMAPE. It is never estimated from a sample."
     },
     {
      "id": "D14",
      "term": "preferred candidate",
      "kind": "concept",
      "definition": "Of the candidate forecasts for one distribution, the one a measure gives the lowest expected score, or, where every expected score is infinite but their differences are finite, the one the differences rank first (C8); none where no forecast is ranked first."
     },
     {
      "id": "D15",
      "term": "derived value",
      "kind": "concept",
      "definition": "A value computed from the outcome distribution itself by numerical integration of its density (Gauss-Legendre quadrature and a convolution on a grid of 524288 points of $\\ln y$), never from a sample: the expected scores, the local minima of expected sMAPE, the formulas of the hypotheses."
     },
     {
      "id": "D16",
      "term": "selection design",
      "kind": "method",
      "definition": "For each outcome distribution: derive the candidate forecasts and their expected scores; then, for each test-set size $n$, draw test sets of $n$ independent outcomes, score every candidate by every measure on each test set, and count the candidate with the lowest score as the measure's choice (a tie of $j$ candidates counts $1/j$ each; a test set in which every candidate scores infinite chooses none)."
     },
     {
      "id": "D17",
      "term": "efficiency design",
      "kind": "method",
      "definition": "For each outcome distribution and test-set size $n$: draw test sets of $n$ independent outcomes; in each, compute the sample median and the exact sMAPE minimiser (an outcome of the test set, since the test set's sMAPE is concave in $\\ln f$ between consecutive outcomes); over test sets, compare the variances of their logarithms."
     },
     {
      "id": "D18",
      "term": "flatness design",
      "kind": "method",
      "definition": "For families whose spread $\\sigma$ of $\\ln y$ grows from 0.1 to 10: derive the sMAPE target, the expected sMAPE of forecasts at fixed quantile levels and the large-spread approximation of C10; on test sets, count how often each measure scores the target lower than a forecast away from it, and where the test set's sMAPE minimiser falls."
     },
     {
      "id": "D31",
      "term": "formula checks",
      "kind": "method",
      "definition": "Numerical checks of the formulas of the formal claims (claimset-code/smape-target/check_formulas.py, seed 6; check_jump.py for C11): identities and inequalities on 200000 random pairs, derivatives against finite differences, minimisers, balances and thresholds against quadrature and bisection on every setting of the selection design, and the large-spread expansion against its bound. For the claims new or reworded in version 3 (C3, C4, C12, C17 to C20): check_v3.py (seed 8), check_h39.py and check_corrected.py (seed 8), in the same way."
     },
     {
      "id": "D21",
      "term": "inverse Gaussian outcomes",
      "kind": "population",
      "definition": "Inverse Gaussian outcomes with mean 100 and shape 1000, 200, 100, 50, 20 or 10 (coefficient of variation 0.32 to 3.16): positive, skewed right, every moment and $\\mathrm E[1/y]$ finite."
     },
     {
      "id": "D22",
      "term": "log-skewed outcomes",
      "kind": "population",
      "definition": "$y = 100\\,e^x$ with $x = \\pm\\sigma(G-\\alpha)/\\sqrt\\alpha$, $G \\sim \\mathrm{Gamma}(\\alpha, 1)$: $\\ln y$ with standard deviation $\\sigma$ (0.5 or 1; sd in the settings' names) and skewness $\\pm 2/\\sqrt\\alpha$ (-2, -1, -0.5, +0.5, +1, +2), and the lognormal of the same $\\sigma$ (skewness 0). Some have infinite moments: $\\mathrm E[y]$ infinite ($\\sigma = 1$, skewness +2), $\\mathrm E[y^2]$ infinite ($\\sigma = 0.5$, skewness +2; $\\sigma = 1$, skewness +1), $\\mathrm E[1/y]$ infinite ($\\sigma = 1$, skewness -2)."
     },
     {
      "id": "D23",
      "term": "bimodal outcomes",
      "kind": "population",
      "definition": "$y = 100\\,e^x$, $x$ normal with mean $-1$ (probability $p$) or $+1$, standard deviation 0.25; $p = 0.5$ (equal modes), 0.65 (left mode higher), 0.35 (right mode higher)."
     },
     {
      "id": "D24",
      "term": "outcomes straddling zero",
      "kind": "population",
      "definition": "Normal outcomes with standard deviation 1 and mean 0, 0.5 or 1: 50%, 31% and 16% negative; $\\mathrm E[1/|y|]$ infinite."
     },
     {
      "id": "D25",
      "term": "zero-inflated outcomes",
      "kind": "population",
      "definition": "Lognormal outcomes (median 100, $\\sigma$ 0.5 or 1) replaced by 0 with probability $p_0$: 0.25, 0.42, 0.47, 0.6 for $\\sigma = 0.5$ and 0.25, 0.37, 0.42, 0.6 for $\\sigma = 1$, on both sides of the zero threshold of C7 (0.4469 and 0.3940)."
     },
     {
      "id": "D26",
      "term": "log-symmetric outcomes",
      "kind": "population",
      "definition": "Lognormal outcomes with $\\sigma$, and log-Laplace outcomes with scale $b$, each 0.25, 0.5, 1, 1.5, 2, 3, 4 or 6, median 100: $\\ln y$ symmetric and strictly unimodal, so that sMAPE and MAE both target the median (C5)."
     },
     {
      "id": "D27",
      "term": "widening outcomes",
      "kind": "population",
      "definition": "$\\ln y = \\ln 100 + \\sigma\\varepsilon$ with $\\varepsilon$ standard normal (lognormal), standard Gumbel (log-Gumbel right, $\\ln y$ skewed right) or minus standard Gumbel (log-Gumbel left), for $\\sigma$ = 0.1, 0.25, 0.5, 1, 1.5, 2, 3, 4, 6 and 10."
     },
     {
      "id": "D30",
      "term": "Poisson counts",
      "kind": "population",
      "definition": "Poisson outcomes with means 0.3, 0.5, 1, 1.5, 3 and 10: counts whose share of zeros falls from 0.74 to almost 0."
     },
     {
      "id": "D35",
      "term": "intermittent demand settings",
      "kind": "population",
      "definition": "Demand with many zeros. Counts: Poisson with means 0.1, 0.3, 0.5, 0.7, 1, 1.5 and 3 (share of zeros from 0.90 to 0.05) and, overdispersed, negative binomial with size 0.5 and means 0.5, 1 and 2 (share of zeros 0.71, 0.58 and 0.45). Non-integer demand: lognormal demand sizes (median 1, $\\sigma = 1$) replaced by 0 with probability 0.2, 0.4, 0.6 or 0.8."
     }
    ],
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "remove",
    "path": "/evidence/6",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/evidence/0/description",
    "value": "Numerical checks of the formulas of the formal claims ([[D31:formula checks]]), lab run RN8 under the formal plan that decided them; every check is within the plan's tolerances. They check the formulas on concrete numbers; the proofs are the evidence for the claims. The rows are labelled by the hypotheses of the record's first plan, whose wording kept the factor 200: H1 checks C1; H2 and H10 check C2; H5 checks C5; H7 checks C6; H8 checks C7; H12 checks C8; H13 checks C9; H14 checks C10. H3 and H9 checked C3, and H4 and H6 checked C4, as version 2 worded them; C3, C4 and C12 as now worded are checked in E31. The two results E1.R1 and E1.R2 compare a simulated variance with the large-sample variance that C19 now proves.",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/evidence/5/description",
    "value": "Lab runs RN10 under plan PL11. Design: Exact computation, with no sampling, of the forecasts that minimise the expected Suilin term and the expected sMAPE term (claimset-code/smape-target/msmape_targets.py, run once): by Gauss-Legendre quadrature of the density of ln y for continuous distributions (with zeros added as a point mass) and by sums for counts, minimised over forecasts f >= 0 on a grid of 0 and 4000 log-spaced points refined by golden section. Settings: away from zero, lognormal sigma 0.5 and 1, inverse Gaussian shape 10 and 100 and log-skewed sd 1 with skewness -1 and +1, all at median about 100; near zero, lognormal sigma 1 at median 1 and at median 100 with shares of zeros 0, 0.2, 0.4, 0.6 and 0.8; counts, Poisson with means 0.3, 0.5, 1, 1.5, 3 and 10. For the lognormal sigma 1 at medians 1 and 100, the share of zeros at which each measure switches to the forecast 0, by bisection. Analysis: For each setting and measure: the target, its quantile level and the expected score there; for each setting, the relative difference of the Suilin target from the sMAPE target; and the two zero thresholds. Reported as tables, with no decision rule. The plan, the run's script and its results file name the Suilin loss msMAPE, as version 2 of this paper did.",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/evidence/5/tables",
    "value": [
     {
      "id": "T1",
      "description": "The Suilin target beside the sMAPE target, with their quantile levels and the relative difference of the Suilin target from the sMAPE target",
      "row_labels": [
       "setting"
      ],
      "columns": [
       {
        "variable": "target",
        "conditions": {
         "measure": "sMAPE"
        }
       },
       {
        "variable": "target",
        "conditions": {
         "measure": "Suilin"
        }
       },
       {
        "variable": "quantile level of the target",
        "conditions": {
         "measure": "sMAPE"
        }
       },
       {
        "variable": "quantile level of the target",
        "conditions": {
         "measure": "Suilin"
        }
       },
       "relative difference of the Suilin target from the sMAPE target",
       "share of zero outcomes"
      ],
      "rows": [
       {
        "id": "r1",
        "label": [
         "lognormal sigma=0.5"
        ],
        "values": [
         100.0,
         100.0,
         0.5,
         0.5,
         -1.403e-05,
         0
        ]
       },
       {
        "id": "r2",
        "label": [
         "lognormal sigma=1"
        ],
        "values": [
         100.0,
         99.98,
         0.5,
         0.4999,
         -0.0001834,
         0
        ]
       },
       {
        "id": "r3",
        "label": [
         "inverse Gaussian shape=10"
        ],
        "values": [
         14.02,
         13.98,
         0.4387,
         0.438,
         -0.002701,
         0
        ]
       },
       {
        "id": "r4",
        "label": [
         "inverse Gaussian shape=100"
        ],
        "values": [
         66.53,
         66.52,
         0.4929,
         0.4929,
         -0.0001284,
         0
        ]
       },
       {
        "id": "r5",
        "label": [
         "log-skewed left sd=1 skew=-1"
        ],
        "values": [
         128.3,
         128.2,
         0.5362,
         0.5361,
         -0.0002099,
         0
        ]
       },
       {
        "id": "r6",
        "label": [
         "log-skewed right sd=1 skew=+1"
        ],
        "values": [
         77.96,
         77.96,
         0.4638,
         0.4638,
         -4.995e-05,
         0
        ]
       },
       {
        "id": "r7",
        "label": [
         "lognormal median=1 sigma=1"
        ],
        "values": [
         1.0,
         0.9846,
         0.5,
         0.4938,
         -0.01536,
         0
        ]
       },
       {
        "id": "r8",
        "label": [
         "lognormal median=1 sigma=1 with zeros p0=0.2"
        ],
        "values": [
         1.0,
         0.9054,
         0.6,
         0.5683,
         -0.0946,
         0.2
        ]
       },
       {
        "id": "r9",
        "label": [
         "lognormal median=1 sigma=1 with zeros p0=0.4"
        ],
        "values": [
         0,
         0.1702,
         0.4,
         0.423,
         null,
         0.4
        ]
       },
       {
        "id": "r10",
        "label": [
         "lognormal median=1 sigma=1 with zeros p0=0.6"
        ],
        "values": [
         0,
         0,
         0.6,
         0.6,
         0,
         0.6
        ]
       },
       {
        "id": "r11",
        "label": [
         "lognormal median=1 sigma=1 with zeros p0=0.8"
        ],
        "values": [
         0,
         0,
         0.8,
         0.8,
         0,
         0.8
        ]
       },
       {
        "id": "r12",
        "label": [
         "lognormal median=100 sigma=1"
        ],
        "values": [
         100.0,
         99.98,
         0.5,
         0.4999,
         -0.0001834,
         0
        ]
       },
       {
        "id": "r13",
        "label": [
         "lognormal median=100 sigma=1 with zeros p0=0.2"
        ],
        "values": [
         100.0,
         99.89,
         0.6,
         0.5997,
         -0.001093,
         0.2
        ]
       },
       {
        "id": "r14",
        "label": [
         "lognormal median=100 sigma=1 with zeros p0=0.4"
        ],
        "values": [
         0,
         0,
         0.4,
         0.4,
         0,
         0.4
        ]
       },
       {
        "id": "r15",
        "label": [
         "lognormal median=100 sigma=1 with zeros p0=0.6"
        ],
        "values": [
         0,
         0,
         0.6,
         0.6,
         0,
         0.6
        ]
       },
       {
        "id": "r16",
        "label": [
         "lognormal median=100 sigma=1 with zeros p0=0.8"
        ],
        "values": [
         0,
         0,
         0.8,
         0.8,
         0,
         0.8
        ]
       },
       {
        "id": "r17",
        "label": [
         "Poisson mean=0.3"
        ],
        "values": [
         0,
         0,
         0.7408,
         0.7408,
         0,
         0.7408
        ]
       },
       {
        "id": "r18",
        "label": [
         "Poisson mean=0.5"
        ],
        "values": [
         0,
         0,
         0.6065,
         0.6065,
         0,
         0.6065
        ]
       },
       {
        "id": "r19",
        "label": [
         "Poisson mean=1"
        ],
        "values": [
         1,
         1,
         0.7358,
         0.7358,
         0,
         0.3679
        ]
       },
       {
        "id": "r20",
        "label": [
         "Poisson mean=1.5"
        ],
        "values": [
         2,
         2,
         0.8088,
         0.8088,
         0,
         0.2231
        ]
       },
       {
        "id": "r21",
        "label": [
         "Poisson mean=3"
        ],
        "values": [
         3,
         3,
         0.6472,
         0.6472,
         0,
         0.04979
        ]
       },
       {
        "id": "r22",
        "label": [
         "Poisson mean=10"
        ],
        "values": [
         10,
         10,
         0.583,
         0.583,
         0,
         4.54e-05
        ]
       }
      ],
      "artefact": "A52"
     },
     {
      "id": "T2",
      "description": "The share of zeros at which each measure's target switches to the forecast 0, for lognormal outcomes with sigma 1",
      "row_labels": [
       "setting"
      ],
      "columns": [
       {
        "variable": "zero threshold",
        "conditions": {
         "measure": "sMAPE"
        }
       },
       {
        "variable": "zero threshold",
        "conditions": {
         "measure": "Suilin"
        }
       }
      ],
      "rows": [
       {
        "id": "r1",
        "label": [
         "lognormal median=1 sigma=1"
        ],
        "values": [
         0.394,
         0.47
        ]
       },
       {
        "id": "r2",
        "label": [
         "lognormal median=100 sigma=1"
        ],
        "values": [
         0.394,
         0.3937
        ]
       }
      ],
      "artefact": "A52"
     }
    ],
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/evidence/8/steps",
    "value": [
     {
      "id": "S1",
      "statement": "Write $z = \\ln y - \\ln f$, so that $r = y/f = e^z$. By C1, $s = \\dfrac{|e^z - 1|}{e^z + 1}$.",
      "rule": "definition",
      "by": [
       "C1"
      ]
     },
     {
      "id": "S2",
      "statement": "Multiplying the numerator and the denominator of $\\dfrac{e^z - 1}{e^z + 1}$ by $e^{-z/2}$ gives $\\dfrac{e^{z/2} - e^{-z/2}}{e^{z/2} + e^{-z/2}} = \\tanh(z/2)$. The denominator is positive, so $s = |\\tanh(z/2)|$.",
      "rule": "algebra",
      "by": [
       "S1"
      ]
     },
     {
      "id": "S3",
      "statement": "$\\tanh$ changes only its sign when its argument changes sign: $\\tanh(-x) = -\\tanh(x)$, as the definition shows. So $|\\tanh(z/2)| = \\tanh(|z|/2)$, which is the stated form.",
      "rule": "algebra",
      "by": [
       "S2"
      ]
     },
     {
      "id": "S4",
      "statement": "For (a): $\\tanh$ is 0 at 0. On $[0, \\infty)$ it is strictly increasing and concave, and it approaches 1 as its argument grows; its derivative $1 - \\tanh^2 x$ is positive and decreasing for $x > 0$. With the argument $|\\ln y - \\ln f|/2$, $s$ has these properties in $|\\ln y - \\ln f|$.",
      "rule": "monotonicity",
      "by": [
       "S3"
      ]
     },
     {
      "id": "S5",
      "statement": "For (b): for every $x \\ge 0$, $x - x^3/3 \\le \\tanh x \\le x$. The upper bound holds because $\\tanh$ has slope at most 1; the lower one because the derivative of $\\tanh x - x + x^3/3$ is $x^2 - \\tanh^2 x \\ge 0$, and the difference is 0 at $x = 0$. Putting $x = |\\ln y - \\ln f|/2$ gives the bounds.",
      "rule": "algebra",
      "by": [
       "S4"
      ]
     },
     {
      "id": "S6",
      "statement": "For (c), the median: the mean of $|\\ln y_t - \\ln f|$ over a test set is minimised when $\\ln f$ is a median of the $\\ln y_t$. Since the logarithm is increasing, this is when $f$ is a median $m$ of the $y_t$.",
      "rule": "monotonicity",
      "by": [
       "S5"
      ]
     },
     {
      "id": "S7",
      "statement": "For (c), the bound: with $z_t = \\ln y_t - \\ln f$ the log errors of the test set, write $\\overline{|z|}(f) = \\frac1n\\sum_t|\\ln y_t - \\ln f|$, $\\overline{|z|^3}(f) = \\frac1n\\sum_t|\\ln y_t - \\ln f|^3$ and $\\bar s(f)$ for the mean of the sMAPE terms of the test set. Averaging the bounds of step 5 over the test set gives $\\bar s(f) \\le \\tfrac12\\overline{|z|}(f) \\le \\bar s(f) + \\tfrac1{24}\\overline{|z|^3}(f)$ for every $f > 0$. Let $f^*$ minimise $\\bar s$, so that $\\bar s(f^*) \\le \\bar s(m)$. Then\n$$\\tfrac12\\overline{|z|}(f^*) \\le \\bar s(f^*) + \\tfrac1{24}\\overline{|z|^3}(f^*) \\le \\bar s(m) + \\tfrac1{24}\\overline{|z|^3}(f^*) \\le \\tfrac12\\overline{|z|}(m) + \\tfrac1{24}\\overline{|z|^3}(f^*).$$\nMultiplying by 2 gives $\\overline{|z|}(f^*) - \\overline{|z|}(m) \\le \\tfrac1{12}\\overline{|z|^3}(f^*)$, the stated inequality.",
      "rule": "algebra",
      "by": [
       "S5"
      ]
     },
     {
      "id": "S8",
      "statement": "For (c), not the median in general: take the five actuals $1, 1, e, e^4, e^4$, whose median is $e$. With the forecast at $e$ the log errors are $1, 1, 0, 3, 3$ in size and the sMAPE terms sum to $2\\tanh(0.5) + 2\\tanh(1.5) = 2.734$. With the forecast at 1 the log errors are $0, 0, 1, 4, 4$ and the terms sum to $\\tanh(0.5) + 2\\tanh(2) = 2.390$, which is lower. So the median does not minimise the sMAPE of this test set, although it minimises its log-MAE (step 6).",
      "rule": "algebra",
      "by": [
       "S6"
      ]
     },
     {
      "id": "S9",
      "statement": "So $s = \\tanh(|\\ln y - \\ln f|/2)$. It rises from 0 like half the absolute log error and levels off at 1. For small errors sMAPE is close to the log-MAE, and its minimiser is close to the median without in general being the median.",
      "rule": "conclusion",
      "by": [
       "S8"
      ]
     }
    ],
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/evidence/10/description",
    "value": "Proof of C4 (section H5 of the proof document A1, rendered in A2). For any distribution, the best forecast balances the actuals below it and above it. Each actual is weighted by how close it is to the forecast by ratio.",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/evidence/10/steps",
    "value": [
     {
      "id": "S1",
      "statement": "The derivative of $s$. When $y > f > 0$, $s = (y - f)/(y + f)$. By the quotient rule its derivative in $f$ is $\\dfrac{-(y+f) - (y-f)}{(y+f)^2} = -2w(y,f)$. When $f > y > 0$, $s = (f - y)/(y + f)$, and in the same way its derivative is $2w(y,f)$. Together: for $y \\ne f$, the derivative is $2w(y,f)(\\mathbf 1\\{y < f\\} - \\mathbf 1\\{y > f\\})$.",
      "rule": "quotient rule",
      "by": [
       "D1"
      ]
     },
     {
      "id": "S2",
      "statement": "A bound on it. Since $(y+f)^2 \\ge 4yf$, $w(y,f) \\le 1/(4f)$. So the derivative of $s$ is at most $1/(2f)$ in size, whatever $y$, and at most $1/f$ for every forecast above $f/2$. The only actual at which $s$ has no derivative is $y = f$, which has probability zero.",
      "rule": "algebra",
      "by": [
       "S1"
      ]
     },
     {
      "id": "S3",
      "statement": "The derivative of the expectation. With this bound, the derivative may be taken inside the expectation (dominated convergence):\n$$\\frac{d}{df}\\,\\mathrm E[s(y,f)] = 2\\,\\mathrm E\\big[w(y,f)\\,(\\mathbf 1\\{y < f\\} - \\mathbf 1\\{y > f\\})\\big].$$",
      "rule": "dominated convergence",
      "by": [
       "S2"
      ]
     },
     {
      "id": "S4",
      "statement": "The balance. A forecast with the lowest expected sMAPE exists (C9). There this derivative is 0, which is the stated balance: half the total weight lies below $f$ and half above.",
      "rule": "earlier result",
      "by": [
       "C9"
      ]
     },
     {
      "id": "S5",
      "statement": "The weight. With $r = y/f$, $w(y,f) = \\dfrac1f\\cdot\\dfrac{r}{(1+r)^2} = \\dfrac1f\\cdot\\dfrac{1}{r + 2 + 1/r}$. The sum $r + 1/r$ is the same for $r$ and $1/r$. It is smallest, 2, at $r = 1$, and grows as $r$ moves away from 1. So $w$ is largest at $y = f$ and falls as $y/f$ moves away from 1, equally for $r$ and $1/r$.",
      "rule": "algebra",
      "by": [
       "S4"
      ]
     },
     {
      "id": "S6",
      "statement": "So sMAPE targets a weighted median of the actuals. Each actual is weighted by how close it is to the forecast by ratio.",
      "rule": "conclusion",
      "by": [
       "S5"
      ]
     }
    ],
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/evidence/12/steps",
    "value": [
     {
      "id": "S1",
      "statement": "Write forecasts on the log scale as $\\ln f = \\ln m + c$. By C3, the expected contribution is $\\mathrm E[\\tanh(|\\ln y - \\ln m - c|/2)]$, a function of $c$.",
      "rule": "setup",
      "by": [
       "C3"
      ]
     },
     {
      "id": "S2",
      "statement": "Why the median is bad. Take the groups to be single points first. A forecast at the median is $a$ away, on the log scale, from every actual. So its expected $s$ is $\\tanh(a/2)$.",
      "rule": "case analysis",
      "by": [
       "S1"
      ]
     },
     {
      "id": "S3",
      "statement": "A forecast at the mode, $\\ln f = \\ln m + a$, is exact for half the actuals and $2a$ away from the other half. So its expected $s$ is $\\tfrac12\\tanh(a)$. Since $\\tanh(a) = \\dfrac{2\\tanh(a/2)}{1 + \\tanh^2(a/2)}$, this is $\\dfrac{\\tanh(a/2)}{1 + \\tanh^2(a/2)} < \\tanh(a/2)$. The mode is better.",
      "rule": "algebra",
      "by": [
       "S2"
      ]
     },
     {
      "id": "S4",
      "statement": "The expected $s$ changes continuously as the groups widen from points, because $s$ is bounded and continuous. So the mode is still better for narrow groups.",
      "rule": "continuity",
      "by": [
       "S3"
      ]
     },
     {
      "id": "S5",
      "statement": "Why the median is a local maximum. By C12, step 4, the second derivative of the expected contribution at the median is\n$$g(\\ln m) - \\tfrac12\\,\\mathrm E\\Big[\\operatorname{sech}^2\\Big(\\frac{\\ln y - \\ln m}{2}\\Big)\\tanh\\Big(\\frac{|\\ln y - \\ln m|}{2}\\Big)\\Big].$$",
      "rule": "earlier result",
      "by": [
       "C12"
      ]
     },
     {
      "id": "S6",
      "statement": "By symmetry the first derivative at the median is 0. When the groups narrow, few actuals lie near the median, so $g(\\ln m)$ tends to 0. Meanwhile the expectation tends to $\\operatorname{sech}^2(a/2)\\tanh(a/2) > 0$. So the second derivative becomes negative, and the median is a local maximum.",
      "rule": "symmetry",
      "by": [
       "S5"
      ]
     },
     {
      "id": "S7",
      "statement": "A minimiser on each side. The expected sMAPE has a lowest point (C9). It is not at the local maximum $c = 0$. By symmetry, if $c^*$ is a minimiser, so is $-c^*$.",
      "rule": "symmetry",
      "by": [
       "C9"
      ]
     },
     {
      "id": "S8",
      "statement": "Near the modes. Write $\\tanh(|z|/2) = 1 - 2\\lambda(z)$, with $\\lambda(z) = 1/(1 + e^{|z|})$. The function $\\lambda$ is largest at 0 and falls as $|z|$ grows, more slowly the larger $|z|$ is. Minimising the expected $s$ is maximising $\\mathrm E[\\lambda(\\ln y - \\ln m - c)]$. As the groups narrow, this approaches $p\\,\\lambda(c + a) + (1 - p)\\,\\lambda(c - a)$, uniformly in $c$.",
      "rule": "definition",
      "by": [
       "S7"
      ]
     },
     {
      "id": "S9",
      "statement": "With equal weights, take $0 < c < a$. The derivative of this limit is half the rate at which $\\lambda$ falls at distance $a - c$, minus half the rate at distance $a + c$. It is positive, because $\\lambda$ falls faster nearer 0. For $c > a$ the limit decreases. So on $c > 0$ its only maximum is at $c = a$, and the minimisers on that side approach $a$. By symmetry, those on the other side approach $-a$. This proves (a).",
      "rule": "symmetry",
      "by": [
       "S8"
      ]
     },
     {
      "id": "S10",
      "statement": "Unequal weights. The limit increases up to $c = -a$ and decreases after $c = a$. In between it is convex, by the argument of step 9. So its maximum is at $-a$ or at $a$. The difference between these two values is $(2p - 1)\\big(\\lambda(0) - \\lambda(2a)\\big)$, which is positive when $p > 1/2$. So the maximum, and with it the best forecast for narrow groups, is at the mode of the larger group. This proves (b).",
      "rule": "convexity",
      "by": [
       "S9"
      ]
     },
     {
      "id": "S11",
      "statement": "So with two separate groups of actuals, the median is a hump of the expected sMAPE. The best forecasts are near the modes, and with unequal groups the larger one wins. As the weights pass through equality, the target jumps from one mode to the other.",
      "rule": "conclusion",
      "by": [
       "S10"
      ]
     }
    ],
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/evidence/17/description",
    "value": "Proof of C12 (section H12 of the proof document A1, rendered in A2). The target is above the ordinary median when the actuals above it carry more weight, and below it when those below carry more. When the weights balance and enough actuals lie near the median, the median is the target. This is so when the log of the actual is symmetric with a single peak.",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/evidence/17/steps",
    "value": [
     {
      "id": "S1",
      "statement": "The derivative at the median. By C4, step 3, the derivative of the expected sMAPE term in $f$ is $2\\,\\mathrm E[w(y,f)(\\mathbf 1\\{y < f\\} - \\mathbf 1\\{y > f\\})]$. At $f = m$ it is twice the total weight of the actuals below $m$ minus that of the actuals above $m$.",
      "rule": "earlier result",
      "by": [
       "C4"
      ]
     },
     {
      "id": "S2",
      "statement": "The shape of expected sMAPE. With a single local minimum, the expected sMAPE decreases up to it and increases after it. Otherwise it would turn again, at a local maximum. After that turn it would have to fall and then rise once more, because it returns towards 1 at both ends (C9). That would give a second local minimum.",
      "rule": "earlier result",
      "by": [
       "C9"
      ]
     },
     {
      "id": "S3",
      "statement": "(a). If the actuals above $m$ carry more weight, the derivative at the median is negative (step 1): the expected sMAPE is decreasing at the median. By step 2, the median then lies before the minimum, so the target is above the median. The case where the actuals below carry more weight is the mirror image.",
      "rule": "symmetry",
      "by": [
       "S1",
       "S2"
      ]
     },
     {
      "id": "S4",
      "statement": "The second derivative. Measure the forecast on the log scale, $c = \\ln f$, so that the expected contribution is $\\mathrm E[\\tanh(|\\ln y - c|/2)]$ (C3). The derivative of $\\tanh(|x - c|/2)$ in $c$ is $-\\tfrac12\\operatorname{sech}^2\\big(\\tfrac{x-c}{2}\\big)$ for $c < x$ and $+\\tfrac12\\operatorname{sech}^2\\big(\\tfrac{x-c}{2}\\big)$ for $c > x$: it jumps by 1 at $c = x$. Differentiating again and taking the expectation over $x = \\ln y$, the jumps add the density $g$ of $\\ln y$ at $c$, and elsewhere the second derivative is $-\\tfrac12\\operatorname{sech}^2\\big(\\tfrac{x-c}{2}\\big)\\tanh\\big(\\tfrac{|x-c|}{2}\\big)$. So the second derivative of the expected contribution is\n$$g(c) - \\tfrac12\\,\\mathrm E\\Big[\\operatorname{sech}^2\\Big(\\frac{\\ln y - c}{2}\\Big)\\tanh\\Big(\\frac{|\\ln y - c|}{2}\\Big)\\Big].$$\nThis holds when $g$ is continuous at $c$: since $\\operatorname{sech}^2 = 1 - \\tanh^2$, the first derivative is $\\mathrm P(\\ln y \\le c) - \\tfrac12$, whose derivative at $c$ is $g(c)$, plus the expectation of $-\\tfrac12\\tanh^2\\big(\\tfrac{x-c}{2}\\big)$ with the sign of $c - x$, a function of $c$ with a bounded continuous derivative, which may be differentiated inside the expectation.",
      "rule": "continuity",
      "by": [
       "C3"
      ]
     },
     {
      "id": "S5",
      "statement": "(b), the sufficient condition. If the two weights are equal, the first derivative at the median is 0 (step 1): the expected sMAPE is flat there. If, in addition, the stated inequality holds, the second derivative at $c = \\ln m$ is positive (step 4), so the median is a strict local minimum. With a single local minimum, it is the target.",
      "rule": "algebra",
      "by": [
       "S1",
       "S4"
      ]
     },
     {
      "id": "S6",
      "statement": "(b), without the inequality. A point where the first derivative is 0 can be a flat point of inflection, through which the expected sMAPE keeps rising (or keeps falling) without turning. An example, built by the independent checker of this section: let $\\ln y - \\ln m$ be a mixture of three normal distributions, with mean $-1.5$ and standard deviation $0.3$ (weight 0.406125), mean 0 and standard deviation $0.5$ (weight 0.054825), and mean $1.158664$ and standard deviation 1 (weight 0.539050); these values are rounded from the exact solution of the three conditions. Its median is $m$, the two weights are equal there, and the second derivative of step 4 is 0 at $\\ln m$; the expected sMAPE rises through $m$, where it is flat, and its only minimum is near $\\ln f = \\ln m - 1.25$. So equal weights alone do not make the median the target.",
      "rule": "algebra",
      "by": [
       "S4"
      ]
     },
     {
      "id": "S7",
      "statement": "(c). By C5 the median is then the only forecast with the lowest expected sMAPE (a density cannot be constant, so a symmetric unimodal density is strictly lower somewhere away from its centre, as C5 asks). For lognormal actuals $\\ln y$ has a normal density, and for log-Laplace actuals the Laplace density $e^{-|x - \\ln m|/b}/(2b)$; both are symmetric about $\\ln m$ and strictly decreasing in the distance from it.",
      "rule": "earlier result",
      "by": [
       "C5"
      ]
     },
     {
      "id": "S8",
      "statement": "So the target is above the ordinary median when the actuals above it carry more of the weight, and below it when those below carry more. When the two balance, the median is the target if, in addition, enough actuals lie near it, as the stated inequality says; this is the case when the log of the actual is symmetric with a single peak.",
      "rule": "conclusion",
      "by": [
       "S7"
      ]
     }
    ],
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/evidence/18/description",
    "value": "Proof of C17 (the first proof of the proof document A5). The Suilin loss behaves like sMAPE except near zero, where it becomes an absolute error.",
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/evidence/18/steps",
    "value": [
     {
      "id": "S1",
      "statement": "Which branch of the maximum. $\\ell + \\delta \\ge 0.5 + \\delta$ exactly when $\\ell \\ge 0.5$. So the denominator of $\\tilde s$ is $\\ell + \\delta$ when $\\ell \\ge 0.5$, and $0.5 + \\delta$ when $\\ell < 0.5$.",
      "rule": "algebra",
      "by": [
       "D28"
      ]
     },
     {
      "id": "S2",
      "statement": "(a), the identity. When $\\ell \\ge 0.5$, $\\ell > 0$, so $s = |y-f|/\\ell$ and $|y-f| = s\\,\\ell$. By step 1, $\\tilde s = |y-f|/(\\ell+\\delta) = s\\,\\ell/(\\ell+\\delta)$.",
      "rule": "algebra",
      "by": [
       "S1"
      ]
     },
     {
      "id": "S3",
      "statement": "(a), the factor. The function $\\ell \\mapsto \\ell/(\\ell+\\delta) = 1 - \\delta/(\\ell+\\delta)$ is increasing in $\\ell$, below 1, and tends to 1 as $\\ell$ grows. At $\\ell = 0.5$ it equals $0.5/0.6 = 5/6$. So for $\\ell \\ge 0.5$ it lies in $[5/6, 1)$.",
      "rule": "monotonicity",
      "by": [
       "S2"
      ]
     },
     {
      "id": "S4",
      "statement": "(a), close to $s$, but not a function of the ratio. By steps 2 and 3, $0 \\le s - \\tilde s = s\\,\\delta/(\\ell+\\delta) \\le \\delta/(\\ell+\\delta)$, since $s \\le 1$ (step 7). So $\\tilde s$ is within $\\delta/(\\ell+\\delta)$ of $s$, which shrinks as $\\ell$ grows. But $\\tilde s$ depends on the scale: $(y,f) = (1,2)$ and $(10,20)$ have the same ratio and the same $s = 1/3$, while $\\tilde s = 1/3.1$ and $10/30.1$. So $\\tilde s$ is not a function of $y/f$ alone.",
      "rule": "algebra",
      "by": [
       "S2",
       "S3"
      ]
     },
     {
      "id": "S5",
      "statement": "(a), counts. If $y \\ge 1$, then $\\ell \\ge 1 \\ge 0.5$. If $y = 0$, then $\\ell = |f|$, which is at least 0.5 unless $|f| < 0.5$. So on counts only a zero actual with $|f| < 0.5$ has $\\ell < 0.5$.",
      "rule": "algebra",
      "by": [
       "S4"
      ]
     },
     {
      "id": "S6",
      "statement": "(b). When $\\ell < 0.5$, by step 1, $\\tilde s = |y-f|/(0.5+\\delta)$. For $y = f = 0$ this is 0. For $y = 0$ and $|f| < 0.5$ it is $|f|/(0.5+\\delta)$. And $s(0,f) = |f|/|f| = 1$ for $f \\ne 0$, the largest value of $s$ (step 7).",
      "rule": "algebra",
      "by": [
       "S1"
      ]
     },
     {
      "id": "S7",
      "statement": "(c). $|y - f| \\le |y| + |f| = \\ell$ (the triangle inequality), so $s \\le 1$. When $\\ell \\ge 0.5$, $\\tilde s \\le \\ell/(\\ell+\\delta) < 1$ (step 3). When $\\ell < 0.5$, $\\tilde s \\le \\ell/(0.5+\\delta) < 0.5/0.6 < 1$. And $\\tilde s \\ge 0$ because it is a ratio of a non-negative number to a positive one. A quantity between 0 and 1 has an expectation between 0 and 1, whatever the joint distribution of actuals and forecasts, so the expected Suilin loss is between 0 and 200.",
      "rule": "algebra",
      "by": [
       "S3"
      ]
     },
     {
      "id": "S8",
      "statement": "So the Suilin term is the sMAPE term shrunk by a factor between 5/6 and 1 wherever $|y| + |f| \\ge 0.5$, and an absolute error where $|y| + |f| < 0.5$: it differs from sMAPE in kind only near zero.",
      "rule": "conclusion",
      "by": [
       "S7"
      ]
     }
    ],
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "add",
    "path": "/evidence/6",
    "value": {
     "id": "E31",
     "profile": "quant",
     "design": "other",
     "description": "Numerical checks of the formulas of the claims new or reworded in version 3 ([[D31:formula checks]]), lab run RN13 (check_v3.py, seed 8) under the formal plan that decided C3, C4, C12, C17 and C20; every check is within the plan's tolerances. They check the formulas on concrete numbers; the proofs are the evidence for the claims. The tables are labelled by the hypotheses the run's script names: H3 (T1 and T2) checks C3; H4 (T3) checks C4; H34 (T4 and T5) checks C12; H35 (T6 and T7) checks C17; H42 (T8 and T9) checks C20. The tables of H39 (T10 to T12) and H40 (T13 to T17) check the formulas of C18 and C19, whose wording was decided after later checks (E32, E33).",
     "availability": {
      "level": "public"
     },
     "lab_runs": [
      "RN13"
     ],
     "tables": [
      {
       "id": "T1",
       "description": "violations of the bound on the log-MAE of the sMAPE minimiser, largest excess of the minimiser's log-MAE over the median's, minus the bound; hypothesis: H3, test_sets: 2000 (lab runs RN13)",
       "conditions": {
        "hypothesis": "H3",
        "test_sets": 2000
       },
       "row_labels": [],
       "columns": [
        "violations of the bound on the log-MAE of the sMAPE minimiser",
        "largest excess of the minimiser's log-MAE over the median's, minus the bound"
       ],
       "rows": [
        {
         "id": "r1",
         "label": [],
         "values": [
          0,
          -1.263e-06
         ]
        }
       ],
       "artefact": "A6"
      },
      {
       "description": "sum of sMAPE terms, forecast at the median, actuals 1, 1, e, e^4, e^4, sum of sMAPE terms, forecast at 1, actuals 1, 1, e, e^4, e^4; hypothesis: H3 (lab runs RN13)",
       "row_labels": [],
       "columns": [
        "sum of sMAPE terms, forecast at the median, actuals 1, 1, e, e^4, e^4",
        "sum of sMAPE terms, forecast at 1, actuals 1, 1, e, e^4, e^4"
       ],
       "rows": [
        {
         "id": "r1",
         "label": [],
         "values": [
          2.735,
          2.39
         ]
        }
       ],
       "conditions": {
        "hypothesis": "H3"
       },
       "artefact": "A6",
       "id": "T2"
      },
      {
       "description": "largest difference of the weights below and above the target, over their sum; hypothesis: H4, distributions: 6 (lab runs RN13)",
       "row_labels": [],
       "columns": [
        "largest difference of the weights below and above the target, over their sum"
       ],
       "rows": [
        {
         "id": "r1",
         "label": [],
         "values": [
          2.089e-08
         ]
        }
       ],
       "conditions": {
        "hypothesis": "H4",
        "distributions": 6
       },
       "artefact": "A6",
       "id": "T3"
      },
      {
       "description": "distributions with more than one local minimum, distributions where the target is not on the side of the median that carries more weight; hypothesis: H34, distributions: 6 (lab runs RN13)",
       "row_labels": [],
       "columns": [
        "distributions with more than one local minimum",
        "distributions where the target is not on the side of the median that carries more weight"
       ],
       "rows": [
        {
         "id": "r1",
         "label": [],
         "values": [
          0,
          0
         ]
        }
       ],
       "conditions": {
        "hypothesis": "H34",
        "distributions": 6
       },
       "artefact": "A6",
       "id": "T4"
      },
      {
       "description": "largest distance in logs between the target and the median, lognormal and log-Laplace; hypothesis: H34, distributions: 16 (lab runs RN13)",
       "row_labels": [],
       "columns": [
        "largest distance in logs between the target and the median, lognormal and log-Laplace"
       ],
       "rows": [
        {
         "id": "r1",
         "label": [],
         "values": [
          1.753e-07
         ]
        }
       ],
       "conditions": {
        "hypothesis": "H34",
        "distributions": 16
       },
       "artefact": "A6",
       "id": "T5"
      },
      {
       "description": "Suilin term of a zero forecast of a zero actual; hypothesis: H35 (lab runs RN13)",
       "row_labels": [],
       "columns": [
        "Suilin term of a zero forecast of a zero actual"
       ],
       "rows": [
        {
         "id": "r1",
         "label": [],
         "values": [
          0.0
         ]
        }
       ],
       "conditions": {
        "hypothesis": "H35"
       },
       "artefact": "A6",
       "id": "T6"
      },
      {
       "id": "T7",
       "description": "largest error of the identity of (a), largest error of the identity of (b), values of the Suilin term outside [0, 1); hypothesis: H35, pairs: 200000 (lab runs RN13)",
       "conditions": {
        "hypothesis": "H35",
        "pairs": 200000
       },
       "row_labels": [],
       "columns": [
        "largest error of the identity of (a)",
        "largest error of the identity of (b)",
        "values of the Suilin term outside [0, 1)"
       ],
       "rows": [
        {
         "id": "r1",
         "label": [],
         "values": [
          1.11e-16,
          0.0,
          0
         ]
        }
       ],
       "artefact": "A6"
      },
      {
       "id": "T8",
       "description": "largest error of the identity of (a), largest error of the identity of (b), values of the msMAPE term outside [0, 1], values of the msMAPE term not below 1 for actuals below nu, pairs with a count of at least 1 where the msMAPE and sMAPE terms differ by nu; hypothesis: H42, pairs: 200000 (lab runs RN13)",
       "conditions": {
        "hypothesis": "H42",
        "pairs": 200000
       },
       "row_labels": [
        "nu"
       ],
       "columns": [
        "largest error of the identity of (a)",
        "largest error of the identity of (b)",
        "values of the msMAPE term outside [0, 1]",
        "values of the msMAPE term not below 1 for actuals below nu",
        "pairs with a count of at least 1 where the msMAPE and sMAPE terms differ"
       ],
       "rows": [
        {
         "id": "r1",
         "label": [
          0.1
         ],
         "values": [
          0.0,
          0.0,
          0,
          0,
          0
         ]
        },
        {
         "id": "r2",
         "label": [
          0.5
         ],
         "values": [
          0.0,
          0.0,
          0,
          0,
          0
         ]
        },
        {
         "id": "r3",
         "label": [
          1.0
         ],
         "values": [
          0.0,
          0.0,
          0,
          0,
          0
         ]
        }
       ],
       "artefact": "A6"
      },
      {
       "id": "T9",
       "description": "msMAPE term of a forecast of 1e-9 for a zero actual by nu; hypothesis: H42 (lab runs RN13)",
       "conditions": {
        "hypothesis": "H42"
       },
       "row_labels": [
        "nu"
       ],
       "columns": [
        "msMAPE term of a forecast of 1e-9 for a zero actual"
       ],
       "rows": [
        {
         "id": "r1",
         "label": [
          0.1
         ],
         "values": [
          1e-08
         ]
        },
        {
         "id": "r2",
         "label": [
          0.5
         ],
         "values": [
          2e-09
         ]
        },
        {
         "id": "r3",
         "label": [
          1.0
         ],
         "values": [
          1e-09
         ]
        }
       ],
       "artefact": "A6"
      },
      {
       "id": "T10",
       "description": "signal-to-noise ratio of sMAPE over its large-spread formula, signal-to-noise ratio of the absolute error over its limit by error_distribution, c1, c2; hypothesis: H39, sigma: 200 (lab runs RN13)",
       "conditions": {
        "hypothesis": "H39",
        "sigma": 200
       },
       "row_labels": [
        "error_distribution",
        "c1",
        "c2"
       ],
       "columns": [
        "signal-to-noise ratio of sMAPE over its large-spread formula",
        "signal-to-noise ratio of the absolute error over its limit"
       ],
       "rows": [
        {
         "id": "r1",
         "label": [
          "normal",
          0.0,
          0.5
         ],
         "values": [
          0.9999,
          0.9913
         ]
        },
        {
         "id": "r2",
         "label": [
          "normal",
          0.0,
          1.0
         ],
         "values": [
          1.0,
          0.9956
         ]
        },
        {
         "id": "r3",
         "label": [
          "normal",
          -0.5,
          1.5
         ],
         "values": [
          1.001,
          0.9966
         ]
        },
        {
         "id": "r4",
         "label": [
          "Gumbel",
          0.0,
          0.5
         ],
         "values": [
          0.9998,
          0.9647
         ]
        },
        {
         "id": "r5",
         "label": [
          "Gumbel",
          0.0,
          1.0
         ],
         "values": [
          1.0,
          0.9937
         ]
        },
        {
         "id": "r6",
         "label": [
          "Gumbel",
          -0.5,
          1.5
         ],
         "values": [
          1.0,
          0.9967
         ]
        }
       ],
       "artefact": "A6"
      },
      {
       "id": "T11",
       "description": "relative spread of the absolute log error's signal-to-noise ratio over sigma = 2, 10, 50 by error_distribution, c1, c2; hypothesis: H39 (lab runs RN13)",
       "conditions": {
        "hypothesis": "H39"
       },
       "row_labels": [
        "error_distribution",
        "c1",
        "c2"
       ],
       "columns": [
        "relative spread of the absolute log error's signal-to-noise ratio over sigma = 2, 10, 50"
       ],
       "rows": [
        {
         "id": "r1",
         "label": [
          "normal",
          0.0,
          0.5
         ],
         "values": [
          1.686e-15
         ]
        },
        {
         "id": "r2",
         "label": [
          "normal",
          0.0,
          1.0
         ],
         "values": [
          8.388e-16
         ]
        },
        {
         "id": "r3",
         "label": [
          "normal",
          -0.5,
          1.5
         ],
         "values": [
          4.51e-16
         ]
        },
        {
         "id": "r4",
         "label": [
          "Gumbel",
          0.0,
          0.5
         ],
         "values": [
          4.186e-15
         ]
        },
        {
         "id": "r5",
         "label": [
          "Gumbel",
          0.0,
          1.0
         ],
         "values": [
          4.885e-15
         ]
        },
        {
         "id": "r6",
         "label": [
          "Gumbel",
          -0.5,
          1.5
         ],
         "values": [
          1.356e-14
         ]
        }
       ],
       "artefact": "A6"
      },
      {
       "id": "T12",
       "description": "signal-to-noise ratio of the squared error over its formula, lognormal, exp(sigma^2 / 2) times the squared error's ratio, lognormal, formula by c1, c2, sigma; hypothesis: H39 (lab runs RN13)",
       "conditions": {
        "hypothesis": "H39"
       },
       "row_labels": [
        "c1",
        "c2",
        "sigma"
       ],
       "columns": [
        "signal-to-noise ratio of the squared error over its formula, lognormal",
        "exp(sigma^2 / 2) times the squared error's ratio, lognormal, formula"
       ],
       "rows": [
        {
         "id": "r1",
         "label": [
          0.0,
          0.5,
          0.5
         ],
         "values": [
          1.0,
          null
         ]
        },
        {
         "id": "r2",
         "label": [
          0.0,
          1.0,
          0.5
         ],
         "values": [
          1.0,
          null
         ]
        },
        {
         "id": "r3",
         "label": [
          -0.5,
          1.5,
          0.5
         ],
         "values": [
          1.0,
          null
         ]
        },
        {
         "id": "r4",
         "label": [
          0.0,
          0.5,
          1.0
         ],
         "values": [
          1.0,
          null
         ]
        },
        {
         "id": "r5",
         "label": [
          0.0,
          1.0,
          1.0
         ],
         "values": [
          1.0,
          null
         ]
        },
        {
         "id": "r6",
         "label": [
          -0.5,
          1.5,
          1.0
         ],
         "values": [
          1.0,
          null
         ]
        },
        {
         "id": "r7",
         "label": [
          0.0,
          0.5,
          2.0
         ],
         "values": [
          1.0,
          null
         ]
        },
        {
         "id": "r8",
         "label": [
          0.0,
          1.0,
          2.0
         ],
         "values": [
          1.0,
          null
         ]
        },
        {
         "id": "r9",
         "label": [
          -0.5,
          1.5,
          2.0
         ],
         "values": [
          1.0,
          null
         ]
        },
        {
         "id": "r10",
         "label": [
          0.0,
          1.0,
          4.0
         ],
         "values": [
          null,
          0.9907
         ]
        },
        {
         "id": "r11",
         "label": [
          0.0,
          1.0,
          6.0
         ],
         "values": [
          null,
          1.0
         ]
        }
       ],
       "artefact": "A6"
      },
      {
       "description": "distributions compared with the asymptotic values recorded by RN7; hypothesis: H40 (lab runs RN13)",
       "row_labels": [],
       "columns": [
        "distributions compared with the asymptotic values recorded by RN7"
       ],
       "rows": [
        {
         "id": "r1",
         "label": [],
         "values": [
          16
         ]
        }
       ],
       "conditions": {
        "hypothesis": "H40"
       },
       "artefact": "A6",
       "id": "T13"
      },
      {
       "description": "distributions where A is not below 2 g at the median, largest difference from the asymptotic values recorded by RN7; hypothesis: H40, distributions: 16 (lab runs RN13)",
       "row_labels": [],
       "columns": [
        "distributions where A is not below 2 g at the median",
        "largest difference from the asymptotic values recorded by RN7"
       ],
       "rows": [
        {
         "id": "r1",
         "label": [],
         "values": [
          0,
          0.00014
         ]
        }
       ],
       "conditions": {
        "hypothesis": "H40",
        "distributions": 16
       },
       "artefact": "A6",
       "id": "T14"
      },
      {
       "id": "T15",
       "description": "large-sample relative efficiency, formula by family, spread; hypothesis: H40 (lab runs RN13)",
       "conditions": {
        "hypothesis": "H40"
       },
       "row_labels": [
        "family",
        "spread"
       ],
       "columns": [
        "large-sample relative efficiency, formula"
       ],
       "rows": [
        {
         "id": "r1",
         "label": [
          "lognormal",
          0.25
         ],
         "values": [
          0.9696
         ]
        },
        {
         "id": "r2",
         "label": [
          "lognormal",
          0.5
         ],
         "values": [
          0.8881
         ]
        },
        {
         "id": "r3",
         "label": [
          "lognormal",
          1.0
         ],
         "values": [
          0.6601
         ]
        },
        {
         "id": "r4",
         "label": [
          "lognormal",
          1.5
         ],
         "values": [
          0.454
         ]
        },
        {
         "id": "r5",
         "label": [
          "lognormal",
          2.0
         ],
         "values": [
          0.3073
         ]
        },
        {
         "id": "r6",
         "label": [
          "lognormal",
          3.0
         ],
         "values": [
          0.1474
         ]
        },
        {
         "id": "r7",
         "label": [
          "lognormal",
          4.0
         ],
         "values": [
          0.07778
         ]
        },
        {
         "id": "r8",
         "label": [
          "lognormal",
          6.0
         ],
         "values": [
          0.0279
         ]
        },
        {
         "id": "r9",
         "label": [
          "log-Laplace",
          0.25
         ],
         "values": [
          0.9969
         ]
        },
        {
         "id": "r10",
         "label": [
          "log-Laplace",
          0.5
         ],
         "values": [
          0.9765
         ]
        },
        {
         "id": "r11",
         "label": [
          "log-Laplace",
          1.0
         ],
         "values": [
          0.8953
         ]
        },
        {
         "id": "r12",
         "label": [
          "log-Laplace",
          1.5
         ],
         "values": [
          0.8036
         ]
        },
        {
         "id": "r13",
         "label": [
          "log-Laplace",
          2.0
         ],
         "values": [
          0.7207
         ]
        },
        {
         "id": "r14",
         "label": [
          "log-Laplace",
          3.0
         ],
         "values": [
          0.5903
         ]
        },
        {
         "id": "r15",
         "label": [
          "log-Laplace",
          4.0
         ],
         "values": [
          0.4969
         ]
        },
        {
         "id": "r16",
         "label": [
          "log-Laplace",
          6.0
         ],
         "values": [
          0.3755
         ]
        }
       ],
       "artefact": "A6"
      },
      {
       "id": "T16",
       "description": "distance of the large-sample relative efficiency from 1 at spread 0.01 by family; hypothesis: H40 (lab runs RN13)",
       "conditions": {
        "hypothesis": "H40"
       },
       "row_labels": [
        "family"
       ],
       "columns": [
        "distance of the large-sample relative efficiency from 1 at spread 0.01"
       ],
       "rows": [
        {
         "id": "r1",
         "label": [
          "lognormal"
         ],
         "values": [
          5e-05
         ]
        },
        {
         "id": "r2",
         "label": [
          "log-Laplace"
         ],
         "values": [
          1.249e-08
         ]
        }
       ],
       "artefact": "A6"
      },
      {
       "id": "T17",
       "description": "large-sample relative efficiency times sigma^3 over 6 (ln 2)^2 sqrt(2 pi), lognormal, large-sample relative efficiency times b over 3, log-Laplace; hypothesis: H40, spread: 200 (lab runs RN13)",
       "conditions": {
        "hypothesis": "H40",
        "spread": 200
       },
       "row_labels": [],
       "columns": [
        "large-sample relative efficiency times sigma^3 over 6 (ln 2)^2 sqrt(2 pi), lognormal",
        "large-sample relative efficiency times b over 3, log-Laplace"
       ],
       "rows": [
        {
         "id": "r1",
         "label": [],
         "values": [
          0.9998,
          0.9906
         ]
        }
       ],
       "artefact": "A6"
      }
     ],
     "artefacts": [
      "A6"
     ],
     "setup": {
      "studied": [
       "D1",
       "D28",
       "D33",
       "D6",
       "D8"
      ],
      "lens": [
       "D32",
       "D9",
       "D5",
       "D4",
       "D19",
       "D11",
       "D10"
      ],
      "procedure": [
       "D31"
      ]
     },
     "execution": {
      "code_version": "2f0d2fca48b049d8004fec7228e47b8857f6ace4",
      "software": {
       "python": "3.13.18",
       "platform": "Linux-6.18.44-fc-v595-x86_64-with-glibc2.39"
      }
     }
    },
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "add",
    "path": "/evidence/7",
    "value": {
     "id": "E32",
     "profile": "quant",
     "design": "other",
     "description": "Numerical checks of the formulas of C18 ([[D31:formula checks]]), lab run RN18 (check_h39.py) under the formal plan that decided C18; every check is within the plan's tolerances. They check the formulas on concrete numbers; the proof is the evidence for the claim. The run's script labels the checks H39, the record's first wording of C18; the formulas are those of C18.",
     "availability": {
      "level": "public"
     },
     "lab_runs": [
      "RN18"
     ],
     "tables": [
      {
       "id": "T1",
       "description": "distance of sMAPE's signal-to-noise ratio over its large-spread formula from 1, distance of the absolute error's signal-to-noise ratio over its limit from 1 by error_distribution, c1, c2, sigma; hypothesis: H39 (lab runs RN18)",
       "conditions": {
        "hypothesis": "H39"
       },
       "row_labels": [
        "error_distribution",
        "c1",
        "c2",
        "sigma"
       ],
       "columns": [
        "distance of sMAPE's signal-to-noise ratio over its large-spread formula from 1",
        "distance of the absolute error's signal-to-noise ratio over its limit from 1"
       ],
       "rows": [
        {
         "id": "r1",
         "label": [
          "normal",
          0.0,
          0.5,
          200
         ],
         "values": [
          5.266e-05,
          0.008727
         ]
        },
        {
         "id": "r2",
         "label": [
          "normal",
          0.0,
          0.5,
          2000
         ],
         "values": [
          2.749e-06,
          0.0008713
         ]
        },
        {
         "id": "r3",
         "label": [
          "normal",
          0.0,
          1.0,
          200
         ],
         "values": [
          0.0003989,
          0.004379
         ]
        },
        {
         "id": "r4",
         "label": [
          "normal",
          0.0,
          1.0,
          2000
         ],
         "values": [
          4.703e-05,
          0.0004373
         ]
        },
        {
         "id": "r5",
         "label": [
          "normal",
          -0.5,
          1.5,
          200
         ],
         "values": [
          0.00122,
          0.003402
         ]
        },
        {
         "id": "r6",
         "label": [
          "normal",
          -0.5,
          1.5,
          2000
         ],
         "values": [
          0.0001273,
          0.0003398
         ]
        },
        {
         "id": "r7",
         "label": [
          "Gumbel",
          0.0,
          0.5,
          200
         ],
         "values": [
          0.0001627,
          0.03529
         ]
        },
        {
         "id": "r8",
         "label": [
          "Gumbel",
          0.0,
          0.5,
          2000
         ],
         "values": [
          5.872e-07,
          0.00352
         ]
        },
        {
         "id": "r9",
         "label": [
          "Gumbel",
          0.0,
          1.0,
          200
         ],
         "values": [
          0.0001512,
          0.006295
         ]
        },
        {
         "id": "r10",
         "label": [
          "Gumbel",
          0.0,
          1.0,
          2000
         ],
         "values": [
          2.457e-05,
          0.0006281
         ]
        },
        {
         "id": "r11",
         "label": [
          "Gumbel",
          -0.5,
          1.5,
          200
         ],
         "values": [
          0.0003779,
          0.00326
         ]
        },
        {
         "id": "r12",
         "label": [
          "Gumbel",
          -0.5,
          1.5,
          2000
         ],
         "values": [
          4.713e-05,
          0.0003255
         ]
        }
       ],
       "artefact": "A6"
      },
      {
       "id": "T2",
       "description": "relative spread of the absolute log error's signal-to-noise ratio over sigma = 2, 10, 50 by error_distribution, c1, c2; hypothesis: H39 (lab runs RN18)",
       "conditions": {
        "hypothesis": "H39"
       },
       "row_labels": [
        "error_distribution",
        "c1",
        "c2"
       ],
       "columns": [
        "relative spread of the absolute log error's signal-to-noise ratio over sigma = 2, 10, 50"
       ],
       "rows": [
        {
         "id": "r1",
         "label": [
          "normal",
          0.0,
          0.5
         ],
         "values": [
          3.761e-15
         ]
        },
        {
         "id": "r2",
         "label": [
          "normal",
          0.0,
          1.0
         ],
         "values": [
          2.397e-16
         ]
        },
        {
         "id": "r3",
         "label": [
          "normal",
          -0.5,
          1.5
         ],
         "values": [
          1.579e-15
         ]
        },
        {
         "id": "r4",
         "label": [
          "Gumbel",
          0.0,
          0.5
         ],
         "values": [
          4.243e-13
         ]
        },
        {
         "id": "r5",
         "label": [
          "Gumbel",
          0.0,
          1.0
         ],
         "values": [
          7.932e-13
         ]
        },
        {
         "id": "r6",
         "label": [
          "Gumbel",
          -0.5,
          1.5
         ],
         "values": [
          1.43e-12
         ]
        }
       ],
       "artefact": "A6"
      },
      {
       "id": "T3",
       "description": "distance of the squared error's signal-to-noise ratio over its formula from 1, lognormal, distance of exp(sigma^2 / 2) times the squared error's ratio from 1, lognormal, formula by c1, c2, sigma; hypothesis: H39 (lab runs RN18)",
       "conditions": {
        "hypothesis": "H39"
       },
       "row_labels": [
        "c1",
        "c2",
        "sigma"
       ],
       "columns": [
        "distance of the squared error's signal-to-noise ratio over its formula from 1, lognormal",
        "distance of exp(sigma^2 / 2) times the squared error's ratio from 1, lognormal, formula"
       ],
       "rows": [
        {
         "id": "r1",
         "label": [
          0.0,
          0.5,
          0.5
         ],
         "values": [
          3.109e-15,
          null
         ]
        },
        {
         "id": "r2",
         "label": [
          0.0,
          1.0,
          0.5
         ],
         "values": [
          6.661e-16,
          null
         ]
        },
        {
         "id": "r3",
         "label": [
          -0.5,
          1.5,
          0.5
         ],
         "values": [
          1.11e-15,
          null
         ]
        },
        {
         "id": "r4",
         "label": [
          0.0,
          0.5,
          1.0
         ],
         "values": [
          1.554e-15,
          null
         ]
        },
        {
         "id": "r5",
         "label": [
          0.0,
          1.0,
          1.0
         ],
         "values": [
          8.882e-16,
          null
         ]
        },
        {
         "id": "r6",
         "label": [
          -0.5,
          1.5,
          1.0
         ],
         "values": [
          5.551e-16,
          null
         ]
        },
        {
         "id": "r7",
         "label": [
          0.0,
          0.5,
          2.0
         ],
         "values": [
          1.409e-13,
          null
         ]
        },
        {
         "id": "r8",
         "label": [
          0.0,
          1.0,
          2.0
         ],
         "values": [
          7.55e-15,
          null
         ]
        },
        {
         "id": "r9",
         "label": [
          -0.5,
          1.5,
          2.0
         ],
         "values": [
          1.887e-15,
          null
         ]
        },
        {
         "id": "r10",
         "label": [
          0.0,
          1.0,
          4.0
         ],
         "values": [
          null,
          0.009325
         ]
        },
        {
         "id": "r11",
         "label": [
          0.0,
          1.0,
          6.0
         ],
         "values": [
          null,
          3.08e-06
         ]
        },
        {
         "id": "r12",
         "label": [
          0.0,
          1.0,
          8.0
         ],
         "values": [
          null,
          1.888e-11
         ]
        }
       ],
       "artefact": "A6"
      }
     ],
     "artefacts": [
      "A6"
     ],
     "setup": {
      "studied": [
       "D1"
      ],
      "lens": [
       "D32",
       "D19",
       "D11",
       "D10"
      ],
      "procedure": [
       "D31"
      ]
     },
     "execution": {
      "code_version": "6aa7a7ab87bf5c4e40d704a7492cd9af9ef481db",
      "software": {
       "python": "3.13.18",
       "platform": "Linux-6.18.44-fc-v595-x86_64-with-glibc2.39"
      }
     }
    },
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "add",
    "path": "/evidence/8",
    "value": {
     "id": "E33",
     "profile": "quant",
     "design": "other",
     "description": "Numerical checks of the formulas of C19 ([[D31:formula checks]]), lab run RN16 (check_corrected.py, seed 8) under the formal plan that decided C19; every check is within the plan's tolerances. They check the formulas on concrete numbers; the proof is the evidence for the claim. The run's script labels the checks by the record's first wordings: H40 (T4 to T7 and the result R1) checks C19; H39 (T1 to T3) checks the formulas of C18, which the same plan tested in an earlier wording.",
     "availability": {
      "level": "public"
     },
     "lab_runs": [
      "RN16"
     ],
     "results": [
      {
       "id": "R1",
       "variable": "distributions compared with the asymptotic values recorded by RN7",
       "value": 16,
       "conditions": {
        "hypothesis": "H40"
       }
      }
     ],
     "tables": [
      {
       "id": "T1",
       "description": "distance of sMAPE's signal-to-noise ratio over its large-spread formula from 1, distance of the absolute error's signal-to-noise ratio over its limit from 1 by error_distribution, c1, c2, sigma; hypothesis: H39 (lab runs RN16)",
       "conditions": {
        "hypothesis": "H39"
       },
       "row_labels": [
        "error_distribution",
        "c1",
        "c2",
        "sigma"
       ],
       "columns": [
        "distance of sMAPE's signal-to-noise ratio over its large-spread formula from 1",
        "distance of the absolute error's signal-to-noise ratio over its limit from 1"
       ],
       "rows": [
        {
         "id": "r1",
         "label": [
          "normal",
          0.0,
          0.5,
          200
         ],
         "values": [
          5.266e-05,
          0.008727
         ]
        },
        {
         "id": "r2",
         "label": [
          "normal",
          0.0,
          0.5,
          2000
         ],
         "values": [
          2.749e-06,
          0.0008713
         ]
        },
        {
         "id": "r3",
         "label": [
          "normal",
          0.0,
          1.0,
          200
         ],
         "values": [
          0.0003989,
          0.004379
         ]
        },
        {
         "id": "r4",
         "label": [
          "normal",
          0.0,
          1.0,
          2000
         ],
         "values": [
          4.703e-05,
          0.0004373
         ]
        },
        {
         "id": "r5",
         "label": [
          "normal",
          -0.5,
          1.5,
          200
         ],
         "values": [
          0.00122,
          0.003402
         ]
        },
        {
         "id": "r6",
         "label": [
          "normal",
          -0.5,
          1.5,
          2000
         ],
         "values": [
          0.0001273,
          0.0003398
         ]
        },
        {
         "id": "r7",
         "label": [
          "Gumbel",
          0.0,
          0.5,
          200
         ],
         "values": [
          0.0001627,
          0.03529
         ]
        },
        {
         "id": "r8",
         "label": [
          "Gumbel",
          0.0,
          0.5,
          2000
         ],
         "values": [
          5.872e-07,
          0.00352
         ]
        },
        {
         "id": "r9",
         "label": [
          "Gumbel",
          0.0,
          1.0,
          200
         ],
         "values": [
          0.0001512,
          0.006295
         ]
        },
        {
         "id": "r10",
         "label": [
          "Gumbel",
          0.0,
          1.0,
          2000
         ],
         "values": [
          2.457e-05,
          0.0006281
         ]
        },
        {
         "id": "r11",
         "label": [
          "Gumbel",
          -0.5,
          1.5,
          200
         ],
         "values": [
          0.0003779,
          0.00326
         ]
        },
        {
         "id": "r12",
         "label": [
          "Gumbel",
          -0.5,
          1.5,
          2000
         ],
         "values": [
          4.713e-05,
          0.0003255
         ]
        }
       ],
       "artefact": "A6"
      },
      {
       "id": "T2",
       "description": "relative spread of the absolute log error's signal-to-noise ratio over sigma = 2, 10, 50 by error_distribution, c1, c2; hypothesis: H39 (lab runs RN16)",
       "conditions": {
        "hypothesis": "H39"
       },
       "row_labels": [
        "error_distribution",
        "c1",
        "c2"
       ],
       "columns": [
        "relative spread of the absolute log error's signal-to-noise ratio over sigma = 2, 10, 50"
       ],
       "rows": [
        {
         "id": "r1",
         "label": [
          "normal",
          0.0,
          0.5
         ],
         "values": [
          3.761e-15
         ]
        },
        {
         "id": "r2",
         "label": [
          "normal",
          0.0,
          1.0
         ],
         "values": [
          2.397e-16
         ]
        },
        {
         "id": "r3",
         "label": [
          "normal",
          -0.5,
          1.5
         ],
         "values": [
          1.579e-15
         ]
        },
        {
         "id": "r4",
         "label": [
          "Gumbel",
          0.0,
          0.5
         ],
         "values": [
          4.243e-13
         ]
        },
        {
         "id": "r5",
         "label": [
          "Gumbel",
          0.0,
          1.0
         ],
         "values": [
          7.932e-13
         ]
        },
        {
         "id": "r6",
         "label": [
          "Gumbel",
          -0.5,
          1.5
         ],
         "values": [
          1.43e-12
         ]
        }
       ],
       "artefact": "A6"
      },
      {
       "id": "T3",
       "description": "distance of the squared error's signal-to-noise ratio over its formula from 1, lognormal, distance of exp(sigma^2 / 2) times the squared error's ratio from 1, lognormal, formula by c1, c2, sigma; hypothesis: H39 (lab runs RN16)",
       "conditions": {
        "hypothesis": "H39"
       },
       "row_labels": [
        "c1",
        "c2",
        "sigma"
       ],
       "columns": [
        "distance of the squared error's signal-to-noise ratio over its formula from 1, lognormal",
        "distance of exp(sigma^2 / 2) times the squared error's ratio from 1, lognormal, formula"
       ],
       "rows": [
        {
         "id": "r1",
         "label": [
          0.0,
          0.5,
          0.5
         ],
         "values": [
          3.109e-15,
          null
         ]
        },
        {
         "id": "r2",
         "label": [
          0.0,
          1.0,
          0.5
         ],
         "values": [
          6.661e-16,
          null
         ]
        },
        {
         "id": "r3",
         "label": [
          -0.5,
          1.5,
          0.5
         ],
         "values": [
          1.11e-15,
          null
         ]
        },
        {
         "id": "r4",
         "label": [
          0.0,
          0.5,
          1.0
         ],
         "values": [
          1.554e-15,
          null
         ]
        },
        {
         "id": "r5",
         "label": [
          0.0,
          1.0,
          1.0
         ],
         "values": [
          8.882e-16,
          null
         ]
        },
        {
         "id": "r6",
         "label": [
          -0.5,
          1.5,
          1.0
         ],
         "values": [
          5.551e-16,
          null
         ]
        },
        {
         "id": "r7",
         "label": [
          0.0,
          0.5,
          2.0
         ],
         "values": [
          1.409e-13,
          null
         ]
        },
        {
         "id": "r8",
         "label": [
          0.0,
          1.0,
          2.0
         ],
         "values": [
          7.55e-15,
          null
         ]
        },
        {
         "id": "r9",
         "label": [
          -0.5,
          1.5,
          2.0
         ],
         "values": [
          1.887e-15,
          null
         ]
        },
        {
         "id": "r10",
         "label": [
          0.0,
          1.0,
          4.0
         ],
         "values": [
          null,
          0.009325
         ]
        },
        {
         "id": "r11",
         "label": [
          0.0,
          1.0,
          6.0
         ],
         "values": [
          null,
          3.08e-06
         ]
        },
        {
         "id": "r12",
         "label": [
          0.0,
          1.0,
          8.0
         ],
         "values": [
          null,
          1.888e-11
         ]
        }
       ],
       "artefact": "A6"
      },
      {
       "id": "T4",
       "description": "large-sample relative efficiency, formula by family, spread; hypothesis: H40 (lab runs RN16)",
       "conditions": {
        "hypothesis": "H40"
       },
       "row_labels": [
        "family",
        "spread"
       ],
       "columns": [
        "large-sample relative efficiency, formula"
       ],
       "rows": [
        {
         "id": "r1",
         "label": [
          "lognormal",
          0.25
         ],
         "values": [
          0.9696
         ]
        },
        {
         "id": "r2",
         "label": [
          "lognormal",
          0.5
         ],
         "values": [
          0.8881
         ]
        },
        {
         "id": "r3",
         "label": [
          "lognormal",
          1.0
         ],
         "values": [
          0.6601
         ]
        },
        {
         "id": "r4",
         "label": [
          "lognormal",
          1.5
         ],
         "values": [
          0.454
         ]
        },
        {
         "id": "r5",
         "label": [
          "lognormal",
          2.0
         ],
         "values": [
          0.3073
         ]
        },
        {
         "id": "r6",
         "label": [
          "lognormal",
          3.0
         ],
         "values": [
          0.1474
         ]
        },
        {
         "id": "r7",
         "label": [
          "lognormal",
          4.0
         ],
         "values": [
          0.07778
         ]
        },
        {
         "id": "r8",
         "label": [
          "lognormal",
          6.0
         ],
         "values": [
          0.0279
         ]
        },
        {
         "id": "r9",
         "label": [
          "log-Laplace",
          0.25
         ],
         "values": [
          0.9969
         ]
        },
        {
         "id": "r10",
         "label": [
          "log-Laplace",
          0.5
         ],
         "values": [
          0.9765
         ]
        },
        {
         "id": "r11",
         "label": [
          "log-Laplace",
          1.0
         ],
         "values": [
          0.8953
         ]
        },
        {
         "id": "r12",
         "label": [
          "log-Laplace",
          1.5
         ],
         "values": [
          0.8036
         ]
        },
        {
         "id": "r13",
         "label": [
          "log-Laplace",
          2.0
         ],
         "values": [
          0.7207
         ]
        },
        {
         "id": "r14",
         "label": [
          "log-Laplace",
          3.0
         ],
         "values": [
          0.5903
         ]
        },
        {
         "id": "r15",
         "label": [
          "log-Laplace",
          4.0
         ],
         "values": [
          0.4969
         ]
        },
        {
         "id": "r16",
         "label": [
          "log-Laplace",
          6.0
         ],
         "values": [
          0.3755
         ]
        }
       ],
       "artefact": "A6"
      },
      {
       "id": "T5",
       "description": "distance of the large-sample relative efficiency from 1 at spread 0.01 by family; hypothesis: H40 (lab runs RN16)",
       "conditions": {
        "hypothesis": "H40"
       },
       "row_labels": [
        "family"
       ],
       "columns": [
        "distance of the large-sample relative efficiency from 1 at spread 0.01"
       ],
       "rows": [
        {
         "id": "r1",
         "label": [
          "lognormal"
         ],
         "values": [
          5e-05
         ]
        },
        {
         "id": "r2",
         "label": [
          "log-Laplace"
         ],
         "values": [
          1.249e-08
         ]
        }
       ],
       "artefact": "A6"
      },
      {
       "id": "T6",
       "description": "distributions where A is not below 2 g at the median, largest difference from the asymptotic values recorded by RN7; hypothesis: H40, distributions: 16 (lab runs RN16)",
       "conditions": {
        "hypothesis": "H40",
        "distributions": 16
       },
       "row_labels": [],
       "columns": [
        "distributions where A is not below 2 g at the median",
        "largest difference from the asymptotic values recorded by RN7"
       ],
       "rows": [
        {
         "id": "r1",
         "label": [],
         "values": [
          0,
          0.00014
         ]
        }
       ],
       "artefact": "A6"
      },
      {
       "id": "T7",
       "description": "large-sample relative efficiency times sigma^3 over 6 (ln 2)^2 sqrt(2 pi), lognormal, large-sample relative efficiency times b over 3, log-Laplace; hypothesis: H40, spread: 200 (lab runs RN16)",
       "conditions": {
        "hypothesis": "H40",
        "spread": 200
       },
       "row_labels": [],
       "columns": [
        "large-sample relative efficiency times sigma^3 over 6 (ln 2)^2 sqrt(2 pi), lognormal",
        "large-sample relative efficiency times b over 3, log-Laplace"
       ],
       "rows": [
        {
         "id": "r1",
         "label": [],
         "values": [
          0.9998,
          0.9906
         ]
        }
       ],
       "artefact": "A6"
      }
     ],
     "artefacts": [
      "A6"
     ],
     "setup": {
      "studied": [
       "D1",
       "D8"
      ],
      "lens": [
       "D32",
       "D9",
       "D19",
       "D11",
       "D10"
      ],
      "procedure": [
       "D31"
      ]
     },
     "execution": {
      "code_version": "6aa7a7ab87bf5c4e40d704a7492cd9af9ef481db",
      "software": {
       "python": "3.13.18",
       "platform": "Linux-6.18.44-fc-v595-x86_64-with-glibc2.39"
      }
     }
    },
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "add",
    "path": "/evidence/9",
    "value": {
     "id": "E34",
     "profile": "quant",
     "design": "other",
     "description": "Lab runs RN15 under plan PL14. Design: Which forecast does msMAPE reward on intermittent demand, beside sMAPE and the Suilin loss? One run of python3 claimset-code/smape-target/msmape_intermittent.py --seed 9, in two parts, for five measures: sMAPE, msMAPE with nu = 0.1, 0.5 and 1, and the Suilin loss, on the 14 intermittent demand settings. (1) Exact, with no sampling: the forecast that minimises the expected term of each measure (sums for counts; Gauss-Legendre quadrature of the density of ln y with the zeros as a point mass for the lognormal settings), minimised over forecasts f >= 0 on a grid of 0, the integers and 4000 log-spaced points, refined by golden section; and, for the lognormal demand sizes, the share of zeros at which each measure switches to the forecast 0, by bisection. (2) Simulated: test sets of n = 12, 52 and 365 independent actuals (a year of months, of weeks and of days), 2000 test sets per setting and n; in each test set, the forecast with the lowest value of each measure, searched over 0, the test set's own values and a grid (step 0.05 up to the largest value plus 1 for counts; 400 log-spaced points for the lognormal settings). Analysis: Part 1, per setting and measure: the target, its quantile level and the expected score there, beside the setting's share of zeros, median, mode and mean, so that the functional the target coincides with can be read; per measure, the zero threshold. Part 2, per setting, measure and n: the share of test sets whose minimiser is 0, the share whose minimiser is the exact target of part 1 (counts), and the median of the minimisers. Reported as tables and one figure (the targets against the share of zeros, a panel per measure); no test is made and no threshold is set. The figure as drawn (A56) has a panel per family of settings and a line per measure.",
     "availability": {
      "level": "public"
     },
     "lab_runs": [
      "RN15"
     ],
     "tables": [
      {
       "id": "T1",
       "description": "The forecast each measure rewards (its target, computed from the distribution itself) on each intermittent demand setting, beside the setting's share of zeros, median, mode and mean; the measures are sMAPE, msMAPE with nu = 0.1, 0.5 and 1, and the Suilin loss",
       "row_labels": [
        "setting"
       ],
       "columns": [
        "share of zero outcomes",
        "median of the outcomes",
        "mode of the outcomes",
        "mean of the outcomes",
        {
         "variable": "target",
         "conditions": {
          "measure": "sMAPE"
         }
        },
        {
         "variable": "target",
         "conditions": {
          "measure": "msMAPE nu=0.1"
         }
        },
        {
         "variable": "target",
         "conditions": {
          "measure": "msMAPE nu=0.5"
         }
        },
        {
         "variable": "target",
         "conditions": {
          "measure": "msMAPE nu=1"
         }
        },
        {
         "variable": "target",
         "conditions": {
          "measure": "Suilin"
         }
        }
       ],
       "rows": [
        {
         "id": "r1",
         "label": [
          "Poisson mean=0.1"
         ],
         "values": [
          0.9048,
          0,
          0,
          0.1,
          0,
          0,
          0,
          0,
          0
         ]
        },
        {
         "id": "r2",
         "label": [
          "Poisson mean=0.3"
         ],
         "values": [
          0.7408,
          0,
          0,
          0.3,
          0,
          0,
          0,
          0,
          0
         ]
        },
        {
         "id": "r3",
         "label": [
          "Poisson mean=0.5"
         ],
         "values": [
          0.6065,
          0,
          0,
          0.5,
          0,
          0,
          0,
          1,
          0
         ]
        },
        {
         "id": "r4",
         "label": [
          "Poisson mean=0.7"
         ],
         "values": [
          0.4966,
          1,
          0,
          0.7,
          0,
          0,
          1,
          1,
          0
         ]
        },
        {
         "id": "r5",
         "label": [
          "Poisson mean=1"
         ],
         "values": [
          0.3679,
          1,
          0,
          1,
          1,
          1,
          1,
          1,
          1
         ]
        },
        {
         "id": "r6",
         "label": [
          "Poisson mean=1.5"
         ],
         "values": [
          0.2231,
          1,
          1,
          1.5,
          2,
          2,
          1,
          1,
          2
         ]
        },
        {
         "id": "r7",
         "label": [
          "Poisson mean=3"
         ],
         "values": [
          0.04979,
          3,
          2,
          3.0,
          3,
          3,
          3,
          3,
          3
         ]
        },
        {
         "id": "r8",
         "label": [
          "negative binomial mean=0.5 size=0.5"
         ],
         "values": [
          0.7071,
          0,
          0,
          0.5,
          0,
          0,
          0,
          0,
          0
         ]
        },
        {
         "id": "r9",
         "label": [
          "negative binomial mean=1 size=0.5"
         ],
         "values": [
          0.5774,
          0,
          0,
          1.0,
          0,
          0,
          0,
          1,
          0
         ]
        },
        {
         "id": "r10",
         "label": [
          "negative binomial mean=2 size=0.5"
         ],
         "values": [
          0.4472,
          1,
          0,
          2.0,
          0,
          0,
          1,
          1,
          0
         ]
        },
        {
         "id": "r11",
         "label": [
          "lognormal sizes with zeros p0=0.2"
         ],
         "values": [
          0.2,
          0.7271,
          null,
          1.319,
          1.0,
          0.922,
          0.8021,
          0.8672,
          0.9054
         ]
        },
        {
         "id": "r12",
         "label": [
          "lognormal sizes with zeros p0=0.4"
         ],
         "values": [
          0.4,
          0.3801,
          null,
          0.9892,
          0,
          0.7791,
          0.5271,
          0.5932,
          0.1702
         ]
        },
        {
         "id": "r13",
         "label": [
          "lognormal sizes with zeros p0=0.6"
         ],
         "values": [
          0.6,
          0,
          null,
          0.6595,
          0,
          0,
          0,
          0,
          0
         ]
        },
        {
         "id": "r14",
         "label": [
          "lognormal sizes with zeros p0=0.8"
         ],
         "values": [
          0.8,
          0,
          null,
          0.3297,
          0,
          0,
          0,
          0,
          0
         ]
        }
       ],
       "artefact": "A55"
      },
      {
       "id": "T2",
       "description": "The share of zeros at which each measure's target switches to the forecast 0, for lognormal demand sizes (median 1, sigma 1) with zeros",
       "row_labels": [
        "setting"
       ],
       "columns": [
        {
         "variable": "zero threshold",
         "conditions": {
          "measure": "sMAPE"
         }
        },
        {
         "variable": "zero threshold",
         "conditions": {
          "measure": "msMAPE nu=0.1"
         }
        },
        {
         "variable": "zero threshold",
         "conditions": {
          "measure": "msMAPE nu=0.5"
         }
        },
        {
         "variable": "zero threshold",
         "conditions": {
          "measure": "msMAPE nu=1"
         }
        },
        {
         "variable": "zero threshold",
         "conditions": {
          "measure": "Suilin"
         }
        }
       ],
       "rows": [
        {
         "id": "r1",
         "label": [
          "lognormal sizes with zeros"
         ],
         "values": [
          0.394,
          0.4191,
          0.5192,
          0.5647,
          0.47
         ]
        }
       ],
       "artefact": "A55"
      },
      {
       "id": "T3",
       "description": "The quantile level of each measure's target and the expected score there (200 times the expected term), for each setting",
       "row_labels": [
        "setting"
       ],
       "columns": [
        {
         "variable": "quantile level of the target",
         "conditions": {
          "measure": "sMAPE"
         }
        },
        {
         "variable": "quantile level of the target",
         "conditions": {
          "measure": "msMAPE nu=0.1"
         }
        },
        {
         "variable": "quantile level of the target",
         "conditions": {
          "measure": "msMAPE nu=0.5"
         }
        },
        {
         "variable": "quantile level of the target",
         "conditions": {
          "measure": "msMAPE nu=1"
         }
        },
        {
         "variable": "quantile level of the target",
         "conditions": {
          "measure": "Suilin"
         }
        },
        {
         "variable": "expected score at the target",
         "conditions": {
          "measure": "sMAPE"
         }
        },
        {
         "variable": "expected score at the target",
         "conditions": {
          "measure": "msMAPE nu=0.1"
         }
        },
        {
         "variable": "expected score at the target",
         "conditions": {
          "measure": "msMAPE nu=0.5"
         }
        },
        {
         "variable": "expected score at the target",
         "conditions": {
          "measure": "msMAPE nu=1"
         }
        },
        {
         "variable": "expected score at the target",
         "conditions": {
          "measure": "Suilin"
         }
        }
       ],
       "rows": [
        {
         "id": "r1",
         "label": [
          "Poisson mean=0.1"
         ],
         "values": [
          0.9048,
          0.9048,
          0.9048,
          0.9048,
          0.9048,
          19.03,
          19.03,
          19.03,
          19.03,
          17.34
         ]
        },
        {
         "id": "r2",
         "label": [
          "Poisson mean=0.3"
         ],
         "values": [
          0.7408,
          0.7408,
          0.7408,
          0.7408,
          0.7408,
          51.84,
          51.84,
          51.84,
          51.84,
          47.46
         ]
        },
        {
         "id": "r3",
         "label": [
          "Poisson mean=0.5"
         ],
         "values": [
          0.6065,
          0.6065,
          0.6065,
          0.9098,
          0.6065,
          78.69,
          78.69,
          78.69,
          67.18,
          72.37
         ]
        },
        {
         "id": "r4",
         "label": [
          "Poisson mean=0.7"
         ],
         "values": [
          0.4966,
          0.4966,
          0.8442,
          0.8442,
          0.4966,
          100.7,
          100.7,
          77.86,
          61.31,
          92.99
         ]
        },
        {
         "id": "r5",
         "label": [
          "Poisson mean=1"
         ],
         "values": [
          0.7358,
          0.7358,
          0.7358,
          0.7358,
          0.7358,
          94.3,
          87.61,
          69.78,
          57.52,
          87.03
         ]
        },
        {
         "id": "r6",
         "label": [
          "Poisson mean=1.5"
         ],
         "values": [
          0.8088,
          0.8088,
          0.5578,
          0.5578,
          0.8088,
          76.77,
          74.64,
          67.21,
          59.77,
          73.75
         ]
        },
        {
         "id": "r7",
         "label": [
          "Poisson mean=3"
         ],
         "values": [
          0.6472,
          0.6472,
          0.6472,
          0.6472,
          0.6472,
          49.91,
          49.59,
          48.49,
          47.42,
          48.86
         ]
        },
        {
         "id": "r8",
         "label": [
          "negative binomial mean=0.5 size=0.5"
         ],
         "values": [
          0.7071,
          0.7071,
          0.7071,
          0.7071,
          0.7071,
          58.58,
          58.58,
          58.58,
          58.58,
          54.46
         ]
        },
        {
         "id": "r9",
         "label": [
          "negative binomial mean=1 size=0.5"
         ],
         "values": [
          0.5774,
          0.5774,
          0.5774,
          0.7698,
          0.5774,
          84.53,
          84.53,
          84.53,
          80.38,
          79.46
         ]
        },
        {
         "id": "r10",
         "label": [
          "negative binomial mean=2 size=0.5"
         ],
         "values": [
          0.4472,
          0.4472,
          0.6261,
          0.6261,
          0.4472,
          110.6,
          110.6,
          101.9,
          86.95,
          105.2
         ]
        },
        {
         "id": "r11",
         "label": [
          "lognormal sizes with zeros p0=0.2"
         ],
         "values": [
          0.6,
          0.5741,
          0.5302,
          0.5547,
          0.5683,
          95.95,
          92.15,
          78.37,
          66.53,
          89.49
         ]
        },
        {
         "id": "r12",
         "label": [
          "lognormal sizes with zeros p0=0.4"
         ],
         "values": [
          0.4,
          0.6409,
          0.5566,
          0.5805,
          0.423,
          120,
          113.9,
          87.15,
          70.9,
          94.45
         ]
        },
        {
         "id": "r13",
         "label": [
          "lognormal sizes with zeros p0=0.6"
         ],
         "values": [
          0.6,
          0.6,
          0.6,
          0.6,
          0.6,
          80,
          79.78,
          72.4,
          60.93,
          65.9
         ]
        },
        {
         "id": "r14",
         "label": [
          "lognormal sizes with zeros p0=0.8"
         ],
         "values": [
          0.8,
          0.8,
          0.8,
          0.8,
          0.8,
          40.0,
          39.89,
          36.2,
          30.46,
          32.95
         ]
        }
       ],
       "artefact": "A55"
      },
      {
       "id": "T4",
       "description": "Test sets: the share of 2000 test sets of n independent actuals in which the forecast with the lowest score is 0, for each setting, n and measure",
       "row_labels": [
        "setting",
        "n"
       ],
       "columns": [
        {
         "variable": "share of test sets whose minimiser is 0",
         "conditions": {
          "measure": "sMAPE"
         }
        },
        {
         "variable": "share of test sets whose minimiser is 0",
         "conditions": {
          "measure": "msMAPE nu=0.1"
         }
        },
        {
         "variable": "share of test sets whose minimiser is 0",
         "conditions": {
          "measure": "msMAPE nu=0.5"
         }
        },
        {
         "variable": "share of test sets whose minimiser is 0",
         "conditions": {
          "measure": "msMAPE nu=1"
         }
        },
        {
         "variable": "share of test sets whose minimiser is 0",
         "conditions": {
          "measure": "Suilin"
         }
        }
       ],
       "rows": [
        {
         "id": "r1",
         "label": [
          "Poisson mean=0.1",
          12
         ],
         "values": [
          1,
          0.9995,
          0.996,
          0.975,
          1
         ]
        },
        {
         "id": "r2",
         "label": [
          "Poisson mean=0.1",
          52
         ],
         "values": [
          1,
          1,
          1,
          1,
          1
         ]
        },
        {
         "id": "r3",
         "label": [
          "Poisson mean=0.1",
          365
         ],
         "values": [
          1,
          1,
          1,
          1,
          1
         ]
        },
        {
         "id": "r4",
         "label": [
          "Poisson mean=0.3",
          12
         ],
         "values": [
          0.979,
          0.937,
          0.831,
          0.7145,
          0.979
         ]
        },
        {
         "id": "r5",
         "label": [
          "Poisson mean=0.3",
          52
         ],
         "values": [
          1,
          1,
          0.9915,
          0.9155,
          1
         ]
        },
        {
         "id": "r6",
         "label": [
          "Poisson mean=0.3",
          365
         ],
         "values": [
          1,
          1,
          1,
          0.9995,
          1
         ]
        },
        {
         "id": "r7",
         "label": [
          "Poisson mean=0.5",
          12
         ],
         "values": [
          0.848,
          0.7525,
          0.539,
          0.378,
          0.848
         ]
        },
        {
         "id": "r8",
         "label": [
          "Poisson mean=0.5",
          52
         ],
         "values": [
          0.9685,
          0.923,
          0.6515,
          0.288,
          0.966
         ]
        },
        {
         "id": "r9",
         "label": [
          "Poisson mean=0.5",
          365
         ],
         "values": [
          1,
          1,
          0.8445,
          0.067,
          1
         ]
        },
        {
         "id": "r10",
         "label": [
          "Poisson mean=0.7",
          12
         ],
         "values": [
          0.6045,
          0.515,
          0.2915,
          0.1555,
          0.6045
         ]
        },
        {
         "id": "r11",
         "label": [
          "Poisson mean=0.7",
          52
         ],
         "values": [
          0.6575,
          0.523,
          0.1525,
          0.0265,
          0.6455
         ]
        },
        {
         "id": "r12",
         "label": [
          "Poisson mean=0.7",
          365
         ],
         "values": [
          0.844,
          0.56,
          0.0025,
          0,
          0.8295
         ]
        },
        {
         "id": "r13",
         "label": [
          "Poisson mean=1",
          12
         ],
         "values": [
          0.256,
          0.2265,
          0.0895,
          0.037,
          0.256
         ]
        },
        {
         "id": "r14",
         "label": [
          "Poisson mean=1",
          52
         ],
         "values": [
          0.1015,
          0.051,
          0.003,
          0,
          0.1
         ]
        },
        {
         "id": "r15",
         "label": [
          "Poisson mean=1",
          365
         ],
         "values": [
          0,
          0,
          0,
          0,
          0
         ]
        },
        {
         "id": "r16",
         "label": [
          "Poisson mean=1.5",
          12
         ],
         "values": [
          0.0325,
          0.03,
          0.0075,
          0.002,
          0.0325
         ]
        },
        {
         "id": "r17",
         "label": [
          "Poisson mean=1.5",
          52
         ],
         "values": [
          0.0005,
          0.0005,
          0,
          0,
          0.0005
         ]
        },
        {
         "id": "r18",
         "label": [
          "Poisson mean=1.5",
          365
         ],
         "values": [
          0,
          0,
          0,
          0,
          0
         ]
        },
        {
         "id": "r19",
         "label": [
          "Poisson mean=3",
          12
         ],
         "values": [
          0,
          0,
          0,
          0,
          0
         ]
        },
        {
         "id": "r20",
         "label": [
          "Poisson mean=3",
          52
         ],
         "values": [
          0,
          0,
          0,
          0,
          0
         ]
        },
        {
         "id": "r21",
         "label": [
          "Poisson mean=3",
          365
         ],
         "values": [
          0,
          0,
          0,
          0,
          0
         ]
        },
        {
         "id": "r22",
         "label": [
          "negative binomial mean=0.5 size=0.5",
          12
         ],
         "values": [
          0.971,
          0.952,
          0.857,
          0.721,
          0.9705
         ]
        },
        {
         "id": "r23",
         "label": [
          "negative binomial mean=0.5 size=0.5",
          52
         ],
         "values": [
          1,
          1,
          0.9895,
          0.9095,
          1
         ]
        },
        {
         "id": "r24",
         "label": [
          "negative binomial mean=0.5 size=0.5",
          365
         ],
         "values": [
          1,
          1,
          1,
          1,
          1
         ]
        },
        {
         "id": "r25",
         "label": [
          "negative binomial mean=1 size=0.5",
          12
         ],
         "values": [
          0.8035,
          0.78,
          0.621,
          0.4435,
          0.808
         ]
        },
        {
         "id": "r26",
         "label": [
          "negative binomial mean=1 size=0.5",
          52
         ],
         "values": [
          0.975,
          0.9625,
          0.7675,
          0.405,
          0.974
         ]
        },
        {
         "id": "r27",
         "label": [
          "negative binomial mean=1 size=0.5",
          365
         ],
         "values": [
          1,
          1,
          0.975,
          0.256,
          1
         ]
        },
        {
         "id": "r28",
         "label": [
          "negative binomial mean=2 size=0.5",
          12
         ],
         "values": [
          0.567,
          0.5265,
          0.384,
          0.2405,
          0.5765
         ]
        },
        {
         "id": "r29",
         "label": [
          "negative binomial mean=2 size=0.5",
          52
         ],
         "values": [
          0.707,
          0.635,
          0.3055,
          0.068,
          0.7255
         ]
        },
        {
         "id": "r30",
         "label": [
          "negative binomial mean=2 size=0.5",
          365
         ],
         "values": [
          0.932,
          0.8475,
          0.1115,
          0,
          0.9475
         ]
        },
        {
         "id": "r31",
         "label": [
          "lognormal sizes with zeros p0=0.2",
          12
         ],
         "values": [
          0.046,
          0.026,
          0.0055,
          0.002,
          0.016
         ]
        },
        {
         "id": "r32",
         "label": [
          "lognormal sizes with zeros p0=0.2",
          52
         ],
         "values": [
          0,
          0,
          0,
          0,
          0
         ]
        },
        {
         "id": "r33",
         "label": [
          "lognormal sizes with zeros p0=0.2",
          365
         ],
         "values": [
          0,
          0,
          0,
          0,
          0
         ]
        },
        {
         "id": "r34",
         "label": [
          "lognormal sizes with zeros p0=0.4",
          12
         ],
         "values": [
          0.4655,
          0.374,
          0.1735,
          0.119,
          0.303
         ]
        },
        {
         "id": "r35",
         "label": [
          "lognormal sizes with zeros p0=0.4",
          52
         ],
         "values": [
          0.5235,
          0.36,
          0.045,
          0.012,
          0.172
         ]
        },
        {
         "id": "r36",
         "label": [
          "lognormal sizes with zeros p0=0.4",
          365
         ],
         "values": [
          0.5665,
          0.207,
          0,
          0,
          0.004
         ]
        },
        {
         "id": "r37",
         "label": [
          "lognormal sizes with zeros p0=0.6",
          12
         ],
         "values": [
          0.896,
          0.8565,
          0.663,
          0.5805,
          0.8055
         ]
        },
        {
         "id": "r38",
         "label": [
          "lognormal sizes with zeros p0=0.6",
          52
         ],
         "values": [
          0.9975,
          0.993,
          0.844,
          0.6765,
          0.9705
         ]
        },
        {
         "id": "r39",
         "label": [
          "lognormal sizes with zeros p0=0.6",
          365
         ],
         "values": [
          1,
          1,
          0.999,
          0.892,
          1
         ]
        },
        {
         "id": "r40",
         "label": [
          "lognormal sizes with zeros p0=0.8",
          12
         ],
         "values": [
          0.9975,
          0.9955,
          0.9725,
          0.959,
          0.991
         ]
        },
        {
         "id": "r41",
         "label": [
          "lognormal sizes with zeros p0=0.8",
          52
         ],
         "values": [
          1,
          1,
          1,
          0.9995,
          1
         ]
        },
        {
         "id": "r42",
         "label": [
          "lognormal sizes with zeros p0=0.8",
          365
         ],
         "values": [
          1,
          1,
          1,
          1,
          1
         ]
        }
       ],
       "artefact": "A55"
      },
      {
       "id": "T5",
       "description": "Test sets: the share of 2000 test sets of n independent actuals in which the forecast with the lowest score is the measure's target of T1, for each count setting, n and measure",
       "row_labels": [
        "setting",
        "n"
       ],
       "columns": [
        {
         "variable": "share of test sets whose minimiser is the target",
         "conditions": {
          "measure": "sMAPE"
         }
        },
        {
         "variable": "share of test sets whose minimiser is the target",
         "conditions": {
          "measure": "msMAPE nu=0.1"
         }
        },
        {
         "variable": "share of test sets whose minimiser is the target",
         "conditions": {
          "measure": "msMAPE nu=0.5"
         }
        },
        {
         "variable": "share of test sets whose minimiser is the target",
         "conditions": {
          "measure": "msMAPE nu=1"
         }
        },
        {
         "variable": "share of test sets whose minimiser is the target",
         "conditions": {
          "measure": "Suilin"
         }
        }
       ],
       "rows": [
        {
         "id": "r1",
         "label": [
          "Poisson mean=0.1",
          12
         ],
         "values": [
          1,
          0.9995,
          0.996,
          0.975,
          1
         ]
        },
        {
         "id": "r2",
         "label": [
          "Poisson mean=0.1",
          52
         ],
         "values": [
          1,
          1,
          1,
          1,
          1
         ]
        },
        {
         "id": "r3",
         "label": [
          "Poisson mean=0.1",
          365
         ],
         "values": [
          1,
          1,
          1,
          1,
          1
         ]
        },
        {
         "id": "r4",
         "label": [
          "Poisson mean=0.3",
          12
         ],
         "values": [
          0.979,
          0.937,
          0.831,
          0.7145,
          0.979
         ]
        },
        {
         "id": "r5",
         "label": [
          "Poisson mean=0.3",
          52
         ],
         "values": [
          1,
          1,
          0.9915,
          0.9155,
          1
         ]
        },
        {
         "id": "r6",
         "label": [
          "Poisson mean=0.3",
          365
         ],
         "values": [
          1,
          1,
          1,
          0.9995,
          1
         ]
        },
        {
         "id": "r7",
         "label": [
          "Poisson mean=0.5",
          12
         ],
         "values": [
          0.848,
          0.7525,
          0.539,
          0.547,
          0.848
         ]
        },
        {
         "id": "r8",
         "label": [
          "Poisson mean=0.5",
          52
         ],
         "values": [
          0.9685,
          0.923,
          0.6515,
          0.712,
          0.966
         ]
        },
        {
         "id": "r9",
         "label": [
          "Poisson mean=0.5",
          365
         ],
         "values": [
          1,
          1,
          0.8445,
          0.933,
          1
         ]
        },
        {
         "id": "r10",
         "label": [
          "Poisson mean=0.7",
          12
         ],
         "values": [
          0.6045,
          0.515,
          0.691,
          0.8035,
          0.6045
         ]
        },
        {
         "id": "r11",
         "label": [
          "Poisson mean=0.7",
          52
         ],
         "values": [
          0.6575,
          0.523,
          0.8475,
          0.9735,
          0.6455
         ]
        },
        {
         "id": "r12",
         "label": [
          "Poisson mean=0.7",
          365
         ],
         "values": [
          0.844,
          0.56,
          0.9975,
          1,
          0.8295
         ]
        },
        {
         "id": "r13",
         "label": [
          "Poisson mean=1",
          12
         ],
         "values": [
          0.5335,
          0.5665,
          0.775,
          0.8565,
          0.537
         ]
        },
        {
         "id": "r14",
         "label": [
          "Poisson mean=1",
          52
         ],
         "values": [
          0.7705,
          0.8855,
          0.9825,
          0.993,
          0.8415
         ]
        },
        {
         "id": "r15",
         "label": [
          "Poisson mean=1",
          365
         ],
         "values": [
          0.9985,
          1,
          1,
          1,
          1
         ]
        },
        {
         "id": "r16",
         "label": [
          "Poisson mean=1.5",
          12
         ],
         "values": [
          0.5675,
          0.569,
          0.457,
          0.501,
          0.5685
         ]
        },
        {
         "id": "r17",
         "label": [
          "Poisson mean=1.5",
          52
         ],
         "values": [
          0.7395,
          0.657,
          0.5585,
          0.6325,
          0.6565
         ]
        },
        {
         "id": "r18",
         "label": [
          "Poisson mean=1.5",
          365
         ],
         "values": [
          0.9635,
          0.867,
          0.5965,
          0.762,
          0.8665
         ]
        },
        {
         "id": "r19",
         "label": [
          "Poisson mean=3",
          12
         ],
         "values": [
          0.552,
          0.5475,
          0.5515,
          0.554,
          0.548
         ]
        },
        {
         "id": "r20",
         "label": [
          "Poisson mean=3",
          52
         ],
         "values": [
          0.8905,
          0.888,
          0.884,
          0.8795,
          0.889
         ]
        },
        {
         "id": "r21",
         "label": [
          "Poisson mean=3",
          365
         ],
         "values": [
          1,
          1,
          1,
          1,
          1
         ]
        },
        {
         "id": "r22",
         "label": [
          "negative binomial mean=0.5 size=0.5",
          12
         ],
         "values": [
          0.971,
          0.952,
          0.857,
          0.721,
          0.9705
         ]
        },
        {
         "id": "r23",
         "label": [
          "negative binomial mean=0.5 size=0.5",
          52
         ],
         "values": [
          1,
          1,
          0.9895,
          0.9095,
          1
         ]
        },
        {
         "id": "r24",
         "label": [
          "negative binomial mean=0.5 size=0.5",
          365
         ],
         "values": [
          1,
          1,
          1,
          1,
          1
         ]
        },
        {
         "id": "r25",
         "label": [
          "negative binomial mean=1 size=0.5",
          12
         ],
         "values": [
          0.8035,
          0.78,
          0.621,
          0.471,
          0.808
         ]
        },
        {
         "id": "r26",
         "label": [
          "negative binomial mean=1 size=0.5",
          52
         ],
         "values": [
          0.975,
          0.9625,
          0.7675,
          0.592,
          0.974
         ]
        },
        {
         "id": "r27",
         "label": [
          "negative binomial mean=1 size=0.5",
          365
         ],
         "values": [
          1,
          1,
          0.975,
          0.744,
          1
         ]
        },
        {
         "id": "r28",
         "label": [
          "negative binomial mean=2 size=0.5",
          12
         ],
         "values": [
          0.567,
          0.5265,
          0.3345,
          0.4825,
          0.5765
         ]
        },
        {
         "id": "r29",
         "label": [
          "negative binomial mean=2 size=0.5",
          52
         ],
         "values": [
          0.707,
          0.635,
          0.473,
          0.782,
          0.7255
         ]
        },
        {
         "id": "r30",
         "label": [
          "negative binomial mean=2 size=0.5",
          365
         ],
         "values": [
          0.932,
          0.8475,
          0.855,
          0.998,
          0.9475
         ]
        }
       ],
       "artefact": "A55"
      },
      {
       "id": "T6",
       "description": "Test sets: the median, over 2000 test sets of n independent actuals, of the forecast with the lowest score, for each setting, n and measure",
       "row_labels": [
        "setting",
        "n"
       ],
       "columns": [
        {
         "variable": "median of the test-set minimisers",
         "conditions": {
          "measure": "sMAPE"
         }
        },
        {
         "variable": "median of the test-set minimisers",
         "conditions": {
          "measure": "msMAPE nu=0.1"
         }
        },
        {
         "variable": "median of the test-set minimisers",
         "conditions": {
          "measure": "msMAPE nu=0.5"
         }
        },
        {
         "variable": "median of the test-set minimisers",
         "conditions": {
          "measure": "msMAPE nu=1"
         }
        },
        {
         "variable": "median of the test-set minimisers",
         "conditions": {
          "measure": "Suilin"
         }
        }
       ],
       "rows": [
        {
         "id": "r1",
         "label": [
          "Poisson mean=0.1",
          12
         ],
         "values": [
          0,
          0,
          0,
          0,
          0
         ]
        },
        {
         "id": "r2",
         "label": [
          "Poisson mean=0.1",
          52
         ],
         "values": [
          0,
          0,
          0,
          0,
          0
         ]
        },
        {
         "id": "r3",
         "label": [
          "Poisson mean=0.1",
          365
         ],
         "values": [
          0,
          0,
          0,
          0,
          0
         ]
        },
        {
         "id": "r4",
         "label": [
          "Poisson mean=0.3",
          12
         ],
         "values": [
          0,
          0,
          0,
          0,
          0
         ]
        },
        {
         "id": "r5",
         "label": [
          "Poisson mean=0.3",
          52
         ],
         "values": [
          0,
          0,
          0,
          0,
          0
         ]
        },
        {
         "id": "r6",
         "label": [
          "Poisson mean=0.3",
          365
         ],
         "values": [
          0,
          0,
          0,
          0,
          0
         ]
        },
        {
         "id": "r7",
         "label": [
          "Poisson mean=0.5",
          12
         ],
         "values": [
          0,
          0,
          0,
          1,
          0
         ]
        },
        {
         "id": "r8",
         "label": [
          "Poisson mean=0.5",
          52
         ],
         "values": [
          0,
          0,
          0,
          1,
          0
         ]
        },
        {
         "id": "r9",
         "label": [
          "Poisson mean=0.5",
          365
         ],
         "values": [
          0,
          0,
          0,
          1,
          0
         ]
        },
        {
         "id": "r10",
         "label": [
          "Poisson mean=0.7",
          12
         ],
         "values": [
          0,
          0,
          1,
          1,
          0
         ]
        },
        {
         "id": "r11",
         "label": [
          "Poisson mean=0.7",
          52
         ],
         "values": [
          0,
          0,
          1,
          1,
          0
         ]
        },
        {
         "id": "r12",
         "label": [
          "Poisson mean=0.7",
          365
         ],
         "values": [
          0,
          0,
          1,
          1,
          0
         ]
        },
        {
         "id": "r13",
         "label": [
          "Poisson mean=1",
          12
         ],
         "values": [
          1,
          1,
          1,
          1,
          1
         ]
        },
        {
         "id": "r14",
         "label": [
          "Poisson mean=1",
          52
         ],
         "values": [
          1,
          1,
          1,
          1,
          1
         ]
        },
        {
         "id": "r15",
         "label": [
          "Poisson mean=1",
          365
         ],
         "values": [
          1,
          1,
          1,
          1,
          1
         ]
        },
        {
         "id": "r16",
         "label": [
          "Poisson mean=1.5",
          12
         ],
         "values": [
          2,
          2,
          2,
          1,
          2
         ]
        },
        {
         "id": "r17",
         "label": [
          "Poisson mean=1.5",
          52
         ],
         "values": [
          2,
          2,
          1,
          1,
          2
         ]
        },
        {
         "id": "r18",
         "label": [
          "Poisson mean=1.5",
          365
         ],
         "values": [
          2,
          2,
          1,
          1,
          2
         ]
        },
        {
         "id": "r19",
         "label": [
          "Poisson mean=3",
          12
         ],
         "values": [
          3,
          3,
          3,
          3,
          3
         ]
        },
        {
         "id": "r20",
         "label": [
          "Poisson mean=3",
          52
         ],
         "values": [
          3,
          3,
          3,
          3,
          3
         ]
        },
        {
         "id": "r21",
         "label": [
          "Poisson mean=3",
          365
         ],
         "values": [
          3,
          3,
          3,
          3,
          3
         ]
        },
        {
         "id": "r22",
         "label": [
          "negative binomial mean=0.5 size=0.5",
          12
         ],
         "values": [
          0,
          0,
          0,
          0,
          0
         ]
        },
        {
         "id": "r23",
         "label": [
          "negative binomial mean=0.5 size=0.5",
          52
         ],
         "values": [
          0,
          0,
          0,
          0,
          0
         ]
        },
        {
         "id": "r24",
         "label": [
          "negative binomial mean=0.5 size=0.5",
          365
         ],
         "values": [
          0,
          0,
          0,
          0,
          0
         ]
        },
        {
         "id": "r25",
         "label": [
          "negative binomial mean=1 size=0.5",
          12
         ],
         "values": [
          0,
          0,
          0,
          1,
          0
         ]
        },
        {
         "id": "r26",
         "label": [
          "negative binomial mean=1 size=0.5",
          52
         ],
         "values": [
          0,
          0,
          0,
          1,
          0
         ]
        },
        {
         "id": "r27",
         "label": [
          "negative binomial mean=1 size=0.5",
          365
         ],
         "values": [
          0,
          0,
          0,
          1,
          0
         ]
        },
        {
         "id": "r28",
         "label": [
          "negative binomial mean=2 size=0.5",
          12
         ],
         "values": [
          0,
          0,
          1,
          1,
          0
         ]
        },
        {
         "id": "r29",
         "label": [
          "negative binomial mean=2 size=0.5",
          52
         ],
         "values": [
          0,
          0,
          1,
          1,
          0
         ]
        },
        {
         "id": "r30",
         "label": [
          "negative binomial mean=2 size=0.5",
          365
         ],
         "values": [
          0,
          0,
          1,
          1,
          0
         ]
        },
        {
         "id": "r31",
         "label": [
          "lognormal sizes with zeros p0=0.2",
          12
         ],
         "values": [
          0.9791,
          0.917,
          0.8181,
          0.8745,
          0.8357
         ]
        },
        {
         "id": "r32",
         "label": [
          "lognormal sizes with zeros p0=0.2",
          52
         ],
         "values": [
          1.003,
          0.9223,
          0.7943,
          0.8633,
          0.8825
         ]
        },
        {
         "id": "r33",
         "label": [
          "lognormal sizes with zeros p0=0.2",
          365
         ],
         "values": [
          1.003,
          0.9257,
          0.8006,
          0.8684,
          0.907
         ]
        },
        {
         "id": "r34",
         "label": [
          "lognormal sizes with zeros p0=0.4",
          12
         ],
         "values": [
          0.4646,
          0.5008,
          0.5196,
          0.6003,
          0.1931
         ]
        },
        {
         "id": "r35",
         "label": [
          "lognormal sizes with zeros p0=0.4",
          52
         ],
         "values": [
          0,
          0.63,
          0.5224,
          0.5913,
          0.1681
         ]
        },
        {
         "id": "r36",
         "label": [
          "lognormal sizes with zeros p0=0.4",
          365
         ],
         "values": [
          0,
          0.7535,
          0.5278,
          0.5938,
          0.1719
         ]
        },
        {
         "id": "r37",
         "label": [
          "lognormal sizes with zeros p0=0.6",
          12
         ],
         "values": [
          0,
          0,
          0,
          0,
          0
         ]
        },
        {
         "id": "r38",
         "label": [
          "lognormal sizes with zeros p0=0.6",
          52
         ],
         "values": [
          0,
          0,
          0,
          0,
          0
         ]
        },
        {
         "id": "r39",
         "label": [
          "lognormal sizes with zeros p0=0.6",
          365
         ],
         "values": [
          0,
          0,
          0,
          0,
          0
         ]
        },
        {
         "id": "r40",
         "label": [
          "lognormal sizes with zeros p0=0.8",
          12
         ],
         "values": [
          0,
          0,
          0,
          0,
          0
         ]
        },
        {
         "id": "r41",
         "label": [
          "lognormal sizes with zeros p0=0.8",
          52
         ],
         "values": [
          0,
          0,
          0,
          0,
          0
         ]
        },
        {
         "id": "r42",
         "label": [
          "lognormal sizes with zeros p0=0.8",
          365
         ],
         "values": [
          0,
          0,
          0,
          0,
          0
         ]
        }
       ],
       "artefact": "A55"
      }
     ],
     "artefacts": [
      "A6",
      "A55",
      "A56"
     ],
     "setup": {
      "studied": [
       "D33",
       "D1",
       "D28"
      ],
      "cases": [
       "D35"
      ],
      "procedure": [
       "D15"
      ],
      "lens": [
       "D34",
       "D6",
       "D29"
      ]
     },
     "execution": {
      "code_version": "d49b5fd17718af3efd407e840ee50bc9665e4300",
      "software": {
       "python": "3.13.18",
       "platform": "Linux-6.18.44-fc-v595-x86_64-with-glibc2.39"
      }
     },
     "results": []
    },
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "add",
    "path": "/evidence/22",
    "value": {
     "id": "E35",
     "profile": "formal",
     "description": "Proof of C18 (section H13 of the proof document A1, rendered in A2). For widely spread actuals, sMAPE loses its power to tell two forecasts apart, at the rate $\\sigma^{-1/2}$. The absolute log error keeps its power, and the absolute error keeps it unless the higher forecast is the median; the squared error on lognormal actuals loses it faster than any power of $\\sigma$.",
     "foundation": "Kolmogorov probability axioms and real analysis",
     "steps": [
      {
       "id": "S1",
       "statement": "The sMAPE difference. Write $q(x) = 2/(1 + e^{|x|})$, so that $\\tanh(|x|/2) = 1 - q(x)$, and $q_i = q(\\sigma(\\varepsilon - c_i))$. By C3 the sMAPE term of the forecast $f_i$ is $1 - q_i$, because $\\ln y - \\ln f_i = \\sigma(\\varepsilon - c_i)$. So $D = q_1 - q_2$.",
       "rule": "definition",
       "by": [
        "C3"
       ]
      },
      {
       "id": "S2",
       "statement": "Its mean. By C10, $\\mathrm E[q_i] = 4\\ln 2\\,h(c_i)/\\sigma + O(\\sigma^{-3})$. So $\\mathrm E[D] = 4\\ln 2\\,(h(c_1) - h(c_2))/\\sigma + O(\\sigma^{-3})$.",
       "rule": "earlier result",
       "by": [
        "C10"
       ]
      },
      {
       "id": "S3",
       "statement": "The mean of $q_i^2$. Substituting $v = \\sigma(\\varepsilon - c_i)$, as in C10, step 1, $\\mathrm E[q_i^2] = \\frac1\\sigma\\int q(v)^2\\,h(c_i + v/\\sigma)\\,dv$. Expanding $h$ by Taylor's theorem, as in C10, step 2: the first term integrates to $h(c_i)\\int q(v)^2\\,dv$; the second integrates to 0, because its factor $v$ changes sign with $v$ while the rest does not; the third is at most $\\tfrac12\\max|h''|\\int v^2 q(v)^2\\,dv$ divided by $\\sigma^2$, and this integral is finite because $q(v) \\le 2e^{-|v|}$. Since $\\dfrac{1}{(1+e^v)^2} = \\dfrac{1}{1+e^v} - \\dfrac{e^v}{(1+e^v)^2}$, whose two parts integrate over $v > 0$ to $\\ln 2$ and $\\tfrac12$, $\\int q(v)^2\\,dv = 8(\\ln 2 - \\tfrac12) = 4(2\\ln 2 - 1)$. So $\\mathrm E[q_i^2] = 4(2\\ln 2 - 1)\\,h(c_i)/\\sigma + O(\\sigma^{-3})$.",
       "rule": "Taylor's theorem",
       "by": [
        "C10"
       ]
      },
      {
       "id": "S4",
       "statement": "The cross term. Since $q(x) \\le 2e^{-|x|}$, $q_1 q_2 \\le 4e^{-\\sigma(|\\varepsilon - c_1| + |\\varepsilon - c_2|)} \\le 4e^{-\\sigma(c_2 - c_1)}$, because $|\\varepsilon - c_1| + |\\varepsilon - c_2| \\ge c_2 - c_1$. So $\\mathrm E[q_1 q_2] \\le 4e^{-\\sigma(c_2 - c_1)}$, which is smaller than every power of $1/\\sigma$.",
       "rule": "algebra",
       "by": [
        "S3"
       ]
      },
      {
       "id": "S5",
       "statement": "The variance. $\\operatorname{Var}(D) = \\mathrm E[q_1^2] + \\mathrm E[q_2^2] - 2\\,\\mathrm E[q_1 q_2] - (\\mathrm E[D])^2$. By steps 2 to 4 this is $4(2\\ln 2 - 1)(h(c_1) + h(c_2))/\\sigma + O(\\sigma^{-2})$; the term $(\\mathrm E[D])^2$ is of order $\\sigma^{-2}$.",
       "rule": "algebra",
       "by": [
        "S2",
        "S3",
        "S4"
       ]
      },
      {
       "id": "S6",
       "statement": "(a). When $h(c_1) \\ne h(c_2)$, $h(c_1) + h(c_2) > 0$. Dividing the size of the mean (step 2) by the square root of the variance (step 5) gives the stated ratio. To keep $\\sqrt n$ times the ratio at a given value, $n$ must be that value squared over the ratio squared, which grows in proportion to $\\sigma$. When $h(c_1) = h(c_2) > 0$ instead, the mean is $O(\\sigma^{-3})$ (step 2) while the variance is still of order $1/\\sigma$ (step 5), so the ratio is $O(\\sigma^{-5/2})$: at most of that order.",
       "rule": "algebra",
       "by": [
        "S2",
        "S5"
       ]
      },
      {
       "id": "S7",
       "statement": "(b). For the absolute log error, $|\\ln y - \\ln f_i| = \\sigma|\\varepsilon - c_i|$, so $D = \\sigma(|\\varepsilon - c_2| - |\\varepsilon - c_1|)$. It is bounded by $\\sigma(c_2 - c_1)$, so its mean and standard deviation exist. Both are $\\sigma$ times those of $|\\varepsilon - c_2| - |\\varepsilon - c_1|$, whose distribution does not depend on $\\sigma$. So their ratio, wherever the standard deviation is not zero, does not depend on $\\sigma$.",
       "rule": "algebra",
       "by": [
        "S6"
       ]
      },
      {
       "id": "S8",
       "statement": "(c). For the absolute error, $D = |y - f_2| - |y - f_1|$ with $f_1 < f_2$. It equals $f_2 - f_1$ when $y \\le f_1$, $f_1 + f_2 - 2y$ when $f_1 < y < f_2$, and $-(f_2 - f_1)$ when $y \\ge f_2$; so $|D| \\le f_2$. Now $f_1/f_2 = e^{-\\sigma(c_2 - c_1)}$ tends to 0, and for every $\\varepsilon < c_2$, $y/f_2 = e^{\\sigma(\\varepsilon - c_2)}$ tends to 0. So $D/f_2$ tends to 1 for $\\varepsilon < c_2$ and to $-1$ for $\\varepsilon > c_2$. By bounded convergence, $\\mathrm E[D/f_2] \\to H(c_2) - (1 - H(c_2)) = 2H(c_2) - 1$ and $\\mathrm E[(D/f_2)^2] \\to 1$. So the variance of $D/f_2$ tends to $1 - (2H(c_2) - 1)^2 = 4H(c_2)(1 - H(c_2))$, and the ratio, which is the same for $D$ and $D/f_2$, tends to the stated limit when $0 < H(c_2) < 1$. The limit is 0 only when $H(c_2) = 1/2$.",
       "rule": "bounded convergence",
       "by": [
        "S7"
       ]
      },
      {
       "id": "S9",
       "statement": "(d). For the squared error, $D = (y - f_2)^2 - (y - f_1)^2 = (f_2 - f_1)(f_1 + f_2 - 2y)$. So $\\mathrm E[D] = (f_2 - f_1)(f_1 + f_2 - 2\\,\\mathrm E[y])$ and $\\mathrm{sd}(D) = 2(f_2 - f_1)\\,\\mathrm{sd}(y)$, finite when $\\mathrm E[y^2]$ is. Their ratio is the stated one. For lognormal actuals $\\mathrm E[y] = e^{\\mu + \\sigma^2/2}$, $\\mathrm{sd}(y) = \\mathrm E[y]\\sqrt{e^{\\sigma^2} - 1}$ and $f_i/\\mathrm E[y] = e^{\\sigma c_i - \\sigma^2/2}$, which tends to 0. So $e^{\\sigma^2/2}$ times the ratio is $|1 - (f_1 + f_2)/(2\\,\\mathrm E[y])|\\,e^{\\sigma^2/2}/\\sqrt{e^{\\sigma^2} - 1}$, which tends to 1.",
       "rule": "algebra",
       "by": [
        "S8"
       ]
      },
      {
       "id": "S10",
       "statement": "So for widely spread actuals the absolute log error keeps its power to tell two forecasts apart wherever its ratio is defined. The absolute error keeps it when $0 < H(c_2) < 1$ and the higher forecast is not the median. sMAPE loses it at the rate $\\sigma^{-1/2}$ when $h(c_1) \\ne h(c_2)$. It loses it faster, with a ratio of at most order $\\sigma^{-5/2}$, when $h(c_1) = h(c_2) > 0$. The squared error on lognormal actuals loses it faster than any power of $\\sigma$.",
       "rule": "conclusion",
       "by": [
        "S9"
       ]
      }
     ],
     "artefacts": [
      "A1",
      "A2"
     ],
     "availability": {
      "level": "public"
     }
    },
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "add",
    "path": "/evidence/23",
    "value": {
     "id": "E36",
     "profile": "formal",
     "description": "Proof of C19 (section H14 of the proof document A1, rendered in A2). Where the log of the actual is symmetric with a single peak, the forecast that minimises the sMAPE of a test set is asymptotically normal about the median. Compared with the sample median it is asymptotically as precise for narrow distributions and asymptotically loses all of its precision as the actuals spread out.",
     "foundation": "Kolmogorov probability axioms and real analysis",
     "steps": [
      {
       "id": "S1",
       "statement": "Write $x_t = \\ln y_t - \\ln m$ for the actuals of a test set and $\\theta = \\ln f - \\ln m$ for a forecast $f$. By C3 the mean sMAPE term of the test set is $M_n(\\theta) = \\frac1n\\sum_t\\tanh(|x_t - \\theta|/2)$, and its expectation is $M(\\theta) = \\mathrm E[\\tanh(|x - \\theta|/2)]$. So $\\hat\\theta_n = \\ln\\hat f_n - \\ln m$ minimises $M_n$; a minimiser exists, because $M_n$ is continuous, tends to 1 as $|\\theta|$ grows, and is below 1 at each $x_t$.",
       "rule": "setup",
       "by": [
        "C3"
       ]
      },
      {
       "id": "S2",
       "statement": "(a). Write $\\kappa(u) = \\operatorname{sech}^2(u/2)\\tanh(|u|/2) \\ge 0$. Since $\\kappa(u)$ is the derivative of $\\tanh^2(u/2)$ for $u > 0$, $\\int\\kappa(u)\\,du = 2$. By unimodality $g(\\ln m + u) \\le g(\\ln m)$, so $A = \\int\\kappa(u)\\,g(\\ln m + u)\\,du \\le 2g(\\ln m)$. A density cannot equal $g(\\ln m)$ everywhere, and $g(\\ln m + u)$ does not increase as $|u|$ grows; so $g(\\ln m + u) < g(\\ln m)$ for every $|u|$ beyond some point, where $\\kappa > 0$. So the inequality is strict.",
       "rule": "definition",
       "by": [
        "S1"
       ]
      },
      {
       "id": "S3",
       "statement": "The expected loss is lowest at 0 only. $M(\\theta)$ is the expected sMAPE term of the forecast $m\\,e^\\theta$. By C5, $M(\\theta) > M(0)$ for every $\\theta \\ne 0$. The function $\\tanh(|u|/2)$ has slope at most $\\tfrac12$, so $M$ and every $M_n$ change by at most $\\tfrac12|\\theta_1 - \\theta_2|$ between $\\theta_1$ and $\\theta_2$; in particular $M$ is continuous. And $M(\\theta) \\to 1 > M(0)$ as $|\\theta| \\to \\infty$, by bounded convergence. So for every $\\eta > 0$ the lowest value of $M$ over $|\\theta| \\ge \\eta$ is above $M(0)$.",
       "rule": "bounded convergence",
       "by": [
        "C5"
       ]
      },
      {
       "id": "S4",
       "statement": "$M_n$ is close to $M$ for all $\\theta$ at once. For each $\\theta$, $M_n(\\theta) \\to M(\\theta)$ with probability 1 (law of large numbers; the terms are bounded). Take a grid of spacing $\\gamma$ on $[-R, R]$: by the slope bound of step 3, once $M_n$ is within $\\gamma$ of $M$ at the grid points it is within $2\\gamma$ of $M$ on all of $[-R, R]$. Outside $[-R, R]$: an actual with $|x_t| \\le R/2$ has $\\tanh(|x_t - \\theta|/2) \\ge \\tanh(R/4)$ when $|\\theta| > R$, so $M_n(\\theta)$ is at least $\\tanh(R/4)$ times the share of such actuals, which tends to $\\mathrm P(|x| \\le R/2)$. Since $M(0) < 1$, for $R$ large this is above $M(0)$ by a fixed margin.",
       "rule": "case analysis",
       "by": [
        "S3"
       ]
      },
      {
       "id": "S5",
       "statement": "Consistency. $M_n(\\hat\\theta_n) \\le M_n(0)$, which tends to $M(0)$. By step 4, with probability tending to 1, $|\\hat\\theta_n| \\le R$ and $M(\\hat\\theta_n)$ is within any given margin of $M(0)$; by step 3, $|\\hat\\theta_n| < \\eta$. So $\\hat\\theta_n \\to 0$ in probability, which is $\\hat f_n \\to m$. This is the argument of van der Vaart (1998, Theorem 5.7).",
       "rule": "algebra",
       "by": [
        "S4",
        "S3",
        "B15"
       ]
      },
      {
       "id": "S6",
       "statement": "The curvature at 0. By C12, step 4, with $c = \\ln m$, the second derivative of $M$ at 0 is $g(\\ln m) - \\tfrac12 A = \\tfrac12(2g(\\ln m) - A)$, which is positive by step 2. The first derivative at 0 is 0, because 0 is the minimum (step 3).",
       "rule": "earlier result",
       "by": [
        "S2",
        "S3",
        "C12"
       ]
      },
      {
       "id": "S7",
       "statement": "Asymptotic normality. For every $x \\ne 0$, $\\theta \\mapsto \\tanh(|x - \\theta|/2)$ has a derivative at $\\theta = 0$, $-\\psi(x)$ with $\\psi(x) = \\tfrac12\\operatorname{sech}^2(x/2)$ for $x > 0$ and $-\\tfrac12\\operatorname{sech}^2(x/2)$ for $x < 0$; and $x = 0$ has probability zero. For every $x$ it changes by at most $\\tfrac12|\\theta_1 - \\theta_2|$ (step 3). $M$ has a second-order expansion at its minimum 0 with a positive second derivative (step 6). And $\\hat\\theta_n$ minimises $M_n$ and tends to 0 in probability (step 5). These are the conditions of the theorem on estimators that minimise a criterion which is Lipschitz in the parameter (van der Vaart, 1998, Theorem 5.23). It gives that $\\sqrt n\\,\\hat\\theta_n$ is asymptotically normal with mean 0 and variance $\\mathrm E[\\psi(x)^2]$ divided by the square of the second derivative of $M$ at 0.",
       "rule": "cited theorem",
       "by": [
        "S3",
        "S6",
        "S5",
        "B15"
       ]
      },
      {
       "id": "S8",
       "statement": "The variance. $\\mathrm E[\\psi(x)^2] = \\tfrac14\\mathrm E[\\operatorname{sech}^4(x/2)] = B/4$, and the square of the second derivative is $\\tfrac14(2g(\\ln m) - A)^2$ (step 6). So the variance is $B/(2g(\\ln m) - A)^2$.",
       "rule": "algebra",
       "by": [
        "S6"
       ]
      },
      {
       "id": "S9",
       "statement": "The sample median. The logarithm of the sample median of the $y_t$ is the sample median of the $\\ln y_t$, which minimises the mean of $|x_t - \\theta|$. The same theorem applies with $|x - \\theta|$ in place of $\\tanh(|x - \\theta|/2)$: the derivative at 0 is $-1$ for $x > 0$ and $+1$ for $x < 0$, with square 1; the criterion changes by at most $|\\theta_1 - \\theta_2|$; and the second derivative of $\\mathrm E|x - \\theta|$ at 0 is $2g(\\ln m)$, positive because $g$ is largest at $\\ln m$. So the variance is $1/(4g(\\ln m)^2)$.",
       "rule": "cited theorem",
       "by": [
        "S8",
        "B15"
       ]
      },
      {
       "id": "S10",
       "statement": "(b), the ratio. Dividing the variance of step 9 by that of step 8 gives $\\mathrm{ARE} = (2g(\\ln m) - A)^2/(4g(\\ln m)^2 B)$.",
       "rule": "algebra",
       "by": [
        "S9",
        "S8"
       ]
      },
      {
       "id": "S11",
       "statement": "(c), narrow distributions. Write $\\ln y = \\ln m + \\sigma\\varepsilon$ with $\\varepsilon$ standard normal, or Laplace with scale 1 and $\\sigma = b$; then $g(\\ln m) = h(0)/\\sigma$ with $h$ the density of $\\varepsilon$, and $x = \\sigma\\varepsilon$. As $\\sigma \\to 0$, $A \\to 0$ and $B \\to 1$ by bounded convergence, and $A/(2g(\\ln m)) = \\sigma A/(2h(0)) \\to 0$. So $\\mathrm{ARE} = (1 - A/(2g(\\ln m)))^2/B \\to 1$.",
       "rule": "bounded convergence",
       "by": [
        "S10"
       ]
      },
      {
       "id": "S12",
       "statement": "(c), wide lognormal distributions. Since $\\int\\kappa = 2$ (step 2), $2g(\\ln m) - A = \\int\\kappa(u)\\,[g(\\ln m) - g(\\ln m + u)]\\,du$. For the normal density, $g(\\ln m) - g(\\ln m + u) = \\big(1 - e^{-u^2/(2\\sigma^2)}\\big)/(\\sigma\\sqrt{2\\pi})$, and $\\tfrac{u^2}{2\\sigma^2} - \\tfrac{u^4}{8\\sigma^4} \\le 1 - e^{-u^2/(2\\sigma^2)} \\le \\tfrac{u^2}{2\\sigma^2}$. Integrating by parts, $\\int u^2\\kappa(u)\\,du = 4\\int_0^\\infty u\\operatorname{sech}^2(u/2)\\,du = 16\\ln 2$; and $\\int u^4\\kappa(u)\\,du$ is finite. So $2g(\\ln m) - A = 8\\ln 2/(\\sigma^3\\sqrt{2\\pi}) + O(\\sigma^{-5})$. In the same way, with $\\int\\operatorname{sech}^4(u/2)\\,du = 8/3$, $B = 8/(3\\sigma\\sqrt{2\\pi}) + O(\\sigma^{-3})$. With $g(\\ln m) = 1/(\\sigma\\sqrt{2\\pi})$, $\\mathrm{ARE}\\,\\sigma^3$ tends to $\\dfrac{64(\\ln 2)^2}{2\\pi}\\cdot\\dfrac{2\\pi}{4}\\cdot\\dfrac{3\\sqrt{2\\pi}}{8} = 6(\\ln 2)^2\\sqrt{2\\pi}$.",
       "rule": "algebra",
       "by": [
        "S2"
       ]
      },
      {
       "id": "S13",
       "statement": "(c), wide log-Laplace distributions. Here $g(\\ln m + u) = e^{-|u|/b}/(2b)$, so $g(\\ln m) - g(\\ln m + u) = (1 - e^{-|u|/b})/(2b)$, which lies between $\\big(|u|/b - u^2/(2b^2)\\big)/(2b)$ and $|u|/(2b^2)$. Integrating by parts, $\\int |u|\\,\\kappa(u)\\,du = 2\\int_0^\\infty\\operatorname{sech}^2(u/2)\\,du = 4$. So $2g(\\ln m) - A = 2/b^2 + O(b^{-3})$, and in the same way $B = 4/(3b) + O(b^{-2})$. With $g(\\ln m) = 1/(2b)$, $\\mathrm{ARE}\\,b$ tends to $\\dfrac{4}{b^4}\\cdot b^2\\cdot\\dfrac{3b}{4}\\cdot b = 3$.",
       "rule": "algebra",
       "by": [
        "S12"
       ]
      },
      {
       "id": "S14",
       "statement": "So the sMAPE minimiser of a test set is asymptotically normal about the median, with a variance that the density at the median and two expectations give. On lognormal and log-Laplace actuals it is asymptotically as precise as the sample median for narrow distributions and asymptotically loses all of its precision relative to it as the actuals spread out, at the rates $\\sigma^{-3}$ and $b^{-1}$.",
       "rule": "conclusion",
       "by": [
        "S13"
       ]
      }
     ],
     "artefacts": [
      "A1",
      "A2"
     ],
     "availability": {
      "level": "public"
     }
    },
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "add",
    "path": "/evidence/25",
    "value": {
     "id": "E37",
     "profile": "formal",
     "description": "Proof of C20 (the second proof of the proof document A5). The msMAPE term is the sMAPE term except for actuals below its constant, where a zero actual no longer gives every non-zero forecast the largest value.",
     "foundation": "Kolmogorov probability axioms and real analysis",
     "steps": [
      {
       "id": "S1",
       "statement": "(a). When $y \\ge \\nu$, $\\max\\{y,\\nu\\} = y$, and $y > 0$ because $\\nu > 0$. So $s_\\nu = |f-y|/(f+y)$, which is $s(y,f)$ for non-negative $y$ and $f$. If every actual is at least $\\nu$, the two measures are then the same function of the forecast on every test set and under every distribution of such actuals, so they have the same minimisers.",
       "rule": "algebra",
       "by": [
        "D33"
       ]
      },
      {
       "id": "S2",
       "statement": "(b). When $y < \\nu$, $\\max\\{y,\\nu\\} = \\nu$, so $s_\\nu = |f-y|/(f+\\nu)$. For $y = 0$ this is $f/(f+\\nu) = 1 - \\nu/(f+\\nu)$, which is 0 at $f = 0$, increasing in $f$, below 1, and tends to 1 as $f$ grows. The sMAPE term of a zero actual is $|f|/|f| = 1$ for every $f > 0$.",
       "rule": "monotonicity",
       "by": [
        "S1"
       ]
      },
      {
       "id": "S3",
       "statement": "(c), the bounds. $s_\\nu \\ge 0$ because its numerator is non-negative and its denominator is at least $\\nu > 0$. Where $y \\ge \\nu$, $s_\\nu = s \\le 1$, since $|f-y| \\le f+y$. Where $y < \\nu$: if $f \\ge y$, $|f-y| = f-y \\le f < f+\\nu$; if $f < y$, $|f-y| = y-f \\le y < \\nu \\le f+\\nu$. So $s_\\nu < 1$ there.",
       "rule": "algebra",
       "by": [
        "S2"
       ]
      },
      {
       "id": "S4",
       "statement": "(c), no jump at zero. The denominator $f+\\max\\{y,\\nu\\}$ is at least $\\nu > 0$ for every $f \\ge 0$, so $s_\\nu(y,f)$ is a continuous function of $f$ on $f \\ge 0$, at $f = 0$ too. Since $0 \\le s_\\nu \\le 1$ (step 3), the expected term is continuous in $f$ as well, by bounded convergence. The sMAPE term of a zero actual is 0 at $f = 0$ and 1 for $f > 0$; so when zero actuals have probability $p_0 > 0$, the expected sMAPE term jumps by $p_0$ at $f = 0$.",
       "rule": "bounded convergence",
       "by": [
        "S3"
       ]
      },
      {
       "id": "S5",
       "statement": "(d). A count is 0 or at least 1. With $\\nu \\le 1$, a count of at least 1 is at least $\\nu$, so $s_\\nu = s$ by step 1. A zero count is below $\\nu$, where $s_\\nu(0,f) = f/(f+\\nu)$ and $s(0,f) = 1$ for $f > 0$ (step 2).",
       "rule": "algebra",
       "by": [
        "S1",
        "S2"
       ]
      },
      {
       "id": "S6",
       "statement": "So the msMAPE term is the sMAPE term for actuals of at least $\\nu$. Below $\\nu$ its denominator no longer shrinks with the actual: a zero actual scores $f/(f+\\nu)$, and the expected msMAPE has no jump at the zero forecast.",
       "rule": "conclusion",
       "by": [
        "S5"
       ]
      }
     ],
     "artefacts": [
      "A5"
     ],
     "availability": {
      "level": "public"
     }
    },
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/limitations/2/affects",
    "value": [
     "C14",
     "C15",
     "C21"
    ],
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/limitations/3/affects",
    "value": [
     "C13",
     "C14",
     "C15",
     "C16",
     "C21"
    ],
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "add",
    "path": "/limitations/4",
    "value": {
     "id": "L5",
     "category": "model_specification",
     "affects": [
      "C6"
     ],
     "impact": "reduces_transferability",
     "mitigation": "none",
     "statement": "C6 is proved for two groups that mirror each other about the median on the log scale. It is not proved for groups of different shape or spread."
    },
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "add",
    "path": "/limitations/5",
    "value": {
     "id": "L6",
     "category": "measurement",
     "affects": [
      "C20",
      "C21"
     ],
     "impact": "reduces_transferability",
     "mitigation": "none",
     "statement": "The definition of msMAPE is taken from Kolassa (2026, Definition 4). Smyl's article was not read. Smyl gives no value for the constant; ν = 0.1, 0.5 and 1 are this paper's choices."
    },
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "add",
    "path": "/limitations/6",
    "value": {
     "id": "L7",
     "category": "other",
     "affects": [
      "C19"
     ],
     "impact": "reduces_transferability",
     "mitigation": "partial",
     "mitigation_evidence": "E15",
     "statement": "That the relative efficiency of finite test sets tends to the asymptotic value is not shown; the simulation of C15 agrees with it at n = 10000."
    },
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/paper/classification/keywords",
    "value": [
     "forecast evaluation",
     "error measures",
     "sMAPE",
     "point forecasts",
     "consistent scoring functions",
     "proof",
     "simulation study",
     "msMAPE",
     "Suilin loss",
     "intermittent demand"
    ],
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/paper/contributions",
    "value": [
     "C1",
     "C12",
     "C4",
     "C15",
     "C6",
     "C3",
     "C9",
     "C10",
     "C7",
     "C17",
     "C18",
     "C19",
     "C20"
    ],
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/paper/exploration/hypotheses/supported",
    "value": 16,
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/paper/exploration/hypotheses/total",
    "value": 20,
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/paper/exploration/not_claimed",
    "value": [
     {
      "lab_event": "H1",
      "lab_plan": "PL5",
      "outcome": "supported",
      "because": "reworded",
      "into": "C1",
      "why": "The statement was reworded after this verdict and decided again under PL8.R2, whose verdict is the claim's. This verdict: PL5's rule for supported is met: the proof (paper H1 in proofs.md, AR1) is complete; the independent checker AC4 confirmed it (CK1, full); and the deciding run RN5's numerical checks of its formulas are within the plan's tolerances."
     },
     {
      "lab_event": "H2",
      "lab_plan": "PL5",
      "outcome": "supported",
      "because": "reworded",
      "into": "C2",
      "why": "The statement was reworded after this verdict and decided again under PL8.R2, whose verdict is the claim's. This verdict: PL5's rule for supported is met: the proof (paper H2 in proofs.md, AR1) is complete; the independent checker AC4 confirmed it (CK2, full); and the deciding run RN5's numerical checks of its formulas are within the plan's tolerances."
     },
     {
      "lab_event": "H3",
      "lab_plan": "PL5",
      "outcome": "supported",
      "because": "reworded",
      "into": "C3",
      "why": "The statement was reworded after this verdict and decided again under PL12, whose verdict is the claim's. This verdict: PL5's rule for supported is met: the proof (paper H3 in proofs.md, AR1) is complete; the independent checker AC4 confirmed it (CK3, full); and the deciding run RN5's numerical checks of its formulas are within the plan's tolerances."
     },
     {
      "lab_event": "H23",
      "lab_plan": "PL9",
      "outcome": "supported",
      "because": "reworded",
      "into": "C3",
      "why": "The statement was reworded after this verdict and decided again under PL12, whose verdict is the claim's. This verdict: PL8's rule for supported is met for the current wording: the proof (paper H3 in proofs.md, AR1.R2) is complete; the independent checker AC4 confirmed the current wording (CK3, full), as did the supervisor (CK16); and the deciding run RN8's numerical checks of its formulas are within the plan's tolerances (RN8 repeats RN5 byte for byte). Part (b), the small-error approximation, is checked by RN8's H9 lines (0 violations; the fifth-order coefficient 0.833, as the proof gives 5/6)."
     },
     {
      "lab_event": "H4",
      "lab_plan": "PL5",
      "outcome": "supported",
      "because": "reworded",
      "into": "C4",
      "why": "The statement was reworded after this verdict and decided again under PL12, whose verdict is the claim's. This verdict: Verdict renewed after H4 was corrected twice: H4.R3 (the median with equal weights is a flat point, not necessarily the target), then H4.R4 (a sufficient condition for it to be the target). PL5's rule for supported is met for the current wording: the proof (paper H5, steps 1 to 10) is complete; AC4 confirmed it (CK13, full); and RN5's numerical checks of its formulas are within the plan's tolerances. The earlier checks CK5 (disputed) and CK12 (confirmed) were of earlier wordings."
     },
     {
      "lab_event": "H33",
      "lab_plan": "PL9",
      "outcome": "supported",
      "because": "reworded",
      "into": "C4",
      "why": "The statement was reworded after this verdict and decided again under PL12, whose verdict is the claim's. This verdict: PL8's rule for supported is met for the current wording: the proof (paper H5, steps 1 to 10 in proofs.md, AR1.R2) is complete; the independent checker AC4 confirmed the current wording (CK13, full), as did the supervisor (CK18); and the deciding run RN8's numerical checks of its formulas are within the plan's tolerances (RN8 repeats RN5 byte for byte). Its direction result is checked by RN8's H6 lines (balance residual 8.6e-9; 0 settings misplaced)."
     },
     {
      "lab_event": "H5",
      "lab_plan": "PL5",
      "outcome": "supported",
      "because": "reworded",
      "into": "C5",
      "why": "The statement was reworded after this verdict and decided again under PL8.R2, whose verdict is the claim's. This verdict: PL5's rule for supported is met: the proof (paper H6 in proofs.md, AR1) is complete; the independent checker AC4 confirmed it (CK6, full); and the deciding run RN5's numerical checks of its formulas are within the plan's tolerances."
     },
     {
      "lab_event": "H16",
      "lab_plan": "PL5",
      "outcome": "inconclusive",
      "because": "revised",
      "into": "C4",
      "why": "Merged into H4 by the supervisor's decision DC5: its direction result (the target lies on the side of the median that carries more weight) is now a corollary of H4; its balance in quantile terms and its Newton step are dropped as adding little."
     },
     {
      "lab_event": "H18",
      "lab_plan": "PL5",
      "outcome": "supported",
      "because": "reworded",
      "into": "C6",
      "why": "The statement was reworded after this verdict and decided again under PL8.R2, whose verdict is the claim's. This verdict: PL5's rule for supported is met: the proof (paper H7 in proofs.md, AR1) is complete; the independent checker AC4 confirmed it (CK7, full); and the deciding run RN5's numerical checks of its formulas are within the plan's tolerances. RN5's check that the global minimiser lies on the heavier side (unequal weights) gave True; the lab could not record a boolean as a result, so it is in RN5's output file results.json."
     },
     {
      "lab_event": "H8",
      "lab_plan": "PL5",
      "outcome": "supported",
      "because": "reworded",
      "into": "C7",
      "why": "The statement was reworded after this verdict and decided again under PL8.R2, whose verdict is the claim's. This verdict: PL5's rule for supported is met: the proof (paper H8 in proofs.md, AR1) is complete; the independent checker AC4 confirmed it (CK8, full); and the deciding run RN5's numerical checks of its formulas are within the plan's tolerances."
     },
     {
      "lab_event": "H9",
      "lab_plan": "PL5",
      "outcome": "inconclusive",
      "because": "revised",
      "into": "C3",
      "why": "Merged into H3 by decision DC5: the small-error approximation is part (b) of H3."
     },
     {
      "lab_event": "H10",
      "lab_plan": "PL5",
      "outcome": "inconclusive",
      "because": "revised",
      "into": "C2",
      "why": "Merged into H2 by decision DC5: the comparison of high and low forecasts with equal absolute errors is part (b) of H2."
     },
     {
      "lab_event": "H11",
      "lab_plan": "PL5",
      "outcome": "inconclusive",
      "because": "abandoned",
      "why": "Dropped by decision DC5: it applied a standard large-sample theorem; the efficiency of the sMAPE minimiser against the sample median is reported from the simulation PL2 instead, and part (iii) of H14 is dropped with it."
     },
     {
      "lab_event": "H12",
      "lab_plan": "PL5",
      "outcome": "supported",
      "because": "reworded",
      "into": "C8",
      "why": "The statement was reworded after this verdict and decided again under PL8.R2, whose verdict is the claim's. This verdict: PL5's rule for supported is met: the proof (paper H10 in proofs.md, AR1) is complete; the independent checker AC4 confirmed it (CK10, full); and the deciding run RN5's numerical checks of its formulas are within the plan's tolerances."
     },
     {
      "lab_event": "H13",
      "lab_plan": "PL5",
      "outcome": "supported",
      "because": "reworded",
      "into": "C9",
      "why": "The statement was reworded after this verdict and decided again under PL8.R2, whose verdict is the claim's. This verdict: PL5's rule for supported is met: the proof (paper H4 in proofs.md, AR1) is complete; the independent checker AC4 confirmed it (CK4, full); and the deciding run RN5's numerical checks of its formulas are within the plan's tolerances."
     },
     {
      "lab_event": "H19",
      "lab_plan": "PL5",
      "outcome": "supported",
      "because": "reworded",
      "into": "C10",
      "why": "The statement was reworded after this verdict and decided again under PL8.R2, whose verdict is the claim's. This verdict: PL5's rule for supported is met: the proof (paper H11 in proofs.md, AR1) is complete; the independent checker AC4 confirmed it (CK11, full); and the deciding run RN5's numerical checks of its formulas are within the plan's tolerances. Part (iii), the efficiency rate, was dropped by DC5; RN5's checks of it no longer decide anything."
     },
     {
      "lab_event": "H34",
      "lab_plan": "PL9",
      "outcome": "supported",
      "because": "reworded",
      "into": "C12",
      "why": "The statement was reworded after this verdict and decided again under PL12, whose verdict is the claim's. This verdict: PL8's rule for supported is met: the proof (paper H6 for part (a); paper H5, steps 3, 6, 8 and 9, for part (b), in proofs.md, AR1.R2) is complete; the independent checker AC4 confirmed H34 as worded (CK25, full; recorded 20 seconds after PL8 was first written, as PL8.R2 states); and the deciding run RN8's numerical checks of the formulas it rests on (paper H5 and H6: the minimiser, the median, the balance) are within the plan's tolerances. The supervisor's checks of H4 and H5 (CK18, CK19) cover both parts; no separate supervisor check of H34 was recorded."
     },
     {
      "lab_event": "H35",
      "lab_plan": "PL10",
      "outcome": "supported",
      "because": "reworded",
      "into": "C17",
      "why": "The statement was reworded after this verdict and decided again under PL12, whose verdict is the claim's. This verdict: PL10's rule for supported is met: the proof of H35 (proof-msmape.md, AR8) is complete, every step naming what it rests on; the independent checker AC4 derived H35 from its statement alone, then compared it with the proof, and recorded its check as confirmed, full (CK26); and the deciding run RN9's checks meet the tolerances: the identities of (a) and (b) within 1.2e-16, no violation of the bounds or of 0 <= m < 1, and a zero forecast of a zero actual scoring 0. Steps 4 and 7 of the proof, which the checker suggested and had derived itself, were added after its check."
     },
     {
      "lab_event": "H39",
      "lab_plan": "PL12",
      "outcome": "not_supported",
      "because": "reworded",
      "into": "C18",
      "why": "The statement was reworded after this verdict and decided again under PL17, whose verdict is the claim's. This verdict: As worded, H39 does not hold in every case it allows. The independent checker (CK30) gave a counterexample to part (c): a density with bounded support, for which H(c_2) can be 0 or 1 and the stated limit is 1/0 or the ratio undefined; the proof's step 8 adds the condition 0 < H(c_2) < 1 but the statement did not. Part (b) also needs sd(D) > 0, and the closing sentence claimed more than the parts give. In addition one of RN13's checks of part (c) missed its tolerance (3.5% against 3% at sigma = 200, a slow convergence, not an error). The mathematics of (a), (b) and (d) was confirmed. The statement is corrected as H43."
     },
     {
      "lab_event": "H39",
      "lab_plan": "PL13",
      "outcome": "not_supported",
      "because": "reworded",
      "into": "C18",
      "why": "The statement was reworded after this verdict and decided again under PL17, whose verdict is the claim's. This verdict: PL13's numerical tolerances are all met by RN14 (the absolute error's ratio is within 0.0035 of its limit at sigma = 2000 and closer than at sigma = 200 in all six cases), but the checker's counterexample to the wording of part (c) (CK30) stands whatever the run shows. The statement is corrected as H43, which PL15 decides."
     },
     {
      "lab_event": "H43",
      "lab_plan": "PL15",
      "outcome": "inconclusive",
      "because": "reworded",
      "into": "C18",
      "why": "The statement was reworded after this verdict and decided again under PL17, whose verdict is the claim's. This verdict: Neither rule is met for H43. The run RN16 meets every tolerance and the checker found parts (a) to (d) right as now worded, but disputed the closing sentence again (CK34): 'sMAPE loses it, at the rate sigma^(-1/2) when h(c_1) != h(c_2)' says sMAPE loses its power in every case, which is shown when h(c_1) + h(c_2) > 0 (the ratio is then of order sigma^(-5/2) when the two densities are equal) but not when both are 0, as a density with bounded support allows; no counterexample was found. The sentence is to be narrowed in a further correction."
     },
     {
      "lab_event": "H45",
      "lab_plan": "PL16",
      "outcome": "inconclusive",
      "because": "reworded",
      "into": "C18",
      "why": "The statement was reworded after this verdict and decided again under PL17, whose verdict is the claim's. This verdict: Neither rule is met for H45. The run RN17 meets every tolerance, and the checker found parts (a) to (d) and four of the five closing sentences right, but disputed one (CK36): 'with a ratio of order sigma^(-5/2)' reads as the exact order, and only an upper bound is shown (the ratio is 0 for a symmetric density with c_1 = -c_2). It is to read 'at most of order'."
     },
     {
      "lab_event": "H40",
      "lab_plan": "PL12",
      "outcome": "not_supported",
      "because": "reworded",
      "into": "C19",
      "why": "The statement was reworded after this verdict and decided again under PL15, whose verdict is the claim's. This verdict: As worded, H40 calls the ratio of the two limiting variances the large-sample value of the relative efficiency, which is defined from the variances of finite test sets. The independent checker (CK31) confirmed all of the mathematics (the strict inequality, consistency, the conditions of the theorem, the variance, the constants and both rates) and gave a counterexample to that phrase: log-actuals with very heavy tails, under H40's conditions, for which the sample median has infinite variance at every n. RN13's numerical checks of the formula and of the two rates met their tolerances. The statement is corrected as H44."
     }
    ],
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/paper/exploration/questions",
    "value": 3,
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/paper/exploration/runs/confirmatory",
    "value": 14,
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/paper/exploration/runs/exploratory",
    "value": 4,
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/paper/exploration/runs/total",
    "value": 18,
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/paper/exploration/tests/inconclusive",
    "value": 6,
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/paper/exploration/tests/not_supported",
    "value": 3,
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/paper/exploration/tests/supported",
    "value": 30,
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "replace",
    "path": "/paper/exploration/tests/total",
    "value": 39,
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "add",
    "path": "/paper/sections",
    "value": [
     {
      "title": "What sMAPE measures",
      "claims": [
       "C1",
       "C2",
       "C3"
      ]
     },
     {
      "title": "What sMAPE rewards for positive actuals",
      "claims": [
       "C9",
       "C4",
       "C5",
       "C12",
       "C6"
      ]
     },
     {
      "title": "Zeros and negative actuals",
      "claims": [
       "C7",
       "C11"
      ]
     },
     {
      "title": "Widely spread actuals",
      "claims": [
       "C10",
       "C18"
      ]
     },
     {
      "title": "Other measures",
      "claims": [
       "C8"
      ]
     },
     {
      "title": "sMAPE modifications",
      "claims": [
       "C17",
       "C20",
       "C16",
       "C21"
      ]
     },
     {
      "title": "Precision on a test set",
      "claims": [
       "C15",
       "C19"
      ]
     },
     {
      "title": "Numerical checks",
      "claims": [
       "C13",
       "C14"
      ]
     }
    ],
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "add",
    "path": "/references/14",
    "value": {
     "id": "B15",
     "target": "10.1017/CBO9780511802256",
     "id_type": "doi"
    },
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   },
   {
    "op": "add",
    "path": "/references/15",
    "value": {
     "id": "B16",
     "target": "Smyl, S. (2025). Sparse-Proof sMAPE. Foresight: The International Journal of Applied Forecasting, 78, 29–30. https://ideas.repec.org/a/for/ijafaa/y2025i78p29-30.html",
     "id_type": "reference"
    },
    "reason": "Version 3 answers a reader's comments. Corrected: C3 no longer says sMAPE is minimised by the median for small errors, only near it, with a bound; C4 keeps the weighted median, and C12 now holds what was said about the ordinary median; the measure called msMAPE in version 2 is the Suilin loss. Added: how well a test set tells two forecasts apart as the actuals spread out (C18); the large-sample distribution of the sMAPE minimiser (C19); Smyl's msMAPE (C20) and what it rewards on intermittent demand (C21). Also added: sections, the definition of sMAPE first, and a limitation on C6."
   }
  ]
 },
 "contributors": [
  {
   "id": "P1",
   "type": "agent",
   "name": "Claude Code",
   "positions": [
    "author"
   ],
   "models": [
    {
     "model": "M1",
     "role": "sole"
    }
   ],
   "run_period": {
    "start": "2026-11-16",
    "end": "2026-11-18"
   },
   "tools": [
    "code_execution",
    "file_access"
   ],
   "system": {
    "name": "Claude Code"
   },
   "label": "derivations and simulations",
   "record_source": "self_reported",
   "credit_roles": [
    "methodology",
    "formal_analysis",
    "software",
    "investigation",
    "visualization",
    "writing_original_draft"
   ],
   "ai_tasks": [
    {
     "task": "analysis",
     "extent": "full"
    },
    {
     "task": "code",
     "extent": "full"
    },
    {
     "task": "writing",
     "extent": "full"
    },
    {
     "task": "json_preparation",
     "extent": "full"
    }
   ]
  },
  {
   "id": "P2",
   "type": "human",
   "name": "Ivan Svetunkov",
   "positions": [
    "supervisor",
    "guarantor"
   ],
   "orcid": "0000-0001-7826-0281",
   "credit_roles": [
    "conceptualization",
    "supervision",
    "validation",
    "writing_review_editing"
   ]
  },
  {
   "id": "P3",
   "type": "agent",
   "name": "Claude Code",
   "positions": [
    "contributor"
   ],
   "models": [
    {
     "model": "M2",
     "role": "sole"
    }
   ],
   "run_period": {
    "start": "2026-11-16",
    "end": "2026-11-17"
   },
   "ai_tasks": [
    {
     "task": "review",
     "extent": "full"
    }
   ],
   "system": {
    "name": "Claude Code"
   },
   "label": "proof checking",
   "record_source": "self_reported"
  },
  {
   "id": "P4",
   "type": "agent",
   "name": "Claude Code",
   "positions": [
    "author"
   ],
   "models": [
    {
     "model": "M2",
     "role": "sole"
    }
   ],
   "run_period": {
    "start": "2026-11-18",
    "end": "2026-11-18"
   },
   "tools": [
    "code_execution",
    "file_access"
   ],
   "system": {
    "name": "Claude Code"
   },
   "label": "version 3: corrections and new derivations",
   "record_source": "harness",
   "credit_roles": [
    "methodology",
    "formal_analysis",
    "software",
    "investigation",
    "visualization",
    "writing_original_draft"
   ],
   "ai_tasks": [
    {
     "task": "analysis",
     "extent": "full"
    },
    {
     "task": "code",
     "extent": "full"
    },
    {
     "task": "writing",
     "extent": "full"
    },
    {
     "task": "json_preparation",
     "extent": "full"
    }
   ]
  },
  {
   "id": "P5",
   "type": "agent",
   "name": "Claude Code",
   "positions": [
    "contributor"
   ],
   "models": [
    {
     "model": "M1",
     "role": "sole"
    }
   ],
   "run_period": {
    "start": "2026-11-18",
    "end": "2026-11-18"
   },
   "ai_tasks": [
    {
     "task": "review",
     "extent": "full"
    }
   ],
   "system": {
    "name": "Claude Code"
   },
   "label": "proof checking, version 3",
   "record_source": "self_reported"
  }
 ],
 "models": [
  {
   "id": "M1",
   "provider": "Anthropic",
   "name": "Claude Opus",
   "version": "5.5",
   "identifier": "claude-opus-5-5",
   "accessed": "2026-11-16"
  },
  {
   "id": "M2",
   "provider": "Anthropic",
   "name": "Claude Fable",
   "version": "5.2",
   "identifier": "claude-fable-5-2",
   "accessed": "2026-11-16"
  }
 ],
 "definitions": [
  {
   "id": "D2",
   "term": "sMAPE",
   "kind": "variable",
   "definition": "Symmetric mean absolute percentage error of a point forecast $f$ over a test set of actuals $y_1, \\dots, y_n$: $\\frac{200}{n}\\sum_t |y_t - f|/(|y_t|+|f|)$, where a term with $y_t = f = 0$ counts as 0. Introduced as the symmetric MAPE by Makridakis (1993), whose denominator had no absolute values; for positive actuals and forecasts the two agree.",
   "source": "B8"
  },
  {
   "id": "D1",
   "term": "sMAPE term",
   "kind": "variable",
   "definition": "The contribution of one actual $y$ and one point forecast $f$ to [[D2:sMAPE]]: $s(y,f) = |y-f|/(|y|+|f|)$, between 0 and 1, with $s = 0$ when $y = f$ (including $y = f = 0$, where the fraction is $0/0$).",
   "symbol": "$s(y,f)$"
  },
  {
   "id": "D3",
   "term": "expected sMAPE",
   "kind": "variable",
   "definition": "The expected sMAPE of a fixed forecast $f$ under the distribution of the actual $y$: $200\\,\\mathrm{E}[s(y, f)]$, where $\\mathrm E$ is the expectation over the random actual $y$. Computed from the distribution (by integration or exact sums), never from a sample."
  },
  {
   "id": "D6",
   "term": "sMAPE target",
   "kind": "concept",
   "definition": "The value or values of $f$ that minimise expected sMAPE for a given distribution of the actual $y$, $\\arg\\min_f \\mathrm E[s(y,f)]$: the functional sMAPE rewards. It may be a single value or several."
  },
  {
   "id": "D4",
   "term": "log error",
   "kind": "variable",
   "definition": "For a positive actual $y$ and a positive forecast $f$, $z = \\ln y - \\ln f$.",
   "symbol": "$z$"
  },
  {
   "id": "D5",
   "term": "weight",
   "kind": "variable",
   "definition": "For a positive actual $y$ and a positive forecast $f$, $w(y,f) = y/(y+f)^2$: how much the actual pulls the forecast that minimises expected sMAPE, since the derivative of $\\mathrm E[s(y,f)]$ in $f$ is $2\\,\\mathrm E[w(y,f)(\\mathbf 1\\{y<f\\} - \\mathbf 1\\{y>f\\})]$. Largest at $y = f$, the same for $y/f = r$ and $1/r$.",
   "symbol": "$w(y,f)$"
  },
  {
   "id": "D7",
   "term": "quantile profile",
   "kind": "concept",
   "definition": "For an outcome distribution with quantile function $Q$ and a forecast $f$, the pull of the outcome at each quantile level $u$, $u \\mapsto w(Q(u), f)$ on $(0,1)$; the minimiser's quantile level balances its integrals below and above."
  },
  {
   "id": "D10",
   "term": "RMSE",
   "kind": "variable",
   "definition": "Root mean squared error of a point forecast $f$ over a test set of actuals $y_1, \\dots, y_n$: $\\sqrt{\\frac1n\\sum_t (y_t-f)^2}$."
  },
  {
   "id": "D11",
   "term": "MAE",
   "kind": "variable",
   "definition": "Mean absolute error of a point forecast $f$ over a test set of actuals $y_1, \\dots, y_n$: $\\frac1n\\sum_t |y_t-f|$."
  },
  {
   "id": "D12",
   "term": "MAPE",
   "kind": "variable",
   "definition": "Mean absolute percentage error of a point forecast $f$ over a test set of actuals $y_1, \\dots, y_n$: $\\frac{100}{n}\\sum_t |y_t-f|/|y_t|$; infinite when an actual is zero."
  },
  {
   "id": "D19",
   "term": "log-MAE",
   "kind": "variable",
   "definition": "100 times the mean absolute error of $\\ln f$ as a forecast of $\\ln y$: $\\frac{100}{n}\\sum_t |\\ln y_t - \\ln f|$. Unbounded; sMAPE's small-error limit (C3); minimised by the median."
  },
  {
   "id": "D28",
   "term": "Suilin term",
   "kind": "variable",
   "definition": "The contribution of one actual $y$ and one forecast $f$ to the Suilin loss:\n$$\\tilde s(y,f) = \\frac{|y-f|}{\\max(|y|+|f|+\\delta,\\ 0.5+\\delta)}, \\qquad \\delta = 0.1;$$\nthe Suilin loss is 200 times its mean over a test set. It replaces the denominator $|y|+|f|$ of the [[D1:sMAPE term]]. Proposed by Suilin (2017) in code for a Kaggle competition, as a smoothed variant of sMAPE; defined in print by Hewamalage, Bergmeir and Bandara (2021, Eq. 15), who write the constant $\\delta$ as $\\epsilon$ and the factor 2 into the measure differently (a constant factor changes no property here). Hewamalage, Ackermann and Bergmeir (2023) call it msMAPE; Smyl (2025) gives that name to a different measure, so this study follows Kolassa (2026) and calls it the Suilin loss. Version 2 of this paper called it msMAPE.",
   "source": "B14",
   "symbol": "$\\tilde s(y,f)$"
  },
  {
   "id": "D29",
   "term": "Suilin target",
   "kind": "concept",
   "definition": "The value or values of $f$ that minimise the expected [[D28:Suilin term]] for a given distribution of the actual: the functional the Suilin loss rewards, as the [[D6:sMAPE target]] is the one sMAPE rewards."
  },
  {
   "id": "D33",
   "term": "msMAPE term",
   "kind": "variable",
   "definition": "The contribution of one actual $y \\ge 0$ and one forecast $f \\ge 0$ to the sparse-proof sMAPE (msMAPE) of Smyl (2025), as Kolassa (2026, Definition 4) states it, for a small constant $\\nu > 0$ (Kolassa writes it $m$, which is the median here):\n$$s_\\nu(y,f) = \\frac{|f-y|}{f+\\max\\{y, \\nu\\}},$$\nwhich is 0 when $y = f = 0$; msMAPE is 200 times its mean over a test set (Kolassa writes the factor 2 into the term). It replaces the actual in the denominator of the [[D1:sMAPE term]] by $\\nu$ when the actual is below $\\nu$.",
   "source": "B13",
   "symbol": "$s_\\nu(y,f)$"
  },
  {
   "id": "D34",
   "term": "msMAPE target",
   "kind": "concept",
   "definition": "The value or values of $f$ that minimise the expected [[D33:msMAPE term]] for a given distribution of the actual and a given constant $\\nu$: the functional msMAPE rewards, as the [[D6:sMAPE target]] is the one sMAPE rewards."
  },
  {
   "id": "D8",
   "term": "sMAPE minimiser of a test set",
   "kind": "variable",
   "definition": "The forecast that minimises the sMAPE of one test set $y_1, \\dots, y_n$, $\\hat f_n = \\arg\\min_f \\frac1n\\sum_t s(y_t, f)$, found numerically (the sample counterpart of the [[D6:sMAPE target]]).",
   "symbol": "$\\hat f_n$"
  },
  {
   "id": "D9",
   "term": "relative efficiency",
   "kind": "variable",
   "definition": "For a test-set size $n$ and an outcome distribution where sMAPE and MAE both target the median: $\\mathrm{RE}(n) = \\operatorname{Var}(\\ln \\hat m_n) / \\operatorname{Var}(\\ln \\hat f_n)$ over test sets, where $\\hat m_n$ is the sample median (the MAE minimiser) and $\\hat f_n$ the sMAPE minimiser of the same test set. Above 1, sMAPE's minimiser varies less; below 1, MAE's does.",
   "symbol": "$\\mathrm{RE}(n)$"
  },
  {
   "id": "D20",
   "term": "ranking agreement",
   "kind": "variable",
   "definition": "For one distribution, one measure, a test-set size $n$, the measure's target and a forecast at another quantile level: the probability that a test set of $n$ outcomes gives the target the strictly lower score (ties count as disagreement); 0.5 is a coin toss."
  },
  {
   "id": "D32",
   "term": "signal-to-noise ratio of a comparison",
   "kind": "variable",
   "definition": "For a measure with term $L(y,f)$, two forecasts $f_1$ and $f_2$ and a distribution of the actual $y$: $|\\mathrm E[D]|/\\mathrm{sd}(D)$, where $D = L(y,f_2) - L(y,f_1)$ is the difference of the two forecasts' scores on one actual. Over a test set of $n$ independent actuals the mean of $D$ is $\\sqrt n$ times this many of its standard errors from zero: the larger the ratio, the smaller the test set that tells the two forecasts apart."
  },
  {
   "id": "D13",
   "term": "candidate forecast",
   "kind": "concept",
   "definition": "A point forecast fixed from the outcome distribution before any test set is drawn and the same for every test set: a quantile at a level from 0.01 to 0.99, the mean, the forecast 0, the log-mode, the minimiser of expected MAPE, or a local minimiser of expected sMAPE. It is never estimated from a sample."
  },
  {
   "id": "D14",
   "term": "preferred candidate",
   "kind": "concept",
   "definition": "Of the candidate forecasts for one distribution, the one a measure gives the lowest expected score, or, where every expected score is infinite but their differences are finite, the one the differences rank first (C8); none where no forecast is ranked first."
  },
  {
   "id": "D15",
   "term": "derived value",
   "kind": "concept",
   "definition": "A value computed from the outcome distribution itself by numerical integration of its density (Gauss-Legendre quadrature and a convolution on a grid of 524288 points of $\\ln y$), never from a sample: the expected scores, the local minima of expected sMAPE, the formulas of the hypotheses."
  },
  {
   "id": "D16",
   "term": "selection design",
   "kind": "method",
   "definition": "For each outcome distribution: derive the candidate forecasts and their expected scores; then, for each test-set size $n$, draw test sets of $n$ independent outcomes, score every candidate by every measure on each test set, and count the candidate with the lowest score as the measure's choice (a tie of $j$ candidates counts $1/j$ each; a test set in which every candidate scores infinite chooses none)."
  },
  {
   "id": "D17",
   "term": "efficiency design",
   "kind": "method",
   "definition": "For each outcome distribution and test-set size $n$: draw test sets of $n$ independent outcomes; in each, compute the sample median and the exact sMAPE minimiser (an outcome of the test set, since the test set's sMAPE is concave in $\\ln f$ between consecutive outcomes); over test sets, compare the variances of their logarithms."
  },
  {
   "id": "D18",
   "term": "flatness design",
   "kind": "method",
   "definition": "For families whose spread $\\sigma$ of $\\ln y$ grows from 0.1 to 10: derive the sMAPE target, the expected sMAPE of forecasts at fixed quantile levels and the large-spread approximation of C10; on test sets, count how often each measure scores the target lower than a forecast away from it, and where the test set's sMAPE minimiser falls."
  },
  {
   "id": "D31",
   "term": "formula checks",
   "kind": "method",
   "definition": "Numerical checks of the formulas of the formal claims (claimset-code/smape-target/check_formulas.py, seed 6; check_jump.py for C11): identities and inequalities on 200000 random pairs, derivatives against finite differences, minimisers, balances and thresholds against quadrature and bisection on every setting of the selection design, and the large-spread expansion against its bound. For the claims new or reworded in version 3 (C3, C4, C12, C17 to C20): check_v3.py (seed 8), check_h39.py and check_corrected.py (seed 8), in the same way."
  },
  {
   "id": "D21",
   "term": "inverse Gaussian outcomes",
   "kind": "population",
   "definition": "Inverse Gaussian outcomes with mean 100 and shape 1000, 200, 100, 50, 20 or 10 (coefficient of variation 0.32 to 3.16): positive, skewed right, every moment and $\\mathrm E[1/y]$ finite."
  },
  {
   "id": "D22",
   "term": "log-skewed outcomes",
   "kind": "population",
   "definition": "$y = 100\\,e^x$ with $x = \\pm\\sigma(G-\\alpha)/\\sqrt\\alpha$, $G \\sim \\mathrm{Gamma}(\\alpha, 1)$: $\\ln y$ with standard deviation $\\sigma$ (0.5 or 1; sd in the settings' names) and skewness $\\pm 2/\\sqrt\\alpha$ (-2, -1, -0.5, +0.5, +1, +2), and the lognormal of the same $\\sigma$ (skewness 0). Some have infinite moments: $\\mathrm E[y]$ infinite ($\\sigma = 1$, skewness +2), $\\mathrm E[y^2]$ infinite ($\\sigma = 0.5$, skewness +2; $\\sigma = 1$, skewness +1), $\\mathrm E[1/y]$ infinite ($\\sigma = 1$, skewness -2)."
  },
  {
   "id": "D23",
   "term": "bimodal outcomes",
   "kind": "population",
   "definition": "$y = 100\\,e^x$, $x$ normal with mean $-1$ (probability $p$) or $+1$, standard deviation 0.25; $p = 0.5$ (equal modes), 0.65 (left mode higher), 0.35 (right mode higher)."
  },
  {
   "id": "D24",
   "term": "outcomes straddling zero",
   "kind": "population",
   "definition": "Normal outcomes with standard deviation 1 and mean 0, 0.5 or 1: 50%, 31% and 16% negative; $\\mathrm E[1/|y|]$ infinite."
  },
  {
   "id": "D25",
   "term": "zero-inflated outcomes",
   "kind": "population",
   "definition": "Lognormal outcomes (median 100, $\\sigma$ 0.5 or 1) replaced by 0 with probability $p_0$: 0.25, 0.42, 0.47, 0.6 for $\\sigma = 0.5$ and 0.25, 0.37, 0.42, 0.6 for $\\sigma = 1$, on both sides of the zero threshold of C7 (0.4469 and 0.3940)."
  },
  {
   "id": "D26",
   "term": "log-symmetric outcomes",
   "kind": "population",
   "definition": "Lognormal outcomes with $\\sigma$, and log-Laplace outcomes with scale $b$, each 0.25, 0.5, 1, 1.5, 2, 3, 4 or 6, median 100: $\\ln y$ symmetric and strictly unimodal, so that sMAPE and MAE both target the median (C5)."
  },
  {
   "id": "D27",
   "term": "widening outcomes",
   "kind": "population",
   "definition": "$\\ln y = \\ln 100 + \\sigma\\varepsilon$ with $\\varepsilon$ standard normal (lognormal), standard Gumbel (log-Gumbel right, $\\ln y$ skewed right) or minus standard Gumbel (log-Gumbel left), for $\\sigma$ = 0.1, 0.25, 0.5, 1, 1.5, 2, 3, 4, 6 and 10."
  },
  {
   "id": "D30",
   "term": "Poisson counts",
   "kind": "population",
   "definition": "Poisson outcomes with means 0.3, 0.5, 1, 1.5, 3 and 10: counts whose share of zeros falls from 0.74 to almost 0."
  },
  {
   "id": "D35",
   "term": "intermittent demand settings",
   "kind": "population",
   "definition": "Demand with many zeros. Counts: Poisson with means 0.1, 0.3, 0.5, 0.7, 1, 1.5 and 3 (share of zeros from 0.90 to 0.05) and, overdispersed, negative binomial with size 0.5 and means 0.5, 1 and 2 (share of zeros 0.71, 0.58 and 0.45). Non-integer demand: lognormal demand sizes (median 1, $\\sigma = 1$) replaced by 0 with probability 0.2, 0.4, 0.6 or 0.8."
  }
 ],
 "claims": [
  {
   "id": "C1",
   "name": "sMAPE depends only on the ratio of actual to forecast",
   "summary": "Only the ratio of the actual to the forecast matters to sMAPE, not their size.",
   "statement": "For $y > 0$ and $f > 0$, the [[D1:sMAPE term]] $s(y,f) = |y-f|/(|y|+|f|)$ depends on $y$ and $f$ only through their ratio $r = y/f$: $s = |r - 1|/(r + 1)$.",
   "about": {
    "subject": "D1",
    "relation": "depends only on",
    "object": "the ratio of actual to forecast",
    "comparator": null,
    "conditions": []
   },
   "type": "formal",
   "role": "principal",
   "novelty": "extension",
   "support": [
    {
     "kind": "formal_proof",
     "evidence": "E18"
    },
    {
     "kind": "empirical_quant",
     "evidence": "E1"
    }
   ],
   "provenance": "author_stated",
   "hypothesis": {
    "lab_event": "H21",
    "outcome": "supported"
   },
   "planning": "confirmatory",
   "planned_at": "2026-11-17T09:21:09Z",
   "planned_at_anchor": "git_remote",
   "test": {
    "framework": "formal",
    "supported": "The proof of the hypothesis, as now worded, in the study's final proof document (proofs.md, AR1.R2) is complete, every step naming what it rests on; the independent checker AC4 has a check of the current wording with outcome confirmed; and the deciding run's numerical checks of its formulas meet the tolerances of PL5: identities on random pairs within 1e-9 (on the scale of sMAPE), quantities computed by quadrature within 1e-4 relative (or absolute, for a residual the proof says is zero), counts of violations 0, and the expansion's error times sigma^3 below the proof's bound.",
    "not_supported": "A counterexample to the statement as now worded, found by the proof, a checker, the supervisor or the deciding run, that a correction of the proof cannot remove.",
    "verdict": "judged",
    "runs": 1,
    "lab_plan": "PL9"
   },
   "checks": [
    {
     "by": "P3",
     "outcome": "confirmed",
     "extent": "full",
     "notes": "Derived from the statement alone (first wording, with the factor 200), then compared with the final proofs.md (paper H1), which agrees after rescaling. No conditions needed."
    },
    {
     "by": "P2",
     "outcome": "confirmed",
     "extent": "full",
     "notes": "The supervisor read the proof of paper H1 in proofs.md (AR1.R2; rendered as AR2.R2) and confirmed it: first \"Confirmed\" (DC6), then, after the last corrections of paper H4, H5 and H9, \"All looks fine now. NExt\"."
    }
   ],
   "verification": "agent_checked",
   "relations": [
    {
     "type": "extends",
     "reference": "B12",
     "note": "Tofallis (2015) lists the sMAPE term, |y-f|/[(|y|+|f|)/2], among Tornqvist et al.'s measures of relative change; C1 shows it depends on the actual and the forecast only through their ratio."
    }
   ],
   "verifiability": "open",
   "verifiability_flags": []
  },
  {
   "id": "C2",
   "name": "Too high and too low by the same factor score the same",
   "summary": "A forecast twice too high and one half too low score the same. For equal absolute errors the higher one scores better. This compares forecasts for one known actual. Which forecast is best when the actual is uncertain is claim C4: sMAPE does not in general reward forecasting too high.",
   "statement": "For $y > 0$ and $f > 0$: (a) replacing $r = y/f$ by $1/r$ leaves the [[D1:sMAPE term]] unchanged, so a forecast $k$ times too high ($f = ky$) and one $k$ times too low ($f = y/k$) score the same; (b) of two forecasts the same absolute distance $d$ from $y$ ($0 < d < y$), the higher one scores better:\n$$s(y, y+d) = \\frac{d}{2y+d} < \\frac{d}{2y-d} = s(y, y-d).$$",
   "about": {
    "subject": "D1",
    "relation": "is symmetric in",
    "object": "the ratio of actual to forecast",
    "comparator": null,
    "conditions": []
   },
   "type": "formal",
   "role": "secondary",
   "novelty": "no_prior_found",
   "support": [
    {
     "kind": "formal_proof",
     "evidence": "E19"
    },
    {
     "kind": "empirical_quant",
     "evidence": "E1"
    }
   ],
   "provenance": "author_stated",
   "hypothesis": {
    "lab_event": "H22",
    "outcome": "supported"
   },
   "planning": "confirmatory",
   "planned_at": "2026-11-17T09:21:09Z",
   "planned_at_anchor": "git_remote",
   "test": {
    "framework": "formal",
    "supported": "The proof of the hypothesis, as now worded, in the study's final proof document (proofs.md, AR1.R2) is complete, every step naming what it rests on; the independent checker AC4 has a check of the current wording with outcome confirmed; and the deciding run's numerical checks of its formulas meet the tolerances of PL5: identities on random pairs within 1e-9 (on the scale of sMAPE), quantities computed by quadrature within 1e-4 relative (or absolute, for a residual the proof says is zero), counts of violations 0, and the expansion's error times sigma^3 below the proof's bound.",
    "not_supported": "A counterexample to the statement as now worded, found by the proof, a checker, the supervisor or the deciding run, that a correction of the proof cannot remove.",
    "verdict": "judged",
    "runs": 1,
    "lab_plan": "PL9"
   },
   "checks": [
    {
     "by": "P3",
     "outcome": "confirmed",
     "extent": "full",
     "notes": "Derived from the statement alone (earlier wording: old H2 and H10), then compared with the final proofs.md (paper H2), which agrees. No conditions needed."
    },
    {
     "by": "P2",
     "outcome": "confirmed",
     "extent": "full",
     "notes": "The supervisor read the proof of paper H2 in proofs.md (AR1.R2; rendered as AR2.R2) and confirmed it: first \"Confirmed\" (DC6), then, after the last corrections of paper H4, H5 and H9, \"All looks fine now. NExt\"."
    }
   ],
   "verification": "agent_checked",
   "relations": [
    {
     "type": "qualifies",
     "note": "C33 says sMAPE penalises a forecast below the outcome more than one above it. That holds for two forecasts the same absolute distance away (C2 b); a forecast k times too low and one k times too high score the same (C2 a).",
     "reference": "B3"
    },
    {
     "type": "refines",
     "reference": "B1",
     "note": "Goodwin and Lawton (1999) showed that sMAPE is asymmetric: for the same absolute error it scores a forecast above the actual better than one below. C2 proves this (b) and shows that by ratio sMAPE is symmetric (a)."
    },
    {
     "type": "refines",
     "reference": "B9",
     "note": "Hyndman and Koehler (2006) note that for the same actual sMAPE penalises low forecasts more heavily than high ones; C2 shows this holds for equal absolute errors and not for equal ratios."
    },
    {
     "type": "refines",
     "reference": "B10",
     "note": "Hewamalage et al. (2023) note that sMAPE penalises underestimates more than overestimates for the same actual; C2 shows this holds for equal absolute errors and not for equal ratios."
    },
    {
     "type": "refines",
     "reference": "B11",
     "note": "Chen et al. (2017) show a heavier penalty on a forecast 10% under the actual than on one 10% over; C2 shows that forecasts the same factor too high and too low score the same."
    }
   ],
   "depends_on": [
    "C1"
   ],
   "verifiability": "open",
   "verifiability_flags": []
  },
  {
   "id": "C3",
   "name": "sMAPE is a hyperbolic tangent of the log error",
   "summary": "sMAPE measures the error on the log scale. For small errors it is half the absolute log error; for large errors it levels off. For small errors its best forecast is close to the median, but not in general the median.",
   "statement": "For $y > 0$ and $f > 0$, the [[D1:sMAPE term]] is $s = \\tanh(|\\ln y - \\ln f|/2)$ ([[D4:log error]] $\\ln y - \\ln f$). (a) It is 0 when $y = f$, strictly increasing and concave in $|\\ln y - \\ln f|$, and bounded by 1, which it approaches as $|\\ln y - \\ln f|$ grows. (b) $0 \\le \\tfrac12|\\ln y - \\ln f| - s \\le \\tfrac1{24}|\\ln y - \\ln f|^3$; so when all log errors are small, sMAPE is about 100 times the mean absolute error of $\\ln f$ as a forecast of $\\ln y$ ([[D19:log-MAE]]). (c) The log-MAE of a test set is minimised by its median; sMAPE is then minimised near the median, not in general at it. For positive actuals $y_1, \\dots, y_n$ with a median $m$, a forecast $f^*$ that minimises their sMAPE satisfies\n$$\\frac1n\\sum_t|\\ln y_t - \\ln f^*| - \\frac1n\\sum_t|\\ln y_t - \\ln m| \\le \\frac1{12}\\cdot\\frac1n\\sum_t|\\ln y_t - \\ln f^*|^3:$$\nits mean absolute log error exceeds the smallest possible by at most a twelfth of its mean cubed absolute log error. The minimiser is the median itself only under further conditions on the distribution of the actuals.",
   "about": {
    "subject": "D1",
    "relation": "is a bounded function of",
    "object": "D4",
    "comparator": null,
    "conditions": []
   },
   "type": "formal",
   "role": "principal",
   "novelty": "extension",
   "support": [
    {
     "kind": "formal_proof",
     "evidence": "E20"
    },
    {
     "kind": "empirical_quant",
     "evidence": "E31"
    }
   ],
   "provenance": "author_stated",
   "hypothesis": {
    "lab_event": "H36",
    "outcome": "supported"
   },
   "planning": "confirmatory",
   "planned_at": "2026-11-18T12:02:12Z",
   "planned_at_anchor": "git_remote",
   "test": {
    "framework": "formal",
    "supported": "The proof of the hypothesis as worded here, in the study's proof documents (proofs.md, proof-suilin.md), is complete, every step naming what it rests on; the independent checker (an agent of another model than the author's: Claude Opus 5.5, registered as an actor before it checks) records a check of the hypothesis as worded with outcome confirmed; and the deciding run's numerical checks of its formulas meet the tolerances of the design.",
    "not_supported": "A counterexample to the hypothesis as worded, found by the proof, the checker, the supervisor or the deciding run, that a correction of the proof cannot remove.",
    "verdict": "judged",
    "runs": 1,
    "lab_plan": "PL12"
   },
   "checks": [
    {
     "by": "P3",
     "outcome": "confirmed",
     "extent": "full",
     "notes": "Derived from the statement alone (earlier wording: old H3 and H9), then compared with the final proofs.md (paper H3), which agrees; the bound in (b) follows from x - x^3/3 <= tanh x <= x. Concavity is in the absolute log error on [0, inf); 'minimised by the median' is the small-error approximation, as stated."
    },
    {
     "by": "P2",
     "outcome": "confirmed",
     "extent": "full",
     "notes": "The supervisor read the proof of paper H3 in proofs.md (AR1.R2; rendered as AR2.R2) and confirmed it: first \"Confirmed\" (DC6), then, after the last corrections of paper H4, H5 and H9, \"All looks fine now. NExt\"."
    },
    {
     "by": "P5",
     "outcome": "confirmed",
     "extent": "full",
     "notes": "Derived independently. With $z=\\ln y-\\ln f$, $s=|e^z-1|/(e^z+1)=\\tanh(|z|/2)$. (a) $\\tanh(u/2)$ on $u\\ge0$ is 0 at 0, strictly increasing, concave (derivative $\\tfrac12\\operatorname{sech}^2(u/2)$ positive and decreasing), below 1 with limit 1. (b) $x-x^3/3\\le\\tanh x\\le x$ for $x\\ge0$ (the lower bound: $\\frac{d}{dx}(\\tanh x-x+x^3/3)=x^2-\\tanh^2x\\ge0$); with $x=|z|/2$ this gives $0\\le|z|/2-s\\le|z|^3/24$, so $200\\cdot\\mathrm{mean}(s)\\approx100\\cdot\\mathrm{mean}|z|$. (c) Averaging (b): $\\bar s\\le a/2\\le\\bar s+k/24$; with $\\bar s(f^*)\\le\\bar s(m)$, $a(f^*)/2\\le\\bar s(f^*)+k(f^*)/24\\le\\bar s(m)+k(f^*)/24\\le a(m)/2+k(f^*)/24$, which is the stated bound; the log-MAE is minimised at any median of the $\\ln y_t$, i.e. at any median $m$. The proof (proofs.md H3) does exactly this; its only unproved step is the elementary inequality $x-x^3/3\\le\\tanh x$, closed by the derivative argument above. I recomputed the counterexample of step 8 (actuals $1,1,e,e^4,e^4$: sums 2.734 at the median $e$ against 2.390 at 1), which supports \"not in general at it\". Every part holds as worded."
    }
   ],
   "verification": "agent_checked",
   "depends_on": [
    "C1"
   ],
   "relations": [
    {
     "type": "extends",
     "reference": "B12",
     "note": "Tofallis (2015) relates sMAPE to the log accuracy ratio among relative measures; C3 shows the sMAPE term is exactly the hyperbolic tangent of half the absolute log ratio."
    }
   ],
   "verifiability": "open",
   "verifiability_flags": []
  },
  {
   "id": "C4",
   "name": "sMAPE targets a weighted median",
   "summary": "For any distribution, the forecast sMAPE rewards balances the actuals below it and above it. Each actual is weighted by how close it is to the forecast by ratio.",
   "statement": "Let the actual $y$ be positive and take no single value with positive probability. A forecast $f$ has the lowest expected sMAPE only if the actuals below $f$ and those above it carry equal total [[D5:weight]] $w(y,f) = y/(y+f)^2$:\n$$\\mathrm E[w(y,f)\\,\\mathbf 1\\{y<f\\}] = \\mathrm E[w(y,f)\\,\\mathbf 1\\{y>f\\}].$$\n So the [[D6:sMAPE target]] is a median of the actuals weighted by $w$, where the ordinary median gives every actual the same weight. An actual's weight is largest when it equals the forecast, is the same for $y/f = r$ and $1/r$, and falls as $y/f$ moves away from 1 in either direction.",
   "about": {
    "subject": "D6",
    "relation": "is a weighted median with weights",
    "object": "D5",
    "comparator": null,
    "conditions": []
   },
   "type": "formal",
   "role": "principal",
   "novelty": "no_prior_found",
   "support": [
    {
     "kind": "formal_proof",
     "evidence": "E22"
    },
    {
     "kind": "empirical_quant",
     "evidence": "E31"
    }
   ],
   "provenance": "author_stated",
   "hypothesis": {
    "lab_event": "H37",
    "outcome": "supported"
   },
   "planning": "confirmatory",
   "planned_at": "2026-11-18T12:02:12Z",
   "planned_at_anchor": "git_remote",
   "test": {
    "framework": "formal",
    "supported": "The proof of the hypothesis as worded here, in the study's proof documents (proofs.md, proof-suilin.md), is complete, every step naming what it rests on; the independent checker (an agent of another model than the author's: Claude Opus 5.5, registered as an actor before it checks) records a check of the hypothesis as worded with outcome confirmed; and the deciding run's numerical checks of its formulas meet the tolerances of the design.",
    "not_supported": "A counterexample to the hypothesis as worded, found by the proof, the checker, the supervisor or the deciding run, that a correction of the proof cannot remove.",
    "verdict": "judged",
    "runs": 1,
    "lab_plan": "PL12"
   },
   "checks": [
    {
     "by": "P3",
     "outcome": "disputed",
     "extent": "full",
     "notes": "Derived from the statement alone, then compared with the final proofs.md (paper H5). The balance, the weighted median, the properties of the weight and the 'above'/'below' cases hold. The last clause fails: equal weights at the median make it a stationary point, not necessarily the minimum, and proof step 7 does not show it. Counterexample (numerical): ln Y ~ 0.41408 N(-1.48848, 0.2^2) + 0.03 N(0, 0.3^2) + 0.55592 N(1.13795, 1) has median 1 and equal weights there, a flat inflection at the median, and a single local minimum at ln f = -1.36. Fix: drop 'at the median when the two weights are equal', or add 'if the median is a local minimum'.",
     "resolution": "The statement was corrected, and the checker confirmed the corrected wording in a later check listed here."
    },
    {
     "by": "P3",
     "outcome": "confirmed",
     "extent": "full",
     "notes": "Checked the corrected wording against my own derivation and the revised proofs.md (paper H5, new step 8). The clause disputed in CK5 now says that with equal weights the median is a flat point, the target only if it is a local minimum, and possibly a flat inflection; that is what holds, and my counterexample (single local minimum at ln f = -1.36, median at 1) is an instance. Needs Y positive with no atoms, as stated."
    },
    {
     "by": "P3",
     "outcome": "confirmed",
     "extent": "full",
     "notes": "Checked the added sufficient condition against my own derivation and the revised proofs.md (paper H5, steps 8 to 10). The second derivative of the expected term in ln f is g(c) - E[sech^2 tanh]/2, as I derived independently; zero first derivative plus the stated strict inequality gives a strict local minimum at the median, which is the target when there is a single local minimum. The clause needs ln Y to have a density continuous at ln m, as stated; my counterexample sits at equality, consistent with it."
    },
    {
     "by": "P2",
     "outcome": "confirmed",
     "extent": "full",
     "notes": "The supervisor read the proof of paper H5 in proofs.md (AR1.R2; rendered as AR2.R2) and confirmed it: first \"Confirmed\" (DC6), then, after the last corrections of paper H4, H5 and H9, \"All looks fine now. NExt\"."
    },
    {
     "by": "P5",
     "outcome": "confirmed",
     "extent": "full",
     "notes": "Derived independently. For $y\\ne f$, $\\partial_f s(y,f)=2w(y,f)(\\mathbf 1\\{y<f\\}-\\mathbf 1\\{y>f\\})$ with $w=y/(y+f)^2\\le1/(4f)$; since $P(Y=f)=0$ and the difference quotients are bounded by $1/(2(f-\\epsilon))$ near $f$, dominated convergence gives $\\frac{d}{df}\\mathrm E[s(Y,f)]=2\\mathrm E[w(\\mathbf 1\\{Y<f\\}-\\mathbf 1\\{Y>f\\})]$ at every $f>0$, so a minimiser (necessarily $f>0$: $f\\le0$ scores 1 on every positive actual while $\\mathrm E[s]<1$ for $f>0$) satisfies the stated balance. With $r=y/f$, $w=\\frac1f\\,r/(1+r)^2=\\frac1f/(r+2+1/r)$: symmetric under $r\\to1/r$, largest at $r=1$, decreasing as $r$ moves away from 1 on either side. The proof (proofs.md H5) matches step by step; the local uniformity of the dominating bound in $f$ is left implicit but is immediate as above. Its use of H4 (existence) is not needed for the \"only if\" statement but is harmless. Every part holds as worded."
    }
   ],
   "verification": "agent_checked",
   "relations": [
    {
     "type": "explains",
     "note": "C5 found sMAPE's preferred candidate close to the median on other distributions; C4 shows why: the target is a weighted median, on the side of the median that carries more weight (C12).",
     "reference": "B4"
    },
    {
     "type": "addresses",
     "reference": "B12",
     "note": "Tofallis (2015) states that minimising sMAPE does not produce an established measure of location; C4 identifies the target as a median weighted by y/(y+f)^2."
    },
    {
     "type": "qualifies",
     "reference": "B10",
     "note": "Hewamalage et al. (2023) say sMAPE may tend towards selecting a slightly overestimating model; C4 shows the target is a weighted median, which lies on the side of the median that carries more weight, above or below it (C12)."
    },
    {
     "type": "qualifies",
     "target": "C2",
     "id_type": "local",
     "note": "C2 compares two forecasts of one known actual. For an uncertain actual, the forecast sMAPE rewards is the weighted median of C4, not a forecast that is too high."
    }
   ],
   "depends_on": [
    "C9"
   ],
   "verifiability": "open",
   "verifiability_flags": []
  },
  {
   "id": "C5",
   "name": "Symmetric, unimodal logs: the target is the median",
   "summary": "When the logarithm of the actual is symmetric with a single peak, the weights balance at the median, and the median is the forecast sMAPE rewards. Lognormal and log-Laplace actuals are examples.",
   "statement": "Let $\\ln y$ have a density $g$ symmetric about $\\ln m$, $g(\\ln m+v) = g(\\ln m-v)$, and unimodal: $g(\\ln m+v_1) \\ge g(\\ln m+v_2)$ whenever $|v_1| \\le |v_2|$, strictly on a set of positive probability. Then the median $m$ of $y$ is the only [[D6:sMAPE target]].",
   "about": {
    "subject": "D6",
    "relation": "equals",
    "object": "the median of the actuals",
    "comparator": null,
    "conditions": []
   },
   "type": "formal",
   "role": "supporting",
   "novelty": "replication",
   "support": [
    {
     "kind": "formal_proof",
     "evidence": "E23"
    },
    {
     "kind": "empirical_quant",
     "evidence": "E1"
    }
   ],
   "provenance": "author_stated",
   "hypothesis": {
    "lab_event": "H26",
    "outcome": "supported"
   },
   "planning": "confirmatory",
   "planned_at": "2026-11-17T09:21:09Z",
   "planned_at_anchor": "git_remote",
   "test": {
    "framework": "formal",
    "supported": "The proof of the hypothesis, as now worded, in the study's final proof document (proofs.md, AR1.R2) is complete, every step naming what it rests on; the independent checker AC4 has a check of the current wording with outcome confirmed; and the deciding run's numerical checks of its formulas meet the tolerances of PL5: identities on random pairs within 1e-9 (on the scale of sMAPE), quantities computed by quadrature within 1e-4 relative (or absolute, for a residual the proof says is zero), counts of violations 0, and the expansion's error times sigma^3 below the proof's bound.",
    "not_supported": "A counterexample to the statement as now worded, found by the proof, a checker, the supervisor or the deciding run, that a correction of the proof cannot remove.",
    "verdict": "judged",
    "runs": 1,
    "lab_plan": "PL9"
   },
   "checks": [
    {
     "by": "P3",
     "outcome": "confirmed",
     "extent": "full",
     "notes": "Derived from the statement alone (sign of the derivative of expected sMAPE in ln f), then compared with the final proofs.md (paper H6), which reaches it by reflecting about -t/2; both hold under the unimodality as now defined in the statement."
    },
    {
     "by": "P2",
     "outcome": "confirmed",
     "extent": "full",
     "notes": "The supervisor read the proof of paper H6 in proofs.md (AR1.R2; rendered as AR2.R2) and confirmed it: first \"Confirmed\" (DC6), then, after the last corrections of paper H4, H5 and H9, \"All looks fine now. NExt\"."
    }
   ],
   "verification": "agent_checked",
   "depends_on": [
    "C3"
   ],
   "relations": [
    {
     "type": "replicates",
     "reference": "B2",
     "note": "C5 is an elementary case of Anderson's theorem on symmetric unimodal densities, proved here directly for the expected sMAPE."
    }
   ],
   "verifiability": "open",
   "verifiability_flags": []
  },
  {
   "id": "C6",
   "name": "Two separate groups: the median is a local maximum",
   "summary": "When the actuals form two separate groups, the median falls in the gap between them, far from both. So sMAPE scores it worse than forecasts near either group.",
   "statement": "Let $\\ln y$ have two modes, at $\\ln m - a$ and $\\ln m + a$ ($a > 0$), with the actuals concentrated near them: a mixture of a distribution near $\\ln m - a$ and its mirror image about $\\ln m$, with weights $p$ and $1 - p$. (a) With equal weights the median $m$ lies in the gap between the two groups; when the groups are narrow enough it has a higher [[D3:expected sMAPE]] than forecasts near either mode, and it is a local maximum of expected sMAPE; there is a minimiser on each side of the median, and these approach $m\\,e^{-a}$ and $m\\,e^{a}$ as the groups narrow. (b) With unequal weights and narrow groups, the best forecast is near the mode of the larger group.",
   "about": {
    "subject": "D3",
    "relation": "can have",
    "object": "more than one local minimum",
    "comparator": null,
    "conditions": []
   },
   "type": "formal",
   "role": "principal",
   "novelty": "no_prior_found",
   "support": [
    {
     "kind": "formal_proof",
     "evidence": "E24"
    },
    {
     "kind": "empirical_quant",
     "evidence": "E1"
    }
   ],
   "provenance": "author_stated",
   "hypothesis": {
    "lab_event": "H27",
    "outcome": "supported"
   },
   "planning": "confirmatory",
   "planned_at": "2026-11-17T09:21:09Z",
   "planned_at_anchor": "git_remote",
   "test": {
    "framework": "formal",
    "supported": "The proof of the hypothesis, as now worded, in the study's final proof document (proofs.md, AR1.R2) is complete, every step naming what it rests on; the independent checker AC4 has a check of the current wording with outcome confirmed; and the deciding run's numerical checks of its formulas meet the tolerances of PL5: identities on random pairs within 1e-9 (on the scale of sMAPE), quantities computed by quadrature within 1e-4 relative (or absolute, for a residual the proof says is zero), counts of violations 0, and the expansion's error times sigma^3 below the proof's bound.",
    "not_supported": "A counterexample to the statement as now worded, found by the proof, a checker, the supervisor or the deciding run, that a correction of the proof cannot remove.",
    "verdict": "judged",
    "runs": 1,
    "lab_plan": "PL9"
   },
   "checks": [
    {
     "by": "P3",
     "outcome": "confirmed",
     "extent": "full",
     "notes": "Derived from the statement alone (point-mass limit: tanh(a)/2 < tanh(a/2); second derivative at the median g(m) - E[sech^2 tanh]/2 < 0 for narrow groups; limit convex between the modes), then compared with the final proofs.md (paper H7), which agrees. My earlier dispute concerned the words 'exactly when', which the final wording no longer has; the claims are limits as the groups narrow."
    },
    {
     "by": "P2",
     "outcome": "confirmed",
     "extent": "full",
     "notes": "The supervisor read the proof of paper H7 in proofs.md (AR1.R2; rendered as AR2.R2) and confirmed it: first \"Confirmed\" (DC6), then, after the last corrections of paper H4, H5 and H9, \"All looks fine now. NExt\"."
    }
   ],
   "verification": "agent_checked",
   "depends_on": [
    "C3",
    "C9",
    "C12"
   ],
   "verifiability": "open",
   "verifiability_flags": []
  },
  {
   "id": "C7",
   "name": "Zero actuals: when the target is zero",
   "summary": "The worst possible score goes to forecasts of the wrong sign and to zero forecasts. With enough zeros among the actuals, sMAPE prefers the forecast 0. The share of zeros needed is below one half.",
   "statement": "A forecast of the opposite sign to the actual, a zero forecast of a non-zero actual, and a non-zero forecast of a zero actual all give the [[D1:sMAPE term]] its largest value, 1. Let the actual $y$ be 0 with probability $p_0$ and otherwise positive, and $s_+$ the lowest expected $s$ a positive forecast reaches on the positive actuals. The forecast 0 is the [[D6:sMAPE target]] when $p_0 > (1 - s_+)/(2 - s_+)$, and a positive forecast does when the inequality is reversed; the threshold is below $1/2$ whenever $s_+ > 0$.",
   "about": {
    "subject": "D6",
    "relation": "is zero or a positive forecast, by",
    "object": "the share of zero actuals",
    "comparator": null,
    "conditions": []
   },
   "type": "formal",
   "role": "principal",
   "novelty": "extension",
   "support": [
    {
     "kind": "formal_proof",
     "evidence": "E25"
    },
    {
     "kind": "empirical_quant",
     "evidence": "E1"
    }
   ],
   "provenance": "author_stated",
   "hypothesis": {
    "lab_event": "H28",
    "outcome": "supported"
   },
   "planning": "confirmatory",
   "planned_at": "2026-11-17T09:21:09Z",
   "planned_at_anchor": "git_remote",
   "test": {
    "framework": "formal",
    "supported": "The proof of the hypothesis, as now worded, in the study's final proof document (proofs.md, AR1.R2) is complete, every step naming what it rests on; the independent checker AC4 has a check of the current wording with outcome confirmed; and the deciding run's numerical checks of its formulas meet the tolerances of PL5: identities on random pairs within 1e-9 (on the scale of sMAPE), quantities computed by quadrature within 1e-4 relative (or absolute, for a residual the proof says is zero), counts of violations 0, and the expansion's error times sigma^3 below the proof's bound.",
    "not_supported": "A counterexample to the statement as now worded, found by the proof, a checker, the supervisor or the deciding run, that a correction of the proof cannot remove.",
    "verdict": "judged",
    "runs": 1,
    "lab_plan": "PL9"
   },
   "checks": [
    {
     "by": "P3",
     "outcome": "confirmed",
     "extent": "full",
     "notes": "Derived from the statement alone (1 - p0 against p0 + (1 - p0) m+), then compared with the final proofs.md (paper H8), which agrees; the threshold (1 - m+)/(2 - m+) is my earlier (200 - m+)/(400 - m+) rescaled. The best positive forecast may be more than one value; at equality both are targets."
    },
    {
     "by": "P2",
     "outcome": "confirmed",
     "extent": "full",
     "notes": "The supervisor read the proof of paper H8 in proofs.md (AR1.R2; rendered as AR2.R2) and confirmed it: first \"Confirmed\" (DC6), then, after the last corrections of paper H4, H5 and H9, \"All looks fine now. NExt\"."
    }
   ],
   "verification": "agent_checked",
   "relations": [
    {
     "type": "refines",
     "note": "C6 found sMAPE preferring zero on counts with more than half zeros; C7 gives the exact threshold, which is below one half whenever the positive outcomes vary.",
     "reference": "B5"
    },
    {
     "type": "extends",
     "reference": "B13",
     "note": "Kolassa (2026) shows that sMAPE elicits a zero forecast for sufficiently intermittent actuals; C7 gives the share of zeros at which it does, which is below one half."
    }
   ],
   "depends_on": [
    "C9"
   ],
   "verifiability": "open",
   "verifiability_flags": []
  },
  {
   "id": "C8",
   "name": "Where RMSE, MAE and MAPE fail, and what they then target",
   "summary": "RMSE, MAE and MAPE each have an infinite expected value, for every forecast, when a moment they need is infinite. What each then rewards depends on the measure.",
   "statement": "For a positive random actual $y$: (a) [[D10:RMSE]]: $\\mathrm E[(y-f)^2]$ is finite exactly when $\\mathrm E[y^2]$ is; when $\\mathrm E[y^2]$ is infinite but $\\mathrm E[y]$ finite, every difference between two forecasts $f_1$ and $f_2$,\n$$\\mathrm E[(y-f_1)^2 - (y-f_2)^2] = (f_1-f_2)(f_1+f_2-2\\,\\mathrm E[y]),$$\nis finite and the mean is preferred to every other forecast; when $\\mathrm E[y]$ is infinite no forecast is preferred to all others. (b) [[D11:MAE]]: $\\mathrm E|y-f|$ is finite exactly when $\\mathrm E[y]$ is; otherwise differences are finite and a median is still preferred. (c) [[D12:MAPE]]: when $\\mathrm E[1/y]$ is finite, the expected MAPE is finite and minimised by a median of the actuals weighted by $1/y$; when $\\mathrm E[1/y]$ is infinite, every positive forecast has infinite expected MAPE while the forecast 0 has MAPE 100, so MAPE targets 0; when $y$ is 0 with positive probability, MAPE is infinite or undefined for every forecast and targets nothing.",
   "about": {
    "subject": "the forecast RMSE, MAE and MAPE target",
    "relation": "fails with",
    "object": "infinite moments of the actuals",
    "comparator": null,
    "conditions": []
   },
   "type": "formal",
   "role": "secondary",
   "novelty": "no_prior_found",
   "support": [
    {
     "kind": "formal_proof",
     "evidence": "E27"
    },
    {
     "kind": "empirical_quant",
     "evidence": "E1"
    }
   ],
   "provenance": "author_stated",
   "hypothesis": {
    "lab_event": "H29",
    "outcome": "supported"
   },
   "planning": "confirmatory",
   "planned_at": "2026-11-17T09:21:09Z",
   "planned_at_anchor": "git_remote",
   "test": {
    "framework": "formal",
    "supported": "The proof of the hypothesis, as now worded, in the study's final proof document (proofs.md, AR1.R2) is complete, every step naming what it rests on; the independent checker AC4 has a check of the current wording with outcome confirmed; and the deciding run's numerical checks of its formulas meet the tolerances of PL5: identities on random pairs within 1e-9 (on the scale of sMAPE), quantities computed by quadrature within 1e-4 relative (or absolute, for a residual the proof says is zero), counts of violations 0, and the expansion's error times sigma^3 below the proof's bound.",
    "not_supported": "A counterexample to the statement as now worded, found by the proof, a checker, the supervisor or the deciding run, that a correction of the proof cannot remove.",
    "verdict": "judged",
    "runs": 1,
    "lab_plan": "PL9"
   },
   "checks": [
    {
     "by": "P3",
     "outcome": "confirmed",
     "extent": "full",
     "notes": "Derived from the statement alone, then compared with the final proofs.md (paper H10), which agrees in all three parts. When E[1/Y] is infinite, negative forecasts also have infinite expected MAPE, so 0 is the only finite one."
    },
    {
     "by": "P2",
     "outcome": "confirmed",
     "extent": "full",
     "notes": "The supervisor read the proof of paper H10 in proofs.md (AR1.R2; rendered as AR2.R2) and confirmed it: first \"Confirmed\" (DC6), then, after the last corrections of paper H4, H5 and H9, \"All looks fine now. NExt\"."
    }
   ],
   "verification": "agent_checked",
   "verifiability": "open",
   "verifiability_flags": []
  },
  {
   "id": "C9",
   "name": "Expected sMAPE is always finite and has a minimiser",
   "summary": "The expected sMAPE of a forecast is always finite, even when the actuals have a heavy tail that makes the expected RMSE, MAE or MAPE infinite. And there is always a best forecast.",
   "statement": "For actuals and forecasts of any sign, $0 \\le s \\le 1$. So the [[D3:expected sMAPE]] of every forecast is finite, between 0 and 200, whatever the distribution of the actuals and without any moment of it being finite (unlike RMSE, MAE and MAPE, C8). For a positive random actual $y$, at least one forecast has the lowest expected sMAPE.",
   "about": {
    "subject": "D3",
    "relation": "is always finite and has a minimiser",
    "object": "any distribution of the actuals",
    "comparator": null,
    "conditions": []
   },
   "type": "formal",
   "role": "principal",
   "novelty": "extension",
   "support": [
    {
     "kind": "formal_proof",
     "evidence": "E21"
    },
    {
     "kind": "empirical_quant",
     "evidence": "E1"
    }
   ],
   "provenance": "author_stated",
   "hypothesis": {
    "lab_event": "H24",
    "outcome": "supported"
   },
   "planning": "confirmatory",
   "planned_at": "2026-11-17T09:21:09Z",
   "planned_at_anchor": "git_remote",
   "test": {
    "framework": "formal",
    "supported": "The proof of the hypothesis, as now worded, in the study's final proof document (proofs.md, AR1.R2) is complete, every step naming what it rests on; the independent checker AC4 has a check of the current wording with outcome confirmed; and the deciding run's numerical checks of its formulas meet the tolerances of PL5: identities on random pairs within 1e-9 (on the scale of sMAPE), quantities computed by quadrature within 1e-4 relative (or absolute, for a residual the proof says is zero), counts of violations 0, and the expansion's error times sigma^3 below the proof's bound.",
    "not_supported": "A counterexample to the statement as now worded, found by the proof, a checker, the supervisor or the deciding run, that a correction of the proof cannot remove.",
    "verdict": "judged",
    "runs": 1,
    "lab_plan": "PL9"
   },
   "checks": [
    {
     "by": "P3",
     "outcome": "confirmed",
     "extent": "full",
     "notes": "Derived from the statement alone (triangle inequality; bounded convergence gives continuity in f and the limit 1 at both ends, below 1 between), then compared with the final proofs.md (paper H4), which agrees. No condition on the distribution is needed."
    },
    {
     "by": "P2",
     "outcome": "confirmed",
     "extent": "full",
     "notes": "The supervisor read the proof of paper H4 in proofs.md (AR1.R2; rendered as AR2.R2) and confirmed it: first \"Confirmed\" (DC6), then, after the last corrections of paper H4, H5 and H9, \"All looks fine now. NExt\"."
    }
   ],
   "verification": "agent_checked",
   "depends_on": [
    "C3"
   ],
   "relations": [
    {
     "type": "extends",
     "reference": "B10",
     "note": "Hewamalage et al. (2023) note that sMAPE is bounded by 200; C9 adds that its expected value is therefore finite for every distribution and that a best forecast exists."
    }
   ],
   "verifiability": "open",
   "verifiability_flags": []
  },
  {
   "id": "C10",
   "name": "Expected sMAPE flattens as the actuals spread out",
   "summary": "For widely spread actuals, every forecast scores close to the worst value. The differences between forecasts shrink, and the forecast sMAPE rewards moves from the median to the most likely value on the log scale.",
   "statement": "Let $\\ln y = \\mu + \\sigma\\varepsilon$, where $\\varepsilon$ has a fixed density $h$ with bounded continuous first and second derivatives, and write a forecast as $f = e^{\\mu + \\sigma c}$ (its quantile level, the probability that $\\varepsilon \\le c$, does not depend on $\\sigma$). As $\\sigma$ grows, $\\mathrm E[s(y,f)] = 1 - 4\\ln 2\\,h(c)/\\sigma + O(\\sigma^{-3})$ uniformly in $c$. (a) The [[D3:expected sMAPE]] of every forecast tends to its bound 200, and the difference between forecasts at two fixed quantile levels shrinks like $1/\\sigma$. (b) When $h$ has a single highest point $c_0$, the quantile level of the [[D6:sMAPE target]] tends to that of $c_0$, the mode of $\\ln y$, which differs from 1/2 when the mode of $\\varepsilon$ is not its median. (c) As $\\sigma \\to 0$, when $\\varepsilon$ has a unique median and a finite $\\mathrm E|\\varepsilon|^3$, the quantile level of the target tends to 1/2: sMAPE targets the median for narrow distributions and the mode of $\\ln y$ for wide ones.",
   "about": {
    "subject": "D3",
    "relation": "flattens towards its bound as the spread of log actuals grows, with",
    "object": "D6",
    "comparator": null,
    "conditions": []
   },
   "type": "formal",
   "role": "principal",
   "novelty": "no_prior_found",
   "support": [
    {
     "kind": "formal_proof",
     "evidence": "E28"
    },
    {
     "kind": "empirical_quant",
     "evidence": "E1"
    }
   ],
   "provenance": "author_stated",
   "hypothesis": {
    "lab_event": "H30",
    "outcome": "supported"
   },
   "planning": "confirmatory",
   "planned_at": "2026-11-17T09:21:09Z",
   "planned_at_anchor": "git_remote",
   "test": {
    "framework": "formal",
    "supported": "The proof of the hypothesis, as now worded, in the study's final proof document (proofs.md, AR1.R2) is complete, every step naming what it rests on; the independent checker AC4 has a check of the current wording with outcome confirmed; and the deciding run's numerical checks of its formulas meet the tolerances of PL5: identities on random pairs within 1e-9 (on the scale of sMAPE), quantities computed by quadrature within 1e-4 relative (or absolute, for a residual the proof says is zero), counts of violations 0, and the expansion's error times sigma^3 below the proof's bound.",
    "not_supported": "A counterexample to the statement as now worded, found by the proof, a checker, the supervisor or the deciding run, that a correction of the proof cannot remove.",
    "verdict": "judged",
    "runs": 1,
    "lab_plan": "PL9"
   },
   "checks": [
    {
     "by": "P3",
     "outcome": "confirmed",
     "extent": "full",
     "notes": "Derived from the statement alone (integral of 2/(1+e^|v|) is 4 ln 2, odd term vanishes, remainder at most 3 zeta(3) sup|g''|/sigma^3 uniformly in c; checked numerically for the normal), then compared with the final proofs.md (paper H11), which agrees; the constant 4 ln 2 is my earlier 800 ln 2 rescaled. (b) and (c) are claims about the quantile level, which is what holds in general."
    },
    {
     "by": "P2",
     "outcome": "confirmed",
     "extent": "full",
     "notes": "The supervisor read the proof of paper H11 in proofs.md (AR1.R2; rendered as AR2.R2) and confirmed it: first \"Confirmed\" (DC6), then, after the last corrections of paper H4, H5 and H9, \"All looks fine now. NExt\"."
    }
   ],
   "verification": "agent_checked",
   "relations": [
    {
     "type": "supports",
     "note": "C6 recommends against sMAPE for point forecasts; C10 shows that on widely spread actuals every forecast scores close to the worst value and the target moves to the mode of ln y.",
     "reference": "B6"
    },
    {
     "type": "supports",
     "reference": "B13",
     "note": "Kolassa (2026) concludes that sMAPE should not be used; C10 shows that on widely spread actuals every forecast scores close to the worst value."
    }
   ],
   "depends_on": [
    "C3"
   ],
   "verifiability": "open",
   "verifiability_flags": []
  },
  {
   "id": "C11",
   "name": "Actuals of both signs: the target jumps across zero",
   "summary": "When actuals can be negative or positive, the forecast sMAPE rewards is never near zero. It is on the same side of zero as the mean. So a small shift of the mean across zero makes it jump from a negative value to a positive one.",
   "statement": "Let the actuals be $\\mu + \\varepsilon$, where $\\varepsilon$ is symmetric about 0 and its density decreases away from 0 (for example, normal). If $\\mu > 0$, every forecast with the lowest expected sMAPE is positive; if $\\mu < 0$, every one is negative. These forecasts stay away from 0 as $\\mu$ approaches 0. So as $\\mu$ moves from slightly below 0 to slightly above, the [[D6:sMAPE target]] jumps from a negative forecast to a positive one.",
   "about": {
    "subject": "D6",
    "relation": "jumps across zero with",
    "object": "the mean of the actuals",
    "comparator": null,
    "conditions": []
   },
   "type": "formal",
   "role": "secondary",
   "novelty": "no_prior_found",
   "support": [
    {
     "kind": "formal_proof",
     "evidence": "E26"
    },
    {
     "kind": "empirical_quant",
     "evidence": "E11"
    }
   ],
   "provenance": "author_stated",
   "hypothesis": {
    "lab_event": "H31",
    "outcome": "supported"
   },
   "planning": "confirmatory",
   "planned_at": "2026-11-16T17:35:25Z",
   "planned_at_anchor": "git_remote",
   "test": {
    "framework": "formal",
    "supported": "The proof of H31 in the study's proof document is complete, every step naming what it rests on; the independent checker AC4 records a check with outcome confirmed; and in the deciding run, for normal actuals with standard deviation 1, the global minimiser has the sign of the mean for every mean in the grid, and no local minimiser is within 0.1 of zero.",
    "not_supported": "A counterexample to H31 as recorded, found by the proof, the checker, the supervisor or the deciding run, that a correction of the proof cannot remove.",
    "verdict": "judged",
    "runs": 1,
    "lab_plan": "PL7"
   },
   "checks": [
    {
     "by": "P3",
     "outcome": "confirmed",
     "extent": "full",
     "notes": "Derived from the statement alone: for f > 0, s(y,f) - s(y,-f) is odd in y and negative for y > 0, and the density at y exceeds that at -y when mu > 0, so f beats -f; 0 scores 1; near 0 the expected term tends to 1 uniformly in mu near 0 while a fixed forecast stays below 1. Then compared with the final proofs.md (paper H9), which agrees; its steps 4 and 5 leave this uniformity in mu implicit."
    },
    {
     "by": "P2",
     "outcome": "confirmed",
     "extent": "full",
     "notes": "The supervisor read the proof of paper H9 in proofs.md (AR1.R2; rendered as AR2.R2) and confirmed it: first \"Confirmed\" (DC6), then, after the last corrections of paper H4, H5 and H9, \"All looks fine now. NExt\"."
    }
   ],
   "verification": "agent_checked",
   "depends_on": [
    "C7",
    "C9"
   ],
   "verifiability": "open",
   "verifiability_flags": []
  },
  {
   "id": "C12",
   "name": "When sMAPE targets the ordinary median",
   "summary": "The forecast sMAPE rewards lies above the ordinary median when the actuals above it carry more weight, and below it when those below do. When the weights balance and enough actuals lie near the median, the median is rewarded.",
   "statement": "Let the actual $y$ be positive and take no single value with positive probability, let $m$ be its ordinary (unweighted) median, and let its expected sMAPE have a single local minimum. With the forecast at $m$, compare the total [[D5:weight]] $w(y,m) = y/(y+m)^2$ of the actuals above $m$ with that of the actuals below it. (a) If the actuals above $m$ carry more weight, the [[D6:sMAPE target]] lies above $m$; if those below carry more, it lies below $m$. (b) If the two weights are equal, $m$ is a point where expected sMAPE is flat. It is then the target if, in addition,\n$$2g(\\ln m) > \\mathrm E\\Big[\\operatorname{sech}^2\\Big(\\frac{\\ln y - \\ln m}{2}\\Big)\\tanh\\Big(\\frac{|\\ln y - \\ln m|}{2}\\Big)\\Big],$$\nwhere $g$ is the density of $\\ln y$, continuous at $\\ln m$: this makes $m$ a strict local minimum. Without it, $m$ can be a flat point of inflection, with the target elsewhere. (c) When $\\ln y$ has a density symmetric about $\\ln m$ and unimodal, $m$ is the only target; lognormal and log-Laplace actuals are examples.",
   "about": {
    "subject": "D6",
    "relation": "equals, or lies on a stated side of,",
    "object": "the ordinary median of the actuals",
    "comparator": null,
    "conditions": []
   },
   "type": "formal",
   "role": "principal",
   "novelty": "no_prior_found",
   "support": [
    {
     "kind": "formal_proof",
     "evidence": "E29"
    },
    {
     "kind": "empirical_quant",
     "evidence": "E31"
    }
   ],
   "provenance": "author_stated",
   "hypothesis": {
    "lab_event": "H38",
    "outcome": "supported"
   },
   "planning": "confirmatory",
   "planned_at": "2026-11-18T12:02:12Z",
   "planned_at_anchor": "git_remote",
   "test": {
    "framework": "formal",
    "supported": "The proof of the hypothesis as worded here, in the study's proof documents (proofs.md, proof-suilin.md), is complete, every step naming what it rests on; the independent checker (an agent of another model than the author's: Claude Opus 5.5, registered as an actor before it checks) records a check of the hypothesis as worded with outcome confirmed; and the deciding run's numerical checks of its formulas meet the tolerances of the design.",
    "not_supported": "A counterexample to the hypothesis as worded, found by the proof, the checker, the supervisor or the deciding run, that a correction of the proof cannot remove.",
    "verdict": "judged",
    "runs": 1,
    "lab_plan": "PL12"
   },
   "checks": [
    {
     "by": "P3",
     "outcome": "confirmed",
     "extent": "full",
     "notes": "Checked against my own derivations of H5 and H4 and the proofs in proofs.md (paper H6; paper H5 steps 3, 6, 8, 9). (a) holds even with unimodality in the weak sense, since an integrable density cannot be symmetric about two points; (b) holds because a zero first derivative and a positive second derivative in ln f make the median a strict local minimum, which is the target when there is a single local minimum. Caveat: (b) is 'more general' than (a) only where the density in (a) is continuous at ln m."
    },
    {
     "by": "P5",
     "outcome": "confirmed",
     "extent": "full",
     "notes": "Derived independently, on the log scale $\\phi(t)=\\mathrm E\\tanh(|\\ln Y-t|/2)$. (a) $\\phi'(\\ln m)$ has the sign of (weight below) minus (weight above). With a single local minimum, $\\phi$ is decreasing before it and increasing after it: if it were not monotone on one side, a continuous function that is below 1 and tends to 1 at both ends would have an interior local maximum and then a second local minimum. So a negative derivative at $m$ puts the target above $m$, a positive one below. The proof's step 2 says this tersely but correctly. (b) Writing the kernel $-\\operatorname{sign}(v)\\operatorname{sech}^2(v/2)=-\\operatorname{sign}(v)+\\operatorname{sign}(v)\\tanh^2(v/2)$, the first part gives $\\tfrac12(2G(t)-1)$, which is differentiable at $t$ with derivative $g(t)$ when $g$ is continuous at $t$, and the second has a bounded continuous derivative. So $\\phi''(\\ln m)=g(\\ln m)-\\tfrac12\\mathrm E[\\operatorname{sech}^2(x/2)\\tanh(|x|/2)]$ exists under continuity of $g$ at $\\ln m$ alone. With $\\phi'(\\ln m)=0$ and $\\phi''(\\ln m)>0$, $m$ is a strict local minimum, hence the target. This agrees with H12 step 4, which interchanges derivative and expectation without justification; the decomposition above closes that gap. The existence claim \"Without it, $m$ can be a flat point of inflection, with the target elsewhere\" is supported in the proof only by a reference to an example built by an earlier checker that is not given. I built one myself: $\\ln Y-\\ln m\\sim0.4061\\,N(-1.5,0.3^2)+0.0548\\,N(0,0.5^2)+0.5390\\,N(1.1587,1)$ has median 0, $\\phi'(0)=\\phi''(0)=0$, $\\phi'\\ge0$ on both sides of 0 (a flat inflection) and a single local minimum near $t=-1.25$ (numerical grid). Including such an explicit example in the proof would make step 6 self-contained. (Under the single-minimum hypothesis a failing inequality can only mean equality, $\\phi''=0$; a strict local maximum is impossible, so \"flat point of inflection\" is the right description.) (c) holds via H6. Note that H6 assumes unimodality \"strictly on a set of positive probability\", which H38 (c) does not state; the conclusion holds for every symmetric, weakly unimodal density anyway: $P(|X-t|\\le c)<P(|X|\\le c)$ for $t\\ne0$ on a set of $c$ of positive measure (otherwise $g$ would be constant on $(0,\\infty)$), and $\\tanh$ is strictly increasing, so no correction to the statement is needed. Lognormal and log-Laplace qualify. Every part holds as worded."
    }
   ],
   "verification": "agent_checked",
   "relations": [
    {
     "type": "explains",
     "note": "C5 found by simulation that sMAPE prefers the median on lognormal and log-Laplace outcomes; C12 proves it for every symmetric unimodal ln y.",
     "reference": "B4"
    }
   ],
   "depends_on": [
    "C3",
    "C4",
    "C5",
    "C9"
   ],
   "verifiability": "open",
   "verifiability_flags": []
  },
  {
   "id": "C13",
   "name": "The sMAPE target on 34 distributions",
   "summary": "Computed on 34 distributions, the forecast sMAPE rewards is the median on lognormal data. It moves away from the median on skewed data, sits near a mode on two-group data, and becomes zero once enough of the data are zeros, always fewer than half.",
   "statement": "Computed from the distributions themselves, the [[D6:sMAPE target]] on the 34 outcome distributions of plan PL1 is the median on lognormal outcomes, lies below it when $\\ln y$ is skewed right and above it when skewed left (quantile levels from 0.439 to 0.556 on the unimodal positive distributions), is two forecasts, one near each mode, on equal bimodal outcomes and the larger mode's on unequal ones, lies away from zero on the side of the mean on normal outcomes straddling zero, and switches to 0 at shares of zero outcomes from 0.343 to 0.468, all below one half. On test sets the forecast sMAPE ranks first spreads widely when they are small: on lognormal outcomes with $\\sigma = 1$, its quantile level runs from 0.16 to 0.82 (5th to 95th percentile over test sets) with 6 outcomes and from 0.47 to 0.53 with 1000.",
   "type": "descriptive",
   "role": "secondary",
   "novelty": "no_prior_found",
   "support": [
    {
     "kind": "empirical_quant",
     "evidence": "E13",
     "results": [
      "E13.T1.r6",
      "E13.T1.r14",
      "E13.T1.r1",
      "E13.T3.r20",
      "E13.T3.r6",
      "E13.T1.r29",
      "E13.T3.r27"
     ]
    }
   ],
   "provenance": "author_stated",
   "planning": "planned_estimate",
   "planned_at": "2026-11-16T13:46:18Z",
   "planned_at_anchor": "git_remote",
   "estimate": {
    "estimands": {
     "smape_minima": {
      "variable": "derived local minimum of expected sMAPE",
      "table": true
     },
     "smape_minima_level": {
      "variable": "quantile level of a derived local minimum of expected sMAPE",
      "table": true
     },
     "smape_minima_score": {
      "variable": "expected sMAPE at a derived local minimum",
      "table": true
     },
     "pull_difference": {
      "variable": "pull from above minus pull from below at the median",
      "table": true
     },
     "one_step": {
      "variable": "one-step quantile level of the sMAPE target",
      "table": true
     },
     "zero_threshold": {
      "variable": "zero threshold of sMAPE",
      "table": true
     },
     "log_mode_level": {
      "variable": "quantile level of the log-mode",
      "table": true
     },
     "mape_target": {
      "variable": "derived MAPE target",
      "table": true
     },
     "infinite": {
      "variable": "candidates whose expected score is infinite",
      "table": true
     },
     "preferred_value": {
      "variable": "value of the preferred candidate",
      "table": true
     },
     "preferred_level": {
      "variable": "quantile level of the preferred candidate",
      "table": true
     },
     "chosen_level": {
      "variable": "mean quantile level of the chosen candidates",
      "table": true
     },
     "chosen_level_percentiles": {
      "variable": "percentile of the chosen candidates' quantile levels",
      "table": true
     },
     "preferred_share": {
      "variable": "share of test sets choosing the preferred candidate",
      "table": true
     },
     "all_infinite": {
      "variable": "test sets in which every candidate's score is infinite",
      "table": true
     }
    },
    "runs": 1,
    "lab_plan": "PL1",
    "flags": [
     "code_changed_after_plan"
    ]
   },
   "limitations": [
    "L1"
   ],
   "about": {
    "subject": "D6",
    "relation": "was derived for",
    "object": "34 outcome distributions",
    "comparator": null,
    "conditions": []
   },
   "scope": {
    "data_type": "outcome distributions given exactly (no data)",
    "conditions": [
     "34 distributions: inverse Gaussian, log-skewed, lognormal, bimodal, normal straddling zero, zero-inflated lognormal"
    ]
   },
   "relations": [
    {
     "type": "supports",
     "target": "C4",
     "id_type": "local",
     "note": "Computed on 34 distributions, the target lies on the side of the median that carries more weight, as C4 and C12 prove.",
     "results": [
      "E13.T1.r6",
      "E13.T1.r15",
      "E13.T1.r18",
      "E13.T3.r6",
      "E13.T3.r15",
      "E13.T3.r18"
     ]
    },
    {
     "type": "supports",
     "target": "C5",
     "id_type": "local",
     "note": "Computed on lognormal outcomes, the target is the median, as C5 proves.",
     "results": [
      "E13.T1.r13",
      "E13.T1.r20",
      "E13.T3.r13",
      "E13.T3.r20"
     ]
    },
    {
     "type": "supports",
     "target": "C6",
     "id_type": "local",
     "note": "Computed on bimodal outcomes, the target is two forecasts, one near each mode, when the modes are equal, and the larger mode's when they are not, as C6 proves.",
     "results": [
      "E13.T1.r21",
      "E13.T1.r22",
      "E13.T1.r23",
      "E13.T2.r1",
      "E13.T3.r21",
      "E13.T3.r22",
      "E13.T3.r23"
     ]
    },
    {
     "type": "supports",
     "target": "C7",
     "id_type": "local",
     "note": "Computed on lognormal outcomes with zeros, the target switches to 0 at a share of zeros below one half, as C7 proves.",
     "results": [
      "E13.T1.r27",
      "E13.T1.r28",
      "E13.T1.r29",
      "E13.T1.r30",
      "E13.T1.r31",
      "E13.T1.r32",
      "E13.T1.r33",
      "E13.T1.r34",
      "E13.T3.r27",
      "E13.T3.r28",
      "E13.T3.r29",
      "E13.T3.r30",
      "E13.T3.r31",
      "E13.T3.r32",
      "E13.T3.r33",
      "E13.T3.r34"
     ]
    },
    {
     "type": "supports",
     "target": "C11",
     "id_type": "local",
     "note": "Computed on normal outcomes straddling zero, the target lies away from zero, on the side of the mean, as C11 proves.",
     "results": [
      "E13.T1.r24",
      "E13.T1.r25",
      "E13.T1.r26",
      "E13.T2.r2",
      "E13.T2.r3",
      "E13.T2.r4",
      "E13.T3.r24",
      "E13.T3.r25",
      "E13.T3.r26"
     ]
    }
   ],
   "verifiability": "open",
   "verifiability_flags": []
  },
  {
   "id": "C14",
   "name": "Flattening, computed",
   "summary": "As the data spread out, the best possible sMAPE rises towards its worst value, the forecast it rewards moves from the median towards the most likely value on the log scale, and sMAPE ranks its own best forecast first less often than MAE does.",
   "statement": "As the spread $\\sigma$ of $\\ln y$ grows from 0.1 to 10, the [[D3:expected sMAPE]] at the [[D6:sMAPE target]] rises towards its bound 200 (from 7.97 to 178.2 on lognormal outcomes), the target's quantile level moves from the median to that of the mode of $\\ln y$ (0.372 against the mode's 0.368 at $\\sigma = 10$ on log-Gumbel outcomes skewed right), and the [[D20:ranking agreement]] of sMAPE, for its target against a forecast half a $\\sigma$ away, is lower than that of [[D11:MAE]]: at $\\sigma = 4$ with test sets of 12 outcomes it ranks its target first in 0.61 of the test sets, MAE in 0.82.",
   "type": "descriptive",
   "role": "secondary",
   "novelty": "no_prior_found",
   "support": [
    {
     "kind": "empirical_quant",
     "evidence": "E14",
     "results": [
      "E14.T1.r1",
      "E14.T1.r28",
      "E14.T1.r29",
      "E14.T1.r2",
      "E14.T1.r24",
      "E14.T2.r5"
     ]
    }
   ],
   "provenance": "author_stated",
   "planning": "planned_estimate",
   "planned_at": "2026-11-16T14:20:48Z",
   "planned_at_anchor": "git_remote",
   "estimate": {
    "estimands": {
     "target_level": {
      "variable": "quantile level of the sMAPE target",
      "table": true
     },
     "log_mode_level": {
      "variable": "quantile level of the log-mode",
      "table": true
     },
     "at_target": {
      "variable": "expected sMAPE at the target",
      "table": true
     },
     "approximation": {
      "variable": "large-spread approximation of expected sMAPE at the target",
      "table": true
     },
     "minima": {
      "variable": "local minima of expected sMAPE",
      "table": true
     },
     "relative_excess": {
      "variable": "excess expected score over the target, relative",
      "table": true
     },
     "absolute_excess": {
      "variable": "excess expected sMAPE over the target",
      "table": true
     },
     "agreement": {
      "variable": "D20",
      "table": true
     },
     "minimiser_level": {
      "variable": "mean quantile level of the sMAPE minimiser of a test set",
      "table": true
     },
     "minimiser_percentiles": {
      "variable": "percentile of the quantile level of the sMAPE minimiser of a test set",
      "table": true
     }
    },
    "runs": 1,
    "lab_plan": "PL4"
   },
   "about": {
    "subject": "D3",
    "relation": "flattens as the spread grows, by",
    "object": "D20",
    "comparator": "D11",
    "conditions": []
   },
   "scope": {
    "data_type": "outcome distributions given exactly, and test sets drawn from them",
    "conditions": [
     "lognormal and log-Gumbel outcomes, spread of ln y from 0.1 to 10",
     "test sets of 12 to 1000 outcomes"
    ]
   },
   "relations": [
    {
     "type": "supports",
     "target": "C10",
     "id_type": "local",
     "note": "Computed for spreads of ln y from 0.1 to 10, expected sMAPE flattens and the target moves to the mode of ln y, as C10 proves for large spreads.",
     "results": [
      "E14.T1"
     ]
    }
   ],
   "verifiability": "open",
   "verifiability_flags": []
  },
  {
   "id": "C15",
   "name": "sMAPE finds the median less precisely than the sample median",
   "summary": "Even where sMAPE and MAE both reward the median, the forecast that minimises sMAPE on a test set varies more than the sample median. The loss of precision grows as the data spread out.",
   "statement": "Where sMAPE and MAE both target the median ([[D26:log-symmetric outcomes]]), the [[D8:sMAPE minimiser of a test set]] varies more than the sample median for every spread and test-set size studied: its [[D9:relative efficiency]] is below 1 and falls as the actuals spread out, at $n = 10000$ from 0.97 to 0.03 on lognormal actuals as $\\sigma$ grows from 0.25 to 6, and from 1.00 to 0.34 on log-Laplace actuals as $b$ grows from 0.25 to 6, close to the large-sample values.",
   "type": "descriptive",
   "role": "principal",
   "novelty": "no_prior_found",
   "support": [
    {
     "kind": "empirical_quant",
     "evidence": "E15",
     "results": [
      "E15.T1.r1",
      "E15.T1.r8",
      "E15.T1.r9",
      "E15.T1.r16"
     ]
    }
   ],
   "provenance": "author_stated",
   "planning": "planned_estimate",
   "planned_at": "2026-11-16T13:46:20Z",
   "planned_at_anchor": "git_remote",
   "estimate": {
    "estimands": {
     "re": {
      "variable": "D9",
      "table": true
     },
     "asymptotic_re": {
      "variable": "asymptotic relative efficiency",
      "table": true
     },
     "var_median": {
      "variable": "variance of ln(sample median)",
      "table": true
     },
     "var_smape": {
      "variable": "variance of ln(sMAPE minimiser)",
      "table": true
     },
     "mse_ratio": {
      "variable": "ratio of mean squared errors about the true median",
      "table": true
     },
     "bias": {
      "variable": "mean of ln(sMAPE minimiser) minus ln(median)",
      "table": true
     }
    },
    "runs": 1,
    "lab_plan": "PL2",
    "flags": [
     "code_changed_after_plan"
    ]
   },
   "limitations": [
    "L2"
   ],
   "about": {
    "subject": "D8",
    "relation": "is less efficient than",
    "object": "D9",
    "comparator": "the sample median",
    "conditions": [
     "D26"
    ]
   },
   "scope": {
    "data_type": "test sets of independent draws from known distributions",
    "conditions": [
     "lognormal and log-Laplace outcomes, spread 0.25 to 6",
     "test sets of 6 to 10000 outcomes"
    ]
   },
   "relations": [
    {
     "type": "supports",
     "note": "C11 found sMAPE's choices in short test sets spreading more than MAE's; C15 measures the loss of precision for test sets of 6 to 10000 outcomes.",
     "reference": "B7"
    },
    {
     "type": "supports",
     "note": "C6 recommends against sMAPE for point forecasts; C15 shows that even where it targets the median it estimates it less precisely than the sample median.",
     "reference": "B6"
    },
    {
     "type": "supports",
     "reference": "B13",
     "note": "Kolassa (2026) concludes that sMAPE should not be used; C15 shows that even where it targets the median it estimates it less precisely than the sample median."
    },
    {
     "type": "qualifies",
     "target": "C12",
     "id_type": "local",
     "note": "Where C12 makes the median the target, the forecast that minimises sMAPE on a test set finds it less precisely than the sample median.",
     "results": [
      "E15.T1"
     ]
    }
   ],
   "verifiability": "open",
   "verifiability_flags": []
  },
  {
   "id": "C16",
   "name": "The Suilin target beside the sMAPE target",
   "summary": "Computed on actual distributions, the forecast the Suilin loss rewards is almost the same as sMAPE's away from zero and on counts. It differs near zero, where the Suilin loss needs a larger share of zeros before it rewards a zero forecast.",
   "statement": "Computed from the distributions themselves, the [[D29:Suilin target]] differs little from the [[D6:sMAPE target]] away from zero and on counts, and differs near zero: it is within 0.3% of the sMAPE target on the six distributions with median about 100, and the same as the sMAPE target on every [[D30:Poisson counts]] distribution studied; on lognormal actuals with median 1 and $\\sigma = 1$ it is 0.985 instead of 1, and the share of zeros at which it switches to the forecast 0 is 0.470 instead of sMAPE's 0.394, while at median 100 the two shares are 0.3937 and 0.3940.",
   "type": "descriptive",
   "role": "secondary",
   "novelty": "no_prior_found",
   "support": [
    {
     "kind": "empirical_quant",
     "evidence": "E16",
     "results": [
      "E16.T1.r7",
      "E16.T2.r1",
      "E16.T2.r2"
     ]
    }
   ],
   "provenance": "author_stated",
   "planning": "planned_estimate",
   "planned_at": "2026-11-17T11:22:08Z",
   "planned_at_anchor": "git_remote",
   "estimate": {
    "estimands": {
     "target": {
      "variable": "target",
      "table": true
     },
     "level": {
      "variable": "quantile level of the target",
      "table": true
     },
     "difference": {
      "variable": "relative difference of the Suilin target from the sMAPE target",
      "table": true
     },
     "zero_threshold": {
      "variable": "zero threshold",
      "table": true
     }
    },
    "runs": 1,
    "lab_plan": "PL11"
   },
   "about": {
    "subject": "D29",
    "relation": "differs from",
    "object": "D6",
    "comparator": null,
    "conditions": [
     "D30"
    ]
   },
   "scope": {
    "data_type": "outcome distributions given exactly (no data)",
    "conditions": [
     "positive distributions at median about 100, lognormal with zeros at medians 1 and 100, Poisson counts with means 0.3 to 10"
    ]
   },
   "relations": [
    {
     "type": "supports",
     "target": "C17",
     "id_type": "local",
     "note": "Computed on actual distributions, the Suilin target stays at the sMAPE target away from zero and moves near zero, as C17's two cases imply.",
     "results": [
      "E16.T1",
      "E16.T2"
     ]
    }
   ],
   "verifiability": "open",
   "verifiability_flags": []
  },
  {
   "id": "C17",
   "name": "The Suilin loss behaves like sMAPE except near zero",
   "summary": "The Suilin loss behaves like sMAPE except near zero, where it becomes an absolute error.",
   "statement": "The [[D28:Suilin term]] has the properties of the [[D1:sMAPE term]] except near zero. (a) Where $|y|+|f| \\ge 0.5$,\n$$\\tilde s(y,f) = s(y,f)\\,\\frac{|y|+|f|}{|y|+|f|+\\delta}:$$\nthe sMAPE term shrunk by a factor between $5/6$ and 1 that tends to 1 as $|y|+|f|$ grows. So for actuals and forecasts large compared with $\\delta$, $\\tilde s$ is close to $s$, though no longer exactly a function of the ratio $y/f$; on counts, where every actual is 0 or at least 1, this case covers every pair except a zero actual with $|f| < 0.5$. (b) Where $|y|+|f| < 0.5$, $\\tilde s(y,f) = |y-f|/(0.5+\\delta)$, an absolute error: a zero forecast of a zero actual scores 0, and a forecast $f$ with $|f| < 0.5$ of a zero actual scores $|f|/(0.5+\\delta)$, where the sMAPE term gives every non-zero forecast of a zero actual its largest value, 1. (c) $0 \\le \\tilde s < 1$ everywhere, so the expected Suilin loss of every forecast is finite.",
   "type": "formal",
   "role": "principal",
   "novelty": "extension",
   "support": [
    {
     "kind": "formal_proof",
     "evidence": "E30"
    },
    {
     "kind": "empirical_quant",
     "evidence": "E31"
    }
   ],
   "provenance": "author_stated",
   "hypothesis": {
    "lab_event": "H41",
    "outcome": "supported"
   },
   "planning": "confirmatory",
   "planned_at": "2026-11-18T12:02:12Z",
   "planned_at_anchor": "git_remote",
   "test": {
    "framework": "formal",
    "supported": "The proof of the hypothesis as worded here, in the study's proof documents (proofs.md, proof-suilin.md), is complete, every step naming what it rests on; the independent checker (an agent of another model than the author's: Claude Opus 5.5, registered as an actor before it checks) records a check of the hypothesis as worded with outcome confirmed; and the deciding run's numerical checks of its formulas meet the tolerances of the design.",
    "not_supported": "A counterexample to the hypothesis as worded, found by the proof, the checker, the supervisor or the deciding run, that a correction of the proof cannot remove.",
    "verdict": "judged",
    "runs": 1,
    "lab_plan": "PL12"
   },
   "checks": [
    {
     "by": "P3",
     "outcome": "confirmed",
     "extent": "full",
     "notes": "Derived from the statement alone, then compared with proof-msmape.md. Write t = |y|+|f|. The denominator max(t+eps, 0.5+eps) is t+eps exactly when t >= 0.5, and the two branches agree at t = 0.5 (both 0.6), so m is continuous there. (a) For t >= 0.5, t > 0, so |y-f| = s t and m = s t/(t+eps); the factor t/(t+eps) = 1 - eps/(t+eps) is increasing, equals 5/6 at t = 0.5, is strictly below 1 and tends to 1, so it lies in [5/6, 1): the bound 5/6 is attained, 1 is not. s - m = s eps/(t+eps) <= eps/(t+eps), so m is close to s when t is large compared with eps. m is not a function of y/f: (1,2) and (10,20) have the same ratio and the same s = 1/3, but m = 1/3.1 and 10/30.1. On counts, y >= 1 gives t >= 1, and y = 0 gives t = |f|, so only a zero actual with |f| < 0.5 falls outside this case; this holds for forecasts of any sign and any real value. (b) For t < 0.5, m = |y-f|/0.6; m(0,0) = 0; m(0,f) = |f|/0.6 for |f| < 0.5; s(0,f) = |f|/|f| = 1 for every f != 0, and 1 is the largest value of s because |y-f| <= |y|+|f|. (c) m >= 0, and |y-f| <= t < t+eps <= denominator gives m < 1 strictly in both branches, with supremum 1 not attained (opposite signs, t large); a bounded term has a finite expectation under every joint distribution of actuals and forecasts, so expected msMAPE lies in [0, 200). Checked with exact rational arithmetic on 400000 random pairs of mixed signs and scales, including zeros and the boundary t = 0.5: no violation. Every step of proof-msmape.md (1 to 6) holds. Three clauses of the recorded statement are true but have no step in the proof, whose restated Statement omits them: that m is close to s for values large compared with eps (needs the bound s - m <= eps/(t+eps)), that m is no longer exactly a function of y/f (needs one counterexample such as the one above), and that 1 is the largest value of the sMAPE term (follows from the triangle inequality used in step 6 but is not said). The opening sentence, that m has the properties of the sMAPE term except near zero, is a summary and holds only as parts (a) to (c) qualify it: exact dependence on the ratio y/f alone is lost at every scale, not only near zero, and m never reaches the value 1 that s reaches for opposite signs or a zero actual."
    },
    {
     "by": "P5",
     "outcome": "confirmed",
     "extent": "full",
     "notes": "Derived independently. With $\\ell=|y|+|f|$, the denominator $\\max(\\ell+\\delta,0.5+\\delta)$ is $\\ell+\\delta$ exactly when $\\ell\\ge0.5$. (a) Then $\\ell>0$, $|y-f|=s\\ell$, $\\tilde s=s\\,\\ell/(\\ell+\\delta)$, and $\\ell/(\\ell+\\delta)$ is increasing from $0.5/0.6=5/6$ towards 1, never reaching 1. It depends on scale: $(1,2)$ and $(10,20)$ have equal $s$ but different $\\tilde s$. On counts, $y\\ge1$ gives $\\ell\\ge1$, and $y=0$ gives $\\ell=|f|$, so only a zero actual with $|f|<0.5$ falls outside. (b) For $\\ell<0.5$, $\\tilde s=|y-f|/0.6$; $(0,0)$ scores 0, $(0,f)$ scores $|f|/0.6$, and $s(0,f)=1=\\max s$ for $f\\ne0$. (c) In case (a), $\\tilde s\\le s\\cdot\\ell/(\\ell+\\delta)<1$ since $s\\le1$. In case (b), $\\tilde s\\le\\ell/0.6<5/6$. And $\\tilde s\\ge0$, so every expectation is finite. The proof (proof-suilin.md, first proof) matches each step. Every part holds as worded."
    }
   ],
   "verification": "agent_checked",
   "relations": [
    {
     "type": "extends",
     "reference": "B14",
     "note": "Hewamalage, Bergmeir and Bandara (2021) evaluate forecasts with the Suilin loss, Suilin's modification of sMAPE for series with values near zero; C17 shows that it keeps sMAPE's behaviour away from zero and becomes an absolute error near zero."
    }
   ],
   "about": {
    "subject": "D28",
    "relation": "behaves like, except near zero,",
    "object": "D1",
    "comparator": null,
    "conditions": []
   },
   "verifiability": "open",
   "verifiability_flags": []
  },
  {
   "id": "C18",
   "name": "sMAPE loses its power to tell forecasts apart as the actuals spread out",
   "summary": "For widely spread actuals a test set tells two forecasts apart less and less well under sMAPE: the signal-to-noise ratio of their score difference falls like one over the square root of the spread. The absolute log error and MAE keep their power.",
   "statement": "Let $\\ln y = \\mu + \\sigma\\varepsilon$, where $\\varepsilon$ has a fixed density $h$ with bounded continuous first and second derivatives and distribution function $H$, and compare two forecasts at fixed quantile levels, $f_1 = e^{\\mu+\\sigma c_1}$ and $f_2 = e^{\\mu+\\sigma c_2}$ with $c_1 < c_2$, by their [[D32:signal-to-noise ratio of a comparison]]. (a) For the [[D1:sMAPE term]], as $\\sigma$ grows,\n$$\\mathrm E[D] = \\frac{4\\ln 2\\,(h(c_1) - h(c_2))}{\\sigma} + O(\\sigma^{-3}), \\qquad \\operatorname{Var}(D) = \\frac{4(2\\ln 2 - 1)(h(c_1)+h(c_2))}{\\sigma} + O(\\sigma^{-2}).$$\nSo when $h(c_1) \\ne h(c_2)$ the ratio is\n$$\\frac{2\\ln 2}{\\sqrt{2\\ln 2 - 1}}\\,\\frac{|h(c_1)-h(c_2)|}{\\sqrt{h(c_1)+h(c_2)}}\\,\\sigma^{-1/2}\\,(1 + O(1/\\sigma)):$$\nit falls to zero like $\\sigma^{-1/2}$, and the size of the test set that keeps the mean difference a given number of standard errors from zero grows in proportion to $\\sigma$. (b) For the absolute log error $|\\ln y - \\ln f|$ ([[D19:log-MAE]]), $D = \\sigma(|\\varepsilon - c_2| - |\\varepsilon - c_1|)$, so the ratio, wherever $\\mathrm{sd}(D) > 0$, does not depend on $\\sigma$. (c) For the absolute error $|y - f|$ ([[D11:MAE]]), as $\\sigma$ grows the ratio tends to\n$$\\frac{|2H(c_2) - 1|}{2\\sqrt{H(c_2)(1 - H(c_2))}}$$\nwhen $0 < H(c_2) < 1$; the limit is positive unless the higher forecast is the median. (d) For the squared error $(y-f)^2$ ([[D10:RMSE]]), when $\\mathrm E[y^2]$ is finite the ratio is $|2\\,\\mathrm E[y] - f_1 - f_2|/(2\\,\\mathrm{sd}(y))$; for lognormal actuals, $e^{\\sigma^2/2}$ times the ratio tends to 1 as $\\sigma$ grows. For widely spread actuals, then: the absolute log error keeps its power to tell two forecasts apart wherever its ratio is defined. The absolute error keeps it when $0 < H(c_2) < 1$ and the higher forecast is not the median. sMAPE loses it at the rate $\\sigma^{-1/2}$ when $h(c_1) \\ne h(c_2)$. It loses it faster, with a ratio of at most order $\\sigma^{-5/2}$, when $h(c_1) = h(c_2) > 0$. The squared error on lognormal actuals loses it faster than any power of $\\sigma$.",
   "about": {
    "subject": "D32",
    "relation": "falls like the inverse square root of the spread of log actuals for",
    "object": "D1",
    "comparator": "D19",
    "conditions": []
   },
   "type": "formal",
   "role": "principal",
   "novelty": "no_prior_found",
   "support": [
    {
     "kind": "formal_proof",
     "evidence": "E35"
    },
    {
     "kind": "empirical_quant",
     "evidence": "E32"
    }
   ],
   "provenance": "author_stated",
   "hypothesis": {
    "lab_event": "H46",
    "outcome": "supported"
   },
   "planning": "confirmatory",
   "planned_at": "2026-11-18T12:29:29Z",
   "planned_at_anchor": "git_remote",
   "test": {
    "framework": "formal",
    "supported": "The proof of the hypothesis as worded here (proofs.md, section H13, revised to it) is complete; the independent checker AC6 (Claude Opus 5.5) records a check of this wording with outcome confirmed; and the deciding run's checks meet the tolerances of PL13.",
    "not_supported": "A counterexample to the hypothesis as worded here, found by the proof, the checker, the supervisor or the deciding run, that a correction of the proof cannot remove.",
    "verdict": "judged",
    "runs": 1,
    "lab_plan": "PL17"
   },
   "checks": [
    {
     "by": "P5",
     "outcome": "disputed",
     "extent": "full",
     "notes": "Derived independently. (a) With $q(v)=1-\\tanh(|v|/2)=2/(1+e^{|v|})$, $\\mathrm E[q(\\sigma(\\varepsilon-c))]=\\frac1\\sigma\\int q(u)h(c+u/\\sigma)du=4\\ln2\\,h(c)/\\sigma+O(\\sigma^{-3})$ (the $h'$ term vanishes by symmetry; the remainder is bounded by $\\tfrac12\\sup|h''|\\int u^2q/\\sigma^3$). Likewise $\\mathrm E[q^2]=4(2\\ln2-1)h(c)/\\sigma+O(\\sigma^{-3})$ with $\\int q^2=8(\\ln2-\\tfrac12)$; the cross term is at most $4e^{-\\sigma(c_2-c_1)}$; $(\\mathrm E D)^2=O(\\sigma^{-2})$. So both expansions and their remainder orders hold, and the ratio is as stated with relative error $O(1/\\sigma)$ (from the variance), with $h(c_1)+h(c_2)>0$ when $h(c_1)\\ne h(c_2)$. A quadrature check (normal $h$, $c_1=-0.3$, $c_2=0.8$, $\\sigma$ up to 160) agrees. (b) is immediate; $D$ is bounded by $\\sigma(c_2-c_1)$. (c) $D/f_2\\to\\mathbf 1\\{\\varepsilon<c_2\\}-\\mathbf 1\\{\\varepsilon>c_2\\}$ a.s. with $|D/f_2|\\le1$, so $\\mathrm E[D/f_2]\\to2H(c_2)-1$ and $\\operatorname{Var}(D/f_2)\\to4H(c_2)(1-H(c_2))$. The stated limit therefore holds only when $0<H(c_2)<1$. The statement does not give this condition, and the proof (H13 step 8) adds it. The hypotheses allow it to fail: $h$ may have bounded support, e.g. $h\\propto(1-x^2)^3$ on $[-1,1]$, which has bounded continuous $h'$ and $h''$. If $H(c_2)=1$ the formula is $1/0$ (when $c_1$ is inside the support the ratio tends to infinity; when it is not, $D$ is constant and the ratio is undefined for every $\\sigma$). If $H(c_2)=0$, $D$ is constant and the ratio is undefined for every $\\sigma$. (d) $D=(f_2-f_1)(f_1+f_2-2Y)$ gives the ratio exactly; for lognormal actuals $e^{\\sigma^2/2}$ times the ratio is $|1-(f_1+f_2)/(2\\mathrm EY)|\\,e^{\\sigma^2/2}/\\sqrt{e^{\\sigma^2}-1}\\to1$. Every computation in the proof is correct. The dispute concerns only the missing condition in (c). Suggested wording: \"(c) ... as $\\sigma$ grows the ratio tends to $|2H(c_2)-1|/(2\\sqrt{H(c_2)(1-H(c_2))})$ when $0<H(c_2)<1$, which is positive unless the higher forecast is the median.\" Optionally, also in (b): \"wherever $\\mathrm{sd}(D)>0$\" (it is 0 when $\\varepsilon\\le c_1$ a.s. or $\\varepsilon\\ge c_2$ a.s.). The closing sentence is also loose: sMAPE loses power at the rate $\\sigma^{-1/2}$ only when $h(c_1)\\ne h(c_2)$ (otherwise faster), and the absolute error keeps it only when $H(c_2)\\ne1/2$. It would be safer as \"for widely spread actuals, comparisons of forecasts with $h(c_1)\\ne h(c_2)$ ...\".",
     "resolution": "The statement was corrected, and the checker confirmed the corrected wording in a later check listed here."
    },
    {
     "by": "P5",
     "outcome": "disputed",
     "extent": "full",
     "notes": "Re-checked every part as now worded, against the revised H13. (a) is unchanged from H39 and correct: $\\mathrm E[q]=4\\ln2\\,h(c)/\\sigma+O(\\sigma^{-3})$ and $\\mathrm E[q^2]=4(2\\ln2-1)h(c)/\\sigma+O(\\sigma^{-3})$; the cross term is at most $4e^{-\\sigma(c_2-c_1)}$; the ratio is as stated with relative error $O(1/\\sigma)$. (b), \"wherever $\\mathrm{sd}(D)>0$\", is correct. (c), now with \"when $0<H(c_2)<1$\", is correct: $D/f_2\\to\\mathbf 1\\{\\varepsilon<c_2\\}-\\mathbf 1\\{\\varepsilon>c_2\\}$ boundedly, and quadrature agrees. (d) is correct. One point remains, in the closing sentence: \"sMAPE loses it, at the rate $\\sigma^{-1/2}$ when $h(c_1)\\ne h(c_2)$\". This asserts that sMAPE loses its power in every case, but neither the statement nor the proof (whose closing paragraph repeats it) shows this when $h(c_1)=h(c_2)$. If $h(c_1)=h(c_2)>0$, it follows from (a) in one line: $|\\mathrm E D|=O(\\sigma^{-3})$ and $\\operatorname{Var}D\\sim8(2\\ln2-1)h(c_1)/\\sigma$, so the ratio is $O(\\sigma^{-5/2})$. The proof should say this. If $h(c_1)=h(c_2)=0$ (possible when $h$ vanishes at a point or has bounded support), (a) gives only $\\mathrm E D=O(\\sigma^{-3})$ and $\\operatorname{Var}D=O(\\sigma^{-2})$, which do not bound the ratio. I could not derive the claim in that case. It is true in the cases I computed: $h\\propto(1-x^2)^3$ with $(c_1,c_2)=(1.5,2)$ or $(-1.5,2)$ gives a ratio falling like $\\sigma^{-2}$; $(-1.3,1.3)$ gives 0. The gap is my own doing: my note on H39 suggested \"otherwise faster\" without having proved the zero-density case. Suggested wording: \"sMAPE loses it when $h(c_1)+h(c_2)>0$, at the rate $\\sigma^{-1/2}$ when $h(c_1)\\ne h(c_2)$ and faster when $h(c_1)=h(c_2)$\", with the one-line argument above added to the proof. Alternatively, keep only \"sMAPE loses it at the rate $\\sigma^{-1/2}$ when $h(c_1)\\ne h(c_2)$\". Smaller and optional: \"the absolute error keeps it unless the higher forecast is the median\" omits the condition $0<H(c_2)<1$ of (c) (with $H(c_2)=0$ the ratio is undefined); adding \"when $0<H(c_2)<1$\" would make it exact. With these changes I would confirm.",
     "resolution": "The statement was corrected, and the checker confirmed the corrected wording in a later check listed here."
    },
    {
     "by": "P5",
     "outcome": "disputed",
     "extent": "full",
     "notes": "Parts (a)-(d) are unchanged from H43 and correct (checked in CK30 and CK34). I judged each sentence of the new closing against them. (1) The absolute log error keeps its power wherever its ratio is defined: this follows from (b). (2) The absolute error keeps it when $0<H(c_2)<1$ and the higher forecast is not the median: this follows from (c), whose limit is then positive. (3) sMAPE loses it at the rate $\\sigma^{-1/2}$ when $h(c_1)\\ne h(c_2)$: this follows from (a). (5) The squared error on lognormal actuals: (d) gives $e^{\\sigma^2/2}\\cdot\\text{ratio}\\to1$, so the ratio is $e^{-\\sigma^2/2}(1+o(1))=o(\\sigma^{-k})$ for every $k$, for every $c_1<c_2$; this is fully covered by (d), and lognormal actuals (normal $h$) meet the hypotheses. (4), \"It loses it faster, with a ratio of order $\\sigma^{-5/2}$, when $h(c_1)=h(c_2)>0$\", is not established as worded. (a) gives $\\mathrm E[D]=O(\\sigma^{-3})$, an upper bound only, and $\\operatorname{Var}(D)=8(2\\ln2-1)h(c_1)/\\sigma+O(\\sigma^{-2})$, which is exactly of order $1/\\sigma$. So the ratio is $O(\\sigma^{-5/2})$: at most of order $\\sigma^{-5/2}$, which does establish \"loses it faster\". But \"a ratio of order $\\sigma^{-5/2}$\" naturally reads as the exact order, and that is false in general. The $\\sigma^{-3}$ coefficient of $\\mathrm E[D]$ is proportional to $h''(c_1)-h''(c_2)$, which can vanish: for a symmetric $h$ with $c_1=-c_2$, $\\mathrm E[D]=0$ exactly and the ratio is 0 for every $\\sigma$. The proof's new sentence in step 6 has the same slip: \"the mean is of order $\\sigma^{-3}$ (step 2)\", whereas step 2 gives only $O(\\sigma^{-3})$. Suggested wording: \"It loses it faster, with a ratio of at most order $\\sigma^{-5/2}$ ($O(\\sigma^{-5/2})$), when $h(c_1)=h(c_2)>0$\", and in proof step 6 \"the mean is $O(\\sigma^{-3})$ ... so the ratio is $O(\\sigma^{-5/2})$\". With that change I would confirm. Separately, I checked the revised H12 step 6 (behind H38): the weights 0.406125, 0.054825 and 0.539050 now sum to 1; the parameters are correctly marked as rounded; and \"rises through $m$, where it is flat\" matches my computation ($\\phi'\\ge0$ on both sides of 0). It is now right.",
     "resolution": "The statement was corrected, and the checker confirmed the corrected wording in a later check listed here."
    },
    {
     "by": "P5",
     "outcome": "confirmed",
     "extent": "full",
     "notes": "A word diff against H45 shows that the only change is \"at most\" in the sentence I disputed in CK36. Parts (a)-(d) are correct, as checked in CK30, CK34 and CK36. I judged each closing sentence against them. The absolute log error keeps its power wherever its ratio is defined: this follows from (b). The absolute error keeps it when $0<H(c_2)<1$ and the higher forecast is not the median: this follows from (c). sMAPE loses it at the rate $\\sigma^{-1/2}$ when $h(c_1)\\ne h(c_2)$: this follows from (a). \"It loses it faster, with a ratio of at most order $\\sigma^{-5/2}$, when $h(c_1)=h(c_2)>0$\" is now exactly what (a) gives. The mean is $O(\\sigma^{-3})$ and the variance is $8(2\\ln2-1)h(c_1)/\\sigma+O(\\sigma^{-2})$, exactly of order $1/\\sigma$, so the ratio is $O(\\sigma^{-5/2})$; the bound allows the ratio to be smaller, even 0 (symmetric $h$, $c_1=-c_2$). The squared error on lognormal actuals loses its power faster than any power of $\\sigma$: by (d) the ratio is $e^{-\\sigma^2/2}(1+o(1))$, for every $c_1<c_2$. The revised proof step 6 writes $O(\\sigma^{-3})$ and $O(\\sigma^{-5/2})$ (\"at most of that order\"), and the closing paragraph matches the statement. Every part and every closing sentence holds as worded, and the proof establishes them."
    }
   ],
   "verification": "agent_checked",
   "relations": [
    {
     "type": "supports",
     "note": "C6 recommends against sMAPE for point forecasts; C18 shows that on widely spread actuals a test set tells two forecasts apart less and less well under sMAPE.",
     "reference": "B6"
    },
    {
     "type": "supports",
     "reference": "B13",
     "note": "Kolassa (2026) concludes that sMAPE should not be used; C18 shows that on widely spread actuals a test set tells two forecasts apart less and less well under sMAPE."
    },
    {
     "type": "refines",
     "target": "C10",
     "id_type": "local",
     "note": "C10 shows that the scores of forecasts at fixed quantile levels move together like 1/σ. C18 adds the noise and shows that the signal-to-noise ratio falls like σ^(-1/2), while that of the absolute log error does not change."
    }
   ],
   "depends_on": [
    "C3",
    "C10"
   ],
   "verifiability": "open",
   "verifiability_flags": []
  },
  {
   "id": "C19",
   "name": "The sMAPE minimiser in large samples",
   "summary": "Where the log of the actual is symmetric with one peak, the forecast that minimises sMAPE on a test set is asymptotically normal about the median. It is asymptotically as precise as the sample median for narrow distributions and loses its precision as the actuals spread out.",
   "statement": "Let $\\ln y$ have a density $g$ that is symmetric about $\\ln m$, unimodal ($g(\\ln m + v_1) \\ge g(\\ln m + v_2)$ whenever $|v_1| \\le |v_2|$, strictly on a set of positive probability) and continuous at $\\ln m$. Write $x = \\ln y - \\ln m$, $A = \\mathrm E[\\operatorname{sech}^2(x/2)\\tanh(|x|/2)]$ and $B = \\mathrm E[\\operatorname{sech}^4(x/2)]$. (a) $A < 2g(\\ln m)$. (b) Over test sets of $n$ independent actuals, the [[D8:sMAPE minimiser of a test set]] $\\hat f_n$ (a forecast with the lowest sMAPE of the test set) tends to $m$ in probability, and $\\sqrt n(\\ln\\hat f_n - \\ln m)$ is asymptotically normal with mean 0 and variance $B/(2g(\\ln m) - A)^2$; for the sample median the variance is $1/(4g(\\ln m)^2)$. The ratio of these two limiting variances, the asymptotic relative efficiency of the sMAPE minimiser to the sample median, is\n$$\\mathrm{ARE} = \\frac{(2g(\\ln m) - A)^2}{4\\,g(\\ln m)^2\\,B};$$\nthat the [[D9:relative efficiency]] of finite test sets tends to it is not shown here, since the variances of finite test sets need not tend to those of the limits. (c) For lognormal actuals with standard deviation $\\sigma$ of $\\ln y$: $\\mathrm{ARE} \\to 1$ as $\\sigma \\to 0$ and $\\mathrm{ARE}\\,\\sigma^3 \\to 6(\\ln 2)^2\\sqrt{2\\pi} \\approx 7.23$ as $\\sigma \\to \\infty$. For log-Laplace actuals with scale $b$ of $\\ln y$: $\\mathrm{ARE} \\to 1$ as $b \\to 0$ and $\\mathrm{ARE}\\,b \\to 3$ as $b \\to \\infty$. So on both families the sMAPE minimiser is asymptotically as precise as the sample median for narrow distributions and asymptotically loses all of its precision relative to it as the actuals spread out.",
   "about": {
    "subject": "D8",
    "relation": "is asymptotically normal about the median, with a stated variance and",
    "object": "D9",
    "comparator": "the sample median",
    "conditions": []
   },
   "type": "formal",
   "role": "principal",
   "novelty": "no_prior_found",
   "support": [
    {
     "kind": "formal_proof",
     "evidence": "E36"
    },
    {
     "kind": "empirical_quant",
     "evidence": "E33"
    },
    {
     "kind": "cited",
     "reference": "B15",
     "locator": "Theorems 5.7 and 5.23"
    }
   ],
   "provenance": "author_stated",
   "hypothesis": {
    "lab_event": "H44",
    "outcome": "supported"
   },
   "planning": "confirmatory",
   "planned_at": "2026-11-18T12:19:07Z",
   "planned_at_anchor": "git_remote",
   "test": {
    "framework": "formal",
    "supported": "The proof of the hypothesis as worded here (proofs.md, section H13 for the signal-to-noise ratios, H14 for the large-sample efficiency, revised to the corrected wording) is complete; the independent checker AC6 (Claude Opus 5.5) records a check of the corrected wording with outcome confirmed; and the deciding run's checks meet the tolerances of the design.",
    "not_supported": "A counterexample to the hypothesis as worded here, found by the proof, the checker, the supervisor or the deciding run, that a correction of the proof cannot remove.",
    "verdict": "judged",
    "runs": 1,
    "lab_plan": "PL15"
   },
   "checks": [
    {
     "by": "P5",
     "outcome": "disputed",
     "extent": "full",
     "notes": "Derived independently, with $x=\\ln Y-\\ln m$, $\\theta=\\ln f-\\ln m$, $M(\\theta)=\\mathrm E\\tanh(|x-\\theta|/2)$. (a) $k(u)=\\operatorname{sech}^2(u/2)\\tanh(|u|/2)$ is positive for $u\\ne0$ with $\\int k=2$, and $g(\\ln m+u)\\le g(\\ln m)$ with strict inequality on a set of positive Lebesgue measure (a density cannot be constant on the whole line), so $A<2g(\\ln m)$ strictly. This holds even without the \"positive probability\" clause (e.g. for a uniform $g$), so the hypothesis is stronger than needed. (b) Consistency: the proof's direct argument (step 3: a grid on $[-R,R]$ plus the $\\tfrac12$-Lipschitz bound; outside, $M_n\\ge\\tanh(R/4)P_n(|x|\\le R/2)$) is valid. A minimiser of $M_n$ exists because $M_n$ is continuous, tends to 1 at $\\pm\\infty$ and is below 1 at any data point; this is not stated. van der Vaart Thm 5.23 conditions: $|m_{\\theta_1}-m_{\\theta_2}|\\le\\tfrac12|\\theta_1-\\theta_2|$ (constant, square integrable); differentiable at 0 for every $x\\ne0$; second-order expansion $M(\\theta)=M(0)+\\tfrac12V\\theta^2+o(\\theta^2)$ with $V=g(\\ln m)-A/2>0$. I re-derived H12 step 4: writing the kernel as $-\\operatorname{sign}+\\operatorname{sign}\\cdot\\tanh^2$, $M'$ exists everywhere and is differentiable at 0 with $M''(0)=g(\\ln m)-A/2$ whenever $g$ is continuous at $\\ln m$, and $M(\\theta)-M(0)=\\int_0^\\theta M'$ gives the expansion. So continuity at $\\ln m$ suffices. Variance $(B/4)/(V^2)=B/(2g-A)^2$; sample median $1/(4g^2)$; ratio $(2g-A)^2/(4g^2B)$: all correct. (c) Recomputed: $\\int u^2k=16\\ln2$, $\\int\\operatorname{sech}^4(u/2)=8/3$, $\\int|u|k=4$; lognormal $\\mathrm{ARE}\\,\\sigma^3\\to6(\\ln2)^2\\sqrt{2\\pi}\\approx7.226$ with relative error $O(\\sigma^{-2})$; log-Laplace $2g-A=2/b^2+O(b^{-3})$, $B=4/(3b)+O(b^{-2})$, so $\\mathrm{ARE}\\,b=3+O(1/b)$ (the proof's orders are right; quadrature gives $b(\\mathrm{ARE}\\,b-3)\\approx-5.6$). Both narrow limits are 1. Quadrature agrees with all four limits. The one statement I cannot confirm as worded is the appositive in (b): \"The ratio of the two, the large-sample value of the [[TM9:relative efficiency]], is ARE\". TM9 defines $\\mathrm{RE}(n)$ as a ratio of finite-sample variances, so its \"large-sample value\" reads as $\\lim_n\\mathrm{RE}(n)$. The proof establishes only the ratio of the variances of the limiting normal distributions. Convergence in distribution does not give convergence of $n\\operatorname{Var}$, and under the hypotheses of H40 the limit can fail. For example, a symmetric, decreasing, continuous $g$ with $P(|x|>t)\\sim1/\\ln t$ makes the sample median's variance infinite for every $n$, so $\\mathrm{RE}(n)$ never approaches the finite ARE. For lognormal and log-Laplace actuals the identification is very likely true, but it would need a uniform-integrability argument that is not given. Suggested wording: \"The ratio of the two asymptotic variances, the asymptotic relative efficiency (the ratio of the variances of the limiting distributions; it is the limit of the relative efficiency only where $n$ times each estimator's variance converges to its asymptotic variance, which is not shown here), is ARE = ...\". The closing \"as precise as ... loses all of its precision\" should then say \"asymptotically\". With that rewording I would confirm every part.",
     "resolution": "The statement was corrected, and the checker confirmed the corrected wording in a later check listed here."
    },
    {
     "by": "P5",
     "outcome": "confirmed",
     "extent": "full",
     "notes": "Re-checked every part as now worded, against the revised H14. (a): $\\int k=2$, $k>0$ off 0, and $g(\\ln m+u)<g(\\ln m)$ on a set of positive measure, so $A<2g(\\ln m)$ strictly. (b): a minimiser of $M_n$ exists (now stated in the proof: $M_n$ is continuous, tends to 1 at both ends, and is below 1 at each $x_t$). The consistency argument (grid on $[-R,R]$, Lipschitz $\\tfrac12$, lower bound outside) is valid. The conditions of van der Vaart Thm 5.23 hold: constant Lipschitz bound $\\tfrac12$; differentiability at 0 for $x\\ne0$; second-order expansion with $M''(0)=g(\\ln m)-A/2>0$, which exists under continuity of $g$ at $\\ln m$ by the decomposition now in H12 step 4. The asymptotic variances are $B/(2g-A)^2$ and $1/(4g^2)$, and their ratio is the stated ARE. The corrected sentence now calls this the asymptotic relative efficiency and says explicitly that convergence of the finite-sample relative efficiency (TM9) to it is not shown; that is exactly what the proof establishes. (c): recomputed $\\int u^2k=16\\ln2$, $\\int\\operatorname{sech}^4(u/2)=8/3$, $\\int|u|k=4$. The limits $\\mathrm{ARE}\\to1$ (both narrow cases), $\\mathrm{ARE}\\,\\sigma^3\\to6(\\ln2)^2\\sqrt{2\\pi}\\approx7.226$ and $\\mathrm{ARE}\\,b\\to3$ check by hand and by quadrature. The remainder orders in steps 11–12 are right (the log-Laplace one is a relative $O(1/b)$). The closing sentence, now with \"asymptotically\", follows from (c). Every part holds as worded and the proof establishes it."
    }
   ],
   "verification": "agent_checked",
   "depends_on": [
    "C3",
    "C5",
    "C12"
   ],
   "relations": [
    {
     "type": "explains",
     "target": "C15",
     "id_type": "local",
     "note": "C15 found by simulation that the sMAPE minimiser of a test set varies more than the sample median, the more so the wider the spread. C19 gives the limiting variances of both and the rates at which their ratio falls.",
     "results": [
      "E33.T4",
      "E33.T6"
     ]
    }
   ],
   "verifiability": "open",
   "verifiability_flags": []
  },
  {
   "id": "C20",
   "name": "msMAPE is sMAPE except for small actuals",
   "summary": "Smyl's msMAPE equals sMAPE for actuals of at least its constant ν. Below ν its denominator stops shrinking. A zero actual then no longer gives every non-zero forecast the worst score, and the expected score has no jump at the zero forecast.",
   "statement": "The [[D33:msMAPE term]] $s_\\nu$ is the [[D1:sMAPE term]] $s$ except for actuals below $\\nu$. For $y \\ge 0$, $f \\ge 0$ and $\\nu > 0$: (a) where $y \\ge \\nu$, $s_\\nu(y,f) = s(y,f)$; so when every actual is at least $\\nu$, msMAPE is sMAPE and has the same target. (b) Where $y < \\nu$, $s_\\nu(y,f) = |f-y|/(f+\\nu)$: for a zero actual it is $f/(f+\\nu)$, which is 0 at $f = 0$ and rises towards 1 as $f$ grows, where the sMAPE term is 1 for every $f > 0$. (c) $0 \\le s_\\nu \\le 1$, with $s_\\nu < 1$ where $y < \\nu$; and for every $y$, $s_\\nu(y,f)$ is continuous in $f$ at $f = 0$, so the expected msMAPE term has no jump at the zero forecast, where the expected sMAPE term jumps when zero actuals have positive probability. (d) On counts with $\\nu \\le 1$, $s_\\nu$ differs from $s$ only for zero actuals.",
   "type": "formal",
   "role": "principal",
   "novelty": "extension",
   "support": [
    {
     "kind": "formal_proof",
     "evidence": "E37"
    },
    {
     "kind": "empirical_quant",
     "evidence": "E31"
    }
   ],
   "provenance": "author_stated",
   "hypothesis": {
    "lab_event": "H42",
    "outcome": "supported"
   },
   "planning": "confirmatory",
   "planned_at": "2026-11-18T12:02:12Z",
   "planned_at_anchor": "git_remote",
   "test": {
    "framework": "formal",
    "supported": "The proof of the hypothesis as worded here, in the study's proof documents (proofs.md, proof-suilin.md), is complete, every step naming what it rests on; the independent checker (an agent of another model than the author's: Claude Opus 5.5, registered as an actor before it checks) records a check of the hypothesis as worded with outcome confirmed; and the deciding run's numerical checks of its formulas meet the tolerances of the design.",
    "not_supported": "A counterexample to the hypothesis as worded, found by the proof, the checker, the supervisor or the deciding run, that a correction of the proof cannot remove.",
    "verdict": "judged",
    "runs": 1,
    "lab_plan": "PL12"
   },
   "checks": [
    {
     "by": "P5",
     "outcome": "confirmed",
     "extent": "full",
     "notes": "Derived independently from the definition of TM38 (I did not check Kolassa's text; the check is of the statement about the term as TM38 defines it). (a) For $y\\ge\\nu>0$, $\\max\\{y,\\nu\\}=y$ and $p=|f-y|/(f+y)=s$ for $y,f\\ge0$. So with every actual at least $\\nu$ the two expected losses coincide for $f\\ge0$; negative forecasts score 1 under sMAPE on positive actuals and cannot be better, so the targets agree. (b) For $y<\\nu$, $p=|f-y|/(f+\\nu)$; at $y=0$ it is $f/(f+\\nu)=1-\\nu/(f+\\nu)$: 0 at $f=0$, increasing, tending to 1, whereas $s(0,f)=1$ for $f>0$. (c) $p\\ge0$; for $y\\ge\\nu$, $p=s\\le1$; for $y<\\nu$, $f\\ge y$ gives $f-y\\le f<f+\\nu$ and $f<y$ gives $y-f<\\nu\\le f+\\nu$, so $p<1$. The denominator is at least $\\nu>0$, so $p$ is continuous in $f\\ge0$, and bounded convergence makes the expectation continuous. For sMAPE, $\\mathrm E[s(Y,0)]=1-P(Y=0)$ while $\\lim_{f\\downarrow0}\\mathrm E[s(Y,f)]=1$: a jump of $P(Y=0)$. (d) A count below $\\nu\\le1$ is 0. The proof (proof-suilin.md, second proof) matches step by step. Every part holds as worded."
    }
   ],
   "verification": "agent_checked",
   "about": {
    "subject": "D33",
    "relation": "equals, except for actuals below its constant,",
    "object": "D1",
    "comparator": null,
    "conditions": []
   },
   "relations": [
    {
     "type": "extends",
     "reference": "B13",
     "note": "Kolassa (2026, Proposition 3) shows that the minimisers of expected msMAPE on counts are integers; C20 shows where the msMAPE term equals the sMAPE term and that it removes the jump at the zero forecast."
    },
    {
     "type": "extends",
     "reference": "B16",
     "note": "Smyl (2025) proposed msMAPE, the sparse-proof sMAPE, as Kolassa (2026) reports; Smyl's article was not read. C20 states properties of its term as Kolassa (2026, Definition 4) defines it."
    }
   ],
   "verifiability": "open",
   "verifiability_flags": []
  },
  {
   "id": "C21",
   "name": "What msMAPE rewards on intermittent demand",
   "summary": "On counts, msMAPE with ν = 0.1 and the Suilin loss reward the same forecast as sMAPE. With ν = 0.5 or 1 it is 0 less often. On non-integer demand the zero forecast needs a larger share of zeros: sMAPE 0.394; msMAPE 0.419, 0.519, 0.565; Suilin 0.470.",
   "statement": "The forecasts that sMAPE, msMAPE with $\\nu = 0.1$, 0.5 and 1, and the Suilin loss reward were computed from the distributions themselves on the 14 [[D35:intermittent demand settings]]. (a) On each of the 10 count distributions, the [[D34:msMAPE target]] with $\\nu = 0.1$ and the [[D29:Suilin target]] equal the [[D6:sMAPE target]]. With $\\nu = 0.5$ the msMAPE target is 1 instead of 0 on Poisson counts with mean 0.7 and on negative binomial counts with mean 2, and 1 instead of 2 on Poisson counts with mean 1.5. With $\\nu = 1$ it is also 1 instead of 0 on Poisson counts with mean 0.5 and on negative binomial counts with mean 1. (b) For lognormal demand sizes with zeros, the share of zeros at which the target switches to the forecast 0 is 0.394 for sMAPE, 0.419 for msMAPE with $\\nu = 0.1$, 0.470 for the Suilin loss, 0.519 for msMAPE with $\\nu = 0.5$ and 0.565 with $\\nu = 1$. (c) In test sets of 12, 52 and 365 actuals, the forecast with the lowest score is 0 no more often under msMAPE than under sMAPE, and no more often under a larger $\\nu$. With 52 actuals from Poisson counts with mean 0.7, it is 0 in 0.6575 of the test sets under sMAPE, 0.523 with $\\nu = 0.1$, 0.1525 with $\\nu = 0.5$ and 0.0265 with $\\nu = 1$.",
   "type": "descriptive",
   "role": "secondary",
   "novelty": "no_prior_found",
   "support": [
    {
     "kind": "empirical_quant",
     "evidence": "E34",
     "results": [
      "E34.T1.r1",
      "E34.T1.r2",
      "E34.T1.r3",
      "E34.T1.r4",
      "E34.T1.r5",
      "E34.T1.r6",
      "E34.T1.r7",
      "E34.T1.r8",
      "E34.T1.r9",
      "E34.T1.r10",
      "E34.T2.r1",
      "E34.T4",
      "E34.T4.r11"
     ]
    }
   ],
   "provenance": "author_stated",
   "planning": "planned_estimate",
   "planned_at": "2026-11-18T12:14:53Z",
   "planned_at_anchor": "git_remote",
   "estimate": {
    "estimands": {
     "target": {
      "variable": "target",
      "table": true
     },
     "level": {
      "variable": "quantile level of the target",
      "table": true
     },
     "score": {
      "variable": "expected score at the target",
      "table": true
     },
     "zero_threshold": {
      "variable": "zero threshold",
      "table": true
     },
     "share_zero": {
      "variable": "share of test sets whose minimiser is 0",
      "table": true
     },
     "share_target": {
      "variable": "share of test sets whose minimiser is the target",
      "table": true
     },
     "median_minimiser": {
      "variable": "median of the test-set minimisers",
      "table": true
     }
    },
    "runs": 1,
    "lab_plan": "PL14"
   },
   "about": {
    "subject": "D34",
    "relation": "differs from",
    "object": "D6",
    "comparator": "D29",
    "conditions": [
     "D35"
    ]
   },
   "scope": {
    "data_type": "outcome distributions given exactly, and test sets drawn from them",
    "conditions": [
     "14 settings: 10 count distributions (Poisson with means 0.1 to 3, negative binomial with size 0.5 and means 0.5 to 2) and lognormal demand sizes (median 1, sigma 1) with shares of zeros 0.2 to 0.8",
     "msMAPE with nu = 0.1, 0.5 and 1",
     "test sets of 12, 52 and 365 independent actuals, 2000 test sets per setting and size"
    ]
   },
   "relations": [
    {
     "type": "supports",
     "target": "C20",
     "id_type": "local",
     "note": "Computed on intermittent demand, msMAPE needs a larger share of zeros than sMAPE before its target is 0, in line with C20 (b): a zero actual no longer gives every non-zero forecast the worst score.",
     "results": [
      "E34.T1",
      "E34.T2"
     ]
    }
   ],
   "verifiability": "open",
   "verifiability_flags": []
  }
 ],
 "evidence": [
  {
   "id": "E1",
   "profile": "quant",
   "design": "other",
   "description": "Numerical checks of the formulas of the formal claims ([[D31:formula checks]]), lab run RN8 under the formal plan that decided them; every check is within the plan's tolerances. They check the formulas on concrete numbers; the proofs are the evidence for the claims. The rows are labelled by the hypotheses of the record's first plan, whose wording kept the factor 200: H1 checks C1; H2 and H10 check C2; H5 checks C5; H7 checks C6; H8 checks C7; H12 checks C8; H13 checks C9; H14 checks C10. H3 and H9 checked C3, and H4 and H6 checked C4, as version 2 worded them; C3, C4 and C12 as now worded are checked in E31. The two results E1.R1 and E1.R2 compare a simulated variance with the large-sample variance that C19 now proves.",
   "availability": {
    "level": "public"
   },
   "lab_runs": [
    "RN8"
   ],
   "results": [
    {
     "id": "R1",
     "variable": "simulated n Var(ln sMAPE minimiser) / large-sample value, lognormal sigma=1",
     "value": 0.9944,
     "interval": {
      "level": 0.99,
      "lower": 0.9353981515994695,
      "upper": 1.0598078456840363,
      "kind": "confidence"
     },
     "conditions": {
      "hypothesis": "H11",
      "test-set size": 1000,
      "test sets": 4000
     }
    },
    {
     "id": "R2",
     "variable": "simulated n Var(ln sMAPE minimiser) / large-sample value, lognormal sigma=1",
     "value": 1.022,
     "interval": {
      "level": 0.99,
      "lower": 0.9396347019298308,
      "upper": 1.0981626206556903,
      "kind": "confidence"
     },
     "conditions": {
      "hypothesis": "H11",
      "test-set size": 10000,
      "test sets": 2000
     }
    },
    {
     "id": "R3",
     "variable": "maximum absolute difference from the ratio form",
     "value": 8.527e-14,
     "conditions": {
      "hypothesis": "H1"
     }
    },
    {
     "id": "R4",
     "variable": "maximum absolute difference between the ratio and its reciprocal",
     "value": 8.527e-14,
     "conditions": {
      "hypothesis": "H2"
     }
    },
    {
     "id": "R5",
     "variable": "maximum absolute difference from 200 tanh(|z|/2)",
     "value": 8.527e-14,
     "conditions": {
      "hypothesis": "H3"
     }
    },
    {
     "id": "R6",
     "variable": "violations of monotonicity, concavity or the bound 200 on 0 <= |z| <= 12",
     "value": 0,
     "conditions": {
      "hypothesis": "H3"
     }
    },
    {
     "id": "R7",
     "variable": "violations of 0 <= 100|z| - s <= 25|z|^3/3",
     "value": 0,
     "conditions": {
      "hypothesis": "H9"
     }
    },
    {
     "id": "R8",
     "variable": "maximum of |s - 100|z| + 25|z|^3/3| / |z|^5 for 0.05 < |z| < 0.5",
     "value": 0.8331,
     "conditions": {
      "hypothesis": "H9"
     }
    },
    {
     "id": "R9",
     "variable": "violations of s(y, y+e) < s(y, y-e)",
     "value": 0,
     "conditions": {
      "hypothesis": "H10"
     }
    },
    {
     "id": "R10",
     "variable": "maximum absolute difference between s(y, ky) and s(y, y/k)",
     "value": 1.137e-13,
     "conditions": {
      "hypothesis": "H10"
     }
    },
    {
     "id": "R11",
     "variable": "maximum relative difference between the formula for S' and a numerical derivative",
     "value": 4.823e-08,
     "conditions": {
      "hypothesis": "H4"
     }
    },
    {
     "id": "R12",
     "variable": "maximum distance in ln f between the minimiser and the median",
     "value": 8.641e-08,
     "conditions": {
      "hypothesis": "H5"
     }
    },
    {
     "id": "R13",
     "variable": "number of local minima, lognormal sigma=1 and log-Laplace b=1",
     "value": 1,
     "conditions": {
      "hypothesis": "H5"
     }
    },
    {
     "id": "R14",
     "variable": "maximum balance residual at the minimiser (pull below minus pull above)",
     "value": 8.633e-09,
     "conditions": {
      "hypothesis": "H6"
     }
    },
    {
     "id": "R15",
     "variable": "settings with one local minimum where the sign of D misplaces the target",
     "value": 0,
     "conditions": {
      "hypothesis": "H6"
     }
    },
    {
     "id": "R16",
     "variable": "maximum distance between the one-step and the exact quantile level",
     "value": 0.002514,
     "conditions": {
      "hypothesis": "H6"
     }
    },
    {
     "id": "R17",
     "variable": "2 g(m) - E[sech^2(Z/2) tanh(|Z|/2)] for the equal-mode bimodal setting (negative: median a local maximum)",
     "value": -0.3475,
     "conditions": {
      "hypothesis": "H7"
     }
    },
    {
     "id": "R18",
     "variable": "number of local minima of the equal-mode bimodal setting",
     "value": 2,
     "conditions": {
      "hypothesis": "H7"
     }
    },
    {
     "id": "R19",
     "variable": "difference between the two minimal expected sMAPEs, equal-mode bimodal setting",
     "value": 1.421e-14,
     "conditions": {
      "hypothesis": "H7"
     }
    },
    {
     "id": "R20",
     "variable": "maximum distance between the formula and the bisected threshold",
     "value": 5.551e-17,
     "conditions": {
      "hypothesis": "H8"
     }
    },
    {
     "id": "R21",
     "variable": "opposite-sign or one-zero pairs not scoring 200",
     "value": 0,
     "conditions": {
      "hypothesis": "H8"
     }
    },
    {
     "id": "R22",
     "variable": "relative distance between the 1/y-weighted median and the minimiser of expected MAPE",
     "value": 6.745e-05,
     "conditions": {
      "hypothesis": "H12"
     }
    },
    {
     "id": "R23",
     "variable": "difference of expected squared errors minus (f1-f2)(f1+f2-2E[y])",
     "value": 0.0,
     "conditions": {
      "hypothesis": "H12"
     }
    },
    {
     "id": "R24",
     "variable": "sMAPE terms outside [0, 200]",
     "value": 0,
     "conditions": {
      "hypothesis": "H13"
     }
    },
    {
     "id": "R25",
     "variable": "expected sMAPE at f = 1e-6 and f = 1e10 (both near 200)",
     "value": 200.0,
     "conditions": {
      "hypothesis": "H13"
     }
    },
    {
     "id": "R26",
     "variable": "the proof's bound on the error times sigma^3, 600 zeta(3) max|g''|, log-Gumbel",
     "value": 309.9,
     "conditions": {
      "hypothesis": "H14"
     }
    },
    {
     "id": "R27",
     "variable": "limit of sigma^3 times the relative efficiency for normal log-outcomes, from the proof",
     "value": 7.226,
     "conditions": {
      "hypothesis": "H14"
     }
    }
   ],
   "tables": [
    {
     "id": "T1",
     "description": "zero threshold by setting; hypothesis: H8 (lab runs RN8)",
     "conditions": {
      "hypothesis": "H8"
     },
     "row_labels": [
      "setting"
     ],
     "columns": [
      "zero threshold"
     ],
     "rows": [
      {
       "id": "r1",
       "label": [
        "lognormal sigma=0.3"
       ],
       "values": [
        0.4687
       ]
      },
      {
       "id": "r2",
       "label": [
        "lognormal sigma=0.5"
       ],
       "values": [
        0.4469
       ]
      },
      {
       "id": "r3",
       "label": [
        "lognormal sigma=1"
       ],
       "values": [
        0.394
       ]
      },
      {
       "id": "r4",
       "label": [
        "lognormal sigma=2"
       ],
       "values": [
        0.3075
       ]
      }
     ],
     "artefact": "A6"
    },
    {
     "id": "T2",
     "description": "maximum |S - (200 - 800 ln2 h)| times sigma^3, log-Gumbel right, sigma^3 times the asymptotic relative efficiency, lognormal by sigma; hypothesis: H14 (lab runs RN8)",
     "conditions": {
      "hypothesis": "H14"
     },
     "row_labels": [
      "sigma"
     ],
     "columns": [
      "maximum |S - (200 - 800 ln2 h)| times sigma^3, log-Gumbel right",
      "sigma^3 times the asymptotic relative efficiency, lognormal"
     ],
     "rows": [
      {
       "id": "r1",
       "label": [
        5.0
       ],
       "values": [
        264.5,
        5.613
       ]
      },
      {
       "id": "r2",
       "label": [
        10.0
       ],
       "values": [
        292.3,
        6.743
       ]
      },
      {
       "id": "r3",
       "label": [
        20.0
       ],
       "values": [
        300.1,
        7.099
       ]
      },
      {
       "id": "r4",
       "label": [
        40.0
       ],
       "values": [
        302.1,
        7.193
       ]
      }
     ],
     "artefact": "A6"
    }
   ],
   "artefacts": [
    "A6"
   ],
   "execution": {
    "code_version": "fa0ce626f18bd1ca22076d62452d2b68c484a297",
    "software": {
     "python": "3.13.18",
     "platform": "Linux-6.18.44-fc-v553-x86_64-with-glibc2.39"
    }
   },
   "setup": {
    "studied": [
     "D1",
     "D3",
     "D4",
     "D5",
     "D6"
    ],
    "procedure": [
     "D31"
    ]
   }
  },
  {
   "id": "E11",
   "profile": "quant",
   "design": "other",
   "description": "Numerical checks of the formulas of C11 ([[D31:formula checks]]), lab run RN6 under the formal plan that decided C11; every check is within the plan's tolerances. They check the formulas on concrete numbers; the proof is the evidence for the claim.",
   "availability": {
    "level": "public"
   },
   "lab_runs": [
    "RN6"
   ],
   "tables": [
    {
     "id": "T1",
     "description": "sign of the global minimiser of expected sMAPE, distance of the global minimiser from 0, lowest expected sMAPE term, local minima of expected sMAPE, smallest distance of any local minimiser from 0 by mean; hypothesis: H31 (lab runs RN6)",
     "conditions": {
      "hypothesis": "H31"
     },
     "row_labels": [
      "mean"
     ],
     "columns": [
      "sign of the global minimiser of expected sMAPE",
      "distance of the global minimiser from 0",
      "lowest expected sMAPE term",
      "local minima of expected sMAPE",
      "smallest distance of any local minimiser from 0"
     ],
     "rows": [
      {
       "id": "r1",
       "label": [
        -0.5
       ],
       "values": [
        -1,
        0.9838,
        0.5227,
        2,
        0.5849
       ]
      },
      {
       "id": "r2",
       "label": [
        -0.2
       ],
       "values": [
        -1,
        0.8371,
        0.609,
        2,
        0.679
       ]
      },
      {
       "id": "r3",
       "label": [
        -0.05
       ],
       "values": [
        -1,
        0.773,
        0.6525,
        2,
        0.7335
       ]
      },
      {
       "id": "r4",
       "label": [
        -0.01
       ],
       "values": [
        -1,
        0.7569,
        0.664,
        2,
        0.7491
       ]
      },
      {
       "id": "r5",
       "label": [
        0.01
       ],
       "values": [
        1,
        0.7569,
        0.664,
        2,
        0.7491
       ]
      },
      {
       "id": "r6",
       "label": [
        0.05
       ],
       "values": [
        1,
        0.773,
        0.6525,
        2,
        0.7336
       ]
      },
      {
       "id": "r7",
       "label": [
        0.2
       ],
       "values": [
        1,
        0.8371,
        0.609,
        2,
        0.679
       ]
      },
      {
       "id": "r8",
       "label": [
        0.5
       ],
       "values": [
        1,
        0.9838,
        0.5227,
        2,
        0.5849
       ]
      }
     ],
     "artefact": "A6"
    },
    {
     "id": "T2",
     "description": "expected sMAPE term at a forecast near 0 by mean, forecast; hypothesis: H31 (lab runs RN6)",
     "conditions": {
      "hypothesis": "H31"
     },
     "row_labels": [
      "mean",
      "forecast"
     ],
     "columns": [
      "expected sMAPE term at a forecast near 0"
     ],
     "rows": [
      {
       "id": "r1",
       "label": [
        -0.5,
        -0.001
       ],
       "values": [
        0.9948
       ]
      },
      {
       "id": "r2",
       "label": [
        -0.5,
        0.001
       ],
       "values": [
        0.9957
       ]
      },
      {
       "id": "r3",
       "label": [
        -0.2,
        -0.001
       ],
       "values": [
        0.9946
       ]
      },
      {
       "id": "r4",
       "label": [
        -0.2,
        0.001
       ],
       "values": [
        0.995
       ]
      },
      {
       "id": "r5",
       "label": [
        -0.05,
        -0.001
       ],
       "values": [
        0.9947
       ]
      },
      {
       "id": "r6",
       "label": [
        -0.05,
        0.001
       ],
       "values": [
        0.9948
       ]
      },
      {
       "id": "r7",
       "label": [
        -0.01,
        -0.001
       ],
       "values": [
        0.9947
       ]
      },
      {
       "id": "r8",
       "label": [
        -0.01,
        0.001
       ],
       "values": [
        0.9948
       ]
      },
      {
       "id": "r9",
       "label": [
        0.01,
        -0.001
       ],
       "values": [
        0.9948
       ]
      },
      {
       "id": "r10",
       "label": [
        0.01,
        0.001
       ],
       "values": [
        0.9947
       ]
      },
      {
       "id": "r11",
       "label": [
        0.05,
        -0.001
       ],
       "values": [
        0.9948
       ]
      },
      {
       "id": "r12",
       "label": [
        0.05,
        0.001
       ],
       "values": [
        0.9947
       ]
      },
      {
       "id": "r13",
       "label": [
        0.2,
        -0.001
       ],
       "values": [
        0.995
       ]
      },
      {
       "id": "r14",
       "label": [
        0.2,
        0.001
       ],
       "values": [
        0.9946
       ]
      },
      {
       "id": "r15",
       "label": [
        0.5,
        -0.001
       ],
       "values": [
        0.9957
       ]
      },
      {
       "id": "r16",
       "label": [
        0.5,
        0.001
       ],
       "values": [
        0.9948
       ]
      }
     ],
     "artefact": "A6"
    }
   ],
   "artefacts": [
    "A6"
   ],
   "execution": {
    "code_version": "2da728ac88adda9aa5654f1797cf3b98451aa5cc",
    "software": {
     "python": "3.13.18",
     "platform": "Linux-6.18.44-fc-v470-x86_64-with-glibc2.39"
    }
   },
   "setup": {
    "studied": [
     "D6"
    ],
    "procedure": [
     "D31"
    ]
   }
  },
  {
   "id": "E13",
   "profile": "quant",
   "design": "other",
   "description": "Lab runs RN1 under plan PL1. Design: Selection design (D16) on 33 outcome distributions: inverse Gaussian (6), log-skewed (12) and lognormal (2), bimodal (3), straddling zero (3), zero-inflated lognormal (8). For each: the derived values (D15) of every local minimum of expected sMAPE with its quantile level and expected score; the one-step quantile level of H6 and the pull difference D at the median; the zero threshold of H8; the quantile level of the log-mode; which moments are finite (H12); the MAPE target (H12); each measure's preferred candidate (D14), ranked by finite differences where expected scores are infinite. Candidates: quantiles at levels 0.01 to 0.99, the mean (when finite), 0, the log-mode, the MAPE minimiser (when it exists and is not 0) and every local minimiser of expected sMAPE, merged when within 1e-4 (relative). Test sets of n = 6, 8, 12, 13, 14, 18, 24, 36, 48, 100, 250, 500, 1000 and 10000 outcomes, 1000 test sets per size, seed 3 with the setting's index and n. Measures RMSE, MAE, MAPE, sMAPE on every setting. python3 claimset-code/smape-target/targets.py --seed 3 --reps 1000 (numpy 2.5.3, matplotlib 3.11.2). Analysis: Per distribution: the derived values listed in the design. Per distribution, measure and n, over test sets, of the candidate ranked first: its quantile level's mean with a 99% confidence interval and 5th, 25th, 50th, 75th and 95th percentiles; the share choosing each candidate; the share choosing the preferred candidate with its 99% Wilson interval; the number of test sets in which every candidate scores infinite. Figures: one per distribution, a panel per measure, violins of the chosen quantile levels over n with boxplots and the mean, lines at every measure's preferred candidate and at further local minima of expected sMAPE, and a note in the panel of a measure that targets nothing. These are compared with what H4 to H8, H12 and H13 predict; no test is made and no threshold is set.",
   "availability": {
    "level": "public"
   },
   "lab_runs": [
    "RN1"
   ],
   "results": [],
   "tables": [
    {
     "id": "T1",
     "description": "The sMAPE target and related values derived from each distribution: the lowest local minimum of expected sMAPE (rank 1), its quantile level and expected sMAPE, its one-step approximation, the pull from above minus the pull from below at the median, the share of zeros at which the target switches to 0, the quantile level of the log-mode, and the MAPE target",
     "row_labels": [
      "setting"
     ],
     "columns": [
      {
       "variable": "derived local minimum of expected sMAPE",
       "conditions": {
        "rank": 1
       }
      },
      {
       "variable": "quantile level of a derived local minimum of expected sMAPE",
       "conditions": {
        "rank": 1
       }
      },
      {
       "variable": "expected sMAPE at a derived local minimum",
       "conditions": {
        "rank": 1
       }
      },
      "one-step quantile level of the sMAPE target",
      "pull from above minus pull from below at the median",
      "zero threshold of sMAPE",
      "quantile level of the log-mode",
      "median of the outcome distribution",
      "mean of the outcome distribution"
     ],
     "rows": [
      {
       "id": "r1",
       "label": [
        "inverse Gaussian shape=1000"
       ],
       "values": [
        95.27,
        0.4999,
        24.36,
        0.4999,
        -7.522e-05,
        0.4676,
        0.498,
        95.27,
        100
       ]
      },
      {
       "id": "r2",
       "label": [
        "inverse Gaussian shape=200"
       ],
       "values": [
        80.14,
        0.4978,
        48.68,
        0.4978,
        -0.001851,
        0.4307,
        0.4822,
        80.43,
        100
       ]
      },
      {
       "id": "r3",
       "label": [
        "inverse Gaussian shape=100"
       ],
       "values": [
        66.53,
        0.4929,
        61.97,
        0.4929,
        -0.005184,
        0.4083,
        0.4597,
        67.58,
        100
       ]
      },
      {
       "id": "r4",
       "label": [
        "inverse Gaussian shape=50"
       ],
       "values": [
        48.86,
        0.482,
        75.05,
        0.482,
        -0.01136,
        0.3845,
        0.4233,
        51.42,
        100
       ]
      },
      {
       "id": "r5",
       "label": [
        "inverse Gaussian shape=20"
       ],
       "values": [
        25.9,
        0.4578,
        88.96,
        0.457,
        -0.02237,
        0.357,
        0.3726,
        30.11,
        100
       ]
      },
      {
       "id": "r6",
       "label": [
        "inverse Gaussian shape=10"
       ],
       "values": [
        14.02,
        0.4387,
        95.75,
        0.4362,
        -0.03001,
        0.3426,
        0.3471,
        17.85,
        100
       ]
      },
      {
       "id": "r7",
       "label": [
        "log-skewed left sd=0.5 skew=-2"
       ],
       "values": [
        119.0,
        0.5211,
        32.31,
        0.5211,
        0.01967,
        0.4561,
        1,
        116.6,
        109.9
       ]
      },
      {
       "id": "r8",
       "label": [
        "log-skewed left sd=0.5 skew=-1"
       ],
       "values": [
        110.2,
        0.513,
        36.74,
        0.513,
        0.01171,
        0.4494,
        0.6472,
        108.5,
        111.3
       ]
      },
      {
       "id": "r9",
       "label": [
        "log-skewed left sd=0.5 skew=-0.5"
       ],
       "values": [
        105.1,
        0.5068,
        37.96,
        0.5068,
        0.006115,
        0.4476,
        0.5681,
        104.2,
        112.2
       ]
      },
      {
       "id": "r10",
       "label": [
        "log-skewed right sd=0.5 skew=+2"
       ],
       "values": [
        84.02,
        0.4789,
        32.31,
        0.4789,
        -0.01967,
        0.4561,
        0,
        85.78,
        121.3
       ]
      },
      {
       "id": "r11",
       "label": [
        "log-skewed right sd=0.5 skew=+1"
       ],
       "values": [
        90.72,
        0.487,
        36.74,
        0.487,
        -0.01171,
        0.4494,
        0.3528,
        92.13,
        116.3
       ]
      },
      {
       "id": "r12",
       "label": [
        "log-skewed right sd=0.5 skew=+0.5"
       ],
       "values": [
        95.13,
        0.4932,
        37.96,
        0.4932,
        -0.006115,
        0.4476,
        0.4319,
        95.93,
        114.6
       ]
      },
      {
       "id": "r13",
       "label": [
        "lognormal sigma=0.5"
       ],
       "values": [
        100.0,
        0.5,
        38.38,
        0.5,
        -8.965e-15,
        0.4469,
        0.5,
        100.0,
        113.3
       ]
      },
      {
       "id": "r14",
       "label": [
        "log-skewed left sd=1 skew=-2"
       ],
       "values": [
        151.2,
        0.5562,
        57.04,
        0.5572,
        0.04564,
        0.4168,
        1,
        135.9,
        135.9
       ]
      },
      {
       "id": "r15",
       "label": [
        "log-skewed left sd=1 skew=-1"
       ],
       "values": [
        128.3,
        0.5362,
        66.42,
        0.5365,
        0.02609,
        0.4004,
        0.6472,
        117.8,
        146.0
       ]
      },
      {
       "id": "r16",
       "label": [
        "log-skewed left sd=1 skew=-0.5"
       ],
       "values": [
        113.9,
        0.5191,
        69.05,
        0.5192,
        0.01332,
        0.3957,
        0.5681,
        108.7,
        153.7
       ]
      },
      {
       "id": "r17",
       "label": [
        "log-skewed right sd=1 skew=+2"
       ],
       "values": [
        66.14,
        0.4438,
        57.04,
        0.4428,
        -0.04564,
        0.4168,
        0,
        73.58,
        null
       ]
      },
      {
       "id": "r18",
       "label": [
        "log-skewed right sd=1 skew=+1"
       ],
       "values": [
        77.96,
        0.4638,
        66.42,
        0.4635,
        -0.02609,
        0.4004,
        0.3528,
        84.88,
        216.5
       ]
      },
      {
       "id": "r19",
       "label": [
        "log-skewed right sd=1 skew=+0.5"
       ],
       "values": [
        87.79,
        0.4809,
        69.05,
        0.4808,
        -0.01332,
        0.3957,
        0.4319,
        92.03,
        182.7
       ]
      },
      {
       "id": "r20",
       "label": [
        "lognormal sigma=1"
       ],
       "values": [
        100.0,
        0.5,
        69.94,
        0.5,
        -2.606e-14,
        0.394,
        0.5,
        100.0,
        164.9
       ]
      },
      {
       "id": "r21",
       "label": [
        "bimodal equal modes"
       ],
       "values": [
        43.49,
        0.3742,
        83.84,
        0.5,
        -1.532e-11,
        0.3674,
        0.25,
        100.0,
        159.2
       ]
      },
      {
       "id": "r22",
       "label": [
        "bimodal left mode higher"
       ],
       "values": [
        39.84,
        0.4063,
        65.21,
        0.4042,
        -0.08387,
        0.4026,
        0.325,
        44.22,
        122.8
       ]
      },
      {
       "id": "r23",
       "label": [
        "bimodal right mode higher"
       ],
       "values": [
        251.0,
        0.5937,
        65.21,
        0.5958,
        0.08387,
        0.4026,
        0.675,
        226.1,
        195.6
       ]
      },
      {
       "id": "r24",
       "label": [
        "normal straddling zero mean=0"
       ],
       "values": [
        -0.753,
        0.2257,
        133.4,
        null,
        null,
        null,
        null,
        -3e-05,
        0
       ]
      },
      {
       "id": "r25",
       "label": [
        "normal straddling zero mean=0.5"
       ],
       "values": [
        0.9838,
        0.6858,
        104.5,
        null,
        null,
        null,
        null,
        0.5,
        0.5
       ]
      },
      {
       "id": "r26",
       "label": [
        "normal straddling zero mean=1"
       ],
       "values": [
        1.287,
        0.6128,
        78.51,
        null,
        null,
        null,
        null,
        1.0,
        1
       ]
      },
      {
       "id": "r27",
       "label": [
        "lognormal sigma=0.5 with zeros p0=0.25"
       ],
       "values": [
        100.0,
        0.625,
        78.78,
        0.5,
        -8.965e-15,
        0.4469,
        0.625,
        80.62,
        84.99
       ]
      },
      {
       "id": "r28",
       "label": [
        "lognormal sigma=0.5 with zeros p0=0.42"
       ],
       "values": [
        100.0,
        0.71,
        106.3,
        0.5,
        -8.965e-15,
        0.4469,
        0.71,
        57.99,
        65.72
       ]
      },
      {
       "id": "r29",
       "label": [
        "lognormal sigma=0.5 with zeros p0=0.47"
       ],
       "values": [
        0,
        0.47,
        106,
        0.5,
        -8.965e-15,
        0.4469,
        0.735,
        45.3,
        60.06
       ]
      },
      {
       "id": "r30",
       "label": [
        "lognormal sigma=0.5 with zeros p0=0.6"
       ],
       "values": [
        0,
        0.6,
        80,
        0.5,
        -8.965e-15,
        0.4469,
        0.8,
        0,
        45.33
       ]
      },
      {
       "id": "r31",
       "label": [
        "lognormal sigma=1 with zeros p0=0.25"
       ],
       "values": [
        100.0,
        0.625,
        102.5,
        0.5,
        -2.606e-14,
        0.394,
        0.625,
        65.0,
        123.7
       ]
      },
      {
       "id": "r32",
       "label": [
        "lognormal sigma=1 with zeros p0=0.37"
       ],
       "values": [
        100.0,
        0.685,
        118.1,
        0.5,
        -2.606e-14,
        0.394,
        0.685,
        44.08,
        103.9
       ]
      },
      {
       "id": "r33",
       "label": [
        "lognormal sigma=1 with zeros p0=0.42"
       ],
       "values": [
        0,
        0.42,
        116.0,
        0.5,
        -2.606e-14,
        0.394,
        0.71,
        33.63,
        95.63
       ]
      },
      {
       "id": "r34",
       "label": [
        "lognormal sigma=1 with zeros p0=0.6"
       ],
       "values": [
        0,
        0.6,
        80,
        0.5,
        -2.606e-14,
        0.394,
        0.8,
        0,
        65.95
       ]
      }
     ],
     "artefact": "A8"
    },
    {
     "id": "T2",
     "description": "The second local minimum of expected sMAPE, where there is one (rank 2)",
     "row_labels": [
      "setting"
     ],
     "columns": [
      {
       "variable": "derived local minimum of expected sMAPE",
       "conditions": {
        "rank": 2
       }
      },
      {
       "variable": "quantile level of a derived local minimum of expected sMAPE",
       "conditions": {
        "rank": 2
       }
      },
      {
       "variable": "expected sMAPE at a derived local minimum",
       "conditions": {
        "rank": 2
       }
      }
     ],
     "rows": [
      {
       "id": "r1",
       "label": [
        "bimodal equal modes"
       ],
       "values": [
        229.9,
        0.6258,
        83.84
       ]
      },
      {
       "id": "r2",
       "label": [
        "normal straddling zero mean=0"
       ],
       "values": [
        0.753,
        0.7743,
        133.4
       ]
      },
      {
       "id": "r3",
       "label": [
        "normal straddling zero mean=0.5"
       ],
       "values": [
        -0.5849,
        0.139,
        159.9
       ]
      },
      {
       "id": "r4",
       "label": [
        "normal straddling zero mean=1"
       ],
       "values": [
        -0.4649,
        0.07148,
        179.8
       ]
      },
      {
       "id": "r5",
       "label": [
        "lognormal sigma=0.5 with zeros p0=0.25"
       ],
       "values": [
        0,
        0.25,
        150
       ]
      },
      {
       "id": "r6",
       "label": [
        "lognormal sigma=0.5 with zeros p0=0.42"
       ],
       "values": [
        0,
        0.42,
        116.0
       ]
      },
      {
       "id": "r7",
       "label": [
        "lognormal sigma=0.5 with zeros p0=0.47"
       ],
       "values": [
        100.0,
        0.735,
        114.3
       ]
      },
      {
       "id": "r8",
       "label": [
        "lognormal sigma=0.5 with zeros p0=0.6"
       ],
       "values": [
        100.0,
        0.8,
        135.4
       ]
      },
      {
       "id": "r9",
       "label": [
        "lognormal sigma=1 with zeros p0=0.25"
       ],
       "values": [
        0,
        0.25,
        150
       ]
      },
      {
       "id": "r10",
       "label": [
        "lognormal sigma=1 with zeros p0=0.37"
       ],
       "values": [
        0,
        0.37,
        126
       ]
      },
      {
       "id": "r11",
       "label": [
        "lognormal sigma=1 with zeros p0=0.42"
       ],
       "values": [
        100.0,
        0.71,
        124.6
       ]
      },
      {
       "id": "r12",
       "label": [
        "lognormal sigma=1 with zeros p0=0.6"
       ],
       "values": [
        100.0,
        0.8,
        148.0
       ]
      }
     ],
     "artefact": "A8"
    },
    {
     "id": "T3",
     "description": "Each measure's preferred candidate (the candidate forecast with the lowest expected score): its quantile level, and the number of candidates whose expected score is infinite",
     "row_labels": [
      "setting"
     ],
     "columns": [
      {
       "variable": "quantile level of the preferred candidate",
       "conditions": {
        "measure": "RMSE"
       }
      },
      {
       "variable": "quantile level of the preferred candidate",
       "conditions": {
        "measure": "MAE"
       }
      },
      {
       "variable": "quantile level of the preferred candidate",
       "conditions": {
        "measure": "MAPE"
       }
      },
      {
       "variable": "quantile level of the preferred candidate",
       "conditions": {
        "measure": "sMAPE"
       }
      },
      {
       "variable": "candidates whose expected score is infinite",
       "conditions": {
        "measure": "RMSE"
       }
      },
      {
       "variable": "candidates whose expected score is infinite",
       "conditions": {
        "measure": "MAPE"
       }
      }
     ],
     "rows": [
      {
       "id": "r1",
       "label": [
        "inverse Gaussian shape=1000"
       ],
       "values": [
        0.5616,
        0.4999,
        0.3786,
        0.4999,
        0,
        0
       ]
      },
      {
       "id": "r2",
       "label": [
        "inverse Gaussian shape=200"
       ],
       "values": [
        0.6277,
        0.5,
        0.2659,
        0.4978,
        0,
        0
       ]
      },
      {
       "id": "r3",
       "label": [
        "inverse Gaussian shape=100"
       ],
       "values": [
        0.6681,
        0.5,
        0.2158,
        0.4929,
        0,
        0
       ]
      },
      {
       "id": "r4",
       "label": [
        "inverse Gaussian shape=50"
       ],
       "values": [
        0.7138,
        0.5,
        0.1773,
        0.482,
        0,
        0
       ]
      },
      {
       "id": "r5",
       "label": [
        "inverse Gaussian shape=20"
       ],
       "values": [
        0.7768,
        0.5,
        0.1474,
        0.4578,
        0,
        0
       ]
      },
      {
       "id": "r6",
       "label": [
        "inverse Gaussian shape=10"
       ],
       "values": [
        0.8219,
        0.5,
        0.136,
        0.4387,
        0,
        0
       ]
      },
      {
       "id": "r7",
       "label": [
        "log-skewed left sd=0.5 skew=-2"
       ],
       "values": [
        0.4444,
        0.5,
        0.25,
        0.5211,
        0,
        0
       ]
      },
      {
       "id": "r8",
       "label": [
        "log-skewed left sd=0.5 skew=-1"
       ],
       "values": [
        0.5215,
        0.5,
        0.2799,
        0.513,
        0,
        0
       ]
      },
      {
       "id": "r9",
       "label": [
        "log-skewed left sd=0.5 skew=-0.5"
       ],
       "values": [
        0.5603,
        0.5,
        0.2941,
        0.5068,
        0,
        0
       ]
      },
      {
       "id": "r10",
       "label": [
        "log-skewed right sd=0.5 skew=+2"
       ],
       "values": [
        0.75,
        0.5,
        0.37,
        0.4789,
        102,
        0
       ]
      },
      {
       "id": "r11",
       "label": [
        "log-skewed right sd=0.5 skew=+1"
       ],
       "values": [
        0.6748,
        0.5,
        0.3388,
        0.487,
        0,
        0
       ]
      },
      {
       "id": "r12",
       "label": [
        "log-skewed right sd=0.5 skew=+0.5"
       ],
       "values": [
        0.6368,
        0.5,
        0.3234,
        0.4932,
        0,
        0
       ]
      },
      {
       "id": "r13",
       "label": [
        "lognormal sigma=0.5"
       ],
       "values": [
        0.5987,
        0.5,
        0.3085,
        0.5,
        0,
        0
       ]
      },
      {
       "id": "r14",
       "label": [
        "log-skewed left sd=1 skew=-2"
       ],
       "values": [
        0.5,
        0.5,
        0,
        0.5562,
        0,
        101
       ]
      },
      {
       "id": "r15",
       "label": [
        "log-skewed left sd=1 skew=-1"
       ],
       "values": [
        0.5928,
        0.5,
        0.0655,
        0.5362,
        0,
        0
       ]
      },
      {
       "id": "r16",
       "label": [
        "log-skewed left sd=1 skew=-0.5"
       ],
       "values": [
        0.6413,
        0.5,
        0.1155,
        0.5191,
        0,
        0
       ]
      },
      {
       "id": "r17",
       "label": [
        "log-skewed right sd=1 skew=+2"
       ],
       "values": [
        null,
        0.5,
        0.2929,
        0.4438,
        103,
        0
       ]
      },
      {
       "id": "r18",
       "label": [
        "log-skewed right sd=1 skew=+1"
       ],
       "values": [
        0.8034,
        0.5,
        0.2314,
        0.4638,
        104,
        0
       ]
      },
      {
       "id": "r19",
       "label": [
        "log-skewed right sd=1 skew=+0.5"
       ],
       "values": [
        0.7446,
        0.5,
        0.1968,
        0.4809,
        0,
        0
       ]
      },
      {
       "id": "r20",
       "label": [
        "lognormal sigma=1"
       ],
       "values": [
        0.6915,
        0.5,
        0.1587,
        0.5,
        0,
        0
       ]
      },
      {
       "id": "r21",
       "label": [
        "bimodal equal modes"
       ],
       "values": [
        0.5081,
        0.5,
        0.2341,
        0.3742,
        0,
        0
       ]
      },
      {
       "id": "r22",
       "label": [
        "bimodal left mode higher"
       ],
       "values": [
        0.6503,
        0.5,
        0.2841,
        0.4063,
        0,
        0
       ]
      },
      {
       "id": "r23",
       "label": [
        "bimodal right mode higher"
       ],
       "values": [
        0.4111,
        0.5,
        0.1848,
        0.5937,
        0,
        0
       ]
      },
      {
       "id": "r24",
       "label": [
        "normal straddling zero mean=0"
       ],
       "values": [
        0.5,
        0.5,
        0.5,
        0.7743,
        0,
        100
       ]
      },
      {
       "id": "r25",
       "label": [
        "normal straddling zero mean=0.5"
       ],
       "values": [
        0.5,
        0.5,
        0.3085,
        0.6858,
        0,
        101
       ]
      },
      {
       "id": "r26",
       "label": [
        "normal straddling zero mean=1"
       ],
       "values": [
        0.5,
        0.5,
        0.1587,
        0.6128,
        0,
        101
       ]
      },
      {
       "id": "r27",
       "label": [
        "lognormal sigma=0.5 with zeros p0=0.25"
       ],
       "values": [
        0.5293,
        0.5,
        null,
        0.625,
        0,
        77
       ]
      },
      {
       "id": "r28",
       "label": [
        "lognormal sigma=0.5 with zeros p0=0.42"
       ],
       "values": [
        0.5364,
        0.5,
        null,
        0.71,
        0,
        59
       ]
      },
      {
       "id": "r29",
       "label": [
        "lognormal sigma=0.5 with zeros p0=0.47"
       ],
       "values": [
        0.5516,
        0.5,
        null,
        0.47,
        0,
        55
       ]
      },
      {
       "id": "r30",
       "label": [
        "lognormal sigma=0.5 with zeros p0=0.6"
       ],
       "values": [
        0.6227,
        0.6,
        null,
        0.6,
        0,
        41
       ]
      },
      {
       "id": "r31",
       "label": [
        "lognormal sigma=1 with zeros p0=0.25"
       ],
       "values": [
        0.6881,
        0.5,
        null,
        0.625,
        0,
        77
       ]
      },
      {
       "id": "r32",
       "label": [
        "lognormal sigma=1 with zeros p0=0.37"
       ],
       "values": [
        0.6945,
        0.5,
        null,
        0.685,
        0,
        65
       ]
      },
      {
       "id": "r33",
       "label": [
        "lognormal sigma=1 with zeros p0=0.42"
       ],
       "values": [
        0.6997,
        0.5,
        null,
        0.42,
        0,
        59
       ]
      },
      {
       "id": "r34",
       "label": [
        "lognormal sigma=1 with zeros p0=0.6"
       ],
       "values": [
        0.7354,
        0.6,
        null,
        0.6,
        0,
        41
       ]
      }
     ],
     "artefact": "A8"
    },
    {
     "id": "T4",
     "description": "The candidates a measure ranks first over test sets, for four of the distributions at test-set sizes 6, 100 and 1000: the mean and the 5th, 50th and 95th percentiles of their quantile levels, the share of test sets choosing the preferred candidate, and (MAPE) the test sets in which every candidate's score is infinite; every distribution and size is in the run's CSV and its figure",
     "row_labels": [
      "setting",
      "measure",
      "test-set size"
     ],
     "columns": [
      "mean quantile level of the chosen candidates",
      {
       "variable": "percentile of the chosen candidates' quantile levels",
       "conditions": {
        "percentile": 5
       }
      },
      {
       "variable": "percentile of the chosen candidates' quantile levels",
       "conditions": {
        "percentile": 50
       }
      },
      {
       "variable": "percentile of the chosen candidates' quantile levels",
       "conditions": {
        "percentile": 95
       }
      },
      "share of test sets choosing the preferred candidate",
      "test sets in which every candidate's score is infinite"
     ],
     "rows": [
      {
       "id": "r1",
       "label": [
        "lognormal sigma=1",
        "RMSE",
        6
       ],
       "values": [
        0.6287,
        0.3495,
        0.64,
        0.88,
        0.012,
        null
       ]
      },
      {
       "id": "r2",
       "label": [
        "lognormal sigma=1",
        "RMSE",
        100
       ],
       "values": [
        0.6868,
        0.61,
        0.68,
        0.77,
        0.041,
        null
       ]
      },
      {
       "id": "r3",
       "label": [
        "lognormal sigma=1",
        "RMSE",
        1000
       ],
       "values": [
        0.6911,
        0.67,
        0.6915,
        0.7105,
        0.14,
        null
       ]
      },
      {
       "id": "r4",
       "label": [
        "lognormal sigma=1",
        "MAE",
        6
       ],
       "values": [
        0.4987,
        0.2264,
        0.505,
        0.7566,
        0.02217,
        null
       ]
      },
      {
       "id": "r5",
       "label": [
        "lognormal sigma=1",
        "MAE",
        100
       ],
       "values": [
        0.4987,
        0.42,
        0.5,
        0.58,
        0.07333,
        null
       ]
      },
      {
       "id": "r6",
       "label": [
        "lognormal sigma=1",
        "MAE",
        1000
       ],
       "values": [
        0.5002,
        0.47,
        0.5,
        0.53,
        0.256,
        null
       ]
      },
      {
       "id": "r7",
       "label": [
        "lognormal sigma=1",
        "MAPE",
        6
       ],
       "values": [
        0.237,
        0.0195,
        0.21,
        0.5805,
        0.006,
        null
       ]
      },
      {
       "id": "r8",
       "label": [
        "lognormal sigma=1",
        "MAPE",
        100
       ],
       "values": [
        0.1614,
        0.09,
        0.1587,
        0.24,
        0.041,
        null
       ]
      },
      {
       "id": "r9",
       "label": [
        "lognormal sigma=1",
        "MAPE",
        1000
       ],
       "values": [
        0.1594,
        0.13,
        0.16,
        0.19,
        0.123,
        null
       ]
      },
      {
       "id": "r10",
       "label": [
        "lognormal sigma=1",
        "sMAPE",
        6
       ],
       "values": [
        0.4988,
        0.16,
        0.51,
        0.82,
        0.007,
        null
       ]
      },
      {
       "id": "r11",
       "label": [
        "lognormal sigma=1",
        "sMAPE",
        100
       ],
       "values": [
        0.4991,
        0.4,
        0.5,
        0.6,
        0.063,
        null
       ]
      },
      {
       "id": "r12",
       "label": [
        "lognormal sigma=1",
        "sMAPE",
        1000
       ],
       "values": [
        0.5003,
        0.47,
        0.5,
        0.53,
        0.201,
        null
       ]
      },
      {
       "id": "r13",
       "label": [
        "inverse Gaussian shape=10",
        "RMSE",
        6
       ],
       "values": [
        0.7076,
        0.4,
        0.73,
        0.94,
        0.014,
        null
       ]
      },
      {
       "id": "r14",
       "label": [
        "inverse Gaussian shape=10",
        "RMSE",
        100
       ],
       "values": [
        0.8129,
        0.74,
        0.82,
        0.87,
        0.053,
        null
       ]
      },
      {
       "id": "r15",
       "label": [
        "inverse Gaussian shape=10",
        "RMSE",
        1000
       ],
       "values": [
        0.8203,
        0.8,
        0.8219,
        0.84,
        0.175,
        null
       ]
      },
      {
       "id": "r16",
       "label": [
        "inverse Gaussian shape=10",
        "MAE",
        6
       ],
       "values": [
        0.49,
        0.2355,
        0.48,
        0.7677,
        0.01921,
        null
       ]
      },
      {
       "id": "r17",
       "label": [
        "inverse Gaussian shape=10",
        "MAE",
        100
       ],
       "values": [
        0.4979,
        0.415,
        0.5,
        0.58,
        0.08996,
        null
       ]
      },
      {
       "id": "r18",
       "label": [
        "inverse Gaussian shape=10",
        "MAE",
        1000
       ],
       "values": [
        0.5,
        0.47,
        0.5,
        0.52,
        0.25,
        null
       ]
      },
      {
       "id": "r19",
       "label": [
        "inverse Gaussian shape=10",
        "MAPE",
        6
       ],
       "values": [
        0.1991,
        0.01,
        0.17,
        0.49,
        0.008,
        null
       ]
      },
      {
       "id": "r20",
       "label": [
        "inverse Gaussian shape=10",
        "MAPE",
        100
       ],
       "values": [
        0.1397,
        0.08,
        0.136,
        0.2,
        0.049,
        null
       ]
      },
      {
       "id": "r21",
       "label": [
        "inverse Gaussian shape=10",
        "MAPE",
        1000
       ],
       "values": [
        0.1369,
        0.12,
        0.136,
        0.16,
        0.157,
        null
       ]
      },
      {
       "id": "r22",
       "label": [
        "inverse Gaussian shape=10",
        "sMAPE",
        6
       ],
       "values": [
        0.4576,
        0.1398,
        0.4387,
        0.8305,
        0.008,
        null
       ]
      },
      {
       "id": "r23",
       "label": [
        "inverse Gaussian shape=10",
        "sMAPE",
        100
       ],
       "values": [
        0.4392,
        0.32,
        0.4387,
        0.56,
        0.025,
        null
       ]
      },
      {
       "id": "r24",
       "label": [
        "inverse Gaussian shape=10",
        "sMAPE",
        1000
       ],
       "values": [
        0.439,
        0.4,
        0.4387,
        0.48,
        0.084,
        null
       ]
      },
      {
       "id": "r25",
       "label": [
        "bimodal equal modes",
        "RMSE",
        6
       ],
       "values": [
        0.5355,
        0.49,
        0.5081,
        0.69,
        0.223,
        null
       ]
      },
      {
       "id": "r26",
       "label": [
        "bimodal equal modes",
        "RMSE",
        100
       ],
       "values": [
        0.5121,
        0.5081,
        0.5081,
        0.53,
        0.53,
        null
       ]
      },
      {
       "id": "r27",
       "label": [
        "bimodal equal modes",
        "RMSE",
        1000
       ],
       "values": [
        0.5088,
        0.5081,
        0.5081,
        0.51,
        0.65,
        null
       ]
      },
      {
       "id": "r28",
       "label": [
        "bimodal equal modes",
        "MAE",
        6
       ],
       "values": [
        0.4963,
        0.226,
        0.4967,
        0.76,
        0.01697,
        null
       ]
      },
      {
       "id": "r29",
       "label": [
        "bimodal equal modes",
        "MAE",
        100
       ],
       "values": [
        0.5012,
        0.42,
        0.506,
        0.5852,
        0.04309,
        null
       ]
      },
      {
       "id": "r30",
       "label": [
        "bimodal equal modes",
        "MAE",
        1000
       ],
       "values": [
        0.4994,
        0.47,
        0.5,
        0.53,
        0.06,
        null
       ]
      },
      {
       "id": "r31",
       "label": [
        "bimodal equal modes",
        "MAPE",
        6
       ],
       "values": [
        0.255,
        0.04,
        0.25,
        0.47,
        0.011,
        null
       ]
      },
      {
       "id": "r32",
       "label": [
        "bimodal equal modes",
        "MAPE",
        100
       ],
       "values": [
        0.2342,
        0.18,
        0.2341,
        0.29,
        0.056,
        null
       ]
      },
      {
       "id": "r33",
       "label": [
        "bimodal equal modes",
        "MAPE",
        1000
       ],
       "values": [
        0.2344,
        0.21,
        0.2341,
        0.25,
        0.164,
        null
       ]
      },
      {
       "id": "r34",
       "label": [
        "bimodal equal modes",
        "sMAPE",
        6
       ],
       "values": [
        0.4952,
        0.14,
        0.47,
        0.84,
        0.011,
        null
       ]
      },
      {
       "id": "r35",
       "label": [
        "bimodal equal modes",
        "sMAPE",
        100
       ],
       "values": [
        0.5065,
        0.31,
        0.585,
        0.7,
        0.024,
        null
       ]
      },
      {
       "id": "r36",
       "label": [
        "bimodal equal modes",
        "sMAPE",
        1000
       ],
       "values": [
        0.4998,
        0.35,
        0.6,
        0.65,
        0.077,
        null
       ]
      },
      {
       "id": "r37",
       "label": [
        "lognormal sigma=1 with zeros p0=0.42",
        "RMSE",
        6
       ],
       "values": [
        0.6545,
        0.44,
        0.65,
        0.89,
        0.013,
        null
       ]
      },
      {
       "id": "r38",
       "label": [
        "lognormal sigma=1 with zeros p0=0.42",
        "RMSE",
        100
       ],
       "values": [
        0.6942,
        0.62,
        0.69,
        0.77,
        0.052,
        null
       ]
      },
      {
       "id": "r39",
       "label": [
        "lognormal sigma=1 with zeros p0=0.42",
        "RMSE",
        1000
       ],
       "values": [
        0.699,
        0.68,
        0.6997,
        0.72,
        0.124,
        null
       ]
      },
      {
       "id": "r40",
       "label": [
        "lognormal sigma=1 with zeros p0=0.42",
        "MAE",
        6
       ],
       "values": [
        0.5375,
        0.42,
        0.51,
        0.7502,
        0.02274,
        null
       ]
      },
      {
       "id": "r41",
       "label": [
        "lognormal sigma=1 with zeros p0=0.42",
        "MAE",
        100
       ],
       "values": [
        0.4988,
        0.42,
        0.495,
        0.585,
        0.07538,
        null
       ]
      },
      {
       "id": "r42",
       "label": [
        "lognormal sigma=1 with zeros p0=0.42",
        "MAE",
        1000
       ],
       "values": [
        0.4995,
        0.47,
        0.5,
        0.53,
        0.248,
        null
       ]
      },
      {
       "id": "r43",
       "label": [
        "lognormal sigma=1 with zeros p0=0.42",
        "MAPE",
        6
       ],
       "values": [
        0.5524,
        0.43,
        0.56,
        0.688,
        null,
        963
       ]
      },
      {
       "id": "r44",
       "label": [
        "lognormal sigma=1 with zeros p0=0.42",
        "MAPE",
        100
       ],
       "values": [
        null,
        null,
        null,
        null,
        null,
        1000
       ]
      },
      {
       "id": "r45",
       "label": [
        "lognormal sigma=1 with zeros p0=0.42",
        "MAPE",
        1000
       ],
       "values": [
        null,
        null,
        null,
        null,
        null,
        1000
       ]
      },
      {
       "id": "r46",
       "label": [
        "lognormal sigma=1 with zeros p0=0.42",
        "sMAPE",
        6
       ],
       "values": [
        0.5533,
        0.42,
        0.42,
        0.89,
        0.539,
        null
       ]
      },
      {
       "id": "r47",
       "label": [
        "lognormal sigma=1 with zeros p0=0.42",
        "sMAPE",
        100
       ],
       "values": [
        0.5086,
        0.42,
        0.42,
        0.75,
        0.695,
        null
       ]
      },
      {
       "id": "r48",
       "label": [
        "lognormal sigma=1 with zeros p0=0.42",
        "sMAPE",
        1000
       ],
       "values": [
        0.4346,
        0.42,
        0.42,
        0.433,
        0.95,
        null
       ]
      }
     ],
     "artefact": "A8"
    },
    {
     "id": "T5",
     "description": "The MAPE target derived from each distribution (a median of the actuals weighted by 1/y), where it is finite",
     "row_labels": [
      "setting"
     ],
     "columns": [
      "derived MAPE target"
     ],
     "rows": [
      {
       "id": "r1",
       "label": [
        "inverse Gaussian shape=1000"
       ],
       "values": [
        86.51
       ]
      },
      {
       "id": "r2",
       "label": [
        "inverse Gaussian shape=200"
       ],
       "values": [
        53.48
       ]
      },
      {
       "id": "r3",
       "label": [
        "inverse Gaussian shape=100"
       ],
       "values": [
        34.69
       ]
      },
      {
       "id": "r4",
       "label": [
        "inverse Gaussian shape=50"
       ],
       "values": [
        19.7
       ]
      },
      {
       "id": "r5",
       "label": [
        "inverse Gaussian shape=20"
       ],
       "values": [
        8.334
       ]
      },
      {
       "id": "r6",
       "label": [
        "inverse Gaussian shape=10"
       ],
       "values": [
        4.21
       ]
      },
      {
       "id": "r7",
       "label": [
        "log-skewed left sd=0.5 skew=-2"
       ],
       "values": [
        82.43
       ]
      },
      {
       "id": "r8",
       "label": [
        "log-skewed left sd=0.5 skew=-1"
       ],
       "values": [
        79.93
       ]
      },
      {
       "id": "r9",
       "label": [
        "log-skewed left sd=0.5 skew=-0.5"
       ],
       "values": [
        78.8
       ]
      },
      {
       "id": "r10",
       "label": [
        "log-skewed right sd=0.5 skew=+2"
       ],
       "values": [
        76.42
       ]
      },
      {
       "id": "r11",
       "label": [
        "log-skewed right sd=0.5 skew=+1"
       ],
       "values": [
        76.67
       ]
      },
      {
       "id": "r12",
       "label": [
        "log-skewed right sd=0.5 skew=+0.5"
       ],
       "values": [
        77.17
       ]
      },
      {
       "id": "r13",
       "label": [
        "lognormal sigma=0.5"
       ],
       "values": [
        77.88
       ]
      },
      {
       "id": "r14",
       "label": [
        "log-skewed left sd=1 skew=-2"
       ],
       "values": [
        0
       ]
      },
      {
       "id": "r15",
       "label": [
        "log-skewed left sd=1 skew=-1"
       ],
       "values": [
        18.79
       ]
      },
      {
       "id": "r16",
       "label": [
        "log-skewed left sd=1 skew=-0.5"
       ],
       "values": [
        29.44
       ]
      },
      {
       "id": "r17",
       "label": [
        "log-skewed right sd=1 skew=+2"
       ],
       "values": [
        52.02
       ]
      },
      {
       "id": "r18",
       "label": [
        "log-skewed right sd=1 skew=+1"
       ],
       "values": [
        46.02
       ]
      },
      {
       "id": "r19",
       "label": [
        "log-skewed right sd=1 skew=+0.5"
       ],
       "values": [
        42.04
       ]
      },
      {
       "id": "r20",
       "label": [
        "lognormal sigma=1"
       ],
       "values": [
        36.79
       ]
      },
      {
       "id": "r21",
       "label": [
        "bimodal equal modes"
       ],
       "values": [
        36.06
       ]
      },
      {
       "id": "r22",
       "label": [
        "bimodal left mode higher"
       ],
       "values": [
        35.36
       ]
      },
      {
       "id": "r23",
       "label": [
        "bimodal right mode higher"
       ],
       "values": [
        37.44
       ]
      },
      {
       "id": "r24",
       "label": [
        "normal straddling zero mean=0"
       ],
       "values": [
        0
       ]
      },
      {
       "id": "r25",
       "label": [
        "normal straddling zero mean=0.5"
       ],
       "values": [
        0
       ]
      },
      {
       "id": "r26",
       "label": [
        "normal straddling zero mean=1"
       ],
       "values": [
        0
       ]
      }
     ],
     "artefact": "A8"
    },
    {
     "id": "T6",
     "description": "The value of each measure's preferred candidate, for the four distributions of T4",
     "row_labels": [
      "setting",
      "measure"
     ],
     "columns": [
      "value of the preferred candidate"
     ],
     "rows": [
      {
       "id": "r1",
       "label": [
        "lognormal sigma=1",
        "RMSE"
       ],
       "values": [
        164.9
       ]
      },
      {
       "id": "r2",
       "label": [
        "lognormal sigma=1",
        "MAE"
       ],
       "values": [
        100.0
       ]
      },
      {
       "id": "r3",
       "label": [
        "lognormal sigma=1",
        "MAPE"
       ],
       "values": [
        36.79
       ]
      },
      {
       "id": "r4",
       "label": [
        "lognormal sigma=1",
        "sMAPE"
       ],
       "values": [
        100.0
       ]
      },
      {
       "id": "r5",
       "label": [
        "inverse Gaussian shape=10",
        "RMSE"
       ],
       "values": [
        100
       ]
      },
      {
       "id": "r6",
       "label": [
        "inverse Gaussian shape=10",
        "MAE"
       ],
       "values": [
        17.85
       ]
      },
      {
       "id": "r7",
       "label": [
        "inverse Gaussian shape=10",
        "MAPE"
       ],
       "values": [
        4.21
       ]
      },
      {
       "id": "r8",
       "label": [
        "inverse Gaussian shape=10",
        "sMAPE"
       ],
       "values": [
        14.02
       ]
      },
      {
       "id": "r9",
       "label": [
        "bimodal equal modes",
        "RMSE"
       ],
       "values": [
        159.2
       ]
      },
      {
       "id": "r10",
       "label": [
        "bimodal equal modes",
        "MAE"
       ],
       "values": [
        100.0
       ]
      },
      {
       "id": "r11",
       "label": [
        "bimodal equal modes",
        "MAPE"
       ],
       "values": [
        36.06
       ]
      },
      {
       "id": "r12",
       "label": [
        "bimodal equal modes",
        "sMAPE"
       ],
       "values": [
        43.49
       ]
      },
      {
       "id": "r13",
       "label": [
        "lognormal sigma=1 with zeros p0=0.42",
        "RMSE"
       ],
       "values": [
        95.63
       ]
      },
      {
       "id": "r14",
       "label": [
        "lognormal sigma=1 with zeros p0=0.42",
        "MAE"
       ],
       "values": [
        33.63
       ]
      },
      {
       "id": "r15",
       "label": [
        "lognormal sigma=1 with zeros p0=0.42",
        "sMAPE"
       ],
       "values": [
        0
       ]
      }
     ],
     "artefact": "A8"
    }
   ],
   "artefacts": [
    "A6",
    "A8",
    "A9",
    "A10",
    "A11",
    "A12",
    "A13",
    "A14",
    "A15",
    "A16",
    "A17",
    "A18",
    "A19",
    "A20",
    "A21",
    "A22",
    "A23",
    "A24",
    "A25",
    "A26",
    "A27",
    "A28",
    "A29",
    "A30",
    "A31",
    "A32",
    "A33",
    "A34",
    "A35",
    "A36",
    "A37",
    "A38",
    "A39",
    "A40",
    "A41",
    "A42"
   ],
   "setup": {
    "studied": [
     "D10",
     "D11",
     "D12",
     "D2"
    ],
    "cases": [
     "D21",
     "D22",
     "D23",
     "D24",
     "D25"
    ],
    "procedure": [
     "D16",
     "D15"
    ],
    "lens": [
     "D13",
     "D14",
     "D6",
     "D3",
     "D5",
     "D7"
    ]
   },
   "execution": {
    "code_version": "3cb6c7c6045fcbf40b58d8c0ff77a13f704fdc99",
    "software": {
     "python": "3.13.18",
     "platform": "Linux-6.18.44-fc-v470-x86_64-with-glibc2.39"
    }
   }
  },
  {
   "id": "E14",
   "profile": "quant",
   "design": "other",
   "description": "Lab runs RN2 under plan PL4. Design: Flatness design (D18) on 30 distributions (D27): lognormal, log-Gumbel right and log-Gumbel left, with sigma 0.1, 0.25, 0.5, 1, 1.5, 2, 3, 4, 6, 10. Derived (D15) from each: the quantile levels of the sMAPE target and of the log-mode; expected sMAPE at the target and the approximation of H14 there; the number of local minima; for each measure (RMSE, MAE, MAPE, sMAPE) and forecasts e^(k sigma) times its own target (k = -1, -0.5, +0.5, +1: half and one standard deviation of ln y away), the excess expected score over the target's, relative to the target's, where both are finite, and sMAPE's absolute excess; expected sMAPE at quantile levels 0.01 to 0.99. On test sets (10000 per case, seed 5 with the setting's index): the ranking agreement (D20) of each measure between its target and those four forecasts, on the lognormal family (on which every measure has a target), at n = 12, 48, 250 and 1000; the quantile level of the sMAPE minimiser of a test set at n = 48 and 1000 on every distribution. python3 claimset-code/smape-target/flatness.py --seed 5 --reps 10000 (numpy 2.5.3, matplotlib 3.11.2). Analysis: The derived values listed in the design, with 'none' and its reason where a measure targets nothing or its expected score is infinite; the ranking agreement with its 99% Wilson interval; the mean quantile level of the test sets' sMAPE minimisers with a 99% confidence interval and its 5th, 25th, 50th, 75th and 95th percentiles. Figures: the target's quantile level over sigma beside the log-mode's and the median; expected sMAPE over the quantile level of the forecast, a line per sigma; the relative excess over sigma by measure; ranking agreement over sigma by measure and n; the spread of the test sets' sMAPE minimisers over sigma. Compared with H14 (i) and (ii), and with the other measures; no test is made and no threshold is set.",
   "availability": {
    "level": "public"
   },
   "lab_runs": [
    "RN2"
   ],
   "results": [],
   "tables": [
    {
     "id": "T1",
     "description": "The sMAPE target as the spread of ln y grows: its quantile level beside the log-mode's, the expected sMAPE there and its large-spread approximation",
     "row_labels": [
      "setting"
     ],
     "columns": [
      "quantile level of the sMAPE target",
      "quantile level of the log-mode",
      "expected sMAPE at the target",
      "large-spread approximation of expected sMAPE at the target",
      "local minima of expected sMAPE"
     ],
     "rows": [
      {
       "id": "r1",
       "label": [
        "lognormal sigma=0.1"
       ],
       "values": [
        0.5,
        0.5,
        7.966,
        -2012.0,
        1
       ]
      },
      {
       "id": "r2",
       "label": [
        "log-Gumbel right sigma=0.1"
       ],
       "values": [
        0.4989,
        0.3679,
        9.643,
        -1724.0,
        1
       ]
      },
      {
       "id": "r3",
       "label": [
        "log-Gumbel left sigma=0.1"
       ],
       "values": [
        0.5011,
        0.6321,
        9.643,
        -1724.0,
        1
       ]
      },
      {
       "id": "r4",
       "label": [
        "lognormal sigma=0.25"
       ],
       "values": [
        0.5,
        0.5,
        19.74,
        -684.9,
        1
       ]
      },
      {
       "id": "r5",
       "label": [
        "log-Gumbel right sigma=0.25"
       ],
       "values": [
        0.494,
        0.3679,
        23.66,
        -572.7,
        1
       ]
      },
      {
       "id": "r6",
       "label": [
        "log-Gumbel left sigma=0.25"
       ],
       "values": [
        0.506,
        0.6321,
        23.66,
        -572.7,
        1
       ]
      },
      {
       "id": "r7",
       "label": [
        "lognormal sigma=0.5"
       ],
       "values": [
        0.5,
        0.5,
        38.38,
        -242.4,
        1
       ]
      },
      {
       "id": "r8",
       "label": [
        "log-Gumbel right sigma=0.5"
       ],
       "values": [
        0.4811,
        0.3679,
        44.85,
        -190.4,
        1
       ]
      },
      {
       "id": "r9",
       "label": [
        "log-Gumbel left sigma=0.5"
       ],
       "values": [
        0.5189,
        0.6321,
        44.85,
        -190.4,
        1
       ]
      },
      {
       "id": "r10",
       "label": [
        "lognormal sigma=1"
       ],
       "values": [
        0.5,
        0.5,
        69.94,
        -21.22,
        1
       ]
      },
      {
       "id": "r11",
       "label": [
        "log-Gumbel right sigma=1"
       ],
       "values": [
        0.4544,
        0.3679,
        77.99,
        1.246,
        1
       ]
      },
      {
       "id": "r12",
       "label": [
        "log-Gumbel left sigma=1"
       ],
       "values": [
        0.5456,
        0.6321,
        77.99,
        1.246,
        1
       ]
      },
      {
       "id": "r13",
       "label": [
        "lognormal sigma=1.5"
       ],
       "values": [
        0.5,
        0.5,
        93.66,
        52.52,
        1
       ]
      },
      {
       "id": "r14",
       "label": [
        "log-Gumbel right sigma=1.5"
       ],
       "values": [
        0.4337,
        0.3679,
        101.2,
        66.06,
        1
       ]
      },
      {
       "id": "r15",
       "label": [
        "log-Gumbel left sigma=1.5"
       ],
       "values": [
        0.5663,
        0.6321,
        101.2,
        66.06,
        1
       ]
      },
      {
       "id": "r16",
       "label": [
        "lognormal sigma=2"
       ],
       "values": [
        0.5,
        0.5,
        111.2,
        89.39,
        1
       ]
      },
      {
       "id": "r17",
       "label": [
        "log-Gumbel right sigma=2"
       ],
       "values": [
        0.4187,
        0.3679,
        117.9,
        98.93,
        1
       ]
      },
      {
       "id": "r18",
       "label": [
        "log-Gumbel left sigma=2"
       ],
       "values": [
        0.5813,
        0.6321,
        117.9,
        98.93,
        1
       ]
      },
      {
       "id": "r19",
       "label": [
        "lognormal sigma=3"
       ],
       "values": [
        0.5,
        0.5,
        134.4,
        126.3,
        1
       ]
      },
      {
       "id": "r20",
       "label": [
        "log-Gumbel right sigma=3"
       ],
       "values": [
        0.3999,
        0.3679,
        139.5,
        132.3,
        1
       ]
      },
      {
       "id": "r21",
       "label": [
        "log-Gumbel left sigma=3"
       ],
       "values": [
        0.6001,
        0.6321,
        139.5,
        132.3,
        1
       ]
      },
      {
       "id": "r22",
       "label": [
        "lognormal sigma=4"
       ],
       "values": [
        0.5,
        0.5,
        148.5,
        144.7,
        1
       ]
      },
      {
       "id": "r23",
       "label": [
        "log-Gumbel right sigma=4"
       ],
       "values": [
        0.3894,
        0.3679,
        152.5,
        149.1,
        1
       ]
      },
      {
       "id": "r24",
       "label": [
        "log-Gumbel left sigma=4"
       ],
       "values": [
        0.6106,
        0.6321,
        152.5,
        149.1,
        1
       ]
      },
      {
       "id": "r25",
       "label": [
        "lognormal sigma=6"
       ],
       "values": [
        0.5,
        0.5,
        164.4,
        163.1,
        1
       ]
      },
      {
       "id": "r26",
       "label": [
        "log-Gumbel right sigma=6"
       ],
       "values": [
        0.3792,
        0.3679,
        167.1,
        166.0,
        1
       ]
      },
      {
       "id": "r27",
       "label": [
        "log-Gumbel left sigma=6"
       ],
       "values": [
        0.6208,
        0.6321,
        167.1,
        166.0,
        1
       ]
      },
      {
       "id": "r28",
       "label": [
        "lognormal sigma=10"
       ],
       "values": [
        0.5,
        0.5,
        178.2,
        177.9,
        1
       ]
      },
      {
       "id": "r29",
       "label": [
        "log-Gumbel right sigma=10"
       ],
       "values": [
        0.3724,
        0.3679,
        179.9,
        179.6,
        1
       ]
      },
      {
       "id": "r30",
       "label": [
        "log-Gumbel left sigma=10"
       ],
       "values": [
        0.6276,
        0.6321,
        179.9,
        179.6,
        1
       ]
      }
     ],
     "artefact": "A43"
    },
    {
     "id": "T2",
     "description": "Ranking agreement: the share of test sets that give a measure's target a lower score than a forecast half a sigma above it, for three of the settings; every setting, offset and size is in the run's CSV and its figure",
     "row_labels": [
      "setting",
      "test-set size"
     ],
     "columns": [
      {
       "variable": "D20",
       "conditions": {
        "measure": "RMSE",
        "offset in sigma": 0.5
       }
      },
      {
       "variable": "D20",
       "conditions": {
        "measure": "MAE",
        "offset in sigma": 0.5
       }
      },
      {
       "variable": "D20",
       "conditions": {
        "measure": "MAPE",
        "offset in sigma": 0.5
       }
      },
      {
       "variable": "D20",
       "conditions": {
        "measure": "sMAPE",
        "offset in sigma": 0.5
       }
      }
     ],
     "rows": [
      {
       "id": "r1",
       "label": [
        "lognormal sigma=0.5",
        12
       ],
       "values": [
        0.8282,
        0.7772,
        0.7396,
        0.7557
       ]
      },
      {
       "id": "r2",
       "label": [
        "lognormal sigma=0.5",
        48
       ],
       "values": [
        0.9659,
        0.9421,
        0.9118,
        0.9233
       ]
      },
      {
       "id": "r3",
       "label": [
        "lognormal sigma=0.5",
        250
       ],
       "values": [
        0.9998,
        0.9999,
        0.9994,
        0.9994
       ]
      },
      {
       "id": "r4",
       "label": [
        "lognormal sigma=0.5",
        1000
       ],
       "values": [
        1,
        1,
        1,
        1
       ]
      },
      {
       "id": "r5",
       "label": [
        "lognormal sigma=4",
        12
       ],
       "values": [
        0.9826,
        0.8226,
        0.0012,
        0.605
       ]
      },
      {
       "id": "r6",
       "label": [
        "lognormal sigma=4",
        48
       ],
       "values": [
        0.9792,
        0.9705,
        0.0059,
        0.6989
       ]
      },
      {
       "id": "r7",
       "label": [
        "lognormal sigma=4",
        250
       ],
       "values": [
        0.9759,
        1,
        0.0337,
        0.8823
       ]
      },
      {
       "id": "r8",
       "label": [
        "lognormal sigma=4",
        1000
       ],
       "values": [
        0.9771,
        1,
        0.0928,
        0.9909
       ]
      }
     ],
     "artefact": "A43"
    },
    {
     "id": "T3",
     "description": "How much worse a forecast half a sigma or one sigma below or above the target scores, as a share of the score at the target, for lognormal outcomes",
     "row_labels": [
      "setting",
      "offset in sigma"
     ],
     "columns": [
      {
       "variable": "excess expected score over the target, relative",
       "conditions": {
        "measure": "RMSE"
       }
      },
      {
       "variable": "excess expected score over the target, relative",
       "conditions": {
        "measure": "MAE"
       }
      },
      {
       "variable": "excess expected score over the target, relative",
       "conditions": {
        "measure": "MAPE"
       }
      },
      {
       "variable": "excess expected score over the target, relative",
       "conditions": {
        "measure": "sMAPE"
       }
      },
      {
       "variable": "excess expected sMAPE over the target",
       "conditions": {
        "measure": "sMAPE"
       }
      }
     ],
     "rows": [
      {
       "id": "r1",
       "label": [
        "lognormal sigma=0.5",
        -1.0
       ],
       "values": [
        0.243,
        0.3083,
        0.3083,
        0.4212,
        16.17
       ]
      },
      {
       "id": "r2",
       "label": [
        "lognormal sigma=0.5",
        -0.5
       ],
       "values": [
        0.08271,
        0.09554,
        0.09555,
        0.1131,
        4.34
       ]
      },
      {
       "id": "r3",
       "label": [
        "lognormal sigma=0.5",
        0.5
       ],
       "values": [
        0.1331,
        0.1331,
        0.1331,
        0.1131,
        4.34
       ]
      },
      {
       "id": "r4",
       "label": [
        "lognormal sigma=0.5",
        1.0
       ],
       "values": [
        0.5753,
        0.594,
        0.594,
        0.4212,
        16.17
       ]
      },
      {
       "id": "r5",
       "label": [
        "lognormal sigma=2",
        -1.0
       ],
       "values": [
        0.00695,
        0.03174,
        0.03174,
        0.2279,
        25.34
       ]
      },
      {
       "id": "r6",
       "label": [
        "lognormal sigma=2",
        -0.5
       ],
       "values": [
        0.003721,
        0.01468,
        0.01469,
        0.06482,
        7.207
       ]
      },
      {
       "id": "r7",
       "label": [
        "lognormal sigma=2",
        0.5
       ],
       "values": [
        0.02717,
        0.05527,
        0.05527,
        0.06482,
        7.207
       ]
      },
      {
       "id": "r8",
       "label": [
        "lognormal sigma=2",
        1.0
       ],
       "values": [
        0.3273,
        0.4305,
        0.4304,
        0.2279,
        25.34
       ]
      },
      {
       "id": "r9",
       "label": [
        "lognormal sigma=6",
        -1.0
       ],
       "values": [
        0,
        1.945e-09,
        1.986e-09,
        0.08104,
        13.32
       ]
      },
      {
       "id": "r10",
       "label": [
        "lognormal sigma=6",
        -0.5
       ],
       "values": [
        0,
        1.603e-09,
        1.639e-09,
        0.02395,
        3.937
       ]
      },
      {
       "id": "r11",
       "label": [
        "lognormal sigma=6",
        0.5
       ],
       "values": [
        4.241e-14,
        8.113e-08,
        8.152e-08,
        0.02395,
        3.937
       ]
      },
      {
       "id": "r12",
       "label": [
        "lognormal sigma=6",
        1.0
       ],
       "values": [
        1.878e-11,
        3.623e-06,
        3.649e-06,
        0.08104,
        13.32
       ]
      }
     ],
     "artefact": "A43"
    },
    {
     "id": "T4",
     "description": "Where the sMAPE minimiser of a test set falls: the mean and the 5th, 50th and 95th percentiles of its quantile level over test sets",
     "row_labels": [
      "setting",
      "test-set size"
     ],
     "columns": [
      "mean quantile level of the sMAPE minimiser of a test set",
      {
       "variable": "percentile of the quantile level of the sMAPE minimiser of a test set",
       "conditions": {
        "percentile": 5
       }
      },
      {
       "variable": "percentile of the quantile level of the sMAPE minimiser of a test set",
       "conditions": {
        "percentile": 50
       }
      },
      {
       "variable": "percentile of the quantile level of the sMAPE minimiser of a test set",
       "conditions": {
        "percentile": 95
       }
      }
     ],
     "rows": [
      {
       "id": "r1",
       "label": [
        "lognormal sigma=0.1",
        48
       ],
       "values": [
        0.5004,
        0.3764,
        0.4998,
        0.6251
       ]
      },
      {
       "id": "r2",
       "label": [
        "lognormal sigma=0.1",
        1000
       ],
       "values": [
        0.5,
        0.4742,
        0.4997,
        0.5263
       ]
      },
      {
       "id": "r3",
       "label": [
        "log-Gumbel right sigma=0.1",
        48
       ],
       "values": [
        0.4896,
        0.3697,
        0.4893,
        0.6081
       ]
      },
      {
       "id": "r4",
       "label": [
        "log-Gumbel right sigma=0.1",
        1000
       ],
       "values": [
        0.4986,
        0.4724,
        0.4987,
        0.5245
       ]
      },
      {
       "id": "r5",
       "label": [
        "log-Gumbel left sigma=0.1",
        48
       ],
       "values": [
        0.5091,
        0.3897,
        0.5093,
        0.6263
       ]
      },
      {
       "id": "r6",
       "label": [
        "log-Gumbel left sigma=0.1",
        1000
       ],
       "values": [
        0.5012,
        0.4752,
        0.5013,
        0.5275
       ]
      },
      {
       "id": "r7",
       "label": [
        "lognormal sigma=0.25",
        48
       ],
       "values": [
        0.5,
        0.3762,
        0.499,
        0.6242
       ]
      },
      {
       "id": "r8",
       "label": [
        "lognormal sigma=0.25",
        1000
       ],
       "values": [
        0.4997,
        0.4729,
        0.4999,
        0.526
       ]
      },
      {
       "id": "r9",
       "label": [
        "log-Gumbel right sigma=0.25",
        48
       ],
       "values": [
        0.4897,
        0.3741,
        0.4892,
        0.6057
       ]
      },
      {
       "id": "r10",
       "label": [
        "log-Gumbel right sigma=0.25",
        1000
       ],
       "values": [
        0.4942,
        0.4682,
        0.4941,
        0.5206
       ]
      },
      {
       "id": "r11",
       "label": [
        "log-Gumbel left sigma=0.25",
        48
       ],
       "values": [
        0.5116,
        0.3929,
        0.5116,
        0.6293
       ]
      },
      {
       "id": "r12",
       "label": [
        "log-Gumbel left sigma=0.25",
        1000
       ],
       "values": [
        0.5062,
        0.4801,
        0.5064,
        0.5321
       ]
      },
      {
       "id": "r13",
       "label": [
        "lognormal sigma=0.5",
        48
       ],
       "values": [
        0.5003,
        0.3776,
        0.5002,
        0.624
       ]
      },
      {
       "id": "r14",
       "label": [
        "lognormal sigma=0.5",
        1000
       ],
       "values": [
        0.4999,
        0.4728,
        0.4999,
        0.5274
       ]
      },
      {
       "id": "r15",
       "label": [
        "log-Gumbel right sigma=0.5",
        48
       ],
       "values": [
        0.483,
        0.3603,
        0.482,
        0.6085
       ]
      },
      {
       "id": "r16",
       "label": [
        "log-Gumbel right sigma=0.5",
        1000
       ],
       "values": [
        0.4809,
        0.4529,
        0.481,
        0.5087
       ]
      },
      {
       "id": "r17",
       "label": [
        "log-Gumbel left sigma=0.5",
        48
       ],
       "values": [
        0.5175,
        0.3935,
        0.518,
        0.6407
       ]
      },
      {
       "id": "r18",
       "label": [
        "log-Gumbel left sigma=0.5",
        1000
       ],
       "values": [
        0.5187,
        0.4908,
        0.5189,
        0.5467
       ]
      },
      {
       "id": "r19",
       "label": [
        "lognormal sigma=1",
        48
       ],
       "values": [
        0.4996,
        0.359,
        0.4995,
        0.6414
       ]
      },
      {
       "id": "r20",
       "label": [
        "lognormal sigma=1",
        1000
       ],
       "values": [
        0.5001,
        0.4681,
        0.5002,
        0.5316
       ]
      },
      {
       "id": "r21",
       "label": [
        "log-Gumbel right sigma=1",
        48
       ],
       "values": [
        0.4558,
        0.3164,
        0.4545,
        0.6006
       ]
      },
      {
       "id": "r22",
       "label": [
        "log-Gumbel right sigma=1",
        1000
       ],
       "values": [
        0.4546,
        0.4229,
        0.4544,
        0.4874
       ]
      },
      {
       "id": "r23",
       "label": [
        "log-Gumbel left sigma=1",
        48
       ],
       "values": [
        0.5438,
        0.3957,
        0.5461,
        0.684
       ]
      },
      {
       "id": "r24",
       "label": [
        "log-Gumbel left sigma=1",
        1000
       ],
       "values": [
        0.5458,
        0.5131,
        0.5458,
        0.5776
       ]
      },
      {
       "id": "r25",
       "label": [
        "lognormal sigma=1.5",
        48
       ],
       "values": [
        0.5002,
        0.3364,
        0.4996,
        0.6636
       ]
      },
      {
       "id": "r26",
       "label": [
        "lognormal sigma=1.5",
        1000
       ],
       "values": [
        0.4997,
        0.4613,
        0.4995,
        0.5376
       ]
      },
      {
       "id": "r27",
       "label": [
        "log-Gumbel right sigma=1.5",
        48
       ],
       "values": [
        0.4367,
        0.2755,
        0.4325,
        0.6074
       ]
      },
      {
       "id": "r28",
       "label": [
        "log-Gumbel right sigma=1.5",
        1000
       ],
       "values": [
        0.4341,
        0.395,
        0.434,
        0.4726
       ]
      },
      {
       "id": "r29",
       "label": [
        "log-Gumbel left sigma=1.5",
        48
       ],
       "values": [
        0.5617,
        0.3874,
        0.5637,
        0.7247
       ]
      },
      {
       "id": "r30",
       "label": [
        "log-Gumbel left sigma=1.5",
        1000
       ],
       "values": [
        0.566,
        0.5268,
        0.5666,
        0.604
       ]
      },
      {
       "id": "r31",
       "label": [
        "lognormal sigma=2",
        48
       ],
       "values": [
        0.5024,
        0.3106,
        0.5019,
        0.6923
       ]
      },
      {
       "id": "r32",
       "label": [
        "lognormal sigma=2",
        1000
       ],
       "values": [
        0.5001,
        0.4536,
        0.5001,
        0.5468
       ]
      },
      {
       "id": "r33",
       "label": [
        "log-Gumbel right sigma=2",
        48
       ],
       "values": [
        0.4254,
        0.25,
        0.4194,
        0.6192
       ]
      },
      {
       "id": "r34",
       "label": [
        "log-Gumbel right sigma=2",
        1000
       ],
       "values": [
        0.4191,
        0.3732,
        0.4189,
        0.4667
       ]
      },
      {
       "id": "r35",
       "label": [
        "log-Gumbel left sigma=2",
        48
       ],
       "values": [
        0.5719,
        0.3763,
        0.5772,
        0.751
       ]
      },
      {
       "id": "r36",
       "label": [
        "log-Gumbel left sigma=2",
        1000
       ],
       "values": [
        0.5808,
        0.5333,
        0.5815,
        0.6256
       ]
      },
      {
       "id": "r37",
       "label": [
        "lognormal sigma=3",
        48
       ],
       "values": [
        0.4977,
        0.2756,
        0.4992,
        0.7205
       ]
      },
      {
       "id": "r38",
       "label": [
        "lognormal sigma=3",
        1000
       ],
       "values": [
        0.4994,
        0.4346,
        0.4991,
        0.5664
       ]
      },
      {
       "id": "r39",
       "label": [
        "log-Gumbel right sigma=3",
        48
       ],
       "values": [
        0.4154,
        0.2121,
        0.4074,
        0.6461
       ]
      },
      {
       "id": "r40",
       "label": [
        "log-Gumbel right sigma=3",
        1000
       ],
       "values": [
        0.4013,
        0.3385,
        0.4005,
        0.4676
       ]
      },
      {
       "id": "r41",
       "label": [
        "log-Gumbel left sigma=3",
        48
       ],
       "values": [
        0.5845,
        0.3526,
        0.5917,
        0.7901
       ]
      },
      {
       "id": "r42",
       "label": [
        "log-Gumbel left sigma=3",
        1000
       ],
       "values": [
        0.5986,
        0.5307,
        0.5998,
        0.6617
       ]
      },
      {
       "id": "r43",
       "label": [
        "lognormal sigma=4",
        48
       ],
       "values": [
        0.4985,
        0.2581,
        0.4984,
        0.7413
       ]
      },
      {
       "id": "r44",
       "label": [
        "lognormal sigma=4",
        1000
       ],
       "values": [
        0.4998,
        0.4134,
        0.4999,
        0.5844
       ]
      },
      {
       "id": "r45",
       "label": [
        "log-Gumbel right sigma=4",
        48
       ],
       "values": [
        0.4088,
        0.1874,
        0.4003,
        0.6567
       ]
      },
      {
       "id": "r46",
       "label": [
        "log-Gumbel right sigma=4",
        1000
       ],
       "values": [
        0.3917,
        0.3132,
        0.3895,
        0.4769
       ]
      },
      {
       "id": "r47",
       "label": [
        "log-Gumbel left sigma=4",
        48
       ],
       "values": [
        0.5887,
        0.3387,
        0.5973,
        0.8117
       ]
      },
      {
       "id": "r48",
       "label": [
        "log-Gumbel left sigma=4",
        1000
       ],
       "values": [
        0.6082,
        0.5207,
        0.6106,
        0.6879
       ]
      },
      {
       "id": "r49",
       "label": [
        "lognormal sigma=6",
        48
       ],
       "values": [
        0.4987,
        0.2382,
        0.4976,
        0.7646
       ]
      },
      {
       "id": "r50",
       "label": [
        "lognormal sigma=6",
        1000
       ],
       "values": [
        0.5004,
        0.3866,
        0.5013,
        0.613
       ]
      },
      {
       "id": "r51",
       "label": [
        "log-Gumbel right sigma=6",
        48
       ],
       "values": [
        0.4083,
        0.1684,
        0.3992,
        0.6722
       ]
      },
      {
       "id": "r52",
       "label": [
        "log-Gumbel right sigma=6",
        1000
       ],
       "values": [
        0.3839,
        0.2797,
        0.3811,
        0.4981
       ]
      },
      {
       "id": "r53",
       "label": [
        "log-Gumbel left sigma=6",
        48
       ],
       "values": [
        0.5899,
        0.3167,
        0.5986,
        0.8314
       ]
      },
      {
       "id": "r54",
       "label": [
        "log-Gumbel left sigma=6",
        1000
       ],
       "values": [
        0.6157,
        0.5036,
        0.6187,
        0.7188
       ]
      },
      {
       "id": "r55",
       "label": [
        "lognormal sigma=10",
        48
       ],
       "values": [
        0.4999,
        0.2216,
        0.5004,
        0.7798
       ]
      },
      {
       "id": "r56",
       "label": [
        "lognormal sigma=10",
        1000
       ],
       "values": [
        0.5001,
        0.3594,
        0.4997,
        0.6419
       ]
      },
      {
       "id": "r57",
       "label": [
        "log-Gumbel right sigma=10",
        48
       ],
       "values": [
        0.4071,
        0.153,
        0.3956,
        0.698
       ]
      },
      {
       "id": "r58",
       "label": [
        "log-Gumbel right sigma=10",
        1000
       ],
       "values": [
        0.3805,
        0.2529,
        0.3767,
        0.5195
       ]
      },
      {
       "id": "r59",
       "label": [
        "log-Gumbel left sigma=10",
        48
       ],
       "values": [
        0.5915,
        0.304,
        0.5986,
        0.8482
       ]
      },
      {
       "id": "r60",
       "label": [
        "log-Gumbel left sigma=10",
        1000
       ],
       "values": [
        0.6188,
        0.4813,
        0.6227,
        0.7469
       ]
      }
     ],
     "artefact": "A43"
    }
   ],
   "artefacts": [
    "A6",
    "A43",
    "A44",
    "A45",
    "A46",
    "A47",
    "A48"
   ],
   "setup": {
    "studied": [
     "D10",
     "D11",
     "D12",
     "D2"
    ],
    "cases": [
     "D27"
    ],
    "procedure": [
     "D18",
     "D15"
    ],
    "lens": [
     "D3",
     "D6",
     "D20",
     "D8"
    ]
   },
   "execution": {
    "code_version": "3cb6c7c6045fcbf40b58d8c0ff77a13f704fdc99",
    "software": {
     "python": "3.13.18",
     "platform": "Linux-6.18.44-fc-v470-x86_64-with-glibc2.39"
    }
   }
  },
  {
   "id": "E15",
   "profile": "quant",
   "design": "other",
   "description": "Lab runs RN7 under plan PL2. Design: Efficiency design (D17) on 16 log-symmetric distributions (D26): lognormal sigma and log-Laplace b, each 0.25, 0.5, 1, 1.5, 2, 3, 4, 6. Test sets of n = 6, 8, 12, 13, 14, 18, 24, 36, 48, 100, 250, 500, 1000 and 10000 outcomes, 10000 test sets per size, seed 4 with the setting's index and n. In each test set: the sample median and the exact sMAPE minimiser (D8). From each distribution, derived (D15): the large-sample n Var of ln of each, from H11, and their ratio. python3 claimset-code/smape-target/efficiency.py --seed 4 --reps 10000 (numpy 2.5.3, matplotlib 3.11.2). Analysis: Per distribution and n, over test sets: the variances of ln(sample median) and ln(sMAPE minimiser); the relative efficiency (D9) with a 99% bootstrap percentile interval (2000 resamples of test sets); the ratio of their mean squared errors about the true median; the mean of ln(sMAPE minimiser) minus ln(median). Per distribution: the asymptotic relative efficiency. Figure: relative efficiency over n, a panel per distribution, with its band, the asymptotic value dashed and 1 dotted. Compared with H11 and H14 (iii); no test is made and no threshold is set.",
   "availability": {
    "level": "public"
   },
   "lab_runs": [
    "RN7"
   ],
   "results": [],
   "tables": [
    {
     "id": "T1",
     "description": "Relative efficiency of the sMAPE minimiser of a test set against the sample median, at test-set sizes 6, 12, 100, 1000 and 10000, beside its large-sample value",
     "row_labels": [
      "setting"
     ],
     "columns": [
      {
       "variable": "D9",
       "conditions": {
        "test-set size": 6
       }
      },
      {
       "variable": "D9",
       "conditions": {
        "test-set size": 12
       }
      },
      {
       "variable": "D9",
       "conditions": {
        "test-set size": 100
       }
      },
      {
       "variable": "D9",
       "conditions": {
        "test-set size": 1000
       }
      },
      {
       "variable": "D9",
       "conditions": {
        "test-set size": 10000
       }
      },
      "asymptotic relative efficiency"
     ],
     "rows": [
      {
       "id": "r1",
       "label": [
        "lognormal sigma=0.25"
       ],
       "values": [
        0.6059,
        0.7026,
        0.8996,
        0.9662,
        0.9701,
        0.9696
       ]
      },
      {
       "id": "r2",
       "label": [
        "lognormal sigma=0.5"
       ],
       "values": [
        0.611,
        0.7083,
        0.8802,
        0.885,
        0.8906,
        0.8881
       ]
      },
      {
       "id": "r3",
       "label": [
        "lognormal sigma=1"
       ],
       "values": [
        0.6067,
        0.6555,
        0.6524,
        0.6659,
        0.6565,
        0.6601
       ]
      },
      {
       "id": "r4",
       "label": [
        "lognormal sigma=1.5"
       ],
       "values": [
        0.5484,
        0.5123,
        0.4595,
        0.4633,
        0.45,
        0.454
       ]
      },
      {
       "id": "r5",
       "label": [
        "lognormal sigma=2"
       ],
       "values": [
        0.501,
        0.4392,
        0.323,
        0.3077,
        0.3043,
        0.3073
       ]
      },
      {
       "id": "r6",
       "label": [
        "lognormal sigma=3"
       ],
       "values": [
        0.4381,
        0.3646,
        0.2058,
        0.1557,
        0.1513,
        0.1474
       ]
      },
      {
       "id": "r7",
       "label": [
        "lognormal sigma=4"
       ],
       "values": [
        0.4148,
        0.3272,
        0.1561,
        0.08898,
        0.07892,
        0.07778
       ]
      },
      {
       "id": "r8",
       "label": [
        "lognormal sigma=6"
       ],
       "values": [
        0.3933,
        0.3027,
        0.1229,
        0.05076,
        0.03174,
        0.0279
       ]
      },
      {
       "id": "r9",
       "label": [
        "log-Laplace b=0.25"
       ],
       "values": [
        0.7049,
        0.8268,
        0.9769,
        0.9918,
        0.9969,
        0.9969
       ]
      },
      {
       "id": "r10",
       "label": [
        "log-Laplace b=0.5"
       ],
       "values": [
        0.6989,
        0.819,
        0.9448,
        0.9779,
        0.9753,
        0.9765
       ]
      },
      {
       "id": "r11",
       "label": [
        "log-Laplace b=1"
       ],
       "values": [
        0.667,
        0.6887,
        0.8479,
        0.8763,
        0.8853,
        0.8953
       ]
      },
      {
       "id": "r12",
       "label": [
        "log-Laplace b=1.5"
       ],
       "values": [
        0.5766,
        0.5947,
        0.6989,
        0.7789,
        0.7804,
        0.8035
       ]
      },
      {
       "id": "r13",
       "label": [
        "log-Laplace b=2"
       ],
       "values": [
        0.5179,
        0.5036,
        0.612,
        0.683,
        0.7079,
        0.7206
       ]
      },
      {
       "id": "r14",
       "label": [
        "log-Laplace b=3"
       ],
       "values": [
        0.4672,
        0.4365,
        0.4586,
        0.5257,
        0.5628,
        0.5902
       ]
      },
      {
       "id": "r15",
       "label": [
        "log-Laplace b=4"
       ],
       "values": [
        0.4353,
        0.3764,
        0.3496,
        0.4267,
        0.4803,
        0.4968
       ]
      },
      {
       "id": "r16",
       "label": [
        "log-Laplace b=6"
       ],
       "values": [
        0.4167,
        0.3492,
        0.2667,
        0.2896,
        0.3441,
        0.3753
       ]
      }
     ],
     "artefact": "A49"
    },
    {
     "id": "T2",
     "description": "Bias and variance at test-set sizes 12 and 1000, for spreads 0.25, 1 and 6: the mean of ln(sMAPE minimiser) minus ln(median), and the variances of ln(sample median) and ln(sMAPE minimiser); every setting and size is in the run's CSV",
     "row_labels": [
      "setting",
      "test-set size"
     ],
     "columns": [
      "mean of ln(sMAPE minimiser) minus ln(median)",
      "variance of ln(sample median)",
      "variance of ln(sMAPE minimiser)",
      "ratio of mean squared errors about the true median"
     ],
     "rows": [
      {
       "id": "r1",
       "label": [
        "lognormal sigma=0.25",
        12
       ],
       "values": [
        0.0008431,
        0.007228,
        0.01029,
        0.7026
       ]
      },
      {
       "id": "r2",
       "label": [
        "lognormal sigma=0.25",
        1000
       ],
       "values": [
        3.788e-05,
        9.802e-05,
        0.0001015,
        0.9662
       ]
      },
      {
       "id": "r3",
       "label": [
        "lognormal sigma=1",
        12
       ],
       "values": [
        0.0007662,
        0.1159,
        0.1768,
        0.6555
       ]
      },
      {
       "id": "r4",
       "label": [
        "lognormal sigma=1",
        1000
       ],
       "values": [
        9.852e-05,
        0.001587,
        0.002383,
        0.6659
       ]
      },
      {
       "id": "r5",
       "label": [
        "lognormal sigma=6",
        12
       ],
       "values": [
        0.002108,
        4.287,
        14.17,
        0.3027
       ]
      },
      {
       "id": "r6",
       "label": [
        "lognormal sigma=6",
        1000
       ],
       "values": [
        -0.02091,
        0.05714,
        1.126,
        0.05074
       ]
      },
      {
       "id": "r7",
       "label": [
        "log-Laplace b=0.25",
        12
       ],
       "values": [
        -0.0004367,
        0.007304,
        0.008834,
        0.8268
       ]
      },
      {
       "id": "r8",
       "label": [
        "log-Laplace b=0.25",
        1000
       ],
       "values": [
        0.0001574,
        6.678e-05,
        6.733e-05,
        0.9918
       ]
      },
      {
       "id": "r9",
       "label": [
        "log-Laplace b=1",
        12
       ],
       "values": [
        0.001393,
        0.1191,
        0.173,
        0.6887
       ]
      },
      {
       "id": "r10",
       "label": [
        "log-Laplace b=1",
        1000
       ],
       "values": [
        0.0003075,
        0.001055,
        0.001204,
        0.8764
       ]
      },
      {
       "id": "r11",
       "label": [
        "log-Laplace b=6",
        12
       ],
       "values": [
        0.02118,
        4.267,
        12.22,
        0.3492
       ]
      },
      {
       "id": "r12",
       "label": [
        "log-Laplace b=6",
        1000
       ],
       "values": [
        0.002973,
        0.03773,
        0.1303,
        0.2896
       ]
      }
     ],
     "artefact": "A49"
    }
   ],
   "artefacts": [
    "A6",
    "A49",
    "A50",
    "A51"
   ],
   "setup": {
    "studied": [
     "D8",
     "D11"
    ],
    "cases": [
     "D26"
    ],
    "procedure": [
     "D17",
     "D15"
    ],
    "lens": [
     "D9"
    ]
   },
   "execution": {
    "code_version": "fa0ce626f18bd1ca22076d62452d2b68c484a297",
    "software": {
     "python": "3.13.18",
     "platform": "Linux-6.18.44-fc-v470-x86_64-with-glibc2.39"
    }
   }
  },
  {
   "id": "E16",
   "profile": "quant",
   "design": "other",
   "description": "Lab runs RN10 under plan PL11. Design: Exact computation, with no sampling, of the forecasts that minimise the expected Suilin term and the expected sMAPE term (claimset-code/smape-target/msmape_targets.py, run once): by Gauss-Legendre quadrature of the density of ln y for continuous distributions (with zeros added as a point mass) and by sums for counts, minimised over forecasts f >= 0 on a grid of 0 and 4000 log-spaced points refined by golden section. Settings: away from zero, lognormal sigma 0.5 and 1, inverse Gaussian shape 10 and 100 and log-skewed sd 1 with skewness -1 and +1, all at median about 100; near zero, lognormal sigma 1 at median 1 and at median 100 with shares of zeros 0, 0.2, 0.4, 0.6 and 0.8; counts, Poisson with means 0.3, 0.5, 1, 1.5, 3 and 10. For the lognormal sigma 1 at medians 1 and 100, the share of zeros at which each measure switches to the forecast 0, by bisection. Analysis: For each setting and measure: the target, its quantile level and the expected score there; for each setting, the relative difference of the Suilin target from the sMAPE target; and the two zero thresholds. Reported as tables, with no decision rule. The plan, the run's script and its results file name the Suilin loss msMAPE, as version 2 of this paper did.",
   "availability": {
    "level": "public"
   },
   "lab_runs": [
    "RN10"
   ],
   "tables": [
    {
     "id": "T1",
     "description": "The Suilin target beside the sMAPE target, with their quantile levels and the relative difference of the Suilin target from the sMAPE target",
     "row_labels": [
      "setting"
     ],
     "columns": [
      {
       "variable": "target",
       "conditions": {
        "measure": "sMAPE"
       }
      },
      {
       "variable": "target",
       "conditions": {
        "measure": "Suilin"
       }
      },
      {
       "variable": "quantile level of the target",
       "conditions": {
        "measure": "sMAPE"
       }
      },
      {
       "variable": "quantile level of the target",
       "conditions": {
        "measure": "Suilin"
       }
      },
      "relative difference of the Suilin target from the sMAPE target",
      "share of zero outcomes"
     ],
     "rows": [
      {
       "id": "r1",
       "label": [
        "lognormal sigma=0.5"
       ],
       "values": [
        100.0,
        100.0,
        0.5,
        0.5,
        -1.403e-05,
        0
       ]
      },
      {
       "id": "r2",
       "label": [
        "lognormal sigma=1"
       ],
       "values": [
        100.0,
        99.98,
        0.5,
        0.4999,
        -0.0001834,
        0
       ]
      },
      {
       "id": "r3",
       "label": [
        "inverse Gaussian shape=10"
       ],
       "values": [
        14.02,
        13.98,
        0.4387,
        0.438,
        -0.002701,
        0
       ]
      },
      {
       "id": "r4",
       "label": [
        "inverse Gaussian shape=100"
       ],
       "values": [
        66.53,
        66.52,
        0.4929,
        0.4929,
        -0.0001284,
        0
       ]
      },
      {
       "id": "r5",
       "label": [
        "log-skewed left sd=1 skew=-1"
       ],
       "values": [
        128.3,
        128.2,
        0.5362,
        0.5361,
        -0.0002099,
        0
       ]
      },
      {
       "id": "r6",
       "label": [
        "log-skewed right sd=1 skew=+1"
       ],
       "values": [
        77.96,
        77.96,
        0.4638,
        0.4638,
        -4.995e-05,
        0
       ]
      },
      {
       "id": "r7",
       "label": [
        "lognormal median=1 sigma=1"
       ],
       "values": [
        1.0,
        0.9846,
        0.5,
        0.4938,
        -0.01536,
        0
       ]
      },
      {
       "id": "r8",
       "label": [
        "lognormal median=1 sigma=1 with zeros p0=0.2"
       ],
       "values": [
        1.0,
        0.9054,
        0.6,
        0.5683,
        -0.0946,
        0.2
       ]
      },
      {
       "id": "r9",
       "label": [
        "lognormal median=1 sigma=1 with zeros p0=0.4"
       ],
       "values": [
        0,
        0.1702,
        0.4,
        0.423,
        null,
        0.4
       ]
      },
      {
       "id": "r10",
       "label": [
        "lognormal median=1 sigma=1 with zeros p0=0.6"
       ],
       "values": [
        0,
        0,
        0.6,
        0.6,
        0,
        0.6
       ]
      },
      {
       "id": "r11",
       "label": [
        "lognormal median=1 sigma=1 with zeros p0=0.8"
       ],
       "values": [
        0,
        0,
        0.8,
        0.8,
        0,
        0.8
       ]
      },
      {
       "id": "r12",
       "label": [
        "lognormal median=100 sigma=1"
       ],
       "values": [
        100.0,
        99.98,
        0.5,
        0.4999,
        -0.0001834,
        0
       ]
      },
      {
       "id": "r13",
       "label": [
        "lognormal median=100 sigma=1 with zeros p0=0.2"
       ],
       "values": [
        100.0,
        99.89,
        0.6,
        0.5997,
        -0.001093,
        0.2
       ]
      },
      {
       "id": "r14",
       "label": [
        "lognormal median=100 sigma=1 with zeros p0=0.4"
       ],
       "values": [
        0,
        0,
        0.4,
        0.4,
        0,
        0.4
       ]
      },
      {
       "id": "r15",
       "label": [
        "lognormal median=100 sigma=1 with zeros p0=0.6"
       ],
       "values": [
        0,
        0,
        0.6,
        0.6,
        0,
        0.6
       ]
      },
      {
       "id": "r16",
       "label": [
        "lognormal median=100 sigma=1 with zeros p0=0.8"
       ],
       "values": [
        0,
        0,
        0.8,
        0.8,
        0,
        0.8
       ]
      },
      {
       "id": "r17",
       "label": [
        "Poisson mean=0.3"
       ],
       "values": [
        0,
        0,
        0.7408,
        0.7408,
        0,
        0.7408
       ]
      },
      {
       "id": "r18",
       "label": [
        "Poisson mean=0.5"
       ],
       "values": [
        0,
        0,
        0.6065,
        0.6065,
        0,
        0.6065
       ]
      },
      {
       "id": "r19",
       "label": [
        "Poisson mean=1"
       ],
       "values": [
        1,
        1,
        0.7358,
        0.7358,
        0,
        0.3679
       ]
      },
      {
       "id": "r20",
       "label": [
        "Poisson mean=1.5"
       ],
       "values": [
        2,
        2,
        0.8088,
        0.8088,
        0,
        0.2231
       ]
      },
      {
       "id": "r21",
       "label": [
        "Poisson mean=3"
       ],
       "values": [
        3,
        3,
        0.6472,
        0.6472,
        0,
        0.04979
       ]
      },
      {
       "id": "r22",
       "label": [
        "Poisson mean=10"
       ],
       "values": [
        10,
        10,
        0.583,
        0.583,
        0,
        4.54e-05
       ]
      }
     ],
     "artefact": "A52"
    },
    {
     "id": "T2",
     "description": "The share of zeros at which each measure's target switches to the forecast 0, for lognormal outcomes with sigma 1",
     "row_labels": [
      "setting"
     ],
     "columns": [
      {
       "variable": "zero threshold",
       "conditions": {
        "measure": "sMAPE"
       }
      },
      {
       "variable": "zero threshold",
       "conditions": {
        "measure": "Suilin"
       }
      }
     ],
     "rows": [
      {
       "id": "r1",
       "label": [
        "lognormal median=1 sigma=1"
       ],
       "values": [
        0.394,
        0.47
       ]
      },
      {
       "id": "r2",
       "label": [
        "lognormal median=100 sigma=1"
       ],
       "values": [
        0.394,
        0.3937
       ]
      }
     ],
     "artefact": "A52"
    }
   ],
   "artefacts": [
    "A6",
    "A52",
    "A53"
   ],
   "setup": {
    "studied": [
     "D28",
     "D1"
    ],
    "cases": [
     "D25",
     "D30"
    ],
    "procedure": [
     "D15"
    ],
    "lens": [
     "D29",
     "D6"
    ]
   },
   "execution": {
    "code_version": "838c642043625bc4c67dc6226166ad520a5c15ce",
    "software": {
     "python": "3.13.18",
     "platform": "Linux-6.18.44-fc-v553-x86_64-with-glibc2.39"
    }
   },
   "results": []
  },
  {
   "id": "E31",
   "profile": "quant",
   "design": "other",
   "description": "Numerical checks of the formulas of the claims new or reworded in version 3 ([[D31:formula checks]]), lab run RN13 (check_v3.py, seed 8) under the formal plan that decided C3, C4, C12, C17 and C20; every check is within the plan's tolerances. They check the formulas on concrete numbers; the proofs are the evidence for the claims. The tables are labelled by the hypotheses the run's script names: H3 (T1 and T2) checks C3; H4 (T3) checks C4; H34 (T4 and T5) checks C12; H35 (T6 and T7) checks C17; H42 (T8 and T9) checks C20. The tables of H39 (T10 to T12) and H40 (T13 to T17) check the formulas of C18 and C19, whose wording was decided after later checks (E32, E33).",
   "availability": {
    "level": "public"
   },
   "lab_runs": [
    "RN13"
   ],
   "tables": [
    {
     "id": "T1",
     "description": "violations of the bound on the log-MAE of the sMAPE minimiser, largest excess of the minimiser's log-MAE over the median's, minus the bound; hypothesis: H3, test_sets: 2000 (lab runs RN13)",
     "conditions": {
      "hypothesis": "H3",
      "test_sets": 2000
     },
     "row_labels": [],
     "columns": [
      "violations of the bound on the log-MAE of the sMAPE minimiser",
      "largest excess of the minimiser's log-MAE over the median's, minus the bound"
     ],
     "rows": [
      {
       "id": "r1",
       "label": [],
       "values": [
        0,
        -1.263e-06
       ]
      }
     ],
     "artefact": "A6"
    },
    {
     "description": "sum of sMAPE terms, forecast at the median, actuals 1, 1, e, e^4, e^4, sum of sMAPE terms, forecast at 1, actuals 1, 1, e, e^4, e^4; hypothesis: H3 (lab runs RN13)",
     "row_labels": [],
     "columns": [
      "sum of sMAPE terms, forecast at the median, actuals 1, 1, e, e^4, e^4",
      "sum of sMAPE terms, forecast at 1, actuals 1, 1, e, e^4, e^4"
     ],
     "rows": [
      {
       "id": "r1",
       "label": [],
       "values": [
        2.735,
        2.39
       ]
      }
     ],
     "conditions": {
      "hypothesis": "H3"
     },
     "artefact": "A6",
     "id": "T2"
    },
    {
     "description": "largest difference of the weights below and above the target, over their sum; hypothesis: H4, distributions: 6 (lab runs RN13)",
     "row_labels": [],
     "columns": [
      "largest difference of the weights below and above the target, over their sum"
     ],
     "rows": [
      {
       "id": "r1",
       "label": [],
       "values": [
        2.089e-08
       ]
      }
     ],
     "conditions": {
      "hypothesis": "H4",
      "distributions": 6
     },
     "artefact": "A6",
     "id": "T3"
    },
    {
     "description": "distributions with more than one local minimum, distributions where the target is not on the side of the median that carries more weight; hypothesis: H34, distributions: 6 (lab runs RN13)",
     "row_labels": [],
     "columns": [
      "distributions with more than one local minimum",
      "distributions where the target is not on the side of the median that carries more weight"
     ],
     "rows": [
      {
       "id": "r1",
       "label": [],
       "values": [
        0,
        0
       ]
      }
     ],
     "conditions": {
      "hypothesis": "H34",
      "distributions": 6
     },
     "artefact": "A6",
     "id": "T4"
    },
    {
     "description": "largest distance in logs between the target and the median, lognormal and log-Laplace; hypothesis: H34, distributions: 16 (lab runs RN13)",
     "row_labels": [],
     "columns": [
      "largest distance in logs between the target and the median, lognormal and log-Laplace"
     ],
     "rows": [
      {
       "id": "r1",
       "label": [],
       "values": [
        1.753e-07
       ]
      }
     ],
     "conditions": {
      "hypothesis": "H34",
      "distributions": 16
     },
     "artefact": "A6",
     "id": "T5"
    },
    {
     "description": "Suilin term of a zero forecast of a zero actual; hypothesis: H35 (lab runs RN13)",
     "row_labels": [],
     "columns": [
      "Suilin term of a zero forecast of a zero actual"
     ],
     "rows": [
      {
       "id": "r1",
       "label": [],
       "values": [
        0.0
       ]
      }
     ],
     "conditions": {
      "hypothesis": "H35"
     },
     "artefact": "A6",
     "id": "T6"
    },
    {
     "id": "T7",
     "description": "largest error of the identity of (a), largest error of the identity of (b), values of the Suilin term outside [0, 1); hypothesis: H35, pairs: 200000 (lab runs RN13)",
     "conditions": {
      "hypothesis": "H35",
      "pairs": 200000
     },
     "row_labels": [],
     "columns": [
      "largest error of the identity of (a)",
      "largest error of the identity of (b)",
      "values of the Suilin term outside [0, 1)"
     ],
     "rows": [
      {
       "id": "r1",
       "label": [],
       "values": [
        1.11e-16,
        0.0,
        0
       ]
      }
     ],
     "artefact": "A6"
    },
    {
     "id": "T8",
     "description": "largest error of the identity of (a), largest error of the identity of (b), values of the msMAPE term outside [0, 1], values of the msMAPE term not below 1 for actuals below nu, pairs with a count of at least 1 where the msMAPE and sMAPE terms differ by nu; hypothesis: H42, pairs: 200000 (lab runs RN13)",
     "conditions": {
      "hypothesis": "H42",
      "pairs": 200000
     },
     "row_labels": [
      "nu"
     ],
     "columns": [
      "largest error of the identity of (a)",
      "largest error of the identity of (b)",
      "values of the msMAPE term outside [0, 1]",
      "values of the msMAPE term not below 1 for actuals below nu",
      "pairs with a count of at least 1 where the msMAPE and sMAPE terms differ"
     ],
     "rows": [
      {
       "id": "r1",
       "label": [
        0.1
       ],
       "values": [
        0.0,
        0.0,
        0,
        0,
        0
       ]
      },
      {
       "id": "r2",
       "label": [
        0.5
       ],
       "values": [
        0.0,
        0.0,
        0,
        0,
        0
       ]
      },
      {
       "id": "r3",
       "label": [
        1.0
       ],
       "values": [
        0.0,
        0.0,
        0,
        0,
        0
       ]
      }
     ],
     "artefact": "A6"
    },
    {
     "id": "T9",
     "description": "msMAPE term of a forecast of 1e-9 for a zero actual by nu; hypothesis: H42 (lab runs RN13)",
     "conditions": {
      "hypothesis": "H42"
     },
     "row_labels": [
      "nu"
     ],
     "columns": [
      "msMAPE term of a forecast of 1e-9 for a zero actual"
     ],
     "rows": [
      {
       "id": "r1",
       "label": [
        0.1
       ],
       "values": [
        1e-08
       ]
      },
      {
       "id": "r2",
       "label": [
        0.5
       ],
       "values": [
        2e-09
       ]
      },
      {
       "id": "r3",
       "label": [
        1.0
       ],
       "values": [
        1e-09
       ]
      }
     ],
     "artefact": "A6"
    },
    {
     "id": "T10",
     "description": "signal-to-noise ratio of sMAPE over its large-spread formula, signal-to-noise ratio of the absolute error over its limit by error_distribution, c1, c2; hypothesis: H39, sigma: 200 (lab runs RN13)",
     "conditions": {
      "hypothesis": "H39",
      "sigma": 200
     },
     "row_labels": [
      "error_distribution",
      "c1",
      "c2"
     ],
     "columns": [
      "signal-to-noise ratio of sMAPE over its large-spread formula",
      "signal-to-noise ratio of the absolute error over its limit"
     ],
     "rows": [
      {
       "id": "r1",
       "label": [
        "normal",
        0.0,
        0.5
       ],
       "values": [
        0.9999,
        0.9913
       ]
      },
      {
       "id": "r2",
       "label": [
        "normal",
        0.0,
        1.0
       ],
       "values": [
        1.0,
        0.9956
       ]
      },
      {
       "id": "r3",
       "label": [
        "normal",
        -0.5,
        1.5
       ],
       "values": [
        1.001,
        0.9966
       ]
      },
      {
       "id": "r4",
       "label": [
        "Gumbel",
        0.0,
        0.5
       ],
       "values": [
        0.9998,
        0.9647
       ]
      },
      {
       "id": "r5",
       "label": [
        "Gumbel",
        0.0,
        1.0
       ],
       "values": [
        1.0,
        0.9937
       ]
      },
      {
       "id": "r6",
       "label": [
        "Gumbel",
        -0.5,
        1.5
       ],
       "values": [
        1.0,
        0.9967
       ]
      }
     ],
     "artefact": "A6"
    },
    {
     "id": "T11",
     "description": "relative spread of the absolute log error's signal-to-noise ratio over sigma = 2, 10, 50 by error_distribution, c1, c2; hypothesis: H39 (lab runs RN13)",
     "conditions": {
      "hypothesis": "H39"
     },
     "row_labels": [
      "error_distribution",
      "c1",
      "c2"
     ],
     "columns": [
      "relative spread of the absolute log error's signal-to-noise ratio over sigma = 2, 10, 50"
     ],
     "rows": [
      {
       "id": "r1",
       "label": [
        "normal",
        0.0,
        0.5
       ],
       "values": [
        1.686e-15
       ]
      },
      {
       "id": "r2",
       "label": [
        "normal",
        0.0,
        1.0
       ],
       "values": [
        8.388e-16
       ]
      },
      {
       "id": "r3",
       "label": [
        "normal",
        -0.5,
        1.5
       ],
       "values": [
        4.51e-16
       ]
      },
      {
       "id": "r4",
       "label": [
        "Gumbel",
        0.0,
        0.5
       ],
       "values": [
        4.186e-15
       ]
      },
      {
       "id": "r5",
       "label": [
        "Gumbel",
        0.0,
        1.0
       ],
       "values": [
        4.885e-15
       ]
      },
      {
       "id": "r6",
       "label": [
        "Gumbel",
        -0.5,
        1.5
       ],
       "values": [
        1.356e-14
       ]
      }
     ],
     "artefact": "A6"
    },
    {
     "id": "T12",
     "description": "signal-to-noise ratio of the squared error over its formula, lognormal, exp(sigma^2 / 2) times the squared error's ratio, lognormal, formula by c1, c2, sigma; hypothesis: H39 (lab runs RN13)",
     "conditions": {
      "hypothesis": "H39"
     },
     "row_labels": [
      "c1",
      "c2",
      "sigma"
     ],
     "columns": [
      "signal-to-noise ratio of the squared error over its formula, lognormal",
      "exp(sigma^2 / 2) times the squared error's ratio, lognormal, formula"
     ],
     "rows": [
      {
       "id": "r1",
       "label": [
        0.0,
        0.5,
        0.5
       ],
       "values": [
        1.0,
        null
       ]
      },
      {
       "id": "r2",
       "label": [
        0.0,
        1.0,
        0.5
       ],
       "values": [
        1.0,
        null
       ]
      },
      {
       "id": "r3",
       "label": [
        -0.5,
        1.5,
        0.5
       ],
       "values": [
        1.0,
        null
       ]
      },
      {
       "id": "r4",
       "label": [
        0.0,
        0.5,
        1.0
       ],
       "values": [
        1.0,
        null
       ]
      },
      {
       "id": "r5",
       "label": [
        0.0,
        1.0,
        1.0
       ],
       "values": [
        1.0,
        null
       ]
      },
      {
       "id": "r6",
       "label": [
        -0.5,
        1.5,
        1.0
       ],
       "values": [
        1.0,
        null
       ]
      },
      {
       "id": "r7",
       "label": [
        0.0,
        0.5,
        2.0
       ],
       "values": [
        1.0,
        null
       ]
      },
      {
       "id": "r8",
       "label": [
        0.0,
        1.0,
        2.0
       ],
       "values": [
        1.0,
        null
       ]
      },
      {
       "id": "r9",
       "label": [
        -0.5,
        1.5,
        2.0
       ],
       "values": [
        1.0,
        null
       ]
      },
      {
       "id": "r10",
       "label": [
        0.0,
        1.0,
        4.0
       ],
       "values": [
        null,
        0.9907
       ]
      },
      {
       "id": "r11",
       "label": [
        0.0,
        1.0,
        6.0
       ],
       "values": [
        null,
        1.0
       ]
      }
     ],
     "artefact": "A6"
    },
    {
     "description": "distributions compared with the asymptotic values recorded by RN7; hypothesis: H40 (lab runs RN13)",
     "row_labels": [],
     "columns": [
      "distributions compared with the asymptotic values recorded by RN7"
     ],
     "rows": [
      {
       "id": "r1",
       "label": [],
       "values": [
        16
       ]
      }
     ],
     "conditions": {
      "hypothesis": "H40"
     },
     "artefact": "A6",
     "id": "T13"
    },
    {
     "description": "distributions where A is not below 2 g at the median, largest difference from the asymptotic values recorded by RN7; hypothesis: H40, distributions: 16 (lab runs RN13)",
     "row_labels": [],
     "columns": [
      "distributions where A is not below 2 g at the median",
      "largest difference from the asymptotic values recorded by RN7"
     ],
     "rows": [
      {
       "id": "r1",
       "label": [],
       "values": [
        0,
        0.00014
       ]
      }
     ],
     "conditions": {
      "hypothesis": "H40",
      "distributions": 16
     },
     "artefact": "A6",
     "id": "T14"
    },
    {
     "id": "T15",
     "description": "large-sample relative efficiency, formula by family, spread; hypothesis: H40 (lab runs RN13)",
     "conditions": {
      "hypothesis": "H40"
     },
     "row_labels": [
      "family",
      "spread"
     ],
     "columns": [
      "large-sample relative efficiency, formula"
     ],
     "rows": [
      {
       "id": "r1",
       "label": [
        "lognormal",
        0.25
       ],
       "values": [
        0.9696
       ]
      },
      {
       "id": "r2",
       "label": [
        "lognormal",
        0.5
       ],
       "values": [
        0.8881
       ]
      },
      {
       "id": "r3",
       "label": [
        "lognormal",
        1.0
       ],
       "values": [
        0.6601
       ]
      },
      {
       "id": "r4",
       "label": [
        "lognormal",
        1.5
       ],
       "values": [
        0.454
       ]
      },
      {
       "id": "r5",
       "label": [
        "lognormal",
        2.0
       ],
       "values": [
        0.3073
       ]
      },
      {
       "id": "r6",
       "label": [
        "lognormal",
        3.0
       ],
       "values": [
        0.1474
       ]
      },
      {
       "id": "r7",
       "label": [
        "lognormal",
        4.0
       ],
       "values": [
        0.07778
       ]
      },
      {
       "id": "r8",
       "label": [
        "lognormal",
        6.0
       ],
       "values": [
        0.0279
       ]
      },
      {
       "id": "r9",
       "label": [
        "log-Laplace",
        0.25
       ],
       "values": [
        0.9969
       ]
      },
      {
       "id": "r10",
       "label": [
        "log-Laplace",
        0.5
       ],
       "values": [
        0.9765
       ]
      },
      {
       "id": "r11",
       "label": [
        "log-Laplace",
        1.0
       ],
       "values": [
        0.8953
       ]
      },
      {
       "id": "r12",
       "label": [
        "log-Laplace",
        1.5
       ],
       "values": [
        0.8036
       ]
      },
      {
       "id": "r13",
       "label": [
        "log-Laplace",
        2.0
       ],
       "values": [
        0.7207
       ]
      },
      {
       "id": "r14",
       "label": [
        "log-Laplace",
        3.0
       ],
       "values": [
        0.5903
       ]
      },
      {
       "id": "r15",
       "label": [
        "log-Laplace",
        4.0
       ],
       "values": [
        0.4969
       ]
      },
      {
       "id": "r16",
       "label": [
        "log-Laplace",
        6.0
       ],
       "values": [
        0.3755
       ]
      }
     ],
     "artefact": "A6"
    },
    {
     "id": "T16",
     "description": "distance of the large-sample relative efficiency from 1 at spread 0.01 by family; hypothesis: H40 (lab runs RN13)",
     "conditions": {
      "hypothesis": "H40"
     },
     "row_labels": [
      "family"
     ],
     "columns": [
      "distance of the large-sample relative efficiency from 1 at spread 0.01"
     ],
     "rows": [
      {
       "id": "r1",
       "label": [
        "lognormal"
       ],
       "values": [
        5e-05
       ]
      },
      {
       "id": "r2",
       "label": [
        "log-Laplace"
       ],
       "values": [
        1.249e-08
       ]
      }
     ],
     "artefact": "A6"
    },
    {
     "id": "T17",
     "description": "large-sample relative efficiency times sigma^3 over 6 (ln 2)^2 sqrt(2 pi), lognormal, large-sample relative efficiency times b over 3, log-Laplace; hypothesis: H40, spread: 200 (lab runs RN13)",
     "conditions": {
      "hypothesis": "H40",
      "spread": 200
     },
     "row_labels": [],
     "columns": [
      "large-sample relative efficiency times sigma^3 over 6 (ln 2)^2 sqrt(2 pi), lognormal",
      "large-sample relative efficiency times b over 3, log-Laplace"
     ],
     "rows": [
      {
       "id": "r1",
       "label": [],
       "values": [
        0.9998,
        0.9906
       ]
      }
     ],
     "artefact": "A6"
    }
   ],
   "artefacts": [
    "A6"
   ],
   "setup": {
    "studied": [
     "D1",
     "D28",
     "D33",
     "D6",
     "D8"
    ],
    "lens": [
     "D32",
     "D9",
     "D5",
     "D4",
     "D19",
     "D11",
     "D10"
    ],
    "procedure": [
     "D31"
    ]
   },
   "execution": {
    "code_version": "2f0d2fca48b049d8004fec7228e47b8857f6ace4",
    "software": {
     "python": "3.13.18",
     "platform": "Linux-6.18.44-fc-v595-x86_64-with-glibc2.39"
    }
   }
  },
  {
   "id": "E32",
   "profile": "quant",
   "design": "other",
   "description": "Numerical checks of the formulas of C18 ([[D31:formula checks]]), lab run RN18 (check_h39.py) under the formal plan that decided C18; every check is within the plan's tolerances. They check the formulas on concrete numbers; the proof is the evidence for the claim. The run's script labels the checks H39, the record's first wording of C18; the formulas are those of C18.",
   "availability": {
    "level": "public"
   },
   "lab_runs": [
    "RN18"
   ],
   "tables": [
    {
     "id": "T1",
     "description": "distance of sMAPE's signal-to-noise ratio over its large-spread formula from 1, distance of the absolute error's signal-to-noise ratio over its limit from 1 by error_distribution, c1, c2, sigma; hypothesis: H39 (lab runs RN18)",
     "conditions": {
      "hypothesis": "H39"
     },
     "row_labels": [
      "error_distribution",
      "c1",
      "c2",
      "sigma"
     ],
     "columns": [
      "distance of sMAPE's signal-to-noise ratio over its large-spread formula from 1",
      "distance of the absolute error's signal-to-noise ratio over its limit from 1"
     ],
     "rows": [
      {
       "id": "r1",
       "label": [
        "normal",
        0.0,
        0.5,
        200
       ],
       "values": [
        5.266e-05,
        0.008727
       ]
      },
      {
       "id": "r2",
       "label": [
        "normal",
        0.0,
        0.5,
        2000
       ],
       "values": [
        2.749e-06,
        0.0008713
       ]
      },
      {
       "id": "r3",
       "label": [
        "normal",
        0.0,
        1.0,
        200
       ],
       "values": [
        0.0003989,
        0.004379
       ]
      },
      {
       "id": "r4",
       "label": [
        "normal",
        0.0,
        1.0,
        2000
       ],
       "values": [
        4.703e-05,
        0.0004373
       ]
      },
      {
       "id": "r5",
       "label": [
        "normal",
        -0.5,
        1.5,
        200
       ],
       "values": [
        0.00122,
        0.003402
       ]
      },
      {
       "id": "r6",
       "label": [
        "normal",
        -0.5,
        1.5,
        2000
       ],
       "values": [
        0.0001273,
        0.0003398
       ]
      },
      {
       "id": "r7",
       "label": [
        "Gumbel",
        0.0,
        0.5,
        200
       ],
       "values": [
        0.0001627,
        0.03529
       ]
      },
      {
       "id": "r8",
       "label": [
        "Gumbel",
        0.0,
        0.5,
        2000
       ],
       "values": [
        5.872e-07,
        0.00352
       ]
      },
      {
       "id": "r9",
       "label": [
        "Gumbel",
        0.0,
        1.0,
        200
       ],
       "values": [
        0.0001512,
        0.006295
       ]
      },
      {
       "id": "r10",
       "label": [
        "Gumbel",
        0.0,
        1.0,
        2000
       ],
       "values": [
        2.457e-05,
        0.0006281
       ]
      },
      {
       "id": "r11",
       "label": [
        "Gumbel",
        -0.5,
        1.5,
        200
       ],
       "values": [
        0.0003779,
        0.00326
       ]
      },
      {
       "id": "r12",
       "label": [
        "Gumbel",
        -0.5,
        1.5,
        2000
       ],
       "values": [
        4.713e-05,
        0.0003255
       ]
      }
     ],
     "artefact": "A6"
    },
    {
     "id": "T2",
     "description": "relative spread of the absolute log error's signal-to-noise ratio over sigma = 2, 10, 50 by error_distribution, c1, c2; hypothesis: H39 (lab runs RN18)",
     "conditions": {
      "hypothesis": "H39"
     },
     "row_labels": [
      "error_distribution",
      "c1",
      "c2"
     ],
     "columns": [
      "relative spread of the absolute log error's signal-to-noise ratio over sigma = 2, 10, 50"
     ],
     "rows": [
      {
       "id": "r1",
       "label": [
        "normal",
        0.0,
        0.5
       ],
       "values": [
        3.761e-15
       ]
      },
      {
       "id": "r2",
       "label": [
        "normal",
        0.0,
        1.0
       ],
       "values": [
        2.397e-16
       ]
      },
      {
       "id": "r3",
       "label": [
        "normal",
        -0.5,
        1.5
       ],
       "values": [
        1.579e-15
       ]
      },
      {
       "id": "r4",
       "label": [
        "Gumbel",
        0.0,
        0.5
       ],
       "values": [
        4.243e-13
       ]
      },
      {
       "id": "r5",
       "label": [
        "Gumbel",
        0.0,
        1.0
       ],
       "values": [
        7.932e-13
       ]
      },
      {
       "id": "r6",
       "label": [
        "Gumbel",
        -0.5,
        1.5
       ],
       "values": [
        1.43e-12
       ]
      }
     ],
     "artefact": "A6"
    },
    {
     "id": "T3",
     "description": "distance of the squared error's signal-to-noise ratio over its formula from 1, lognormal, distance of exp(sigma^2 / 2) times the squared error's ratio from 1, lognormal, formula by c1, c2, sigma; hypothesis: H39 (lab runs RN18)",
     "conditions": {
      "hypothesis": "H39"
     },
     "row_labels": [
      "c1",
      "c2",
      "sigma"
     ],
     "columns": [
      "distance of the squared error's signal-to-noise ratio over its formula from 1, lognormal",
      "distance of exp(sigma^2 / 2) times the squared error's ratio from 1, lognormal, formula"
     ],
     "rows": [
      {
       "id": "r1",
       "label": [
        0.0,
        0.5,
        0.5
       ],
       "values": [
        3.109e-15,
        null
       ]
      },
      {
       "id": "r2",
       "label": [
        0.0,
        1.0,
        0.5
       ],
       "values": [
        6.661e-16,
        null
       ]
      },
      {
       "id": "r3",
       "label": [
        -0.5,
        1.5,
        0.5
       ],
       "values": [
        1.11e-15,
        null
       ]
      },
      {
       "id": "r4",
       "label": [
        0.0,
        0.5,
        1.0
       ],
       "values": [
        1.554e-15,
        null
       ]
      },
      {
       "id": "r5",
       "label": [
        0.0,
        1.0,
        1.0
       ],
       "values": [
        8.882e-16,
        null
       ]
      },
      {
       "id": "r6",
       "label": [
        -0.5,
        1.5,
        1.0
       ],
       "values": [
        5.551e-16,
        null
       ]
      },
      {
       "id": "r7",
       "label": [
        0.0,
        0.5,
        2.0
       ],
       "values": [
        1.409e-13,
        null
       ]
      },
      {
       "id": "r8",
       "label": [
        0.0,
        1.0,
        2.0
       ],
       "values": [
        7.55e-15,
        null
       ]
      },
      {
       "id": "r9",
       "label": [
        -0.5,
        1.5,
        2.0
       ],
       "values": [
        1.887e-15,
        null
       ]
      },
      {
       "id": "r10",
       "label": [
        0.0,
        1.0,
        4.0
       ],
       "values": [
        null,
        0.009325
       ]
      },
      {
       "id": "r11",
       "label": [
        0.0,
        1.0,
        6.0
       ],
       "values": [
        null,
        3.08e-06
       ]
      },
      {
       "id": "r12",
       "label": [
        0.0,
        1.0,
        8.0
       ],
       "values": [
        null,
        1.888e-11
       ]
      }
     ],
     "artefact": "A6"
    }
   ],
   "artefacts": [
    "A6"
   ],
   "setup": {
    "studied": [
     "D1"
    ],
    "lens": [
     "D32",
     "D19",
     "D11",
     "D10"
    ],
    "procedure": [
     "D31"
    ]
   },
   "execution": {
    "code_version": "6aa7a7ab87bf5c4e40d704a7492cd9af9ef481db",
    "software": {
     "python": "3.13.18",
     "platform": "Linux-6.18.44-fc-v595-x86_64-with-glibc2.39"
    }
   }
  },
  {
   "id": "E33",
   "profile": "quant",
   "design": "other",
   "description": "Numerical checks of the formulas of C19 ([[D31:formula checks]]), lab run RN16 (check_corrected.py, seed 8) under the formal plan that decided C19; every check is within the plan's tolerances. They check the formulas on concrete numbers; the proof is the evidence for the claim. The run's script labels the checks by the record's first wordings: H40 (T4 to T7 and the result R1) checks C19; H39 (T1 to T3) checks the formulas of C18, which the same plan tested in an earlier wording.",
   "availability": {
    "level": "public"
   },
   "lab_runs": [
    "RN16"
   ],
   "results": [
    {
     "id": "R1",
     "variable": "distributions compared with the asymptotic values recorded by RN7",
     "value": 16,
     "conditions": {
      "hypothesis": "H40"
     }
    }
   ],
   "tables": [
    {
     "id": "T1",
     "description": "distance of sMAPE's signal-to-noise ratio over its large-spread formula from 1, distance of the absolute error's signal-to-noise ratio over its limit from 1 by error_distribution, c1, c2, sigma; hypothesis: H39 (lab runs RN16)",
     "conditions": {
      "hypothesis": "H39"
     },
     "row_labels": [
      "error_distribution",
      "c1",
      "c2",
      "sigma"
     ],
     "columns": [
      "distance of sMAPE's signal-to-noise ratio over its large-spread formula from 1",
      "distance of the absolute error's signal-to-noise ratio over its limit from 1"
     ],
     "rows": [
      {
       "id": "r1",
       "label": [
        "normal",
        0.0,
        0.5,
        200
       ],
       "values": [
        5.266e-05,
        0.008727
       ]
      },
      {
       "id": "r2",
       "label": [
        "normal",
        0.0,
        0.5,
        2000
       ],
       "values": [
        2.749e-06,
        0.0008713
       ]
      },
      {
       "id": "r3",
       "label": [
        "normal",
        0.0,
        1.0,
        200
       ],
       "values": [
        0.0003989,
        0.004379
       ]
      },
      {
       "id": "r4",
       "label": [
        "normal",
        0.0,
        1.0,
        2000
       ],
       "values": [
        4.703e-05,
        0.0004373
       ]
      },
      {
       "id": "r5",
       "label": [
        "normal",
        -0.5,
        1.5,
        200
       ],
       "values": [
        0.00122,
        0.003402
       ]
      },
      {
       "id": "r6",
       "label": [
        "normal",
        -0.5,
        1.5,
        2000
       ],
       "values": [
        0.0001273,
        0.0003398
       ]
      },
      {
       "id": "r7",
       "label": [
        "Gumbel",
        0.0,
        0.5,
        200
       ],
       "values": [
        0.0001627,
        0.03529
       ]
      },
      {
       "id": "r8",
       "label": [
        "Gumbel",
        0.0,
        0.5,
        2000
       ],
       "values": [
        5.872e-07,
        0.00352
       ]
      },
      {
       "id": "r9",
       "label": [
        "Gumbel",
        0.0,
        1.0,
        200
       ],
       "values": [
        0.0001512,
        0.006295
       ]
      },
      {
       "id": "r10",
       "label": [
        "Gumbel",
        0.0,
        1.0,
        2000
       ],
       "values": [
        2.457e-05,
        0.0006281
       ]
      },
      {
       "id": "r11",
       "label": [
        "Gumbel",
        -0.5,
        1.5,
        200
       ],
       "values": [
        0.0003779,
        0.00326
       ]
      },
      {
       "id": "r12",
       "label": [
        "Gumbel",
        -0.5,
        1.5,
        2000
       ],
       "values": [
        4.713e-05,
        0.0003255
       ]
      }
     ],
     "artefact": "A6"
    },
    {
     "id": "T2",
     "description": "relative spread of the absolute log error's signal-to-noise ratio over sigma = 2, 10, 50 by error_distribution, c1, c2; hypothesis: H39 (lab runs RN16)",
     "conditions": {
      "hypothesis": "H39"
     },
     "row_labels": [
      "error_distribution",
      "c1",
      "c2"
     ],
     "columns": [
      "relative spread of the absolute log error's signal-to-noise ratio over sigma = 2, 10, 50"
     ],
     "rows": [
      {
       "id": "r1",
       "label": [
        "normal",
        0.0,
        0.5
       ],
       "values": [
        3.761e-15
       ]
      },
      {
       "id": "r2",
       "label": [
        "normal",
        0.0,
        1.0
       ],
       "values": [
        2.397e-16
       ]
      },
      {
       "id": "r3",
       "label": [
        "normal",
        -0.5,
        1.5
       ],
       "values": [
        1.579e-15
       ]
      },
      {
       "id": "r4",
       "label": [
        "Gumbel",
        0.0,
        0.5
       ],
       "values": [
        4.243e-13
       ]
      },
      {
       "id": "r5",
       "label": [
        "Gumbel",
        0.0,
        1.0
       ],
       "values": [
        7.932e-13
       ]
      },
      {
       "id": "r6",
       "label": [
        "Gumbel",
        -0.5,
        1.5
       ],
       "values": [
        1.43e-12
       ]
      }
     ],
     "artefact": "A6"
    },
    {
     "id": "T3",
     "description": "distance of the squared error's signal-to-noise ratio over its formula from 1, lognormal, distance of exp(sigma^2 / 2) times the squared error's ratio from 1, lognormal, formula by c1, c2, sigma; hypothesis: H39 (lab runs RN16)",
     "conditions": {
      "hypothesis": "H39"
     },
     "row_labels": [
      "c1",
      "c2",
      "sigma"
     ],
     "columns": [
      "distance of the squared error's signal-to-noise ratio over its formula from 1, lognormal",
      "distance of exp(sigma^2 / 2) times the squared error's ratio from 1, lognormal, formula"
     ],
     "rows": [
      {
       "id": "r1",
       "label": [
        0.0,
        0.5,
        0.5
       ],
       "values": [
        3.109e-15,
        null
       ]
      },
      {
       "id": "r2",
       "label": [
        0.0,
        1.0,
        0.5
       ],
       "values": [
        6.661e-16,
        null
       ]
      },
      {
       "id": "r3",
       "label": [
        -0.5,
        1.5,
        0.5
       ],
       "values": [
        1.11e-15,
        null
       ]
      },
      {
       "id": "r4",
       "label": [
        0.0,
        0.5,
        1.0
       ],
       "values": [
        1.554e-15,
        null
       ]
      },
      {
       "id": "r5",
       "label": [
        0.0,
        1.0,
        1.0
       ],
       "values": [
        8.882e-16,
        null
       ]
      },
      {
       "id": "r6",
       "label": [
        -0.5,
        1.5,
        1.0
       ],
       "values": [
        5.551e-16,
        null
       ]
      },
      {
       "id": "r7",
       "label": [
        0.0,
        0.5,
        2.0
       ],
       "values": [
        1.409e-13,
        null
       ]
      },
      {
       "id": "r8",
       "label": [
        0.0,
        1.0,
        2.0
       ],
       "values": [
        7.55e-15,
        null
       ]
      },
      {
       "id": "r9",
       "label": [
        -0.5,
        1.5,
        2.0
       ],
       "values": [
        1.887e-15,
        null
       ]
      },
      {
       "id": "r10",
       "label": [
        0.0,
        1.0,
        4.0
       ],
       "values": [
        null,
        0.009325
       ]
      },
      {
       "id": "r11",
       "label": [
        0.0,
        1.0,
        6.0
       ],
       "values": [
        null,
        3.08e-06
       ]
      },
      {
       "id": "r12",
       "label": [
        0.0,
        1.0,
        8.0
       ],
       "values": [
        null,
        1.888e-11
       ]
      }
     ],
     "artefact": "A6"
    },
    {
     "id": "T4",
     "description": "large-sample relative efficiency, formula by family, spread; hypothesis: H40 (lab runs RN16)",
     "conditions": {
      "hypothesis": "H40"
     },
     "row_labels": [
      "family",
      "spread"
     ],
     "columns": [
      "large-sample relative efficiency, formula"
     ],
     "rows": [
      {
       "id": "r1",
       "label": [
        "lognormal",
        0.25
       ],
       "values": [
        0.9696
       ]
      },
      {
       "id": "r2",
       "label": [
        "lognormal",
        0.5
       ],
       "values": [
        0.8881
       ]
      },
      {
       "id": "r3",
       "label": [
        "lognormal",
        1.0
       ],
       "values": [
        0.6601
       ]
      },
      {
       "id": "r4",
       "label": [
        "lognormal",
        1.5
       ],
       "values": [
        0.454
       ]
      },
      {
       "id": "r5",
       "label": [
        "lognormal",
        2.0
       ],
       "values": [
        0.3073
       ]
      },
      {
       "id": "r6",
       "label": [
        "lognormal",
        3.0
       ],
       "values": [
        0.1474
       ]
      },
      {
       "id": "r7",
       "label": [
        "lognormal",
        4.0
       ],
       "values": [
        0.07778
       ]
      },
      {
       "id": "r8",
       "label": [
        "lognormal",
        6.0
       ],
       "values": [
        0.0279
       ]
      },
      {
       "id": "r9",
       "label": [
        "log-Laplace",
        0.25
       ],
       "values": [
        0.9969
       ]
      },
      {
       "id": "r10",
       "label": [
        "log-Laplace",
        0.5
       ],
       "values": [
        0.9765
       ]
      },
      {
       "id": "r11",
       "label": [
        "log-Laplace",
        1.0
       ],
       "values": [
        0.8953
       ]
      },
      {
       "id": "r12",
       "label": [
        "log-Laplace",
        1.5
       ],
       "values": [
        0.8036
       ]
      },
      {
       "id": "r13",
       "label": [
        "log-Laplace",
        2.0
       ],
       "values": [
        0.7207
       ]
      },
      {
       "id": "r14",
       "label": [
        "log-Laplace",
        3.0
       ],
       "values": [
        0.5903
       ]
      },
      {
       "id": "r15",
       "label": [
        "log-Laplace",
        4.0
       ],
       "values": [
        0.4969
       ]
      },
      {
       "id": "r16",
       "label": [
        "log-Laplace",
        6.0
       ],
       "values": [
        0.3755
       ]
      }
     ],
     "artefact": "A6"
    },
    {
     "id": "T5",
     "description": "distance of the large-sample relative efficiency from 1 at spread 0.01 by family; hypothesis: H40 (lab runs RN16)",
     "conditions": {
      "hypothesis": "H40"
     },
     "row_labels": [
      "family"
     ],
     "columns": [
      "distance of the large-sample relative efficiency from 1 at spread 0.01"
     ],
     "rows": [
      {
       "id": "r1",
       "label": [
        "lognormal"
       ],
       "values": [
        5e-05
       ]
      },
      {
       "id": "r2",
       "label": [
        "log-Laplace"
       ],
       "values": [
        1.249e-08
       ]
      }
     ],
     "artefact": "A6"
    },
    {
     "id": "T6",
     "description": "distributions where A is not below 2 g at the median, largest difference from the asymptotic values recorded by RN7; hypothesis: H40, distributions: 16 (lab runs RN16)",
     "conditions": {
      "hypothesis": "H40",
      "distributions": 16
     },
     "row_labels": [],
     "columns": [
      "distributions where A is not below 2 g at the median",
      "largest difference from the asymptotic values recorded by RN7"
     ],
     "rows": [
      {
       "id": "r1",
       "label": [],
       "values": [
        0,
        0.00014
       ]
      }
     ],
     "artefact": "A6"
    },
    {
     "id": "T7",
     "description": "large-sample relative efficiency times sigma^3 over 6 (ln 2)^2 sqrt(2 pi), lognormal, large-sample relative efficiency times b over 3, log-Laplace; hypothesis: H40, spread: 200 (lab runs RN16)",
     "conditions": {
      "hypothesis": "H40",
      "spread": 200
     },
     "row_labels": [],
     "columns": [
      "large-sample relative efficiency times sigma^3 over 6 (ln 2)^2 sqrt(2 pi), lognormal",
      "large-sample relative efficiency times b over 3, log-Laplace"
     ],
     "rows": [
      {
       "id": "r1",
       "label": [],
       "values": [
        0.9998,
        0.9906
       ]
      }
     ],
     "artefact": "A6"
    }
   ],
   "artefacts": [
    "A6"
   ],
   "setup": {
    "studied": [
     "D1",
     "D8"
    ],
    "lens": [
     "D32",
     "D9",
     "D19",
     "D11",
     "D10"
    ],
    "procedure": [
     "D31"
    ]
   },
   "execution": {
    "code_version": "6aa7a7ab87bf5c4e40d704a7492cd9af9ef481db",
    "software": {
     "python": "3.13.18",
     "platform": "Linux-6.18.44-fc-v595-x86_64-with-glibc2.39"
    }
   }
  },
  {
   "id": "E34",
   "profile": "quant",
   "design": "other",
   "description": "Lab runs RN15 under plan PL14. Design: Which forecast does msMAPE reward on intermittent demand, beside sMAPE and the Suilin loss? One run of python3 claimset-code/smape-target/msmape_intermittent.py --seed 9, in two parts, for five measures: sMAPE, msMAPE with nu = 0.1, 0.5 and 1, and the Suilin loss, on the 14 intermittent demand settings. (1) Exact, with no sampling: the forecast that minimises the expected term of each measure (sums for counts; Gauss-Legendre quadrature of the density of ln y with the zeros as a point mass for the lognormal settings), minimised over forecasts f >= 0 on a grid of 0, the integers and 4000 log-spaced points, refined by golden section; and, for the lognormal demand sizes, the share of zeros at which each measure switches to the forecast 0, by bisection. (2) Simulated: test sets of n = 12, 52 and 365 independent actuals (a year of months, of weeks and of days), 2000 test sets per setting and n; in each test set, the forecast with the lowest value of each measure, searched over 0, the test set's own values and a grid (step 0.05 up to the largest value plus 1 for counts; 400 log-spaced points for the lognormal settings). Analysis: Part 1, per setting and measure: the target, its quantile level and the expected score there, beside the setting's share of zeros, median, mode and mean, so that the functional the target coincides with can be read; per measure, the zero threshold. Part 2, per setting, measure and n: the share of test sets whose minimiser is 0, the share whose minimiser is the exact target of part 1 (counts), and the median of the minimisers. Reported as tables and one figure (the targets against the share of zeros, a panel per measure); no test is made and no threshold is set. The figure as drawn (A56) has a panel per family of settings and a line per measure.",
   "availability": {
    "level": "public"
   },
   "lab_runs": [
    "RN15"
   ],
   "tables": [
    {
     "id": "T1",
     "description": "The forecast each measure rewards (its target, computed from the distribution itself) on each intermittent demand setting, beside the setting's share of zeros, median, mode and mean; the measures are sMAPE, msMAPE with nu = 0.1, 0.5 and 1, and the Suilin loss",
     "row_labels": [
      "setting"
     ],
     "columns": [
      "share of zero outcomes",
      "median of the outcomes",
      "mode of the outcomes",
      "mean of the outcomes",
      {
       "variable": "target",
       "conditions": {
        "measure": "sMAPE"
       }
      },
      {
       "variable": "target",
       "conditions": {
        "measure": "msMAPE nu=0.1"
       }
      },
      {
       "variable": "target",
       "conditions": {
        "measure": "msMAPE nu=0.5"
       }
      },
      {
       "variable": "target",
       "conditions": {
        "measure": "msMAPE nu=1"
       }
      },
      {
       "variable": "target",
       "conditions": {
        "measure": "Suilin"
       }
      }
     ],
     "rows": [
      {
       "id": "r1",
       "label": [
        "Poisson mean=0.1"
       ],
       "values": [
        0.9048,
        0,
        0,
        0.1,
        0,
        0,
        0,
        0,
        0
       ]
      },
      {
       "id": "r2",
       "label": [
        "Poisson mean=0.3"
       ],
       "values": [
        0.7408,
        0,
        0,
        0.3,
        0,
        0,
        0,
        0,
        0
       ]
      },
      {
       "id": "r3",
       "label": [
        "Poisson mean=0.5"
       ],
       "values": [
        0.6065,
        0,
        0,
        0.5,
        0,
        0,
        0,
        1,
        0
       ]
      },
      {
       "id": "r4",
       "label": [
        "Poisson mean=0.7"
       ],
       "values": [
        0.4966,
        1,
        0,
        0.7,
        0,
        0,
        1,
        1,
        0
       ]
      },
      {
       "id": "r5",
       "label": [
        "Poisson mean=1"
       ],
       "values": [
        0.3679,
        1,
        0,
        1,
        1,
        1,
        1,
        1,
        1
       ]
      },
      {
       "id": "r6",
       "label": [
        "Poisson mean=1.5"
       ],
       "values": [
        0.2231,
        1,
        1,
        1.5,
        2,
        2,
        1,
        1,
        2
       ]
      },
      {
       "id": "r7",
       "label": [
        "Poisson mean=3"
       ],
       "values": [
        0.04979,
        3,
        2,
        3.0,
        3,
        3,
        3,
        3,
        3
       ]
      },
      {
       "id": "r8",
       "label": [
        "negative binomial mean=0.5 size=0.5"
       ],
       "values": [
        0.7071,
        0,
        0,
        0.5,
        0,
        0,
        0,
        0,
        0
       ]
      },
      {
       "id": "r9",
       "label": [
        "negative binomial mean=1 size=0.5"
       ],
       "values": [
        0.5774,
        0,
        0,
        1.0,
        0,
        0,
        0,
        1,
        0
       ]
      },
      {
       "id": "r10",
       "label": [
        "negative binomial mean=2 size=0.5"
       ],
       "values": [
        0.4472,
        1,
        0,
        2.0,
        0,
        0,
        1,
        1,
        0
       ]
      },
      {
       "id": "r11",
       "label": [
        "lognormal sizes with zeros p0=0.2"
       ],
       "values": [
        0.2,
        0.7271,
        null,
        1.319,
        1.0,
        0.922,
        0.8021,
        0.8672,
        0.9054
       ]
      },
      {
       "id": "r12",
       "label": [
        "lognormal sizes with zeros p0=0.4"
       ],
       "values": [
        0.4,
        0.3801,
        null,
        0.9892,
        0,
        0.7791,
        0.5271,
        0.5932,
        0.1702
       ]
      },
      {
       "id": "r13",
       "label": [
        "lognormal sizes with zeros p0=0.6"
       ],
       "values": [
        0.6,
        0,
        null,
        0.6595,
        0,
        0,
        0,
        0,
        0
       ]
      },
      {
       "id": "r14",
       "label": [
        "lognormal sizes with zeros p0=0.8"
       ],
       "values": [
        0.8,
        0,
        null,
        0.3297,
        0,
        0,
        0,
        0,
        0
       ]
      }
     ],
     "artefact": "A55"
    },
    {
     "id": "T2",
     "description": "The share of zeros at which each measure's target switches to the forecast 0, for lognormal demand sizes (median 1, sigma 1) with zeros",
     "row_labels": [
      "setting"
     ],
     "columns": [
      {
       "variable": "zero threshold",
       "conditions": {
        "measure": "sMAPE"
       }
      },
      {
       "variable": "zero threshold",
       "conditions": {
        "measure": "msMAPE nu=0.1"
       }
      },
      {
       "variable": "zero threshold",
       "conditions": {
        "measure": "msMAPE nu=0.5"
       }
      },
      {
       "variable": "zero threshold",
       "conditions": {
        "measure": "msMAPE nu=1"
       }
      },
      {
       "variable": "zero threshold",
       "conditions": {
        "measure": "Suilin"
       }
      }
     ],
     "rows": [
      {
       "id": "r1",
       "label": [
        "lognormal sizes with zeros"
       ],
       "values": [
        0.394,
        0.4191,
        0.5192,
        0.5647,
        0.47
       ]
      }
     ],
     "artefact": "A55"
    },
    {
     "id": "T3",
     "description": "The quantile level of each measure's target and the expected score there (200 times the expected term), for each setting",
     "row_labels": [
      "setting"
     ],
     "columns": [
      {
       "variable": "quantile level of the target",
       "conditions": {
        "measure": "sMAPE"
       }
      },
      {
       "variable": "quantile level of the target",
       "conditions": {
        "measure": "msMAPE nu=0.1"
       }
      },
      {
       "variable": "quantile level of the target",
       "conditions": {
        "measure": "msMAPE nu=0.5"
       }
      },
      {
       "variable": "quantile level of the target",
       "conditions": {
        "measure": "msMAPE nu=1"
       }
      },
      {
       "variable": "quantile level of the target",
       "conditions": {
        "measure": "Suilin"
       }
      },
      {
       "variable": "expected score at the target",
       "conditions": {
        "measure": "sMAPE"
       }
      },
      {
       "variable": "expected score at the target",
       "conditions": {
        "measure": "msMAPE nu=0.1"
       }
      },
      {
       "variable": "expected score at the target",
       "conditions": {
        "measure": "msMAPE nu=0.5"
       }
      },
      {
       "variable": "expected score at the target",
       "conditions": {
        "measure": "msMAPE nu=1"
       }
      },
      {
       "variable": "expected score at the target",
       "conditions": {
        "measure": "Suilin"
       }
      }
     ],
     "rows": [
      {
       "id": "r1",
       "label": [
        "Poisson mean=0.1"
       ],
       "values": [
        0.9048,
        0.9048,
        0.9048,
        0.9048,
        0.9048,
        19.03,
        19.03,
        19.03,
        19.03,
        17.34
       ]
      },
      {
       "id": "r2",
       "label": [
        "Poisson mean=0.3"
       ],
       "values": [
        0.7408,
        0.7408,
        0.7408,
        0.7408,
        0.7408,
        51.84,
        51.84,
        51.84,
        51.84,
        47.46
       ]
      },
      {
       "id": "r3",
       "label": [
        "Poisson mean=0.5"
       ],
       "values": [
        0.6065,
        0.6065,
        0.6065,
        0.9098,
        0.6065,
        78.69,
        78.69,
        78.69,
        67.18,
        72.37
       ]
      },
      {
       "id": "r4",
       "label": [
        "Poisson mean=0.7"
       ],
       "values": [
        0.4966,
        0.4966,
        0.8442,
        0.8442,
        0.4966,
        100.7,
        100.7,
        77.86,
        61.31,
        92.99
       ]
      },
      {
       "id": "r5",
       "label": [
        "Poisson mean=1"
       ],
       "values": [
        0.7358,
        0.7358,
        0.7358,
        0.7358,
        0.7358,
        94.3,
        87.61,
        69.78,
        57.52,
        87.03
       ]
      },
      {
       "id": "r6",
       "label": [
        "Poisson mean=1.5"
       ],
       "values": [
        0.8088,
        0.8088,
        0.5578,
        0.5578,
        0.8088,
        76.77,
        74.64,
        67.21,
        59.77,
        73.75
       ]
      },
      {
       "id": "r7",
       "label": [
        "Poisson mean=3"
       ],
       "values": [
        0.6472,
        0.6472,
        0.6472,
        0.6472,
        0.6472,
        49.91,
        49.59,
        48.49,
        47.42,
        48.86
       ]
      },
      {
       "id": "r8",
       "label": [
        "negative binomial mean=0.5 size=0.5"
       ],
       "values": [
        0.7071,
        0.7071,
        0.7071,
        0.7071,
        0.7071,
        58.58,
        58.58,
        58.58,
        58.58,
        54.46
       ]
      },
      {
       "id": "r9",
       "label": [
        "negative binomial mean=1 size=0.5"
       ],
       "values": [
        0.5774,
        0.5774,
        0.5774,
        0.7698,
        0.5774,
        84.53,
        84.53,
        84.53,
        80.38,
        79.46
       ]
      },
      {
       "id": "r10",
       "label": [
        "negative binomial mean=2 size=0.5"
       ],
       "values": [
        0.4472,
        0.4472,
        0.6261,
        0.6261,
        0.4472,
        110.6,
        110.6,
        101.9,
        86.95,
        105.2
       ]
      },
      {
       "id": "r11",
       "label": [
        "lognormal sizes with zeros p0=0.2"
       ],
       "values": [
        0.6,
        0.5741,
        0.5302,
        0.5547,
        0.5683,
        95.95,
        92.15,
        78.37,
        66.53,
        89.49
       ]
      },
      {
       "id": "r12",
       "label": [
        "lognormal sizes with zeros p0=0.4"
       ],
       "values": [
        0.4,
        0.6409,
        0.5566,
        0.5805,
        0.423,
        120,
        113.9,
        87.15,
        70.9,
        94.45
       ]
      },
      {
       "id": "r13",
       "label": [
        "lognormal sizes with zeros p0=0.6"
       ],
       "values": [
        0.6,
        0.6,
        0.6,
        0.6,
        0.6,
        80,
        79.78,
        72.4,
        60.93,
        65.9
       ]
      },
      {
       "id": "r14",
       "label": [
        "lognormal sizes with zeros p0=0.8"
       ],
       "values": [
        0.8,
        0.8,
        0.8,
        0.8,
        0.8,
        40.0,
        39.89,
        36.2,
        30.46,
        32.95
       ]
      }
     ],
     "artefact": "A55"
    },
    {
     "id": "T4",
     "description": "Test sets: the share of 2000 test sets of n independent actuals in which the forecast with the lowest score is 0, for each setting, n and measure",
     "row_labels": [
      "setting",
      "n"
     ],
     "columns": [
      {
       "variable": "share of test sets whose minimiser is 0",
       "conditions": {
        "measure": "sMAPE"
       }
      },
      {
       "variable": "share of test sets whose minimiser is 0",
       "conditions": {
        "measure": "msMAPE nu=0.1"
       }
      },
      {
       "variable": "share of test sets whose minimiser is 0",
       "conditions": {
        "measure": "msMAPE nu=0.5"
       }
      },
      {
       "variable": "share of test sets whose minimiser is 0",
       "conditions": {
        "measure": "msMAPE nu=1"
       }
      },
      {
       "variable": "share of test sets whose minimiser is 0",
       "conditions": {
        "measure": "Suilin"
       }
      }
     ],
     "rows": [
      {
       "id": "r1",
       "label": [
        "Poisson mean=0.1",
        12
       ],
       "values": [
        1,
        0.9995,
        0.996,
        0.975,
        1
       ]
      },
      {
       "id": "r2",
       "label": [
        "Poisson mean=0.1",
        52
       ],
       "values": [
        1,
        1,
        1,
        1,
        1
       ]
      },
      {
       "id": "r3",
       "label": [
        "Poisson mean=0.1",
        365
       ],
       "values": [
        1,
        1,
        1,
        1,
        1
       ]
      },
      {
       "id": "r4",
       "label": [
        "Poisson mean=0.3",
        12
       ],
       "values": [
        0.979,
        0.937,
        0.831,
        0.7145,
        0.979
       ]
      },
      {
       "id": "r5",
       "label": [
        "Poisson mean=0.3",
        52
       ],
       "values": [
        1,
        1,
        0.9915,
        0.9155,
        1
       ]
      },
      {
       "id": "r6",
       "label": [
        "Poisson mean=0.3",
        365
       ],
       "values": [
        1,
        1,
        1,
        0.9995,
        1
       ]
      },
      {
       "id": "r7",
       "label": [
        "Poisson mean=0.5",
        12
       ],
       "values": [
        0.848,
        0.7525,
        0.539,
        0.378,
        0.848
       ]
      },
      {
       "id": "r8",
       "label": [
        "Poisson mean=0.5",
        52
       ],
       "values": [
        0.9685,
        0.923,
        0.6515,
        0.288,
        0.966
       ]
      },
      {
       "id": "r9",
       "label": [
        "Poisson mean=0.5",
        365
       ],
       "values": [
        1,
        1,
        0.8445,
        0.067,
        1
       ]
      },
      {
       "id": "r10",
       "label": [
        "Poisson mean=0.7",
        12
       ],
       "values": [
        0.6045,
        0.515,
        0.2915,
        0.1555,
        0.6045
       ]
      },
      {
       "id": "r11",
       "label": [
        "Poisson mean=0.7",
        52
       ],
       "values": [
        0.6575,
        0.523,
        0.1525,
        0.0265,
        0.6455
       ]
      },
      {
       "id": "r12",
       "label": [
        "Poisson mean=0.7",
        365
       ],
       "values": [
        0.844,
        0.56,
        0.0025,
        0,
        0.8295
       ]
      },
      {
       "id": "r13",
       "label": [
        "Poisson mean=1",
        12
       ],
       "values": [
        0.256,
        0.2265,
        0.0895,
        0.037,
        0.256
       ]
      },
      {
       "id": "r14",
       "label": [
        "Poisson mean=1",
        52
       ],
       "values": [
        0.1015,
        0.051,
        0.003,
        0,
        0.1
       ]
      },
      {
       "id": "r15",
       "label": [
        "Poisson mean=1",
        365
       ],
       "values": [
        0,
        0,
        0,
        0,
        0
       ]
      },
      {
       "id": "r16",
       "label": [
        "Poisson mean=1.5",
        12
       ],
       "values": [
        0.0325,
        0.03,
        0.0075,
        0.002,
        0.0325
       ]
      },
      {
       "id": "r17",
       "label": [
        "Poisson mean=1.5",
        52
       ],
       "values": [
        0.0005,
        0.0005,
        0,
        0,
        0.0005
       ]
      },
      {
       "id": "r18",
       "label": [
        "Poisson mean=1.5",
        365
       ],
       "values": [
        0,
        0,
        0,
        0,
        0
       ]
      },
      {
       "id": "r19",
       "label": [
        "Poisson mean=3",
        12
       ],
       "values": [
        0,
        0,
        0,
        0,
        0
       ]
      },
      {
       "id": "r20",
       "label": [
        "Poisson mean=3",
        52
       ],
       "values": [
        0,
        0,
        0,
        0,
        0
       ]
      },
      {
       "id": "r21",
       "label": [
        "Poisson mean=3",
        365
       ],
       "values": [
        0,
        0,
        0,
        0,
        0
       ]
      },
      {
       "id": "r22",
       "label": [
        "negative binomial mean=0.5 size=0.5",
        12
       ],
       "values": [
        0.971,
        0.952,
        0.857,
        0.721,
        0.9705
       ]
      },
      {
       "id": "r23",
       "label": [
        "negative binomial mean=0.5 size=0.5",
        52
       ],
       "values": [
        1,
        1,
        0.9895,
        0.9095,
        1
       ]
      },
      {
       "id": "r24",
       "label": [
        "negative binomial mean=0.5 size=0.5",
        365
       ],
       "values": [
        1,
        1,
        1,
        1,
        1
       ]
      },
      {
       "id": "r25",
       "label": [
        "negative binomial mean=1 size=0.5",
        12
       ],
       "values": [
        0.8035,
        0.78,
        0.621,
        0.4435,
        0.808
       ]
      },
      {
       "id": "r26",
       "label": [
        "negative binomial mean=1 size=0.5",
        52
       ],
       "values": [
        0.975,
        0.9625,
        0.7675,
        0.405,
        0.974
       ]
      },
      {
       "id": "r27",
       "label": [
        "negative binomial mean=1 size=0.5",
        365
       ],
       "values": [
        1,
        1,
        0.975,
        0.256,
        1
       ]
      },
      {
       "id": "r28",
       "label": [
        "negative binomial mean=2 size=0.5",
        12
       ],
       "values": [
        0.567,
        0.5265,
        0.384,
        0.2405,
        0.5765
       ]
      },
      {
       "id": "r29",
       "label": [
        "negative binomial mean=2 size=0.5",
        52
       ],
       "values": [
        0.707,
        0.635,
        0.3055,
        0.068,
        0.7255
       ]
      },
      {
       "id": "r30",
       "label": [
        "negative binomial mean=2 size=0.5",
        365
       ],
       "values": [
        0.932,
        0.8475,
        0.1115,
        0,
        0.9475
       ]
      },
      {
       "id": "r31",
       "label": [
        "lognormal sizes with zeros p0=0.2",
        12
       ],
       "values": [
        0.046,
        0.026,
        0.0055,
        0.002,
        0.016
       ]
      },
      {
       "id": "r32",
       "label": [
        "lognormal sizes with zeros p0=0.2",
        52
       ],
       "values": [
        0,
        0,
        0,
        0,
        0
       ]
      },
      {
       "id": "r33",
       "label": [
        "lognormal sizes with zeros p0=0.2",
        365
       ],
       "values": [
        0,
        0,
        0,
        0,
        0
       ]
      },
      {
       "id": "r34",
       "label": [
        "lognormal sizes with zeros p0=0.4",
        12
       ],
       "values": [
        0.4655,
        0.374,
        0.1735,
        0.119,
        0.303
       ]
      },
      {
       "id": "r35",
       "label": [
        "lognormal sizes with zeros p0=0.4",
        52
       ],
       "values": [
        0.5235,
        0.36,
        0.045,
        0.012,
        0.172
       ]
      },
      {
       "id": "r36",
       "label": [
        "lognormal sizes with zeros p0=0.4",
        365
       ],
       "values": [
        0.5665,
        0.207,
        0,
        0,
        0.004
       ]
      },
      {
       "id": "r37",
       "label": [
        "lognormal sizes with zeros p0=0.6",
        12
       ],
       "values": [
        0.896,
        0.8565,
        0.663,
        0.5805,
        0.8055
       ]
      },
      {
       "id": "r38",
       "label": [
        "lognormal sizes with zeros p0=0.6",
        52
       ],
       "values": [
        0.9975,
        0.993,
        0.844,
        0.6765,
        0.9705
       ]
      },
      {
       "id": "r39",
       "label": [
        "lognormal sizes with zeros p0=0.6",
        365
       ],
       "values": [
        1,
        1,
        0.999,
        0.892,
        1
       ]
      },
      {
       "id": "r40",
       "label": [
        "lognormal sizes with zeros p0=0.8",
        12
       ],
       "values": [
        0.9975,
        0.9955,
        0.9725,
        0.959,
        0.991
       ]
      },
      {
       "id": "r41",
       "label": [
        "lognormal sizes with zeros p0=0.8",
        52
       ],
       "values": [
        1,
        1,
        1,
        0.9995,
        1
       ]
      },
      {
       "id": "r42",
       "label": [
        "lognormal sizes with zeros p0=0.8",
        365
       ],
       "values": [
        1,
        1,
        1,
        1,
        1
       ]
      }
     ],
     "artefact": "A55"
    },
    {
     "id": "T5",
     "description": "Test sets: the share of 2000 test sets of n independent actuals in which the forecast with the lowest score is the measure's target of T1, for each count setting, n and measure",
     "row_labels": [
      "setting",
      "n"
     ],
     "columns": [
      {
       "variable": "share of test sets whose minimiser is the target",
       "conditions": {
        "measure": "sMAPE"
       }
      },
      {
       "variable": "share of test sets whose minimiser is the target",
       "conditions": {
        "measure": "msMAPE nu=0.1"
       }
      },
      {
       "variable": "share of test sets whose minimiser is the target",
       "conditions": {
        "measure": "msMAPE nu=0.5"
       }
      },
      {
       "variable": "share of test sets whose minimiser is the target",
       "conditions": {
        "measure": "msMAPE nu=1"
       }
      },
      {
       "variable": "share of test sets whose minimiser is the target",
       "conditions": {
        "measure": "Suilin"
       }
      }
     ],
     "rows": [
      {
       "id": "r1",
       "label": [
        "Poisson mean=0.1",
        12
       ],
       "values": [
        1,
        0.9995,
        0.996,
        0.975,
        1
       ]
      },
      {
       "id": "r2",
       "label": [
        "Poisson mean=0.1",
        52
       ],
       "values": [
        1,
        1,
        1,
        1,
        1
       ]
      },
      {
       "id": "r3",
       "label": [
        "Poisson mean=0.1",
        365
       ],
       "values": [
        1,
        1,
        1,
        1,
        1
       ]
      },
      {
       "id": "r4",
       "label": [
        "Poisson mean=0.3",
        12
       ],
       "values": [
        0.979,
        0.937,
        0.831,
        0.7145,
        0.979
       ]
      },
      {
       "id": "r5",
       "label": [
        "Poisson mean=0.3",
        52
       ],
       "values": [
        1,
        1,
        0.9915,
        0.9155,
        1
       ]
      },
      {
       "id": "r6",
       "label": [
        "Poisson mean=0.3",
        365
       ],
       "values": [
        1,
        1,
        1,
        0.9995,
        1
       ]
      },
      {
       "id": "r7",
       "label": [
        "Poisson mean=0.5",
        12
       ],
       "values": [
        0.848,
        0.7525,
        0.539,
        0.547,
        0.848
       ]
      },
      {
       "id": "r8",
       "label": [
        "Poisson mean=0.5",
        52
       ],
       "values": [
        0.9685,
        0.923,
        0.6515,
        0.712,
        0.966
       ]
      },
      {
       "id": "r9",
       "label": [
        "Poisson mean=0.5",
        365
       ],
       "values": [
        1,
        1,
        0.8445,
        0.933,
        1
       ]
      },
      {
       "id": "r10",
       "label": [
        "Poisson mean=0.7",
        12
       ],
       "values": [
        0.6045,
        0.515,
        0.691,
        0.8035,
        0.6045
       ]
      },
      {
       "id": "r11",
       "label": [
        "Poisson mean=0.7",
        52
       ],
       "values": [
        0.6575,
        0.523,
        0.8475,
        0.9735,
        0.6455
       ]
      },
      {
       "id": "r12",
       "label": [
        "Poisson mean=0.7",
        365
       ],
       "values": [
        0.844,
        0.56,
        0.9975,
        1,
        0.8295
       ]
      },
      {
       "id": "r13",
       "label": [
        "Poisson mean=1",
        12
       ],
       "values": [
        0.5335,
        0.5665,
        0.775,
        0.8565,
        0.537
       ]
      },
      {
       "id": "r14",
       "label": [
        "Poisson mean=1",
        52
       ],
       "values": [
        0.7705,
        0.8855,
        0.9825,
        0.993,
        0.8415
       ]
      },
      {
       "id": "r15",
       "label": [
        "Poisson mean=1",
        365
       ],
       "values": [
        0.9985,
        1,
        1,
        1,
        1
       ]
      },
      {
       "id": "r16",
       "label": [
        "Poisson mean=1.5",
        12
       ],
       "values": [
        0.5675,
        0.569,
        0.457,
        0.501,
        0.5685
       ]
      },
      {
       "id": "r17",
       "label": [
        "Poisson mean=1.5",
        52
       ],
       "values": [
        0.7395,
        0.657,
        0.5585,
        0.6325,
        0.6565
       ]
      },
      {
       "id": "r18",
       "label": [
        "Poisson mean=1.5",
        365
       ],
       "values": [
        0.9635,
        0.867,
        0.5965,
        0.762,
        0.8665
       ]
      },
      {
       "id": "r19",
       "label": [
        "Poisson mean=3",
        12
       ],
       "values": [
        0.552,
        0.5475,
        0.5515,
        0.554,
        0.548
       ]
      },
      {
       "id": "r20",
       "label": [
        "Poisson mean=3",
        52
       ],
       "values": [
        0.8905,
        0.888,
        0.884,
        0.8795,
        0.889
       ]
      },
      {
       "id": "r21",
       "label": [
        "Poisson mean=3",
        365
       ],
       "values": [
        1,
        1,
        1,
        1,
        1
       ]
      },
      {
       "id": "r22",
       "label": [
        "negative binomial mean=0.5 size=0.5",
        12
       ],
       "values": [
        0.971,
        0.952,
        0.857,
        0.721,
        0.9705
       ]
      },
      {
       "id": "r23",
       "label": [
        "negative binomial mean=0.5 size=0.5",
        52
       ],
       "values": [
        1,
        1,
        0.9895,
        0.9095,
        1
       ]
      },
      {
       "id": "r24",
       "label": [
        "negative binomial mean=0.5 size=0.5",
        365
       ],
       "values": [
        1,
        1,
        1,
        1,
        1
       ]
      },
      {
       "id": "r25",
       "label": [
        "negative binomial mean=1 size=0.5",
        12
       ],
       "values": [
        0.8035,
        0.78,
        0.621,
        0.471,
        0.808
       ]
      },
      {
       "id": "r26",
       "label": [
        "negative binomial mean=1 size=0.5",
        52
       ],
       "values": [
        0.975,
        0.9625,
        0.7675,
        0.592,
        0.974
       ]
      },
      {
       "id": "r27",
       "label": [
        "negative binomial mean=1 size=0.5",
        365
       ],
       "values": [
        1,
        1,
        0.975,
        0.744,
        1
       ]
      },
      {
       "id": "r28",
       "label": [
        "negative binomial mean=2 size=0.5",
        12
       ],
       "values": [
        0.567,
        0.5265,
        0.3345,
        0.4825,
        0.5765
       ]
      },
      {
       "id": "r29",
       "label": [
        "negative binomial mean=2 size=0.5",
        52
       ],
       "values": [
        0.707,
        0.635,
        0.473,
        0.782,
        0.7255
       ]
      },
      {
       "id": "r30",
       "label": [
        "negative binomial mean=2 size=0.5",
        365
       ],
       "values": [
        0.932,
        0.8475,
        0.855,
        0.998,
        0.9475
       ]
      }
     ],
     "artefact": "A55"
    },
    {
     "id": "T6",
     "description": "Test sets: the median, over 2000 test sets of n independent actuals, of the forecast with the lowest score, for each setting, n and measure",
     "row_labels": [
      "setting",
      "n"
     ],
     "columns": [
      {
       "variable": "median of the test-set minimisers",
       "conditions": {
        "measure": "sMAPE"
       }
      },
      {
       "variable": "median of the test-set minimisers",
       "conditions": {
        "measure": "msMAPE nu=0.1"
       }
      },
      {
       "variable": "median of the test-set minimisers",
       "conditions": {
        "measure": "msMAPE nu=0.5"
       }
      },
      {
       "variable": "median of the test-set minimisers",
       "conditions": {
        "measure": "msMAPE nu=1"
       }
      },
      {
       "variable": "median of the test-set minimisers",
       "conditions": {
        "measure": "Suilin"
       }
      }
     ],
     "rows": [
      {
       "id": "r1",
       "label": [
        "Poisson mean=0.1",
        12
       ],
       "values": [
        0,
        0,
        0,
        0,
        0
       ]
      },
      {
       "id": "r2",
       "label": [
        "Poisson mean=0.1",
        52
       ],
       "values": [
        0,
        0,
        0,
        0,
        0
       ]
      },
      {
       "id": "r3",
       "label": [
        "Poisson mean=0.1",
        365
       ],
       "values": [
        0,
        0,
        0,
        0,
        0
       ]
      },
      {
       "id": "r4",
       "label": [
        "Poisson mean=0.3",
        12
       ],
       "values": [
        0,
        0,
        0,
        0,
        0
       ]
      },
      {
       "id": "r5",
       "label": [
        "Poisson mean=0.3",
        52
       ],
       "values": [
        0,
        0,
        0,
        0,
        0
       ]
      },
      {
       "id": "r6",
       "label": [
        "Poisson mean=0.3",
        365
       ],
       "values": [
        0,
        0,
        0,
        0,
        0
       ]
      },
      {
       "id": "r7",
       "label": [
        "Poisson mean=0.5",
        12
       ],
       "values": [
        0,
        0,
        0,
        1,
        0
       ]
      },
      {
       "id": "r8",
       "label": [
        "Poisson mean=0.5",
        52
       ],
       "values": [
        0,
        0,
        0,
        1,
        0
       ]
      },
      {
       "id": "r9",
       "label": [
        "Poisson mean=0.5",
        365
       ],
       "values": [
        0,
        0,
        0,
        1,
        0
       ]
      },
      {
       "id": "r10",
       "label": [
        "Poisson mean=0.7",
        12
       ],
       "values": [
        0,
        0,
        1,
        1,
        0
       ]
      },
      {
       "id": "r11",
       "label": [
        "Poisson mean=0.7",
        52
       ],
       "values": [
        0,
        0,
        1,
        1,
        0
       ]
      },
      {
       "id": "r12",
       "label": [
        "Poisson mean=0.7",
        365
       ],
       "values": [
        0,
        0,
        1,
        1,
        0
       ]
      },
      {
       "id": "r13",
       "label": [
        "Poisson mean=1",
        12
       ],
       "values": [
        1,
        1,
        1,
        1,
        1
       ]
      },
      {
       "id": "r14",
       "label": [
        "Poisson mean=1",
        52
       ],
       "values": [
        1,
        1,
        1,
        1,
        1
       ]
      },
      {
       "id": "r15",
       "label": [
        "Poisson mean=1",
        365
       ],
       "values": [
        1,
        1,
        1,
        1,
        1
       ]
      },
      {
       "id": "r16",
       "label": [
        "Poisson mean=1.5",
        12
       ],
       "values": [
        2,
        2,
        2,
        1,
        2
       ]
      },
      {
       "id": "r17",
       "label": [
        "Poisson mean=1.5",
        52
       ],
       "values": [
        2,
        2,
        1,
        1,
        2
       ]
      },
      {
       "id": "r18",
       "label": [
        "Poisson mean=1.5",
        365
       ],
       "values": [
        2,
        2,
        1,
        1,
        2
       ]
      },
      {
       "id": "r19",
       "label": [
        "Poisson mean=3",
        12
       ],
       "values": [
        3,
        3,
        3,
        3,
        3
       ]
      },
      {
       "id": "r20",
       "label": [
        "Poisson mean=3",
        52
       ],
       "values": [
        3,
        3,
        3,
        3,
        3
       ]
      },
      {
       "id": "r21",
       "label": [
        "Poisson mean=3",
        365
       ],
       "values": [
        3,
        3,
        3,
        3,
        3
       ]
      },
      {
       "id": "r22",
       "label": [
        "negative binomial mean=0.5 size=0.5",
        12
       ],
       "values": [
        0,
        0,
        0,
        0,
        0
       ]
      },
      {
       "id": "r23",
       "label": [
        "negative binomial mean=0.5 size=0.5",
        52
       ],
       "values": [
        0,
        0,
        0,
        0,
        0
       ]
      },
      {
       "id": "r24",
       "label": [
        "negative binomial mean=0.5 size=0.5",
        365
       ],
       "values": [
        0,
        0,
        0,
        0,
        0
       ]
      },
      {
       "id": "r25",
       "label": [
        "negative binomial mean=1 size=0.5",
        12
       ],
       "values": [
        0,
        0,
        0,
        1,
        0
       ]
      },
      {
       "id": "r26",
       "label": [
        "negative binomial mean=1 size=0.5",
        52
       ],
       "values": [
        0,
        0,
        0,
        1,
        0
       ]
      },
      {
       "id": "r27",
       "label": [
        "negative binomial mean=1 size=0.5",
        365
       ],
       "values": [
        0,
        0,
        0,
        1,
        0
       ]
      },
      {
       "id": "r28",
       "label": [
        "negative binomial mean=2 size=0.5",
        12
       ],
       "values": [
        0,
        0,
        1,
        1,
        0
       ]
      },
      {
       "id": "r29",
       "label": [
        "negative binomial mean=2 size=0.5",
        52
       ],
       "values": [
        0,
        0,
        1,
        1,
        0
       ]
      },
      {
       "id": "r30",
       "label": [
        "negative binomial mean=2 size=0.5",
        365
       ],
       "values": [
        0,
        0,
        1,
        1,
        0
       ]
      },
      {
       "id": "r31",
       "label": [
        "lognormal sizes with zeros p0=0.2",
        12
       ],
       "values": [
        0.9791,
        0.917,
        0.8181,
        0.8745,
        0.8357
       ]
      },
      {
       "id": "r32",
       "label": [
        "lognormal sizes with zeros p0=0.2",
        52
       ],
       "values": [
        1.003,
        0.9223,
        0.7943,
        0.8633,
        0.8825
       ]
      },
      {
       "id": "r33",
       "label": [
        "lognormal sizes with zeros p0=0.2",
        365
       ],
       "values": [
        1.003,
        0.9257,
        0.8006,
        0.8684,
        0.907
       ]
      },
      {
       "id": "r34",
       "label": [
        "lognormal sizes with zeros p0=0.4",
        12
       ],
       "values": [
        0.4646,
        0.5008,
        0.5196,
        0.6003,
        0.1931
       ]
      },
      {
       "id": "r35",
       "label": [
        "lognormal sizes with zeros p0=0.4",
        52
       ],
       "values": [
        0,
        0.63,
        0.5224,
        0.5913,
        0.1681
       ]
      },
      {
       "id": "r36",
       "label": [
        "lognormal sizes with zeros p0=0.4",
        365
       ],
       "values": [
        0,
        0.7535,
        0.5278,
        0.5938,
        0.1719
       ]
      },
      {
       "id": "r37",
       "label": [
        "lognormal sizes with zeros p0=0.6",
        12
       ],
       "values": [
        0,
        0,
        0,
        0,
        0
       ]
      },
      {
       "id": "r38",
       "label": [
        "lognormal sizes with zeros p0=0.6",
        52
       ],
       "values": [
        0,
        0,
        0,
        0,
        0
       ]
      },
      {
       "id": "r39",
       "label": [
        "lognormal sizes with zeros p0=0.6",
        365
       ],
       "values": [
        0,
        0,
        0,
        0,
        0
       ]
      },
      {
       "id": "r40",
       "label": [
        "lognormal sizes with zeros p0=0.8",
        12
       ],
       "values": [
        0,
        0,
        0,
        0,
        0
       ]
      },
      {
       "id": "r41",
       "label": [
        "lognormal sizes with zeros p0=0.8",
        52
       ],
       "values": [
        0,
        0,
        0,
        0,
        0
       ]
      },
      {
       "id": "r42",
       "label": [
        "lognormal sizes with zeros p0=0.8",
        365
       ],
       "values": [
        0,
        0,
        0,
        0,
        0
       ]
      }
     ],
     "artefact": "A55"
    }
   ],
   "artefacts": [
    "A6",
    "A55",
    "A56"
   ],
   "setup": {
    "studied": [
     "D33",
     "D1",
     "D28"
    ],
    "cases": [
     "D35"
    ],
    "procedure": [
     "D15"
    ],
    "lens": [
     "D34",
     "D6",
     "D29"
    ]
   },
   "execution": {
    "code_version": "d49b5fd17718af3efd407e840ee50bc9665e4300",
    "software": {
     "python": "3.13.18",
     "platform": "Linux-6.18.44-fc-v595-x86_64-with-glibc2.39"
    }
   },
   "results": []
  },
  {
   "id": "E18",
   "profile": "formal",
   "description": "Proof of C1 (section H1 of the proof document A1, rendered in A2). Only the ratio of the actual to the forecast matters, not their size.",
   "foundation": "Kolmogorov probability axioms and real analysis",
   "steps": [
    {
     "id": "S1",
     "statement": "Since $y$ and $f$ are positive, $|y| + |f| = y + f$. So $s = |y - f|/(y + f)$.",
     "rule": "algebra",
     "by": [
      "D1"
     ]
    },
    {
     "id": "S2",
     "statement": "Dividing the numerator and the denominator by $f > 0$ gives $\\dfrac{|y-f|/f}{(y+f)/f}$. Since $f > 0$, $|y - f|/f = |(y - f)/f| = |r - 1|$. And $(y+f)/f = r + 1$.",
     "rule": "algebra",
     "by": [
      "S1"
     ]
    },
    {
     "id": "S3",
     "statement": "So $s = |r - 1|/(r + 1)$. The contribution depends only on the ratio of actual to forecast.",
     "rule": "conclusion",
     "by": [
      "S2"
     ]
    }
   ],
   "artefacts": [
    "A1",
    "A2"
   ],
   "availability": {
    "level": "public"
   }
  },
  {
   "id": "E19",
   "profile": "formal",
   "description": "Proof of C2 (section H2 of the proof document A1, rendered in A2). sMAPE is symmetric in ratios. It favours a high forecast only when errors are compared in absolute terms.",
   "foundation": "Kolmogorov probability axioms and real analysis",
   "steps": [
    {
     "id": "S1",
     "statement": "By C1, $s = |r - 1|/(r + 1)$. Replacing $r$ by $1/r$ gives $\\dfrac{|1/r - 1|}{1/r + 1}$. Multiplying the numerator and the denominator by $r > 0$ gives $\\dfrac{|1 - r|}{1 + r}$, the same value.",
     "rule": "algebra",
     "by": [
      "C1"
     ]
    },
    {
     "id": "S2",
     "statement": "The forecast $f = ky$ has $r = 1/k$, and $f = y/k$ has $r = k$. By step 1 they score the same. This is (a).",
     "rule": "algebra",
     "by": [
      "S1"
     ]
    },
    {
     "id": "S3",
     "statement": "The forecasts $y + d$ and $y - d$ are both $d$ away from $y$, so both numerators are $d$. The denominators are $y + (y + d) = 2y + d$ and $y + (y - d) = 2y - d > 0$. The larger denominator gives the smaller value. This is (b).",
     "rule": "algebra",
     "by": [
      "S2"
     ]
    },
    {
     "id": "S4",
     "statement": "So sMAPE treats a forecast $k$ times too high and one $k$ times too low alike. It favours the high forecast only when the two are equally wrong in absolute terms.",
     "rule": "conclusion",
     "by": [
      "S3"
     ]
    }
   ],
   "artefacts": [
    "A1",
    "A2"
   ],
   "availability": {
    "level": "public"
   }
  },
  {
   "id": "E20",
   "profile": "formal",
   "description": "Proof of C3 (section H3 of the proof document A1, rendered in A2). sMAPE measures the error on the log scale. For small errors it is half the absolute log error; for large errors it levels off.",
   "foundation": "Kolmogorov probability axioms and real analysis",
   "steps": [
    {
     "id": "S1",
     "statement": "Write $z = \\ln y - \\ln f$, so that $r = y/f = e^z$. By C1, $s = \\dfrac{|e^z - 1|}{e^z + 1}$.",
     "rule": "definition",
     "by": [
      "C1"
     ]
    },
    {
     "id": "S2",
     "statement": "Multiplying the numerator and the denominator of $\\dfrac{e^z - 1}{e^z + 1}$ by $e^{-z/2}$ gives $\\dfrac{e^{z/2} - e^{-z/2}}{e^{z/2} + e^{-z/2}} = \\tanh(z/2)$. The denominator is positive, so $s = |\\tanh(z/2)|$.",
     "rule": "algebra",
     "by": [
      "S1"
     ]
    },
    {
     "id": "S3",
     "statement": "$\\tanh$ changes only its sign when its argument changes sign: $\\tanh(-x) = -\\tanh(x)$, as the definition shows. So $|\\tanh(z/2)| = \\tanh(|z|/2)$, which is the stated form.",
     "rule": "algebra",
     "by": [
      "S2"
     ]
    },
    {
     "id": "S4",
     "statement": "For (a): $\\tanh$ is 0 at 0. On $[0, \\infty)$ it is strictly increasing and concave, and it approaches 1 as its argument grows; its derivative $1 - \\tanh^2 x$ is positive and decreasing for $x > 0$. With the argument $|\\ln y - \\ln f|/2$, $s$ has these properties in $|\\ln y - \\ln f|$.",
     "rule": "monotonicity",
     "by": [
      "S3"
     ]
    },
    {
     "id": "S5",
     "statement": "For (b): for every $x \\ge 0$, $x - x^3/3 \\le \\tanh x \\le x$. The upper bound holds because $\\tanh$ has slope at most 1; the lower one because the derivative of $\\tanh x - x + x^3/3$ is $x^2 - \\tanh^2 x \\ge 0$, and the difference is 0 at $x = 0$. Putting $x = |\\ln y - \\ln f|/2$ gives the bounds.",
     "rule": "algebra",
     "by": [
      "S4"
     ]
    },
    {
     "id": "S6",
     "statement": "For (c), the median: the mean of $|\\ln y_t - \\ln f|$ over a test set is minimised when $\\ln f$ is a median of the $\\ln y_t$. Since the logarithm is increasing, this is when $f$ is a median $m$ of the $y_t$.",
     "rule": "monotonicity",
     "by": [
      "S5"
     ]
    },
    {
     "id": "S7",
     "statement": "For (c), the bound: with $z_t = \\ln y_t - \\ln f$ the log errors of the test set, write $\\overline{|z|}(f) = \\frac1n\\sum_t|\\ln y_t - \\ln f|$, $\\overline{|z|^3}(f) = \\frac1n\\sum_t|\\ln y_t - \\ln f|^3$ and $\\bar s(f)$ for the mean of the sMAPE terms of the test set. Averaging the bounds of step 5 over the test set gives $\\bar s(f) \\le \\tfrac12\\overline{|z|}(f) \\le \\bar s(f) + \\tfrac1{24}\\overline{|z|^3}(f)$ for every $f > 0$. Let $f^*$ minimise $\\bar s$, so that $\\bar s(f^*) \\le \\bar s(m)$. Then\n$$\\tfrac12\\overline{|z|}(f^*) \\le \\bar s(f^*) + \\tfrac1{24}\\overline{|z|^3}(f^*) \\le \\bar s(m) + \\tfrac1{24}\\overline{|z|^3}(f^*) \\le \\tfrac12\\overline{|z|}(m) + \\tfrac1{24}\\overline{|z|^3}(f^*).$$\nMultiplying by 2 gives $\\overline{|z|}(f^*) - \\overline{|z|}(m) \\le \\tfrac1{12}\\overline{|z|^3}(f^*)$, the stated inequality.",
     "rule": "algebra",
     "by": [
      "S5"
     ]
    },
    {
     "id": "S8",
     "statement": "For (c), not the median in general: take the five actuals $1, 1, e, e^4, e^4$, whose median is $e$. With the forecast at $e$ the log errors are $1, 1, 0, 3, 3$ in size and the sMAPE terms sum to $2\\tanh(0.5) + 2\\tanh(1.5) = 2.734$. With the forecast at 1 the log errors are $0, 0, 1, 4, 4$ and the terms sum to $\\tanh(0.5) + 2\\tanh(2) = 2.390$, which is lower. So the median does not minimise the sMAPE of this test set, although it minimises its log-MAE (step 6).",
     "rule": "algebra",
     "by": [
      "S6"
     ]
    },
    {
     "id": "S9",
     "statement": "So $s = \\tanh(|\\ln y - \\ln f|/2)$. It rises from 0 like half the absolute log error and levels off at 1. For small errors sMAPE is close to the log-MAE, and its minimiser is close to the median without in general being the median.",
     "rule": "conclusion",
     "by": [
      "S8"
     ]
    }
   ],
   "artefacts": [
    "A1",
    "A2"
   ],
   "availability": {
    "level": "public"
   }
  },
  {
   "id": "E21",
   "profile": "formal",
   "description": "Proof of C9 (section H4 of the proof document A1, rendered in A2). The expected values of RMSE, MAE and MAPE become infinite for every forecast when the actuals have an infinite moment, such as an infinite mean (a heavy tail, C8). The expected sMAPE never does, and there is always a best forecast.",
   "foundation": "Kolmogorov probability axioms and real analysis",
   "steps": [
    {
     "id": "S1",
     "statement": "Finite. For positive values, $s = \\tanh(|\\ln y - \\ln f|/2)$ (C3), which lies between 0 and 1. For any signs, $|y - f| \\le |y| + |f|$, so again $s \\le 1$. A quantity between 0 and 1 has an expectation between 0 and 1, so the expected sMAPE is between 0 and 200, whatever the actuals.",
     "rule": "earlier result",
     "by": [
      "C3"
     ]
    },
    {
     "id": "S2",
     "statement": "A best forecast. Write $\\mathrm E[s(y,f)]$ for the expectation over the random actual $y$. As $f$ approaches 0 or grows without bound, $|\\ln y - \\ln f|$ grows for every positive value of $y$, so $s$ approaches 1 (C3), and so does its expectation. For every positive $f$ it is below 1. It changes continuously with $f$, because $s$ does and is at most 1 (bounded convergence). A continuous function that is below 1 in between and returns towards 1 at both ends reaches a lowest value at some $f > 0$. A forecast of 0 or a negative forecast scores 1 on every positive actual, so it is never better.",
     "rule": "bounded convergence",
     "by": [
      "C3",
      "D1"
     ]
    },
    {
     "id": "S3",
     "statement": "So the expected sMAPE is always finite, whatever the tails of the actuals. For positive actuals, a forecast with the lowest expected sMAPE always exists.",
     "rule": "conclusion",
     "by": [
      "S2"
     ]
    }
   ],
   "artefacts": [
    "A1",
    "A2"
   ],
   "availability": {
    "level": "public"
   }
  },
  {
   "id": "E22",
   "profile": "formal",
   "description": "Proof of C4 (section H5 of the proof document A1, rendered in A2). For any distribution, the best forecast balances the actuals below it and above it. Each actual is weighted by how close it is to the forecast by ratio.",
   "foundation": "Kolmogorov probability axioms and real analysis",
   "steps": [
    {
     "id": "S1",
     "statement": "The derivative of $s$. When $y > f > 0$, $s = (y - f)/(y + f)$. By the quotient rule its derivative in $f$ is $\\dfrac{-(y+f) - (y-f)}{(y+f)^2} = -2w(y,f)$. When $f > y > 0$, $s = (f - y)/(y + f)$, and in the same way its derivative is $2w(y,f)$. Together: for $y \\ne f$, the derivative is $2w(y,f)(\\mathbf 1\\{y < f\\} - \\mathbf 1\\{y > f\\})$.",
     "rule": "quotient rule",
     "by": [
      "D1"
     ]
    },
    {
     "id": "S2",
     "statement": "A bound on it. Since $(y+f)^2 \\ge 4yf$, $w(y,f) \\le 1/(4f)$. So the derivative of $s$ is at most $1/(2f)$ in size, whatever $y$, and at most $1/f$ for every forecast above $f/2$. The only actual at which $s$ has no derivative is $y = f$, which has probability zero.",
     "rule": "algebra",
     "by": [
      "S1"
     ]
    },
    {
     "id": "S3",
     "statement": "The derivative of the expectation. With this bound, the derivative may be taken inside the expectation (dominated convergence):\n$$\\frac{d}{df}\\,\\mathrm E[s(y,f)] = 2\\,\\mathrm E\\big[w(y,f)\\,(\\mathbf 1\\{y < f\\} - \\mathbf 1\\{y > f\\})\\big].$$",
     "rule": "dominated convergence",
     "by": [
      "S2"
     ]
    },
    {
     "id": "S4",
     "statement": "The balance. A forecast with the lowest expected sMAPE exists (C9). There this derivative is 0, which is the stated balance: half the total weight lies below $f$ and half above.",
     "rule": "earlier result",
     "by": [
      "C9"
     ]
    },
    {
     "id": "S5",
     "statement": "The weight. With $r = y/f$, $w(y,f) = \\dfrac1f\\cdot\\dfrac{r}{(1+r)^2} = \\dfrac1f\\cdot\\dfrac{1}{r + 2 + 1/r}$. The sum $r + 1/r$ is the same for $r$ and $1/r$. It is smallest, 2, at $r = 1$, and grows as $r$ moves away from 1. So $w$ is largest at $y = f$ and falls as $y/f$ moves away from 1, equally for $r$ and $1/r$.",
     "rule": "algebra",
     "by": [
      "S4"
     ]
    },
    {
     "id": "S6",
     "statement": "So sMAPE targets a weighted median of the actuals. Each actual is weighted by how close it is to the forecast by ratio.",
     "rule": "conclusion",
     "by": [
      "S5"
     ]
    }
   ],
   "artefacts": [
    "A1",
    "A2"
   ],
   "availability": {
    "level": "public"
   }
  },
  {
   "id": "E23",
   "profile": "formal",
   "description": "Proof of C5 (section H6 of the proof document A1, rendered in A2). This is the case of C4 in which the weights balance at the median.",
   "foundation": "Kolmogorov probability axioms and real analysis",
   "steps": [
    {
     "id": "S1",
     "statement": "By C3, $\\mathrm E[s(y,f)] = \\mathrm E\\big[\\tanh\\big(|\\ln y - \\ln f|/2\\big)\\big]$. Take a forecast $f = m\\,e^{\\tau}$ with $\\tau > 0$; the case $\\tau < 0$ is the mirror image. We show that its expected sMAPE is larger than that of $m$.",
     "rule": "setup",
     "by": [
      "C3"
     ]
    },
    {
     "id": "S2",
     "statement": "Write both expectations as integrals over $v$, the distance of $\\ln y$ from $\\ln f$. For the forecast $m\\,e^{\\tau}$, $\\ln y = \\ln m + \\tau + v$; for $m$, $\\ln y = \\ln m + v$. Their difference is\n$$\\Delta = \\int \\tanh(|v|/2)\\,\\big[g(\\ln m+\\tau+v) - g(\\ln m+v)\\big]\\,dv.$$",
     "rule": "definition",
     "by": [
      "S1"
     ]
    },
    {
     "id": "S3",
     "statement": "Write $\\phi(v) = g(\\ln m+\\tau+v) - g(\\ln m+v)$. By the symmetry of $g$ about $\\ln m$, $\\phi(-\\tau-v) = g(\\ln m-v) - g(\\ln m-\\tau-v) = g(\\ln m+v) - g(\\ln m+\\tau+v) = -\\phi(v)$. So reflecting $v$ about $-\\tau/2$ reverses the sign of $\\phi$.",
     "rule": "definition",
     "by": [
      "S2"
     ]
    },
    {
     "id": "S4",
     "statement": "Split the integral at $-\\tau/2$ and reflect its left part onto the right. This gives\n$$\\Delta = \\int_{v > -\\tau/2} \\big[\\tanh(|v|/2) - \\tanh(|v+\\tau|/2)\\big]\\,\\phi(v)\\,dv.$$",
     "rule": "algebra",
     "by": [
      "S3"
     ]
    },
    {
     "id": "S5",
     "statement": "For $v > -\\tau/2$, $v$ is closer to 0 than $v + \\tau$ is: $|v| < |v + \\tau|$. So the bracket is negative, because $\\tanh$ is increasing (C3). And $\\phi(v) \\le 0$ by unimodality: $g(\\ln m+\\tau+v) \\le g(\\ln m+v)$. Their product is $\\ge 0$, and $> 0$ on a set of positive probability. So $\\Delta > 0$.",
     "rule": "monotonicity",
     "by": [
      "C3"
     ]
    },
    {
     "id": "S6",
     "statement": "So every forecast other than the median has a larger expected sMAPE. The median is the only forecast that minimises it.",
     "rule": "conclusion",
     "by": [
      "S5"
     ]
    }
   ],
   "artefacts": [
    "A1",
    "A2"
   ],
   "availability": {
    "level": "public"
   }
  },
  {
   "id": "E24",
   "profile": "formal",
   "description": "Proof of C6 (section H7 of the proof document A1, rendered in A2). When the actuals form two separate groups, the median falls in the gap between them, far from both. So sMAPE scores it worse than forecasts near either group.",
   "foundation": "Kolmogorov probability axioms and real analysis",
   "steps": [
    {
     "id": "S1",
     "statement": "Write forecasts on the log scale as $\\ln f = \\ln m + c$. By C3, the expected contribution is $\\mathrm E[\\tanh(|\\ln y - \\ln m - c|/2)]$, a function of $c$.",
     "rule": "setup",
     "by": [
      "C3"
     ]
    },
    {
     "id": "S2",
     "statement": "Why the median is bad. Take the groups to be single points first. A forecast at the median is $a$ away, on the log scale, from every actual. So its expected $s$ is $\\tanh(a/2)$.",
     "rule": "case analysis",
     "by": [
      "S1"
     ]
    },
    {
     "id": "S3",
     "statement": "A forecast at the mode, $\\ln f = \\ln m + a$, is exact for half the actuals and $2a$ away from the other half. So its expected $s$ is $\\tfrac12\\tanh(a)$. Since $\\tanh(a) = \\dfrac{2\\tanh(a/2)}{1 + \\tanh^2(a/2)}$, this is $\\dfrac{\\tanh(a/2)}{1 + \\tanh^2(a/2)} < \\tanh(a/2)$. The mode is better.",
     "rule": "algebra",
     "by": [
      "S2"
     ]
    },
    {
     "id": "S4",
     "statement": "The expected $s$ changes continuously as the groups widen from points, because $s$ is bounded and continuous. So the mode is still better for narrow groups.",
     "rule": "continuity",
     "by": [
      "S3"
     ]
    },
    {
     "id": "S5",
     "statement": "Why the median is a local maximum. By C12, step 4, the second derivative of the expected contribution at the median is\n$$g(\\ln m) - \\tfrac12\\,\\mathrm E\\Big[\\operatorname{sech}^2\\Big(\\frac{\\ln y - \\ln m}{2}\\Big)\\tanh\\Big(\\frac{|\\ln y - \\ln m|}{2}\\Big)\\Big].$$",
     "rule": "earlier result",
     "by": [
      "C12"
     ]
    },
    {
     "id": "S6",
     "statement": "By symmetry the first derivative at the median is 0. When the groups narrow, few actuals lie near the median, so $g(\\ln m)$ tends to 0. Meanwhile the expectation tends to $\\operatorname{sech}^2(a/2)\\tanh(a/2) > 0$. So the second derivative becomes negative, and the median is a local maximum.",
     "rule": "symmetry",
     "by": [
      "S5"
     ]
    },
    {
     "id": "S7",
     "statement": "A minimiser on each side. The expected sMAPE has a lowest point (C9). It is not at the local maximum $c = 0$. By symmetry, if $c^*$ is a minimiser, so is $-c^*$.",
     "rule": "symmetry",
     "by": [
      "C9"
     ]
    },
    {
     "id": "S8",
     "statement": "Near the modes. Write $\\tanh(|z|/2) = 1 - 2\\lambda(z)$, with $\\lambda(z) = 1/(1 + e^{|z|})$. The function $\\lambda$ is largest at 0 and falls as $|z|$ grows, more slowly the larger $|z|$ is. Minimising the expected $s$ is maximising $\\mathrm E[\\lambda(\\ln y - \\ln m - c)]$. As the groups narrow, this approaches $p\\,\\lambda(c + a) + (1 - p)\\,\\lambda(c - a)$, uniformly in $c$.",
     "rule": "definition",
     "by": [
      "S7"
     ]
    },
    {
     "id": "S9",
     "statement": "With equal weights, take $0 < c < a$. The derivative of this limit is half the rate at which $\\lambda$ falls at distance $a - c$, minus half the rate at distance $a + c$. It is positive, because $\\lambda$ falls faster nearer 0. For $c > a$ the limit decreases. So on $c > 0$ its only maximum is at $c = a$, and the minimisers on that side approach $a$. By symmetry, those on the other side approach $-a$. This proves (a).",
     "rule": "symmetry",
     "by": [
      "S8"
     ]
    },
    {
     "id": "S10",
     "statement": "Unequal weights. The limit increases up to $c = -a$ and decreases after $c = a$. In between it is convex, by the argument of step 9. So its maximum is at $-a$ or at $a$. The difference between these two values is $(2p - 1)\\big(\\lambda(0) - \\lambda(2a)\\big)$, which is positive when $p > 1/2$. So the maximum, and with it the best forecast for narrow groups, is at the mode of the larger group. This proves (b).",
     "rule": "convexity",
     "by": [
      "S9"
     ]
    },
    {
     "id": "S11",
     "statement": "So with two separate groups of actuals, the median is a hump of the expected sMAPE. The best forecasts are near the modes, and with unequal groups the larger one wins. As the weights pass through equality, the target jumps from one mode to the other.",
     "rule": "conclusion",
     "by": [
      "S10"
     ]
    }
   ],
   "artefacts": [
    "A1",
    "A2"
   ],
   "availability": {
    "level": "public"
   }
  },
  {
   "id": "E25",
   "profile": "formal",
   "description": "Proof of C7 (section H8 of the proof document A1, rendered in A2). The worst possible score goes to forecasts of the wrong sign and to zero forecasts. With enough zeros among the actuals, sMAPE prefers the forecast 0. The share of zeros needed is below one half.",
   "foundation": "Kolmogorov probability axioms and real analysis",
   "steps": [
    {
     "id": "S1",
     "statement": "When $y$ and $f$ have opposite signs, $|y - f| = |y| + |f|$, so $s = 1$. When exactly one of them is 0, the numerator and the denominator both equal the absolute value of the other one, so $s = 1$.",
     "rule": "algebra",
     "by": [
      "D1"
     ]
    },
    {
     "id": "S2",
     "statement": "A positive forecast scores 1 on every zero actual (step 1). So its expected $s$ is $p_0$ plus $(1 - p_0)$ times its expected $s$ on the positive actuals. This is at least $p_0 + (1-p_0)\\,s_+$, with equality at the best positive forecasts, which exist (C9).",
     "rule": "earlier result",
     "by": [
      "S1",
      "C9"
     ]
    },
    {
     "id": "S3",
     "statement": "The forecast 0 scores 0 on zero actuals and 1 on positive ones. So its expected $s$ is $1 - p_0$. A negative forecast scores 1 on every actual, which is never better than 0.",
     "rule": "algebra",
     "by": [
      "S2"
     ]
    },
    {
     "id": "S4",
     "statement": "The forecast 0 is better when $1 - p_0 < p_0 + (1 - p_0)s_+$. Collecting the terms in $p_0$ gives $1 - s_+ < p_0(2 - s_+)$. Since $2 - s_+ > 0$, this is the stated inequality.",
     "rule": "algebra",
     "by": [
      "S3"
     ]
    },
    {
     "id": "S5",
     "statement": "Compare the threshold with $1/2$: $\\dfrac{1 - s_+}{2 - s_+} < \\dfrac12$ exactly when $2 - 2s_+ < 2 - s_+$, that is, when $s_+ > 0$. And $s_+ > 0$ unless the positive actuals are all one value.",
     "rule": "algebra",
     "by": [
      "S4"
     ]
    },
    {
     "id": "S6",
     "statement": "So sMAPE switches to the forecast 0 once zeros are frequent enough, and the share needed is below one half. At the threshold, the target jumps from a positive forecast to 0.",
     "rule": "conclusion",
     "by": [
      "S5"
     ]
    }
   ],
   "artefacts": [
    "A1",
    "A2"
   ],
   "availability": {
    "level": "public"
   }
  },
  {
   "id": "E26",
   "profile": "formal",
   "description": "Proof of C11 (section H9 of the proof document A1, rendered in A2). When actuals can be negative or positive, sMAPE's target is never near zero. It is on the same side of zero as the mean. So a small shift of the mean across zero makes the target jump from a negative value to a positive one.",
   "foundation": "Kolmogorov probability axioms and real analysis",
   "steps": [
    {
     "id": "S1",
     "statement": "A positive forecast against its mirror image. Take $f > 0$ and compare it with $-f$. An actual $y > 0$ has the sign of $f$, so it scores less than 1 with $f$, and exactly 1 with $-f$ (C7).",
     "rule": "symmetry",
     "by": [
      "C7"
     ]
    },
    {
     "id": "S2",
     "statement": "The mirror actual $-y$ scores 1 with $f$. By symmetry, it scores the same amount less than 1 with $-f$. So each pair of actuals $y$ and $-y$ favours $f$ when $y$ is the more likely of the two, and $-f$ otherwise.",
     "rule": "symmetry",
     "by": [
      "S1"
     ]
    },
    {
     "id": "S3",
     "statement": "When $\\mu > 0$, every positive actual $y$ is more likely than $-y$, because $y$ is closer to $\\mu$ and the density of the actuals falls away from $\\mu$. So every pair favours $f$, and $f$ has the lower expected sMAPE. Every best forecast is therefore positive. The case $\\mu < 0$ is the mirror image.",
     "rule": "symmetry",
     "by": [
      "S2"
     ]
    },
    {
     "id": "S4",
     "statement": "Best forecasts exist on each side. For a forecast $f > 0$, the actuals of its sign score below 1 and have positive probability, so its expected $s$ is below 1. As $f$ approaches 0 or grows without bound, every non-zero actual scores close to 1, so the expected $s$ approaches 1. As in C9, the expected $s$ is continuous in $f$ and so reaches a lowest value for $f > 0$; the same holds for $f < 0$.",
     "rule": "continuity",
     "by": [
      "C9"
     ]
    },
    {
     "id": "S5",
     "statement": "Away from zero, uniformly. Fix a forecast $f_1 > 0$. For every $\\mu$ in a range around 0, say $|\\mu| \\le 1$, its expected $s$ is at most some value $q < 1$, since it changes continuously with $\\mu$. For a forecast $f$ with $|f| \\le \\eta$, every actual with $|y| \\ge \\sqrt\\eta$ scores at least $(\\sqrt\\eta - \\eta)/(\\sqrt\\eta + \\eta)$, which approaches 1 as $\\eta$ shrinks; and the share of actuals with $|y| < \\sqrt\\eta$ is small for every such $\\mu$: since $\\varepsilon$ is symmetric and its density decreases away from 0, an interval of a given length around 0 holds at least as much of its probability as any other interval of that length, so $P(|\\mu + \\varepsilon| < \\sqrt\\eta) \\le P(|\\varepsilon| < \\sqrt\\eta)$, which tends to 0 as $\\eta$ shrinks. So for $\\eta$ small enough, every forecast within $\\eta$ of 0 has an expected $s$ above $q$, for every $\\mu$ in the range: it is worse than $f_1$ (or its mirror image). The best forecasts therefore stay at least $\\eta$ from 0.",
     "rule": "symmetry",
     "by": [
      "S4"
     ]
    },
    {
     "id": "S6",
     "statement": "So the target is always on the same side of zero as the mean, and never near zero. It jumps from negative to positive as the mean crosses zero.",
     "rule": "conclusion",
     "by": [
      "S5"
     ]
    }
   ],
   "artefacts": [
    "A1",
    "A2"
   ],
   "availability": {
    "level": "public"
   }
  },
  {
   "id": "E27",
   "profile": "formal",
   "description": "Proof of C8 (section H10 of the proof document A1, rendered in A2). Each of the other three measures has an infinite expected value, for every forecast, when a moment it needs is infinite. What it then rewards depends on the measure.",
   "foundation": "Kolmogorov probability axioms and real analysis",
   "steps": [
    {
     "id": "S1",
     "statement": "(a) RMSE. If $\\mathrm E[y^2]$ is finite, so is $\\mathrm E[y]$. Expanding the square gives $\\mathrm E[(y-f)^2] = \\mathrm E[y^2] - 2f\\,\\mathrm E[y] + f^2$, which is finite. Conversely, $y^2 \\le 2(y - f)^2 + 2f^2$. So a finite $\\mathrm E[(y-f)^2]$ makes $\\mathrm E[y^2]$ finite.",
     "rule": "algebra",
     "by": [
      "D10"
     ]
    },
    {
     "id": "S2",
     "statement": "Expanding both squares gives $(y - f_1)^2 - (y - f_2)^2 = (f_2 - f_1)(2y - f_1 - f_2)$. Its expectation is finite when $\\mathrm E[y]$ is, and equals $(f_1 - f_2)(f_1 + f_2 - 2\\,\\mathrm E[y]) = (f_1 - \\mathrm E[y])^2 - (f_2 - \\mathrm E[y])^2$.",
     "rule": "algebra",
     "by": [
      "S1"
     ]
    },
    {
     "id": "S3",
     "statement": "For $f_1 = \\mathrm E[y]$ this is negative for every $f_2 \\ne \\mathrm E[y]$. So the mean is preferred.",
     "rule": "algebra",
     "by": [
      "S2"
     ]
    },
    {
     "id": "S4",
     "statement": "If $\\mathrm E[y]$ is infinite, then for $f_2 > f_1$ the expectation in step 2 is $+\\infty$. Every forecast is beaten by any larger one, so none is preferred to all others.",
     "rule": "algebra",
     "by": [
      "S2"
     ]
    },
    {
     "id": "S5",
     "statement": "(b) MAE. Since $|y - f| \\ge y - f$, $\\mathrm E|y - f|$ is infinite when $\\mathrm E[y]$ is. Otherwise it is at most $\\mathrm E[y] + |f|$.",
     "rule": "algebra",
     "by": [
      "D11"
     ]
    },
    {
     "id": "S6",
     "statement": "The difference $|y - f_1| - |y - f_2|$ is at most $|f_1 - f_2|$ in size, so its expectation is finite.",
     "rule": "algebra",
     "by": [
      "S5"
     ]
    },
    {
     "id": "S7",
     "statement": "The derivative of $|y - f|$ in $f$ is $\\mathbf 1\\{y < f\\} - \\mathbf 1\\{y > f\\}$. The expectation of this is $P(y < f) - P(y > f)$, which is negative below a median and positive above it. So a median is preferred.",
     "rule": "algebra",
     "by": [
      "S6"
     ]
    },
    {
     "id": "S8",
     "statement": "(c) MAPE. For $f > 0$, the contribution of an actual to MAPE, divided by 100, is $|y - f|/y = |1 - f/y|$. This is at least $f/y - 1$. So the expected MAPE is infinite when $\\mathrm E[1/y]$ is, and at most $1 + f\\,\\mathrm E[1/y]$ otherwise.",
     "rule": "algebra",
     "by": [
      "D12"
     ]
    },
    {
     "id": "S9",
     "statement": "As a function of $f$, $|1 - f/y|$ is convex. Its derivative is $(1/y)(\\mathbf 1\\{y < f\\} - \\mathbf 1\\{y > f\\})$. Its expectation is 0 where the actuals below $f$ and those above it have equal total weight $1/y$: at a median of the actuals weighted by $1/y$.",
     "rule": "convexity",
     "by": [
      "S8"
     ]
    },
    {
     "id": "S10",
     "statement": "The forecast 0 gives $|1 - 0| = 1$ for every actual, so MAPE 100. A negative forecast gives more than 1. So when $\\mathrm E[1/y]$ is infinite, 0 is the only forecast with a finite expected MAPE.",
     "rule": "algebra",
     "by": [
      "S9"
     ]
    },
    {
     "id": "S11",
     "statement": "If $y = 0$ with positive probability: for $f \\ne 0$ that actual contributes $|f|/0$, which is infinite. For $f = 0$ it contributes $0/0$, which is undefined.",
     "rule": "algebra",
     "by": [
      "S10"
     ]
    },
    {
     "id": "S12",
     "statement": "So each of RMSE, MAE and MAPE has an infinite expected value when a moment it needs is infinite. RMSE and MAE still rank forecasts by their differences. MAPE turns to the forecast 0, and with zero actuals it targets nothing.",
     "rule": "conclusion",
     "by": [
      "S11"
     ]
    }
   ],
   "artefacts": [
    "A1",
    "A2"
   ],
   "availability": {
    "level": "public"
   }
  },
  {
   "id": "E28",
   "profile": "formal",
   "description": "Proof of C10 (section H11 of the proof document A1, rendered in A2). For widely spread actuals, every forecast scores close to the worst value. The differences between forecasts shrink, and the target moves from the median to the mode of $\\ln y$.",
   "foundation": "Kolmogorov probability axioms and real analysis",
   "steps": [
    {
     "id": "S1",
     "statement": "Rewrite the expectation. By C3, and since $\\tanh(|x|/2) = 1 - \\dfrac{2}{1 + e^{|x|}}$, $\\mathrm E[s(y,f)] = 1 - \\mathrm E\\big[2/(1 + e^{|\\ln y - \\ln f|})\\big]$. The density of $\\ln y$ at $x$ is $h\\big((x - \\mu)/\\sigma\\big)/\\sigma$. Substituting $x = \\ln f + v$ in the integral gives\n$$\\mathrm E[s(y,f)] = 1 - \\frac1\\sigma\\int \\frac{2}{1 + e^{|v|}}\\,h\\Big(c + \\frac v\\sigma\\Big)\\,dv.$$",
     "rule": "change of variables",
     "by": [
      "C3"
     ]
    },
    {
     "id": "S2",
     "statement": "Expand $h$. By Taylor's theorem, $h(c + v/\\sigma) = h(c) + h'(c)\\,v/\\sigma + \\tfrac12 h''(\\xi)\\,v^2/\\sigma^2$ for some $\\xi$ between $c$ and $c + v/\\sigma$.",
     "rule": "Taylor's theorem",
     "by": [
      "S1"
     ]
    },
    {
     "id": "S3",
     "statement": "The first term integrates to $4\\ln 2\\,h(c)$, since $\\int \\frac{2}{1+e^{|v|}}\\,dv = 4\\ln 2$. Over $v > 0$, $\\int 2/(1 + e^v)\\,dv = 2\\ln 2$.",
     "rule": "algebra",
     "by": [
      "S2"
     ]
    },
    {
     "id": "S4",
     "statement": "The second term integrates to 0: its factor $v$ changes sign with $v$, while the rest does not.",
     "rule": "algebra",
     "by": [
      "S3"
     ]
    },
    {
     "id": "S5",
     "statement": "The third term is at most $\\tfrac12\\max|h''|$ times $\\int \\frac{2v^2}{1+e^{|v|}}\\,dv = 6\\zeta(3) \\approx 7.21$, divided by $\\sigma^2$. This integral is a standard Fermi–Dirac integral. With the factor $1/\\sigma$ of step 1, this gives the expansion.",
     "rule": "standard integral",
     "by": [
      "S1"
     ]
    },
    {
     "id": "S6",
     "statement": "(a). The expansion tends to 1, so the expected sMAPE tends to 200. Two forecasts at fixed quantile levels have fixed positions $c_1$ and $c_2$. So their difference is $4\\ln 2\\,[h(c_2) - h(c_1)]/\\sigma + O(\\sigma^{-3})$.",
     "rule": "algebra",
     "by": [
      "S5"
     ]
    },
    {
     "id": "S7",
     "statement": "(b). Minimising the expansion over $c$ is maximising $h(c)$, up to a term that shrinks like $\\sigma^{-2}$ for every $c$ at once. A density with a bounded derivative tends to 0 far out. So its single highest point $c_0$ stands clear of every other point, and the minimisers' positions tend to $c_0$. So do their quantile levels.",
     "rule": "algebra",
     "by": [
      "S6"
     ]
    },
    {
     "id": "S8",
     "statement": "(c). By C3 (b), $\\tfrac12\\sigma|\\varepsilon - c| - s$ lies between 0 and $\\tfrac1{24}\\sigma^3|\\varepsilon - c|^3$. So for $c$ in any bounded range, $\\mathrm E[s]/\\sigma = \\tfrac12\\mathrm E|\\varepsilon - c| + O(\\sigma^2)$. This uses $\\mathrm E|\\varepsilon|^3 < \\infty$. And $\\mathrm E|\\varepsilon - c|$ is smallest at the median of $\\varepsilon$.",
     "rule": "earlier result",
     "by": [
      "C3"
     ]
    },
    {
     "id": "S9",
     "statement": "The minimisers also stay in a bounded range. For large $|c|$, most actuals have $\\varepsilon$ at least $|c|/2$ from $c$. And $\\tanh x \\ge x\\tanh 1$ for $0 \\le x \\le 1$. So $\\mathrm E[s]/\\sigma$ grows with $|c|$, or exceeds a multiple of $1/\\sigma$, beyond its value at the median. So the minimisers' positions tend to the median of $\\varepsilon$, whose quantile level is $1/2$.",
     "rule": "algebra",
     "by": [
      "S8"
     ]
    },
    {
     "id": "S10",
     "statement": "So as the actuals spread out, the expected sMAPE flattens towards its bound. Differences between forecasts vanish like $1/\\sigma$. The target moves from the median to the mode of $\\ln y$.",
     "rule": "conclusion",
     "by": [
      "S9"
     ]
    }
   ],
   "artefacts": [
    "A1",
    "A2"
   ],
   "availability": {
    "level": "public"
   }
  },
  {
   "id": "E29",
   "profile": "formal",
   "description": "Proof of C12 (section H12 of the proof document A1, rendered in A2). The target is above the ordinary median when the actuals above it carry more weight, and below it when those below carry more. When the weights balance and enough actuals lie near the median, the median is the target. This is so when the log of the actual is symmetric with a single peak.",
   "foundation": "Kolmogorov probability axioms and real analysis",
   "steps": [
    {
     "id": "S1",
     "statement": "The derivative at the median. By C4, step 3, the derivative of the expected sMAPE term in $f$ is $2\\,\\mathrm E[w(y,f)(\\mathbf 1\\{y < f\\} - \\mathbf 1\\{y > f\\})]$. At $f = m$ it is twice the total weight of the actuals below $m$ minus that of the actuals above $m$.",
     "rule": "earlier result",
     "by": [
      "C4"
     ]
    },
    {
     "id": "S2",
     "statement": "The shape of expected sMAPE. With a single local minimum, the expected sMAPE decreases up to it and increases after it. Otherwise it would turn again, at a local maximum. After that turn it would have to fall and then rise once more, because it returns towards 1 at both ends (C9). That would give a second local minimum.",
     "rule": "earlier result",
     "by": [
      "C9"
     ]
    },
    {
     "id": "S3",
     "statement": "(a). If the actuals above $m$ carry more weight, the derivative at the median is negative (step 1): the expected sMAPE is decreasing at the median. By step 2, the median then lies before the minimum, so the target is above the median. The case where the actuals below carry more weight is the mirror image.",
     "rule": "symmetry",
     "by": [
      "S1",
      "S2"
     ]
    },
    {
     "id": "S4",
     "statement": "The second derivative. Measure the forecast on the log scale, $c = \\ln f$, so that the expected contribution is $\\mathrm E[\\tanh(|\\ln y - c|/2)]$ (C3). The derivative of $\\tanh(|x - c|/2)$ in $c$ is $-\\tfrac12\\operatorname{sech}^2\\big(\\tfrac{x-c}{2}\\big)$ for $c < x$ and $+\\tfrac12\\operatorname{sech}^2\\big(\\tfrac{x-c}{2}\\big)$ for $c > x$: it jumps by 1 at $c = x$. Differentiating again and taking the expectation over $x = \\ln y$, the jumps add the density $g$ of $\\ln y$ at $c$, and elsewhere the second derivative is $-\\tfrac12\\operatorname{sech}^2\\big(\\tfrac{x-c}{2}\\big)\\tanh\\big(\\tfrac{|x-c|}{2}\\big)$. So the second derivative of the expected contribution is\n$$g(c) - \\tfrac12\\,\\mathrm E\\Big[\\operatorname{sech}^2\\Big(\\frac{\\ln y - c}{2}\\Big)\\tanh\\Big(\\frac{|\\ln y - c|}{2}\\Big)\\Big].$$\nThis holds when $g$ is continuous at $c$: since $\\operatorname{sech}^2 = 1 - \\tanh^2$, the first derivative is $\\mathrm P(\\ln y \\le c) - \\tfrac12$, whose derivative at $c$ is $g(c)$, plus the expectation of $-\\tfrac12\\tanh^2\\big(\\tfrac{x-c}{2}\\big)$ with the sign of $c - x$, a function of $c$ with a bounded continuous derivative, which may be differentiated inside the expectation.",
     "rule": "continuity",
     "by": [
      "C3"
     ]
    },
    {
     "id": "S5",
     "statement": "(b), the sufficient condition. If the two weights are equal, the first derivative at the median is 0 (step 1): the expected sMAPE is flat there. If, in addition, the stated inequality holds, the second derivative at $c = \\ln m$ is positive (step 4), so the median is a strict local minimum. With a single local minimum, it is the target.",
     "rule": "algebra",
     "by": [
      "S1",
      "S4"
     ]
    },
    {
     "id": "S6",
     "statement": "(b), without the inequality. A point where the first derivative is 0 can be a flat point of inflection, through which the expected sMAPE keeps rising (or keeps falling) without turning. An example, built by the independent checker of this section: let $\\ln y - \\ln m$ be a mixture of three normal distributions, with mean $-1.5$ and standard deviation $0.3$ (weight 0.406125), mean 0 and standard deviation $0.5$ (weight 0.054825), and mean $1.158664$ and standard deviation 1 (weight 0.539050); these values are rounded from the exact solution of the three conditions. Its median is $m$, the two weights are equal there, and the second derivative of step 4 is 0 at $\\ln m$; the expected sMAPE rises through $m$, where it is flat, and its only minimum is near $\\ln f = \\ln m - 1.25$. So equal weights alone do not make the median the target.",
     "rule": "algebra",
     "by": [
      "S4"
     ]
    },
    {
     "id": "S7",
     "statement": "(c). By C5 the median is then the only forecast with the lowest expected sMAPE (a density cannot be constant, so a symmetric unimodal density is strictly lower somewhere away from its centre, as C5 asks). For lognormal actuals $\\ln y$ has a normal density, and for log-Laplace actuals the Laplace density $e^{-|x - \\ln m|/b}/(2b)$; both are symmetric about $\\ln m$ and strictly decreasing in the distance from it.",
     "rule": "earlier result",
     "by": [
      "C5"
     ]
    },
    {
     "id": "S8",
     "statement": "So the target is above the ordinary median when the actuals above it carry more of the weight, and below it when those below carry more. When the two balance, the median is the target if, in addition, enough actuals lie near it, as the stated inequality says; this is the case when the log of the actual is symmetric with a single peak.",
     "rule": "conclusion",
     "by": [
      "S7"
     ]
    }
   ],
   "artefacts": [
    "A1",
    "A2"
   ],
   "availability": {
    "level": "public"
   }
  },
  {
   "id": "E35",
   "profile": "formal",
   "description": "Proof of C18 (section H13 of the proof document A1, rendered in A2). For widely spread actuals, sMAPE loses its power to tell two forecasts apart, at the rate $\\sigma^{-1/2}$. The absolute log error keeps its power, and the absolute error keeps it unless the higher forecast is the median; the squared error on lognormal actuals loses it faster than any power of $\\sigma$.",
   "foundation": "Kolmogorov probability axioms and real analysis",
   "steps": [
    {
     "id": "S1",
     "statement": "The sMAPE difference. Write $q(x) = 2/(1 + e^{|x|})$, so that $\\tanh(|x|/2) = 1 - q(x)$, and $q_i = q(\\sigma(\\varepsilon - c_i))$. By C3 the sMAPE term of the forecast $f_i$ is $1 - q_i$, because $\\ln y - \\ln f_i = \\sigma(\\varepsilon - c_i)$. So $D = q_1 - q_2$.",
     "rule": "definition",
     "by": [
      "C3"
     ]
    },
    {
     "id": "S2",
     "statement": "Its mean. By C10, $\\mathrm E[q_i] = 4\\ln 2\\,h(c_i)/\\sigma + O(\\sigma^{-3})$. So $\\mathrm E[D] = 4\\ln 2\\,(h(c_1) - h(c_2))/\\sigma + O(\\sigma^{-3})$.",
     "rule": "earlier result",
     "by": [
      "C10"
     ]
    },
    {
     "id": "S3",
     "statement": "The mean of $q_i^2$. Substituting $v = \\sigma(\\varepsilon - c_i)$, as in C10, step 1, $\\mathrm E[q_i^2] = \\frac1\\sigma\\int q(v)^2\\,h(c_i + v/\\sigma)\\,dv$. Expanding $h$ by Taylor's theorem, as in C10, step 2: the first term integrates to $h(c_i)\\int q(v)^2\\,dv$; the second integrates to 0, because its factor $v$ changes sign with $v$ while the rest does not; the third is at most $\\tfrac12\\max|h''|\\int v^2 q(v)^2\\,dv$ divided by $\\sigma^2$, and this integral is finite because $q(v) \\le 2e^{-|v|}$. Since $\\dfrac{1}{(1+e^v)^2} = \\dfrac{1}{1+e^v} - \\dfrac{e^v}{(1+e^v)^2}$, whose two parts integrate over $v > 0$ to $\\ln 2$ and $\\tfrac12$, $\\int q(v)^2\\,dv = 8(\\ln 2 - \\tfrac12) = 4(2\\ln 2 - 1)$. So $\\mathrm E[q_i^2] = 4(2\\ln 2 - 1)\\,h(c_i)/\\sigma + O(\\sigma^{-3})$.",
     "rule": "Taylor's theorem",
     "by": [
      "C10"
     ]
    },
    {
     "id": "S4",
     "statement": "The cross term. Since $q(x) \\le 2e^{-|x|}$, $q_1 q_2 \\le 4e^{-\\sigma(|\\varepsilon - c_1| + |\\varepsilon - c_2|)} \\le 4e^{-\\sigma(c_2 - c_1)}$, because $|\\varepsilon - c_1| + |\\varepsilon - c_2| \\ge c_2 - c_1$. So $\\mathrm E[q_1 q_2] \\le 4e^{-\\sigma(c_2 - c_1)}$, which is smaller than every power of $1/\\sigma$.",
     "rule": "algebra",
     "by": [
      "S3"
     ]
    },
    {
     "id": "S5",
     "statement": "The variance. $\\operatorname{Var}(D) = \\mathrm E[q_1^2] + \\mathrm E[q_2^2] - 2\\,\\mathrm E[q_1 q_2] - (\\mathrm E[D])^2$. By steps 2 to 4 this is $4(2\\ln 2 - 1)(h(c_1) + h(c_2))/\\sigma + O(\\sigma^{-2})$; the term $(\\mathrm E[D])^2$ is of order $\\sigma^{-2}$.",
     "rule": "algebra",
     "by": [
      "S2",
      "S3",
      "S4"
     ]
    },
    {
     "id": "S6",
     "statement": "(a). When $h(c_1) \\ne h(c_2)$, $h(c_1) + h(c_2) > 0$. Dividing the size of the mean (step 2) by the square root of the variance (step 5) gives the stated ratio. To keep $\\sqrt n$ times the ratio at a given value, $n$ must be that value squared over the ratio squared, which grows in proportion to $\\sigma$. When $h(c_1) = h(c_2) > 0$ instead, the mean is $O(\\sigma^{-3})$ (step 2) while the variance is still of order $1/\\sigma$ (step 5), so the ratio is $O(\\sigma^{-5/2})$: at most of that order.",
     "rule": "algebra",
     "by": [
      "S2",
      "S5"
     ]
    },
    {
     "id": "S7",
     "statement": "(b). For the absolute log error, $|\\ln y - \\ln f_i| = \\sigma|\\varepsilon - c_i|$, so $D = \\sigma(|\\varepsilon - c_2| - |\\varepsilon - c_1|)$. It is bounded by $\\sigma(c_2 - c_1)$, so its mean and standard deviation exist. Both are $\\sigma$ times those of $|\\varepsilon - c_2| - |\\varepsilon - c_1|$, whose distribution does not depend on $\\sigma$. So their ratio, wherever the standard deviation is not zero, does not depend on $\\sigma$.",
     "rule": "algebra",
     "by": [
      "S6"
     ]
    },
    {
     "id": "S8",
     "statement": "(c). For the absolute error, $D = |y - f_2| - |y - f_1|$ with $f_1 < f_2$. It equals $f_2 - f_1$ when $y \\le f_1$, $f_1 + f_2 - 2y$ when $f_1 < y < f_2$, and $-(f_2 - f_1)$ when $y \\ge f_2$; so $|D| \\le f_2$. Now $f_1/f_2 = e^{-\\sigma(c_2 - c_1)}$ tends to 0, and for every $\\varepsilon < c_2$, $y/f_2 = e^{\\sigma(\\varepsilon - c_2)}$ tends to 0. So $D/f_2$ tends to 1 for $\\varepsilon < c_2$ and to $-1$ for $\\varepsilon > c_2$. By bounded convergence, $\\mathrm E[D/f_2] \\to H(c_2) - (1 - H(c_2)) = 2H(c_2) - 1$ and $\\mathrm E[(D/f_2)^2] \\to 1$. So the variance of $D/f_2$ tends to $1 - (2H(c_2) - 1)^2 = 4H(c_2)(1 - H(c_2))$, and the ratio, which is the same for $D$ and $D/f_2$, tends to the stated limit when $0 < H(c_2) < 1$. The limit is 0 only when $H(c_2) = 1/2$.",
     "rule": "bounded convergence",
     "by": [
      "S7"
     ]
    },
    {
     "id": "S9",
     "statement": "(d). For the squared error, $D = (y - f_2)^2 - (y - f_1)^2 = (f_2 - f_1)(f_1 + f_2 - 2y)$. So $\\mathrm E[D] = (f_2 - f_1)(f_1 + f_2 - 2\\,\\mathrm E[y])$ and $\\mathrm{sd}(D) = 2(f_2 - f_1)\\,\\mathrm{sd}(y)$, finite when $\\mathrm E[y^2]$ is. Their ratio is the stated one. For lognormal actuals $\\mathrm E[y] = e^{\\mu + \\sigma^2/2}$, $\\mathrm{sd}(y) = \\mathrm E[y]\\sqrt{e^{\\sigma^2} - 1}$ and $f_i/\\mathrm E[y] = e^{\\sigma c_i - \\sigma^2/2}$, which tends to 0. So $e^{\\sigma^2/2}$ times the ratio is $|1 - (f_1 + f_2)/(2\\,\\mathrm E[y])|\\,e^{\\sigma^2/2}/\\sqrt{e^{\\sigma^2} - 1}$, which tends to 1.",
     "rule": "algebra",
     "by": [
      "S8"
     ]
    },
    {
     "id": "S10",
     "statement": "So for widely spread actuals the absolute log error keeps its power to tell two forecasts apart wherever its ratio is defined. The absolute error keeps it when $0 < H(c_2) < 1$ and the higher forecast is not the median. sMAPE loses it at the rate $\\sigma^{-1/2}$ when $h(c_1) \\ne h(c_2)$. It loses it faster, with a ratio of at most order $\\sigma^{-5/2}$, when $h(c_1) = h(c_2) > 0$. The squared error on lognormal actuals loses it faster than any power of $\\sigma$.",
     "rule": "conclusion",
     "by": [
      "S9"
     ]
    }
   ],
   "artefacts": [
    "A1",
    "A2"
   ],
   "availability": {
    "level": "public"
   }
  },
  {
   "id": "E36",
   "profile": "formal",
   "description": "Proof of C19 (section H14 of the proof document A1, rendered in A2). Where the log of the actual is symmetric with a single peak, the forecast that minimises the sMAPE of a test set is asymptotically normal about the median. Compared with the sample median it is asymptotically as precise for narrow distributions and asymptotically loses all of its precision as the actuals spread out.",
   "foundation": "Kolmogorov probability axioms and real analysis",
   "steps": [
    {
     "id": "S1",
     "statement": "Write $x_t = \\ln y_t - \\ln m$ for the actuals of a test set and $\\theta = \\ln f - \\ln m$ for a forecast $f$. By C3 the mean sMAPE term of the test set is $M_n(\\theta) = \\frac1n\\sum_t\\tanh(|x_t - \\theta|/2)$, and its expectation is $M(\\theta) = \\mathrm E[\\tanh(|x - \\theta|/2)]$. So $\\hat\\theta_n = \\ln\\hat f_n - \\ln m$ minimises $M_n$; a minimiser exists, because $M_n$ is continuous, tends to 1 as $|\\theta|$ grows, and is below 1 at each $x_t$.",
     "rule": "setup",
     "by": [
      "C3"
     ]
    },
    {
     "id": "S2",
     "statement": "(a). Write $\\kappa(u) = \\operatorname{sech}^2(u/2)\\tanh(|u|/2) \\ge 0$. Since $\\kappa(u)$ is the derivative of $\\tanh^2(u/2)$ for $u > 0$, $\\int\\kappa(u)\\,du = 2$. By unimodality $g(\\ln m + u) \\le g(\\ln m)$, so $A = \\int\\kappa(u)\\,g(\\ln m + u)\\,du \\le 2g(\\ln m)$. A density cannot equal $g(\\ln m)$ everywhere, and $g(\\ln m + u)$ does not increase as $|u|$ grows; so $g(\\ln m + u) < g(\\ln m)$ for every $|u|$ beyond some point, where $\\kappa > 0$. So the inequality is strict.",
     "rule": "definition",
     "by": [
      "S1"
     ]
    },
    {
     "id": "S3",
     "statement": "The expected loss is lowest at 0 only. $M(\\theta)$ is the expected sMAPE term of the forecast $m\\,e^\\theta$. By C5, $M(\\theta) > M(0)$ for every $\\theta \\ne 0$. The function $\\tanh(|u|/2)$ has slope at most $\\tfrac12$, so $M$ and every $M_n$ change by at most $\\tfrac12|\\theta_1 - \\theta_2|$ between $\\theta_1$ and $\\theta_2$; in particular $M$ is continuous. And $M(\\theta) \\to 1 > M(0)$ as $|\\theta| \\to \\infty$, by bounded convergence. So for every $\\eta > 0$ the lowest value of $M$ over $|\\theta| \\ge \\eta$ is above $M(0)$.",
     "rule": "bounded convergence",
     "by": [
      "C5"
     ]
    },
    {
     "id": "S4",
     "statement": "$M_n$ is close to $M$ for all $\\theta$ at once. For each $\\theta$, $M_n(\\theta) \\to M(\\theta)$ with probability 1 (law of large numbers; the terms are bounded). Take a grid of spacing $\\gamma$ on $[-R, R]$: by the slope bound of step 3, once $M_n$ is within $\\gamma$ of $M$ at the grid points it is within $2\\gamma$ of $M$ on all of $[-R, R]$. Outside $[-R, R]$: an actual with $|x_t| \\le R/2$ has $\\tanh(|x_t - \\theta|/2) \\ge \\tanh(R/4)$ when $|\\theta| > R$, so $M_n(\\theta)$ is at least $\\tanh(R/4)$ times the share of such actuals, which tends to $\\mathrm P(|x| \\le R/2)$. Since $M(0) < 1$, for $R$ large this is above $M(0)$ by a fixed margin.",
     "rule": "case analysis",
     "by": [
      "S3"
     ]
    },
    {
     "id": "S5",
     "statement": "Consistency. $M_n(\\hat\\theta_n) \\le M_n(0)$, which tends to $M(0)$. By step 4, with probability tending to 1, $|\\hat\\theta_n| \\le R$ and $M(\\hat\\theta_n)$ is within any given margin of $M(0)$; by step 3, $|\\hat\\theta_n| < \\eta$. So $\\hat\\theta_n \\to 0$ in probability, which is $\\hat f_n \\to m$. This is the argument of van der Vaart (1998, Theorem 5.7).",
     "rule": "algebra",
     "by": [
      "S4",
      "S3",
      "B15"
     ]
    },
    {
     "id": "S6",
     "statement": "The curvature at 0. By C12, step 4, with $c = \\ln m$, the second derivative of $M$ at 0 is $g(\\ln m) - \\tfrac12 A = \\tfrac12(2g(\\ln m) - A)$, which is positive by step 2. The first derivative at 0 is 0, because 0 is the minimum (step 3).",
     "rule": "earlier result",
     "by": [
      "S2",
      "S3",
      "C12"
     ]
    },
    {
     "id": "S7",
     "statement": "Asymptotic normality. For every $x \\ne 0$, $\\theta \\mapsto \\tanh(|x - \\theta|/2)$ has a derivative at $\\theta = 0$, $-\\psi(x)$ with $\\psi(x) = \\tfrac12\\operatorname{sech}^2(x/2)$ for $x > 0$ and $-\\tfrac12\\operatorname{sech}^2(x/2)$ for $x < 0$; and $x = 0$ has probability zero. For every $x$ it changes by at most $\\tfrac12|\\theta_1 - \\theta_2|$ (step 3). $M$ has a second-order expansion at its minimum 0 with a positive second derivative (step 6). And $\\hat\\theta_n$ minimises $M_n$ and tends to 0 in probability (step 5). These are the conditions of the theorem on estimators that minimise a criterion which is Lipschitz in the parameter (van der Vaart, 1998, Theorem 5.23). It gives that $\\sqrt n\\,\\hat\\theta_n$ is asymptotically normal with mean 0 and variance $\\mathrm E[\\psi(x)^2]$ divided by the square of the second derivative of $M$ at 0.",
     "rule": "cited theorem",
     "by": [
      "S3",
      "S6",
      "S5",
      "B15"
     ]
    },
    {
     "id": "S8",
     "statement": "The variance. $\\mathrm E[\\psi(x)^2] = \\tfrac14\\mathrm E[\\operatorname{sech}^4(x/2)] = B/4$, and the square of the second derivative is $\\tfrac14(2g(\\ln m) - A)^2$ (step 6). So the variance is $B/(2g(\\ln m) - A)^2$.",
     "rule": "algebra",
     "by": [
      "S6"
     ]
    },
    {
     "id": "S9",
     "statement": "The sample median. The logarithm of the sample median of the $y_t$ is the sample median of the $\\ln y_t$, which minimises the mean of $|x_t - \\theta|$. The same theorem applies with $|x - \\theta|$ in place of $\\tanh(|x - \\theta|/2)$: the derivative at 0 is $-1$ for $x > 0$ and $+1$ for $x < 0$, with square 1; the criterion changes by at most $|\\theta_1 - \\theta_2|$; and the second derivative of $\\mathrm E|x - \\theta|$ at 0 is $2g(\\ln m)$, positive because $g$ is largest at $\\ln m$. So the variance is $1/(4g(\\ln m)^2)$.",
     "rule": "cited theorem",
     "by": [
      "S8",
      "B15"
     ]
    },
    {
     "id": "S10",
     "statement": "(b), the ratio. Dividing the variance of step 9 by that of step 8 gives $\\mathrm{ARE} = (2g(\\ln m) - A)^2/(4g(\\ln m)^2 B)$.",
     "rule": "algebra",
     "by": [
      "S9",
      "S8"
     ]
    },
    {
     "id": "S11",
     "statement": "(c), narrow distributions. Write $\\ln y = \\ln m + \\sigma\\varepsilon$ with $\\varepsilon$ standard normal, or Laplace with scale 1 and $\\sigma = b$; then $g(\\ln m) = h(0)/\\sigma$ with $h$ the density of $\\varepsilon$, and $x = \\sigma\\varepsilon$. As $\\sigma \\to 0$, $A \\to 0$ and $B \\to 1$ by bounded convergence, and $A/(2g(\\ln m)) = \\sigma A/(2h(0)) \\to 0$. So $\\mathrm{ARE} = (1 - A/(2g(\\ln m)))^2/B \\to 1$.",
     "rule": "bounded convergence",
     "by": [
      "S10"
     ]
    },
    {
     "id": "S12",
     "statement": "(c), wide lognormal distributions. Since $\\int\\kappa = 2$ (step 2), $2g(\\ln m) - A = \\int\\kappa(u)\\,[g(\\ln m) - g(\\ln m + u)]\\,du$. For the normal density, $g(\\ln m) - g(\\ln m + u) = \\big(1 - e^{-u^2/(2\\sigma^2)}\\big)/(\\sigma\\sqrt{2\\pi})$, and $\\tfrac{u^2}{2\\sigma^2} - \\tfrac{u^4}{8\\sigma^4} \\le 1 - e^{-u^2/(2\\sigma^2)} \\le \\tfrac{u^2}{2\\sigma^2}$. Integrating by parts, $\\int u^2\\kappa(u)\\,du = 4\\int_0^\\infty u\\operatorname{sech}^2(u/2)\\,du = 16\\ln 2$; and $\\int u^4\\kappa(u)\\,du$ is finite. So $2g(\\ln m) - A = 8\\ln 2/(\\sigma^3\\sqrt{2\\pi}) + O(\\sigma^{-5})$. In the same way, with $\\int\\operatorname{sech}^4(u/2)\\,du = 8/3$, $B = 8/(3\\sigma\\sqrt{2\\pi}) + O(\\sigma^{-3})$. With $g(\\ln m) = 1/(\\sigma\\sqrt{2\\pi})$, $\\mathrm{ARE}\\,\\sigma^3$ tends to $\\dfrac{64(\\ln 2)^2}{2\\pi}\\cdot\\dfrac{2\\pi}{4}\\cdot\\dfrac{3\\sqrt{2\\pi}}{8} = 6(\\ln 2)^2\\sqrt{2\\pi}$.",
     "rule": "algebra",
     "by": [
      "S2"
     ]
    },
    {
     "id": "S13",
     "statement": "(c), wide log-Laplace distributions. Here $g(\\ln m + u) = e^{-|u|/b}/(2b)$, so $g(\\ln m) - g(\\ln m + u) = (1 - e^{-|u|/b})/(2b)$, which lies between $\\big(|u|/b - u^2/(2b^2)\\big)/(2b)$ and $|u|/(2b^2)$. Integrating by parts, $\\int |u|\\,\\kappa(u)\\,du = 2\\int_0^\\infty\\operatorname{sech}^2(u/2)\\,du = 4$. So $2g(\\ln m) - A = 2/b^2 + O(b^{-3})$, and in the same way $B = 4/(3b) + O(b^{-2})$. With $g(\\ln m) = 1/(2b)$, $\\mathrm{ARE}\\,b$ tends to $\\dfrac{4}{b^4}\\cdot b^2\\cdot\\dfrac{3b}{4}\\cdot b = 3$.",
     "rule": "algebra",
     "by": [
      "S12"
     ]
    },
    {
     "id": "S14",
     "statement": "So the sMAPE minimiser of a test set is asymptotically normal about the median, with a variance that the density at the median and two expectations give. On lognormal and log-Laplace actuals it is asymptotically as precise as the sample median for narrow distributions and asymptotically loses all of its precision relative to it as the actuals spread out, at the rates $\\sigma^{-3}$ and $b^{-1}$.",
     "rule": "conclusion",
     "by": [
      "S13"
     ]
    }
   ],
   "artefacts": [
    "A1",
    "A2"
   ],
   "availability": {
    "level": "public"
   }
  },
  {
   "id": "E30",
   "profile": "formal",
   "description": "Proof of C17 (the first proof of the proof document A5). The Suilin loss behaves like sMAPE except near zero, where it becomes an absolute error.",
   "foundation": "Kolmogorov probability axioms and real analysis",
   "steps": [
    {
     "id": "S1",
     "statement": "Which branch of the maximum. $\\ell + \\delta \\ge 0.5 + \\delta$ exactly when $\\ell \\ge 0.5$. So the denominator of $\\tilde s$ is $\\ell + \\delta$ when $\\ell \\ge 0.5$, and $0.5 + \\delta$ when $\\ell < 0.5$.",
     "rule": "algebra",
     "by": [
      "D28"
     ]
    },
    {
     "id": "S2",
     "statement": "(a), the identity. When $\\ell \\ge 0.5$, $\\ell > 0$, so $s = |y-f|/\\ell$ and $|y-f| = s\\,\\ell$. By step 1, $\\tilde s = |y-f|/(\\ell+\\delta) = s\\,\\ell/(\\ell+\\delta)$.",
     "rule": "algebra",
     "by": [
      "S1"
     ]
    },
    {
     "id": "S3",
     "statement": "(a), the factor. The function $\\ell \\mapsto \\ell/(\\ell+\\delta) = 1 - \\delta/(\\ell+\\delta)$ is increasing in $\\ell$, below 1, and tends to 1 as $\\ell$ grows. At $\\ell = 0.5$ it equals $0.5/0.6 = 5/6$. So for $\\ell \\ge 0.5$ it lies in $[5/6, 1)$.",
     "rule": "monotonicity",
     "by": [
      "S2"
     ]
    },
    {
     "id": "S4",
     "statement": "(a), close to $s$, but not a function of the ratio. By steps 2 and 3, $0 \\le s - \\tilde s = s\\,\\delta/(\\ell+\\delta) \\le \\delta/(\\ell+\\delta)$, since $s \\le 1$ (step 7). So $\\tilde s$ is within $\\delta/(\\ell+\\delta)$ of $s$, which shrinks as $\\ell$ grows. But $\\tilde s$ depends on the scale: $(y,f) = (1,2)$ and $(10,20)$ have the same ratio and the same $s = 1/3$, while $\\tilde s = 1/3.1$ and $10/30.1$. So $\\tilde s$ is not a function of $y/f$ alone.",
     "rule": "algebra",
     "by": [
      "S2",
      "S3"
     ]
    },
    {
     "id": "S5",
     "statement": "(a), counts. If $y \\ge 1$, then $\\ell \\ge 1 \\ge 0.5$. If $y = 0$, then $\\ell = |f|$, which is at least 0.5 unless $|f| < 0.5$. So on counts only a zero actual with $|f| < 0.5$ has $\\ell < 0.5$.",
     "rule": "algebra",
     "by": [
      "S4"
     ]
    },
    {
     "id": "S6",
     "statement": "(b). When $\\ell < 0.5$, by step 1, $\\tilde s = |y-f|/(0.5+\\delta)$. For $y = f = 0$ this is 0. For $y = 0$ and $|f| < 0.5$ it is $|f|/(0.5+\\delta)$. And $s(0,f) = |f|/|f| = 1$ for $f \\ne 0$, the largest value of $s$ (step 7).",
     "rule": "algebra",
     "by": [
      "S1"
     ]
    },
    {
     "id": "S7",
     "statement": "(c). $|y - f| \\le |y| + |f| = \\ell$ (the triangle inequality), so $s \\le 1$. When $\\ell \\ge 0.5$, $\\tilde s \\le \\ell/(\\ell+\\delta) < 1$ (step 3). When $\\ell < 0.5$, $\\tilde s \\le \\ell/(0.5+\\delta) < 0.5/0.6 < 1$. And $\\tilde s \\ge 0$ because it is a ratio of a non-negative number to a positive one. A quantity between 0 and 1 has an expectation between 0 and 1, whatever the joint distribution of actuals and forecasts, so the expected Suilin loss is between 0 and 200.",
     "rule": "algebra",
     "by": [
      "S3"
     ]
    },
    {
     "id": "S8",
     "statement": "So the Suilin term is the sMAPE term shrunk by a factor between 5/6 and 1 wherever $|y| + |f| \\ge 0.5$, and an absolute error where $|y| + |f| < 0.5$: it differs from sMAPE in kind only near zero.",
     "rule": "conclusion",
     "by": [
      "S7"
     ]
    }
   ],
   "artefacts": [
    "A5"
   ],
   "availability": {
    "level": "public"
   }
  },
  {
   "id": "E37",
   "profile": "formal",
   "description": "Proof of C20 (the second proof of the proof document A5). The msMAPE term is the sMAPE term except for actuals below its constant, where a zero actual no longer gives every non-zero forecast the largest value.",
   "foundation": "Kolmogorov probability axioms and real analysis",
   "steps": [
    {
     "id": "S1",
     "statement": "(a). When $y \\ge \\nu$, $\\max\\{y,\\nu\\} = y$, and $y > 0$ because $\\nu > 0$. So $s_\\nu = |f-y|/(f+y)$, which is $s(y,f)$ for non-negative $y$ and $f$. If every actual is at least $\\nu$, the two measures are then the same function of the forecast on every test set and under every distribution of such actuals, so they have the same minimisers.",
     "rule": "algebra",
     "by": [
      "D33"
     ]
    },
    {
     "id": "S2",
     "statement": "(b). When $y < \\nu$, $\\max\\{y,\\nu\\} = \\nu$, so $s_\\nu = |f-y|/(f+\\nu)$. For $y = 0$ this is $f/(f+\\nu) = 1 - \\nu/(f+\\nu)$, which is 0 at $f = 0$, increasing in $f$, below 1, and tends to 1 as $f$ grows. The sMAPE term of a zero actual is $|f|/|f| = 1$ for every $f > 0$.",
     "rule": "monotonicity",
     "by": [
      "S1"
     ]
    },
    {
     "id": "S3",
     "statement": "(c), the bounds. $s_\\nu \\ge 0$ because its numerator is non-negative and its denominator is at least $\\nu > 0$. Where $y \\ge \\nu$, $s_\\nu = s \\le 1$, since $|f-y| \\le f+y$. Where $y < \\nu$: if $f \\ge y$, $|f-y| = f-y \\le f < f+\\nu$; if $f < y$, $|f-y| = y-f \\le y < \\nu \\le f+\\nu$. So $s_\\nu < 1$ there.",
     "rule": "algebra",
     "by": [
      "S2"
     ]
    },
    {
     "id": "S4",
     "statement": "(c), no jump at zero. The denominator $f+\\max\\{y,\\nu\\}$ is at least $\\nu > 0$ for every $f \\ge 0$, so $s_\\nu(y,f)$ is a continuous function of $f$ on $f \\ge 0$, at $f = 0$ too. Since $0 \\le s_\\nu \\le 1$ (step 3), the expected term is continuous in $f$ as well, by bounded convergence. The sMAPE term of a zero actual is 0 at $f = 0$ and 1 for $f > 0$; so when zero actuals have probability $p_0 > 0$, the expected sMAPE term jumps by $p_0$ at $f = 0$.",
     "rule": "bounded convergence",
     "by": [
      "S3"
     ]
    },
    {
     "id": "S5",
     "statement": "(d). A count is 0 or at least 1. With $\\nu \\le 1$, a count of at least 1 is at least $\\nu$, so $s_\\nu = s$ by step 1. A zero count is below $\\nu$, where $s_\\nu(0,f) = f/(f+\\nu)$ and $s(0,f) = 1$ for $f > 0$ (step 2).",
     "rule": "algebra",
     "by": [
      "S1",
      "S2"
     ]
    },
    {
     "id": "S6",
     "statement": "So the msMAPE term is the sMAPE term for actuals of at least $\\nu$. Below $\\nu$ its denominator no longer shrinks with the actual: a zero actual scores $f/(f+\\nu)$, and the expected msMAPE has no jump at the zero forecast.",
     "rule": "conclusion",
     "by": [
      "S5"
     ]
    }
   ],
   "artefacts": [
    "A5"
   ],
   "availability": {
    "level": "public"
   }
  }
 ],
 "warrants": [],
 "limitations": [
  {
   "id": "L1",
   "category": "analytic_flexibility",
   "statement": "Lab run RN1 deviated from plan PL1: RN1 ran at commit 3cb6c7c, not the plan's 92cd9a5. Between them only flatness.py and plot_flatness.py (plan PL3), test_smape.py and measures.py changed; in measures.py the unused function log_mae was removed (the supervisor dropped log-MAE from PL3, DC3). targets.py, theory.py, dists.py and the measures targets.py uses are unchanged. Reason: The commit of the PL3 revision came between PL1's plan and its run; it does not touch what PL1 runs.",
   "affects": [
    "C13"
   ],
   "impact": "could_bias_magnitude",
   "mitigation": "none"
  },
  {
   "id": "L2",
   "category": "analytic_flexibility",
   "statement": "Lab run RN7 deviated from plan PL2: RN7 ran efficiency.py at commit fa0ce62, not the plan's 92cd9a5: the script runs its 16 settings in parallel processes (one per core) instead of one after another; theory.py also gained a function (balance_exact) that efficiency.py does not use. Each (setting, test-set size) draws from its own seed as before, and on a subset the parallel code and the plan's code gave byte-identical results.json. Reason: Three earlier attempts on one core (over an hour) were killed by container restarts before the lab recorded them (notes NT2.R2, NT5, NT6); in parallel the run finished in about 20 minutes. The analysis, the settings, the seeds and the results are unchanged.",
   "affects": [
    "C15"
   ],
   "impact": "could_bias_magnitude",
   "mitigation": "none"
  },
  {
   "id": "L3",
   "category": "sampling",
   "affects": [
    "C14",
    "C15",
    "C21"
   ],
   "impact": "reduces_transferability",
   "mitigation": "none",
   "statement": "The test sets are independent draws from a fixed distribution. Real forecast errors are often correlated and their distribution can change over a test period, so a real test set carries less information than the same number of outcomes here."
  },
  {
   "id": "L4",
   "category": "model_specification",
   "affects": [
    "C13",
    "C14",
    "C15",
    "C16",
    "C21"
   ],
   "impact": "reduces_transferability",
   "mitigation": "none",
   "statement": "Only the distribution families studied were computed; the formal claims (C4, C10) say what holds beyond them, the simulations only illustrate it."
  },
  {
   "id": "L5",
   "category": "model_specification",
   "affects": [
    "C6"
   ],
   "impact": "reduces_transferability",
   "mitigation": "none",
   "statement": "C6 is proved for two groups that mirror each other about the median on the log scale. It is not proved for groups of different shape or spread."
  },
  {
   "id": "L6",
   "category": "measurement",
   "affects": [
    "C20",
    "C21"
   ],
   "impact": "reduces_transferability",
   "mitigation": "none",
   "statement": "The definition of msMAPE is taken from Kolassa (2026, Definition 4). Smyl's article was not read. Smyl gives no value for the constant; ν = 0.1, 0.5 and 1 are this paper's choices."
  },
  {
   "id": "L7",
   "category": "other",
   "affects": [
    "C19"
   ],
   "impact": "reduces_transferability",
   "mitigation": "partial",
   "mitigation_evidence": "E15",
   "statement": "That the relative efficiency of finite test sets tends to the asymptotic value is not shown; the simulation of C15 agrees with it at n = 10000."
  }
 ],
 "artefacts": [
  {
   "id": "A1",
   "kind": "proof",
   "uri": "https://entrylib.org/p/2026.000013/v3/files/files/proofs.md",
   "availability": {
    "level": "public"
   },
   "version": "sha256:90c89f3395e068a7a2762e2839e86894e4efe488dd09ca3fa988e7ce0cc1df51",
   "licence": "CC-BY-4.0",
   "description": "Proofs of the formal claims, version 3, in Markdown with LaTeX: section H3 with part (c); H5 shortened to the weighted median; new sections H12 (the ordinary median, with the checker's example), H13 (signal-to-noise ratios) and H14 (large-sample efficiency). Checked independently by AC6 (CK27 to CK31, CK34 to CK37) and corrected as it proposed; after its checks the notation was made consistent (no second meaning for k, a, G), which AC6 read and found to change nothing. Sections H1 to H14 prove C1, C2, C3, C9, C4, C5, C6, C7, C11, C8, C10, C12, C18 and C19, in that order. Sections H1 to H11 were also checked independently by lab actor AC4 (proof-check-report.md) and by the supervisor."
  },
  {
   "id": "A2",
   "kind": "proof",
   "uri": "https://entrylib.org/p/2026.000013/v3/files/files/proofs.pdf",
   "availability": {
    "level": "public"
   },
   "version": "sha256:767603774315d5d392a7fed0c7fb35531c52a4f68f836b0d944aa5c83d5cdc35",
   "licence": "CC-BY-4.0",
   "description": "The same proofs, version 3, rendered as PDF."
  },
  {
   "id": "A3",
   "kind": "figure",
   "uri": "https://entrylib.org/p/2026.000013/v3/files/files/smape_weights.png",
   "availability": {
    "level": "public"
   },
   "description": "Figure 1 of the proofs: the weight y/(y+f)^2 of an actual on the forecast that minimises expected sMAPE, against y/f on a log scale, relative to its largest value, beside MAE's constant weight; drawn by claimset-code/smape-target/plot_weights.py.",
   "version": "sha256:6f1c772b4fb9dc31be0041fe791b9143b7ed3b48e22881599e22abe89fb3265c",
   "licence": "CC-BY-4.0",
   "function": "diagram"
  },
  {
   "id": "A4",
   "kind": "other",
   "uri": "https://entrylib.org/p/2026.000013/v3/files/files/proof-check-report.md",
   "availability": {
    "level": "public"
   },
   "description": "The independent checker's reports on the proofs (Claude Fable 5.2, lab actor AC4): derivations from the statements alone, then comparisons with the proofs, with its counterexamples and the corrections they led to.",
   "version": "sha256:9fa4af751c2b7b283ff795cf6899a388bc21f0d9a3f5177f073e45aa4ac7b93e",
   "licence": "CC-BY-4.0"
  },
  {
   "id": "A5",
   "kind": "proof",
   "uri": "https://entrylib.org/p/2026.000013/v3/files/files/proof-suilin.md",
   "availability": {
    "level": "public"
   },
   "version": "sha256:cc1edb21c3cdff150d3bdd5c455a96849e3c04617fad200aad53f32cf63f03ac",
   "licence": "CC-BY-4.0",
   "description": "Proofs of C17 (record hypothesis H41: the Suilin loss beside sMAPE; the measure renamed, content as before) and of C20 (record hypothesis H42: the msMAPE of Smyl beside sMAPE), in Markdown with LaTeX; checked independently by AC6 (CK32, CK33). The msMAPE term is written s_nu here, as in this claimset; the recorded statement writes p. In version 2 the file was proof-msmape.md."
  },
  {
   "id": "A6",
   "kind": "data",
   "uri": "https://entrylib.org/p/2026.000013/v3/files/files/smape-target.results.csv",
   "version": "sha256:dbcda16de8d13b8449906fa1142665fc1efc4b472ebe082233f3e457c2eb65eb",
   "availability": {
    "level": "public"
   },
   "description": "Every result of the lab record (17182 results of 17 runs) at full precision: the lab result's ID, its run, the quantity as the run's script named it, its conditions and its value (both as JSON). The evidence's tables give the values to four significant digits.",
   "columns": [
    {
     "name": "lab_result",
     "dimension": "the result event of the lab record"
    },
    {
     "name": "run",
     "dimension": "the lab run that produced it"
    },
    {
     "name": "quantity",
     "dimension": "the quantity, as the run's script named it"
    },
    {
     "name": "conditions",
     "dimension": "the result's conditions, as JSON"
    },
    {
     "name": "value",
     "variable": "the value of the quantity, as JSON, at full precision"
    }
   ],
   "licence": "CC-BY-4.0"
  },
  {
   "id": "A7",
   "kind": "lab_record",
   "uri": "https://github.com/entrylib/study-planning/blob/032a092a1879211ed0c8e383b5895ce81ab83e7b/lab/lab.jsonl",
   "version": "sha256:26886d9cdd338f2df8c3c9f78b6458897789cf4236b28c36a34946ece8aa2de8",
   "availability": {
    "level": "public"
   },
   "description": "Lab record of the study: questions, hypotheses, plans, runs, results and decisions, as an append-only, hash-chained event log (Claimset lab record specification v0.22, begun 2026-11-16; exported with claimset 0.44.6).",
   "licence": "CC-BY-4.0"
  },
  {
   "kind": "data",
   "uri": "https://entrylib.org/p/2026.000013/v3/files/files/smape-target.RN1.results.csv",
   "version": "sha256:8730088e073035c66e4e2c8fce4454cdf20b22be2429c65f3244cf15447e4d5a",
   "licence": "CC-BY-4.0",
   "availability": {
    "level": "public"
   },
   "columns": [
    {
     "name": "lab_result",
     "dimension": "the result event of the lab record"
    },
    {
     "name": "run",
     "dimension": "the lab run that produced it"
    },
    {
     "name": "quantity",
     "dimension": "the quantity, as the run's script named it"
    },
    {
     "name": "conditions",
     "dimension": "the result's conditions, as JSON"
    },
    {
     "name": "value",
     "variable": "the value of the quantity, as JSON, at full precision"
    }
   ],
   "description": "Every result of lab run RN1 (PL1: the targets of RMSE, MAE, MAPE and sMAPE on 34 distributions) at full precision, one row per result, sorted by setting: the lab result's ID, its run, the quantity, its conditions and its value (both as JSON). The evidence's tables give selected values to four significant digits.",
   "id": "A8"
  },
  {
   "kind": "figure",
   "function": "data_plot",
   "uri": "https://entrylib.org/p/2026.000013/v3/files/files/figures/RN1-inverse_Gaussian_shape_1000.png",
   "version": "sha256:f5d0bd0230ac0e0aac43c1b708f365817da2d02f5a2a27ae8e2cfcc06676ab57",
   "licence": "CC-BY-4.0",
   "availability": {
    "level": "public"
   },
   "shows": [
    "A8",
    "E13.T3.r1"
   ],
   "description": "inverse Gaussian shape=1000: for each measure (RMSE, MAE, MAPE, sMAPE), one violin per test-set size showing how the candidate the measure ranks first is spread over 1000 test sets, as its quantile level, with a boxplot of its median, quartiles and 5th and 95th percentiles and a diamond at its mean; a dashed line marks the measure's preferred candidate and dotted lines the other measures'. Drawn by lab run RN1 from its own results (lab/runs/RN1/figures/inverse_Gaussian_shape=1000.png).",
   "id": "A9"
  },
  {
   "kind": "figure",
   "function": "data_plot",
   "uri": "https://entrylib.org/p/2026.000013/v3/files/files/figures/RN1-inverse_Gaussian_shape_200.png",
   "version": "sha256:af9f9e18e322a45f776994ea9dbdcf1d6d2aa42eaa783b3541ff8f24a738ce92",
   "licence": "CC-BY-4.0",
   "availability": {
    "level": "public"
   },
   "shows": [
    "A8",
    "E13.T3.r2"
   ],
   "description": "inverse Gaussian shape=200: for each measure (RMSE, MAE, MAPE, sMAPE), one violin per test-set size showing how the candidate the measure ranks first is spread over 1000 test sets, as its quantile level, with a boxplot of its median, quartiles and 5th and 95th percentiles and a diamond at its mean; a dashed line marks the measure's preferred candidate and dotted lines the other measures'. Drawn by lab run RN1 from its own results (lab/runs/RN1/figures/inverse_Gaussian_shape=200.png).",
   "id": "A10"
  },
  {
   "kind": "figure",
   "function": "data_plot",
   "uri": "https://entrylib.org/p/2026.000013/v3/files/files/figures/RN1-inverse_Gaussian_shape_100.png",
   "version": "sha256:a53f27c451f7fa9eb8568638888a469ae77f51592d6c30bcf9264c0c79909b9b",
   "licence": "CC-BY-4.0",
   "availability": {
    "level": "public"
   },
   "shows": [
    "A8",
    "E13.T3.r3"
   ],
   "description": "inverse Gaussian shape=100: for each measure (RMSE, MAE, MAPE, sMAPE), one violin per test-set size showing how the candidate the measure ranks first is spread over 1000 test sets, as its quantile level, with a boxplot of its median, quartiles and 5th and 95th percentiles and a diamond at its mean; a dashed line marks the measure's preferred candidate and dotted lines the other measures'. Drawn by lab run RN1 from its own results (lab/runs/RN1/figures/inverse_Gaussian_shape=100.png).",
   "id": "A11"
  },
  {
   "kind": "figure",
   "function": "data_plot",
   "uri": "https://entrylib.org/p/2026.000013/v3/files/files/figures/RN1-inverse_Gaussian_shape_50.png",
   "version": "sha256:ca028c9c07a271dda4c21c8b066f386bc8005874d64b4046519eee918a5b4875",
   "licence": "CC-BY-4.0",
   "availability": {
    "level": "public"
   },
   "shows": [
    "A8",
    "E13.T3.r4"
   ],
   "description": "inverse Gaussian shape=50: for each measure (RMSE, MAE, MAPE, sMAPE), one violin per test-set size showing how the candidate the measure ranks first is spread over 1000 test sets, as its quantile level, with a boxplot of its median, quartiles and 5th and 95th percentiles and a diamond at its mean; a dashed line marks the measure's preferred candidate and dotted lines the other measures'. Drawn by lab run RN1 from its own results (lab/runs/RN1/figures/inverse_Gaussian_shape=50.png).",
   "id": "A12"
  },
  {
   "kind": "figure",
   "function": "data_plot",
   "uri": "https://entrylib.org/p/2026.000013/v3/files/files/figures/RN1-inverse_Gaussian_shape_20.png",
   "version": "sha256:24ad0b8357ca38cf2e87d87f034a199d3ebb0ddae6c61160944bd12c37084629",
   "licence": "CC-BY-4.0",
   "availability": {
    "level": "public"
   },
   "shows": [
    "A8",
    "E13.T3.r5"
   ],
   "description": "inverse Gaussian shape=20: for each measure (RMSE, MAE, MAPE, sMAPE), one violin per test-set size showing how the candidate the measure ranks first is spread over 1000 test sets, as its quantile level, with a boxplot of its median, quartiles and 5th and 95th percentiles and a diamond at its mean; a dashed line marks the measure's preferred candidate and dotted lines the other measures'. Drawn by lab run RN1 from its own results (lab/runs/RN1/figures/inverse_Gaussian_shape=20.png).",
   "id": "A13"
  },
  {
   "kind": "figure",
   "function": "data_plot",
   "uri": "https://entrylib.org/p/2026.000013/v3/files/files/figures/RN1-inverse_Gaussian_shape_10.png",
   "version": "sha256:96a39a12e046bd64f5abcb622c2535ed0be0e5b78fdea9e7529e92b132b9ec70",
   "licence": "CC-BY-4.0",
   "availability": {
    "level": "public"
   },
   "shows": [
    "A8",
    "E13.T3.r6",
    "E13.T4"
   ],
   "description": "inverse Gaussian shape=10: for each measure (RMSE, MAE, MAPE, sMAPE), one violin per test-set size showing how the candidate the measure ranks first is spread over 1000 test sets, as its quantile level, with a boxplot of its median, quartiles and 5th and 95th percentiles and a diamond at its mean; a dashed line marks the measure's preferred candidate and dotted lines the other measures'. Drawn by lab run RN1 from its own results (lab/runs/RN1/figures/inverse_Gaussian_shape=10.png).",
   "id": "A14"
  },
  {
   "kind": "figure",
   "function": "data_plot",
   "uri": "https://entrylib.org/p/2026.000013/v3/files/files/figures/RN1-log-skewed_left_sd_0.5_skew_-2.png",
   "version": "sha256:8056fb70807935f5e62b171a5dfa5eecd745d0c4ae2fa1e1234d714f38a19e35",
   "licence": "CC-BY-4.0",
   "availability": {
    "level": "public"
   },
   "shows": [
    "A8",
    "E13.T3.r7"
   ],
   "description": "log-skewed left sd=0.5 skew=-2: for each measure (RMSE, MAE, MAPE, sMAPE), one violin per test-set size showing how the candidate the measure ranks first is spread over 1000 test sets, as its quantile level, with a boxplot of its median, quartiles and 5th and 95th percentiles and a diamond at its mean; a dashed line marks the measure's preferred candidate and dotted lines the other measures'. Drawn by lab run RN1 from its own results (lab/runs/RN1/figures/log-skewed_left_sd=0.5_skew=-2.png).",
   "id": "A15"
  },
  {
   "kind": "figure",
   "function": "data_plot",
   "uri": "https://entrylib.org/p/2026.000013/v3/files/files/figures/RN1-log-skewed_left_sd_0.5_skew_-1.png",
   "version": "sha256:6318ac70da750fc75ed6ddc068fdaeadb0d4f81959b49713120235025dcfa93f",
   "licence": "CC-BY-4.0",
   "availability": {
    "level": "public"
   },
   "shows": [
    "A8",
    "E13.T3.r8"
   ],
   "description": "log-skewed left sd=0.5 skew=-1: for each measure (RMSE, MAE, MAPE, sMAPE), one violin per test-set size showing how the candidate the measure ranks first is spread over 1000 test sets, as its quantile level, with a boxplot of its median, quartiles and 5th and 95th percentiles and a diamond at its mean; a dashed line marks the measure's preferred candidate and dotted lines the other measures'. Drawn by lab run RN1 from its own results (lab/runs/RN1/figures/log-skewed_left_sd=0.5_skew=-1.png).",
   "id": "A16"
  },
  {
   "kind": "figure",
   "function": "data_plot",
   "uri": "https://entrylib.org/p/2026.000013/v3/files/files/figures/RN1-log-skewed_left_sd_0.5_skew_-0.5.png",
   "version": "sha256:6a96c0e2128b1ff51db864edb656deff49742c14c23b722ebfc8481b4d89c118",
   "licence": "CC-BY-4.0",
   "availability": {
    "level": "public"
   },
   "shows": [
    "A8",
    "E13.T3.r9"
   ],
   "description": "log-skewed left sd=0.5 skew=-0.5: for each measure (RMSE, MAE, MAPE, sMAPE), one violin per test-set size showing how the candidate the measure ranks first is spread over 1000 test sets, as its quantile level, with a boxplot of its median, quartiles and 5th and 95th percentiles and a diamond at its mean; a dashed line marks the measure's preferred candidate and dotted lines the other measures'. Drawn by lab run RN1 from its own results (lab/runs/RN1/figures/log-skewed_left_sd=0.5_skew=-0.5.png).",
   "id": "A17"
  },
  {
   "kind": "figure",
   "function": "data_plot",
   "uri": "https://entrylib.org/p/2026.000013/v3/files/files/figures/RN1-log-skewed_right_sd_0.5_skew__2.png",
   "version": "sha256:5d72e3ee447c7542bb6e83604e2a2bba5927843069cc6a151f3b487bf4a518d2",
   "licence": "CC-BY-4.0",
   "availability": {
    "level": "public"
   },
   "shows": [
    "A8",
    "E13.T3.r10"
   ],
   "description": "log-skewed right sd=0.5 skew=+2: for each measure (RMSE, MAE, MAPE, sMAPE), one violin per test-set size showing how the candidate the measure ranks first is spread over 1000 test sets, as its quantile level, with a boxplot of its median, quartiles and 5th and 95th percentiles and a diamond at its mean; a dashed line marks the measure's preferred candidate and dotted lines the other measures'. Drawn by lab run RN1 from its own results (lab/runs/RN1/figures/log-skewed_right_sd=0.5_skew=_2.png).",
   "id": "A18"
  },
  {
   "kind": "figure",
   "function": "data_plot",
   "uri": "https://entrylib.org/p/2026.000013/v3/files/files/figures/RN1-log-skewed_right_sd_0.5_skew__1.png",
   "version": "sha256:ca1572ce0b42b900562359e8dc8e5b7b479c5556ec4c8f9e05753974fc9f8abc",
   "licence": "CC-BY-4.0",
   "availability": {
    "level": "public"
   },
   "shows": [
    "A8",
    "E13.T3.r11"
   ],
   "description": "log-skewed right sd=0.5 skew=+1: for each measure (RMSE, MAE, MAPE, sMAPE), one violin per test-set size showing how the candidate the measure ranks first is spread over 1000 test sets, as its quantile level, with a boxplot of its median, quartiles and 5th and 95th percentiles and a diamond at its mean; a dashed line marks the measure's preferred candidate and dotted lines the other measures'. Drawn by lab run RN1 from its own results (lab/runs/RN1/figures/log-skewed_right_sd=0.5_skew=_1.png).",
   "id": "A19"
  },
  {
   "kind": "figure",
   "function": "data_plot",
   "uri": "https://entrylib.org/p/2026.000013/v3/files/files/figures/RN1-log-skewed_right_sd_0.5_skew__0.5.png",
   "version": "sha256:e95028fd94e4b16db183cafbe0d822bec0295db9c37140106d2c025ca37bb036",
   "licence": "CC-BY-4.0",
   "availability": {
    "level": "public"
   },
   "shows": [
    "A8",
    "E13.T3.r12"
   ],
   "description": "log-skewed right sd=0.5 skew=+0.5: for each measure (RMSE, MAE, MAPE, sMAPE), one violin per test-set size showing how the candidate the measure ranks first is spread over 1000 test sets, as its quantile level, with a boxplot of its median, quartiles and 5th and 95th percentiles and a diamond at its mean; a dashed line marks the measure's preferred candidate and dotted lines the other measures'. Drawn by lab run RN1 from its own results (lab/runs/RN1/figures/log-skewed_right_sd=0.5_skew=_0.5.png).",
   "id": "A20"
  },
  {
   "kind": "figure",
   "function": "data_plot",
   "uri": "https://entrylib.org/p/2026.000013/v3/files/files/figures/RN1-lognormal_sigma_0.5.png",
   "version": "sha256:a011adfe460bf790b1ebd750205169d4452078ef8bc303cab5762e6dea06c653",
   "licence": "CC-BY-4.0",
   "availability": {
    "level": "public"
   },
   "shows": [
    "A8",
    "E13.T3.r13"
   ],
   "description": "lognormal sigma=0.5: for each measure (RMSE, MAE, MAPE, sMAPE), one violin per test-set size showing how the candidate the measure ranks first is spread over 1000 test sets, as its quantile level, with a boxplot of its median, quartiles and 5th and 95th percentiles and a diamond at its mean; a dashed line marks the measure's preferred candidate and dotted lines the other measures'. Drawn by lab run RN1 from its own results (lab/runs/RN1/figures/lognormal_sigma=0.5.png).",
   "id": "A21"
  },
  {
   "kind": "figure",
   "function": "data_plot",
   "uri": "https://entrylib.org/p/2026.000013/v3/files/files/figures/RN1-log-skewed_left_sd_1_skew_-2.png",
   "version": "sha256:ef4142826b8527f4cb36daeeeb2d1fc0ff9326f162dbd26d0c7636a4c2f01a2b",
   "licence": "CC-BY-4.0",
   "availability": {
    "level": "public"
   },
   "shows": [
    "A8",
    "E13.T3.r14"
   ],
   "description": "log-skewed left sd=1 skew=-2: for each measure (RMSE, MAE, MAPE, sMAPE), one violin per test-set size showing how the candidate the measure ranks first is spread over 1000 test sets, as its quantile level, with a boxplot of its median, quartiles and 5th and 95th percentiles and a diamond at its mean; a dashed line marks the measure's preferred candidate and dotted lines the other measures'. Drawn by lab run RN1 from its own results (lab/runs/RN1/figures/log-skewed_left_sd=1_skew=-2.png).",
   "id": "A22"
  },
  {
   "kind": "figure",
   "function": "data_plot",
   "uri": "https://entrylib.org/p/2026.000013/v3/files/files/figures/RN1-log-skewed_left_sd_1_skew_-1.png",
   "version": "sha256:c6772470669b9b57709494f4e1d608d99038b7955bab36714608011e664d7829",
   "licence": "CC-BY-4.0",
   "availability": {
    "level": "public"
   },
   "shows": [
    "A8",
    "E13.T3.r15"
   ],
   "description": "log-skewed left sd=1 skew=-1: for each measure (RMSE, MAE, MAPE, sMAPE), one violin per test-set size showing how the candidate the measure ranks first is spread over 1000 test sets, as its quantile level, with a boxplot of its median, quartiles and 5th and 95th percentiles and a diamond at its mean; a dashed line marks the measure's preferred candidate and dotted lines the other measures'. Drawn by lab run RN1 from its own results (lab/runs/RN1/figures/log-skewed_left_sd=1_skew=-1.png).",
   "id": "A23"
  },
  {
   "kind": "figure",
   "function": "data_plot",
   "uri": "https://entrylib.org/p/2026.000013/v3/files/files/figures/RN1-log-skewed_left_sd_1_skew_-0.5.png",
   "version": "sha256:86646dc2610775bd413fc30051cfb66394bc169193f2b51d8389895dd1a52d24",
   "licence": "CC-BY-4.0",
   "availability": {
    "level": "public"
   },
   "shows": [
    "A8",
    "E13.T3.r16"
   ],
   "description": "log-skewed left sd=1 skew=-0.5: for each measure (RMSE, MAE, MAPE, sMAPE), one violin per test-set size showing how the candidate the measure ranks first is spread over 1000 test sets, as its quantile level, with a boxplot of its median, quartiles and 5th and 95th percentiles and a diamond at its mean; a dashed line marks the measure's preferred candidate and dotted lines the other measures'. Drawn by lab run RN1 from its own results (lab/runs/RN1/figures/log-skewed_left_sd=1_skew=-0.5.png).",
   "id": "A24"
  },
  {
   "kind": "figure",
   "function": "data_plot",
   "uri": "https://entrylib.org/p/2026.000013/v3/files/files/figures/RN1-log-skewed_right_sd_1_skew__2.png",
   "version": "sha256:10822d590574773a55a86bba5c539f9f67d317ce88c2b6d03b9cac7ff31973f1",
   "licence": "CC-BY-4.0",
   "availability": {
    "level": "public"
   },
   "shows": [
    "A8",
    "E13.T3.r17"
   ],
   "description": "log-skewed right sd=1 skew=+2: for each measure (RMSE, MAE, MAPE, sMAPE), one violin per test-set size showing how the candidate the measure ranks first is spread over 1000 test sets, as its quantile level, with a boxplot of its median, quartiles and 5th and 95th percentiles and a diamond at its mean; a dashed line marks the measure's preferred candidate and dotted lines the other measures'. Drawn by lab run RN1 from its own results (lab/runs/RN1/figures/log-skewed_right_sd=1_skew=_2.png).",
   "id": "A25"
  },
  {
   "kind": "figure",
   "function": "data_plot",
   "uri": "https://entrylib.org/p/2026.000013/v3/files/files/figures/RN1-log-skewed_right_sd_1_skew__1.png",
   "version": "sha256:ceec8d040504718d6c22be2349f418c184e92324dd3fd17220cf3eae123cfe19",
   "licence": "CC-BY-4.0",
   "availability": {
    "level": "public"
   },
   "shows": [
    "A8",
    "E13.T3.r18"
   ],
   "description": "log-skewed right sd=1 skew=+1: for each measure (RMSE, MAE, MAPE, sMAPE), one violin per test-set size showing how the candidate the measure ranks first is spread over 1000 test sets, as its quantile level, with a boxplot of its median, quartiles and 5th and 95th percentiles and a diamond at its mean; a dashed line marks the measure's preferred candidate and dotted lines the other measures'. Drawn by lab run RN1 from its own results (lab/runs/RN1/figures/log-skewed_right_sd=1_skew=_1.png).",
   "id": "A26"
  },
  {
   "kind": "figure",
   "function": "data_plot",
   "uri": "https://entrylib.org/p/2026.000013/v3/files/files/figures/RN1-log-skewed_right_sd_1_skew__0.5.png",
   "version": "sha256:98bda774447672be32d5ec8f6430847fedafff3efae6ce49ad39601b5a7830d1",
   "licence": "CC-BY-4.0",
   "availability": {
    "level": "public"
   },
   "shows": [
    "A8",
    "E13.T3.r19"
   ],
   "description": "log-skewed right sd=1 skew=+0.5: for each measure (RMSE, MAE, MAPE, sMAPE), one violin per test-set size showing how the candidate the measure ranks first is spread over 1000 test sets, as its quantile level, with a boxplot of its median, quartiles and 5th and 95th percentiles and a diamond at its mean; a dashed line marks the measure's preferred candidate and dotted lines the other measures'. Drawn by lab run RN1 from its own results (lab/runs/RN1/figures/log-skewed_right_sd=1_skew=_0.5.png).",
   "id": "A27"
  },
  {
   "kind": "figure",
   "function": "data_plot",
   "uri": "https://entrylib.org/p/2026.000013/v3/files/files/figures/RN1-lognormal_sigma_1.png",
   "version": "sha256:40ed74ee4bdb7769da22899eae6999feacffc838e089fbf558dc8bc7e506b829",
   "licence": "CC-BY-4.0",
   "availability": {
    "level": "public"
   },
   "shows": [
    "A8",
    "E13.T3.r20",
    "E13.T4"
   ],
   "description": "lognormal sigma=1: for each measure (RMSE, MAE, MAPE, sMAPE), one violin per test-set size showing how the candidate the measure ranks first is spread over 1000 test sets, as its quantile level, with a boxplot of its median, quartiles and 5th and 95th percentiles and a diamond at its mean; a dashed line marks the measure's preferred candidate and dotted lines the other measures'. Drawn by lab run RN1 from its own results (lab/runs/RN1/figures/lognormal_sigma=1.png).",
   "id": "A28"
  },
  {
   "kind": "figure",
   "function": "data_plot",
   "uri": "https://entrylib.org/p/2026.000013/v3/files/files/figures/RN1-bimodal_equal_modes.png",
   "version": "sha256:26cb37244e968b3545bcdb281d4892b2ae8b6840240a14f8cd5a4de9f598067e",
   "licence": "CC-BY-4.0",
   "availability": {
    "level": "public"
   },
   "shows": [
    "A8",
    "E13.T3.r21",
    "E13.T4"
   ],
   "description": "bimodal equal modes: for each measure (RMSE, MAE, MAPE, sMAPE), one violin per test-set size showing how the candidate the measure ranks first is spread over 1000 test sets, as its quantile level, with a boxplot of its median, quartiles and 5th and 95th percentiles and a diamond at its mean; a dashed line marks the measure's preferred candidate and dotted lines the other measures'. Drawn by lab run RN1 from its own results (lab/runs/RN1/figures/bimodal_equal_modes.png).",
   "id": "A29"
  },
  {
   "kind": "figure",
   "function": "data_plot",
   "uri": "https://entrylib.org/p/2026.000013/v3/files/files/figures/RN1-bimodal_left_mode_higher.png",
   "version": "sha256:6aa82197f031431440c191c4a101f706fe992af27d3aae7d8107d9004fa76cd1",
   "licence": "CC-BY-4.0",
   "availability": {
    "level": "public"
   },
   "shows": [
    "A8",
    "E13.T3.r22"
   ],
   "description": "bimodal left mode higher: for each measure (RMSE, MAE, MAPE, sMAPE), one violin per test-set size showing how the candidate the measure ranks first is spread over 1000 test sets, as its quantile level, with a boxplot of its median, quartiles and 5th and 95th percentiles and a diamond at its mean; a dashed line marks the measure's preferred candidate and dotted lines the other measures'. Drawn by lab run RN1 from its own results (lab/runs/RN1/figures/bimodal_left_mode_higher.png).",
   "id": "A30"
  },
  {
   "kind": "figure",
   "function": "data_plot",
   "uri": "https://entrylib.org/p/2026.000013/v3/files/files/figures/RN1-bimodal_right_mode_higher.png",
   "version": "sha256:83fcafcc5d9642995542156f4c285ff74477f82e491b1ba1b76e368a4447d14a",
   "licence": "CC-BY-4.0",
   "availability": {
    "level": "public"
   },
   "shows": [
    "A8",
    "E13.T3.r23"
   ],
   "description": "bimodal right mode higher: for each measure (RMSE, MAE, MAPE, sMAPE), one violin per test-set size showing how the candidate the measure ranks first is spread over 1000 test sets, as its quantile level, with a boxplot of its median, quartiles and 5th and 95th percentiles and a diamond at its mean; a dashed line marks the measure's preferred candidate and dotted lines the other measures'. Drawn by lab run RN1 from its own results (lab/runs/RN1/figures/bimodal_right_mode_higher.png).",
   "id": "A31"
  },
  {
   "kind": "figure",
   "function": "data_plot",
   "uri": "https://entrylib.org/p/2026.000013/v3/files/files/figures/RN1-normal_straddling_zero_mean_0.png",
   "version": "sha256:d40bb1a2e829f99a03b65cf9f39b8576c2f0e91d29d6afc24683dd2a6cdd1a85",
   "licence": "CC-BY-4.0",
   "availability": {
    "level": "public"
   },
   "shows": [
    "A8",
    "E13.T3.r24"
   ],
   "description": "normal straddling zero mean=0: for each measure (RMSE, MAE, MAPE, sMAPE), one violin per test-set size showing how the candidate the measure ranks first is spread over 1000 test sets, as its quantile level, with a boxplot of its median, quartiles and 5th and 95th percentiles and a diamond at its mean; a dashed line marks the measure's preferred candidate and dotted lines the other measures'. Drawn by lab run RN1 from its own results (lab/runs/RN1/figures/normal_straddling_zero_mean=0.png).",
   "id": "A32"
  },
  {
   "kind": "figure",
   "function": "data_plot",
   "uri": "https://entrylib.org/p/2026.000013/v3/files/files/figures/RN1-normal_straddling_zero_mean_0.5.png",
   "version": "sha256:e36409efae653fd4f2732b7db28a3f90fabcd465dde513caf58f90c92c0a03cb",
   "licence": "CC-BY-4.0",
   "availability": {
    "level": "public"
   },
   "shows": [
    "A8",
    "E13.T3.r25"
   ],
   "description": "normal straddling zero mean=0.5: for each measure (RMSE, MAE, MAPE, sMAPE), one violin per test-set size showing how the candidate the measure ranks first is spread over 1000 test sets, as its quantile level, with a boxplot of its median, quartiles and 5th and 95th percentiles and a diamond at its mean; a dashed line marks the measure's preferred candidate and dotted lines the other measures'. Drawn by lab run RN1 from its own results (lab/runs/RN1/figures/normal_straddling_zero_mean=0.5.png).",
   "id": "A33"
  },
  {
   "kind": "figure",
   "function": "data_plot",
   "uri": "https://entrylib.org/p/2026.000013/v3/files/files/figures/RN1-normal_straddling_zero_mean_1.png",
   "version": "sha256:d2fbb1ea0cb462e132fcb94e87e596064070c516adf3427a0ed0313c4ce79c1a",
   "licence": "CC-BY-4.0",
   "availability": {
    "level": "public"
   },
   "shows": [
    "A8",
    "E13.T3.r26"
   ],
   "description": "normal straddling zero mean=1: for each measure (RMSE, MAE, MAPE, sMAPE), one violin per test-set size showing how the candidate the measure ranks first is spread over 1000 test sets, as its quantile level, with a boxplot of its median, quartiles and 5th and 95th percentiles and a diamond at its mean; a dashed line marks the measure's preferred candidate and dotted lines the other measures'. Drawn by lab run RN1 from its own results (lab/runs/RN1/figures/normal_straddling_zero_mean=1.png).",
   "id": "A34"
  },
  {
   "kind": "figure",
   "function": "data_plot",
   "uri": "https://entrylib.org/p/2026.000013/v3/files/files/figures/RN1-lognormal_sigma_0.5_with_zeros_p0_0.25.png",
   "version": "sha256:f3c951b394772df8ff25144b38481e88f111ee88d953aebdd99809fe9981060e",
   "licence": "CC-BY-4.0",
   "availability": {
    "level": "public"
   },
   "shows": [
    "A8",
    "E13.T3.r27"
   ],
   "description": "lognormal sigma=0.5 with zeros p0=0.25: for each measure (RMSE, MAE, MAPE, sMAPE), one violin per test-set size showing how the candidate the measure ranks first is spread over 1000 test sets, as its quantile level, with a boxplot of its median, quartiles and 5th and 95th percentiles and a diamond at its mean; a dashed line marks the measure's preferred candidate and dotted lines the other measures'. Drawn by lab run RN1 from its own results (lab/runs/RN1/figures/lognormal_sigma=0.5_with_zeros_p0=0.25.png).",
   "id": "A35"
  },
  {
   "kind": "figure",
   "function": "data_plot",
   "uri": "https://entrylib.org/p/2026.000013/v3/files/files/figures/RN1-lognormal_sigma_0.5_with_zeros_p0_0.42.png",
   "version": "sha256:4ca3938f9aa48a7a364d161708bf7b021a7b3dc4dc430f73e24ce6d5a1065e39",
   "licence": "CC-BY-4.0",
   "availability": {
    "level": "public"
   },
   "shows": [
    "A8",
    "E13.T3.r28"
   ],
   "description": "lognormal sigma=0.5 with zeros p0=0.42: for each measure (RMSE, MAE, MAPE, sMAPE), one violin per test-set size showing how the candidate the measure ranks first is spread over 1000 test sets, as its quantile level, with a boxplot of its median, quartiles and 5th and 95th percentiles and a diamond at its mean; a dashed line marks the measure's preferred candidate and dotted lines the other measures'. Drawn by lab run RN1 from its own results (lab/runs/RN1/figures/lognormal_sigma=0.5_with_zeros_p0=0.42.png).",
   "id": "A36"
  },
  {
   "kind": "figure",
   "function": "data_plot",
   "uri": "https://entrylib.org/p/2026.000013/v3/files/files/figures/RN1-lognormal_sigma_0.5_with_zeros_p0_0.47.png",
   "version": "sha256:29c07116f529100adc3907e01d1a97743cee8b4fe4fa697299cda70962949cc3",
   "licence": "CC-BY-4.0",
   "availability": {
    "level": "public"
   },
   "shows": [
    "A8",
    "E13.T3.r29"
   ],
   "description": "lognormal sigma=0.5 with zeros p0=0.47: for each measure (RMSE, MAE, MAPE, sMAPE), one violin per test-set size showing how the candidate the measure ranks first is spread over 1000 test sets, as its quantile level, with a boxplot of its median, quartiles and 5th and 95th percentiles and a diamond at its mean; a dashed line marks the measure's preferred candidate and dotted lines the other measures'. Drawn by lab run RN1 from its own results (lab/runs/RN1/figures/lognormal_sigma=0.5_with_zeros_p0=0.47.png).",
   "id": "A37"
  },
  {
   "kind": "figure",
   "function": "data_plot",
   "uri": "https://entrylib.org/p/2026.000013/v3/files/files/figures/RN1-lognormal_sigma_0.5_with_zeros_p0_0.6.png",
   "version": "sha256:3a6440016a17d22a7d6c049a915ee5ddc8d66a8d66bc88af004571badb1d8b1c",
   "licence": "CC-BY-4.0",
   "availability": {
    "level": "public"
   },
   "shows": [
    "A8",
    "E13.T3.r30"
   ],
   "description": "lognormal sigma=0.5 with zeros p0=0.6: for each measure (RMSE, MAE, MAPE, sMAPE), one violin per test-set size showing how the candidate the measure ranks first is spread over 1000 test sets, as its quantile level, with a boxplot of its median, quartiles and 5th and 95th percentiles and a diamond at its mean; a dashed line marks the measure's preferred candidate and dotted lines the other measures'. Drawn by lab run RN1 from its own results (lab/runs/RN1/figures/lognormal_sigma=0.5_with_zeros_p0=0.6.png).",
   "id": "A38"
  },
  {
   "kind": "figure",
   "function": "data_plot",
   "uri": "https://entrylib.org/p/2026.000013/v3/files/files/figures/RN1-lognormal_sigma_1_with_zeros_p0_0.25.png",
   "version": "sha256:7e6f60a3683fc89786b0cda4fe15c2218ba64e1b5470f77e073607899adb32ef",
   "licence": "CC-BY-4.0",
   "availability": {
    "level": "public"
   },
   "shows": [
    "A8",
    "E13.T3.r31"
   ],
   "description": "lognormal sigma=1 with zeros p0=0.25: for each measure (RMSE, MAE, MAPE, sMAPE), one violin per test-set size showing how the candidate the measure ranks first is spread over 1000 test sets, as its quantile level, with a boxplot of its median, quartiles and 5th and 95th percentiles and a diamond at its mean; a dashed line marks the measure's preferred candidate and dotted lines the other measures'. Drawn by lab run RN1 from its own results (lab/runs/RN1/figures/lognormal_sigma=1_with_zeros_p0=0.25.png).",
   "id": "A39"
  },
  {
   "kind": "figure",
   "function": "data_plot",
   "uri": "https://entrylib.org/p/2026.000013/v3/files/files/figures/RN1-lognormal_sigma_1_with_zeros_p0_0.37.png",
   "version": "sha256:a6c14a5defbba14fd2a1b2c3b644ba4e25312ee15866efe7e5ee4d0b46856a6e",
   "licence": "CC-BY-4.0",
   "availability": {
    "level": "public"
   },
   "shows": [
    "A8",
    "E13.T3.r32"
   ],
   "description": "lognormal sigma=1 with zeros p0=0.37: for each measure (RMSE, MAE, MAPE, sMAPE), one violin per test-set size showing how the candidate the measure ranks first is spread over 1000 test sets, as its quantile level, with a boxplot of its median, quartiles and 5th and 95th percentiles and a diamond at its mean; a dashed line marks the measure's preferred candidate and dotted lines the other measures'. Drawn by lab run RN1 from its own results (lab/runs/RN1/figures/lognormal_sigma=1_with_zeros_p0=0.37.png).",
   "id": "A40"
  },
  {
   "kind": "figure",
   "function": "data_plot",
   "uri": "https://entrylib.org/p/2026.000013/v3/files/files/figures/RN1-lognormal_sigma_1_with_zeros_p0_0.42.png",
   "version": "sha256:6da889f0f342f2736d15bdc8644991d41876d8b2729d5d693d56d96ccc72c253",
   "licence": "CC-BY-4.0",
   "availability": {
    "level": "public"
   },
   "shows": [
    "A8",
    "E13.T3.r33",
    "E13.T4"
   ],
   "description": "lognormal sigma=1 with zeros p0=0.42: for each measure (RMSE, MAE, MAPE, sMAPE), one violin per test-set size showing how the candidate the measure ranks first is spread over 1000 test sets, as its quantile level, with a boxplot of its median, quartiles and 5th and 95th percentiles and a diamond at its mean; a dashed line marks the measure's preferred candidate and dotted lines the other measures'. Drawn by lab run RN1 from its own results (lab/runs/RN1/figures/lognormal_sigma=1_with_zeros_p0=0.42.png).",
   "id": "A41"
  },
  {
   "kind": "figure",
   "function": "data_plot",
   "uri": "https://entrylib.org/p/2026.000013/v3/files/files/figures/RN1-lognormal_sigma_1_with_zeros_p0_0.6.png",
   "version": "sha256:07829b1bd3bc8046854bf50b2e5445e974a4e893c55ca934ae8fd83da1518007",
   "licence": "CC-BY-4.0",
   "availability": {
    "level": "public"
   },
   "shows": [
    "A8",
    "E13.T3.r34"
   ],
   "description": "lognormal sigma=1 with zeros p0=0.6: for each measure (RMSE, MAE, MAPE, sMAPE), one violin per test-set size showing how the candidate the measure ranks first is spread over 1000 test sets, as its quantile level, with a boxplot of its median, quartiles and 5th and 95th percentiles and a diamond at its mean; a dashed line marks the measure's preferred candidate and dotted lines the other measures'. Drawn by lab run RN1 from its own results (lab/runs/RN1/figures/lognormal_sigma=1_with_zeros_p0=0.6.png).",
   "id": "A42"
  },
  {
   "kind": "data",
   "uri": "https://entrylib.org/p/2026.000013/v3/files/files/smape-target.RN2.results.csv",
   "version": "sha256:68bdbc0f35ce1959ffbd6bfc20074ac0fd99de4c7dc48bf0c20b6dbb27559ce5",
   "licence": "CC-BY-4.0",
   "availability": {
    "level": "public"
   },
   "columns": [
    {
     "name": "lab_result",
     "dimension": "the result event of the lab record"
    },
    {
     "name": "run",
     "dimension": "the lab run that produced it"
    },
    {
     "name": "quantity",
     "dimension": "the quantity, as the run's script named it"
    },
    {
     "name": "conditions",
     "dimension": "the result's conditions, as JSON"
    },
    {
     "name": "value",
     "variable": "the value of the quantity, as JSON, at full precision"
    }
   ],
   "description": "Every result of lab run RN2 (PL3: flatness of expected sMAPE as the spread of ln y grows) at full precision, one row per result, sorted by setting: the lab result's ID, its run, the quantity, its conditions and its value (both as JSON). The evidence's tables give selected values to four significant digits.",
   "id": "A43"
  },
  {
   "kind": "figure",
   "function": "data_plot",
   "uri": "https://entrylib.org/p/2026.000013/v3/files/files/figures/RN2-flatness_target.png",
   "version": "sha256:dc4e6e88aafc1434a3e356b118b2dbca6351976ac70f0cc68dfad5030756409f",
   "licence": "CC-BY-4.0",
   "availability": {
    "level": "public"
   },
   "shows": [
    "A43",
    "E14.T1"
   ],
   "description": "The quantile level of the sMAPE target against the spread sigma of ln y, for lognormal and log-Gumbel outcomes, beside the level of the log-mode: the target moves from the median to the mode of ln y as the spread grows. Drawn by lab run RN2 from its own results (lab/runs/RN2/figures/flatness_target.png).",
   "id": "A44"
  },
  {
   "kind": "figure",
   "function": "data_plot",
   "uri": "https://entrylib.org/p/2026.000013/v3/files/files/figures/RN2-flatness_expected_sMAPE.png",
   "version": "sha256:dc6e08a7028ea86bb04964ccd0d1706fb2c442e525c67516ab9ae320b57dec24",
   "licence": "CC-BY-4.0",
   "availability": {
    "level": "public"
   },
   "shows": [
    "A43",
    "E14.T1"
   ],
   "description": "Expected sMAPE against the quantile level of the forecast, one curve per spread sigma, for each family: the curve flattens towards 200 as sigma grows. Drawn by lab run RN2 from its own results (lab/runs/RN2/figures/flatness_expected_sMAPE.png).",
   "id": "A45"
  },
  {
   "kind": "figure",
   "function": "data_plot",
   "uri": "https://entrylib.org/p/2026.000013/v3/files/files/figures/RN2-flatness_sensitivity.png",
   "version": "sha256:75311a17b81db4fe843fa59bf4904e2f51a161b44990e8319ba6b711bf9137b5",
   "licence": "CC-BY-4.0",
   "availability": {
    "level": "public"
   },
   "shows": [
    "A43",
    "E14.T3"
   ],
   "description": "Lognormal outcomes: how much worse a forecast half a sigma or one sigma away from each measure's target scores, as a share of the score at the target, against sigma. Drawn by lab run RN2 from its own results (lab/runs/RN2/figures/flatness_sensitivity.png).",
   "id": "A46"
  },
  {
   "kind": "figure",
   "function": "data_plot",
   "uri": "https://entrylib.org/p/2026.000013/v3/files/files/figures/RN2-flatness_agreement.png",
   "version": "sha256:afd9a7ed252ca68065400c0f1609731e96542684723d49d9fe7570d734fb23d7",
   "licence": "CC-BY-4.0",
   "availability": {
    "level": "public"
   },
   "shows": [
    "A43",
    "E14.T2"
   ],
   "description": "Ranking agreement: the share of test sets that prefer each measure's target to a forecast away from it, against sigma, for several test-set sizes. Drawn by lab run RN2 from its own results (lab/runs/RN2/figures/flatness_agreement.png).",
   "id": "A47"
  },
  {
   "kind": "figure",
   "function": "data_plot",
   "uri": "https://entrylib.org/p/2026.000013/v3/files/files/figures/RN2-flatness_minimiser.png",
   "version": "sha256:cb05eff70b696b0b528f4a027c0fa631a8910ee25201f93dac17831f6f9c5042",
   "licence": "CC-BY-4.0",
   "availability": {
    "level": "public"
   },
   "shows": [
    "A43",
    "E14.T4"
   ],
   "description": "Where the sMAPE minimiser of a test set falls, and how widely it spreads: its quantile level over test sets against sigma, for each family. Drawn by lab run RN2 from its own results (lab/runs/RN2/figures/flatness_minimiser.png).",
   "id": "A48"
  },
  {
   "kind": "data",
   "uri": "https://entrylib.org/p/2026.000013/v3/files/files/smape-target.RN7.results.csv",
   "version": "sha256:e8dfaaa211d9abab024d15860ada00b42ea3ebf2d5d68d55979d1070085a5bd5",
   "licence": "CC-BY-4.0",
   "availability": {
    "level": "public"
   },
   "columns": [
    {
     "name": "lab_result",
     "dimension": "the result event of the lab record"
    },
    {
     "name": "run",
     "dimension": "the lab run that produced it"
    },
    {
     "name": "quantity",
     "dimension": "the quantity, as the run's script named it"
    },
    {
     "name": "conditions",
     "dimension": "the result's conditions, as JSON"
    },
    {
     "name": "value",
     "variable": "the value of the quantity, as JSON, at full precision"
    }
   ],
   "description": "Every result of lab run RN7 (PL2: efficiency of the sMAPE minimiser against the sample median) at full precision, one row per result, sorted by setting: the lab result's ID, its run, the quantity, its conditions and its value (both as JSON). The evidence's tables give selected values to four significant digits.",
   "id": "A49"
  },
  {
   "kind": "figure",
   "function": "data_plot",
   "uri": "https://entrylib.org/p/2026.000013/v3/files/files/figures/RN7-efficiency_lognormal.png",
   "version": "sha256:51087c8b21b4feaa3c513d5c02b08e2e9005f4ce3bf60c84cfcb04ff5569002a",
   "licence": "CC-BY-4.0",
   "availability": {
    "level": "public"
   },
   "shows": [
    "A49",
    "E15.T1"
   ],
   "description": "lognormal outcomes: relative efficiency of the sMAPE minimiser of a test set against the sample median, against test-set size, one panel per spread, with its 99% band and the large-sample value as a dashed line; below 1 the sMAPE minimiser varies more than the sample median. Drawn by lab run RN7 from its own results (lab/runs/RN7/figures/efficiency_lognormal.png).",
   "id": "A50"
  },
  {
   "kind": "figure",
   "function": "data_plot",
   "uri": "https://entrylib.org/p/2026.000013/v3/files/files/figures/RN7-efficiency_log-Laplace.png",
   "version": "sha256:2fbd7b03699010e6a3d8ef724559f363df2897ba07e2321bd713f69b78f71c44",
   "licence": "CC-BY-4.0",
   "availability": {
    "level": "public"
   },
   "shows": [
    "A49",
    "E15.T1"
   ],
   "description": "log-Laplace outcomes: relative efficiency of the sMAPE minimiser of a test set against the sample median, against test-set size, one panel per spread, with its 99% band and the large-sample value as a dashed line; below 1 the sMAPE minimiser varies more than the sample median. Drawn by lab run RN7 from its own results (lab/runs/RN7/figures/efficiency_log-Laplace.png).",
   "id": "A51"
  },
  {
   "kind": "data",
   "uri": "https://entrylib.org/p/2026.000013/v3/files/files/smape-target.RN10.results.csv",
   "version": "sha256:cb6cbd1d26192f09ce3c9f70de221aa4d2fda0a14fc5c263d327e0dfa8c4ca41",
   "licence": "CC-BY-4.0",
   "availability": {
    "level": "public"
   },
   "columns": [
    {
     "name": "lab_result",
     "dimension": "the result event of the lab record"
    },
    {
     "name": "run",
     "dimension": "the lab run that produced it"
    },
    {
     "name": "quantity",
     "dimension": "the quantity, as the run's script named it"
    },
    {
     "name": "conditions",
     "dimension": "the result's conditions, as JSON"
    },
    {
     "name": "value",
     "variable": "the value of the quantity, as JSON, at full precision"
    }
   ],
   "description": "Every result of lab run RN10 (PL11: the Suilin and sMAPE targets) at full precision, one row per result, sorted by setting: the lab result's ID, its run, the quantity, its conditions and its value (both as JSON). The evidence's tables give selected values to four significant digits. The run named the Suilin loss msMAPE, as version 2 of the paper did: in this file the measure \"msMAPE\" and the quantity \"relative difference of the msMAPE target from the sMAPE target\" are the Suilin loss's.",
   "id": "A52"
  },
  {
   "kind": "figure",
   "function": "data_plot",
   "uri": "https://entrylib.org/p/2026.000013/v3/files/files/figures/RN11-msmape_targets.png",
   "version": "sha256:4ef7aa9d4f706665505aad414777a3b46378f9ea7075afef48b9a1c7c8402d24",
   "licence": "CC-BY-4.0",
   "availability": {
    "level": "public"
   },
   "shows": [
    "E16.T1",
    "E16.T2"
   ],
   "description": "The Suilin loss beside sMAPE: near zero, the two targets as a share of the median against the share of zeros, for lognormal outcomes at medians 1 and 100, with the zero thresholds at median 1; away from zero, the relative difference of the Suilin target from the sMAPE target for each distribution; on Poisson counts, the two targets against the mean. The figure was drawn when the paper called the Suilin loss msMAPE, and labels it so. Drawn by lab run RN11 from the recorded results of RN10 (lab/runs/RN11/figures/msmape_targets.png).",
   "id": "A53"
  },
  {
   "id": "A54",
   "kind": "code",
   "uri": "https://github.com/entrylib/study-planning/tree/6aa7a7ab87bf5c4e40d704a7492cd9af9ef481db/claimset-code/smape-target",
   "version": "6aa7a7ab87bf5c4e40d704a7492cd9af9ef481db",
   "licence": "CC-BY-4.0",
   "availability": {
    "level": "public"
   },
   "description": "The study's code: the distributions, the exact quadrature of expected sMAPE, the formula checks, the three simulations (targets.py, efficiency.py, flatness.py), and the targets of the Suilin loss (msmape_targets.py) and of msMAPE on intermittent demand (msmape_intermittent.py)."
  },
  {
   "kind": "data",
   "uri": "https://entrylib.org/p/2026.000013/v3/files/files/smape-target.RN15.results.csv",
   "version": "sha256:c9c7808e3d002533c95d0f102ad903c1a5d7799ff6d23866507fff9eda260023",
   "licence": "CC-BY-4.0",
   "availability": {
    "level": "public"
   },
   "columns": [
    {
     "name": "lab_result",
     "dimension": "the result event of the lab record"
    },
    {
     "name": "run",
     "dimension": "the lab run that produced it"
    },
    {
     "name": "quantity",
     "dimension": "the quantity, as the run's script named it"
    },
    {
     "name": "conditions",
     "dimension": "the result's conditions, as JSON"
    },
    {
     "name": "value",
     "variable": "the value of the quantity, as JSON, at full precision"
    }
   ],
   "description": "Every result of lab run RN15 (PL14: the targets of sMAPE, msMAPE and the Suilin loss on intermittent demand) at full precision, one row per result, sorted by setting: the lab result's ID, its run, the quantity, its conditions and its value (both as JSON). The evidence's tables give selected values to four significant digits.",
   "id": "A55"
  },
  {
   "kind": "figure",
   "function": "data_plot",
   "uri": "https://entrylib.org/p/2026.000013/v3/files/files/figures/RN15-msmape_intermittent.png",
   "version": "sha256:591a7ac9e2f7b9c1561fc3dfb53c9127c53609687cc9bcac9d33f07540a32390",
   "licence": "CC-BY-4.0",
   "availability": {
    "level": "public"
   },
   "shows": [
    "A55",
    "E34.T1"
   ],
   "description": "The forecast with the lowest expected score (the target) against the share of zero actuals, on intermittent demand: one panel per family of settings (Poisson counts, negative binomial counts, lognormal demand sizes with zeros), one line per measure (sMAPE, msMAPE with nu = 0.1, 0.5 and 1, the Suilin loss). Drawn by lab run RN15 from its own results (lab/runs/RN15/figures/msmape_intermittent.png).",
   "id": "A56"
  }
 ],
 "references": [
  {
   "id": "B1",
   "target": "10.1016/S0169-2070(99)00007-2",
   "id_type": "doi"
  },
  {
   "id": "B2",
   "target": "10.1090/S0002-9939-1955-0069229-1",
   "id_type": "doi"
  },
  {
   "id": "B3",
   "target": "entrylib:2026.000008@v2#C33",
   "id_type": "claimset"
  },
  {
   "id": "B4",
   "target": "entrylib:2026.000009@v1#C5",
   "id_type": "claimset"
  },
  {
   "id": "B5",
   "target": "entrylib:2026.000009@v1#C6",
   "id_type": "claimset"
  },
  {
   "id": "B6",
   "target": "entrylib:2026.000008@v2#C6",
   "id_type": "claimset"
  },
  {
   "id": "B7",
   "target": "entrylib:2026.000009@v1#C11",
   "id_type": "claimset"
  },
  {
   "id": "B8",
   "target": "10.1016/0169-2070(93)90079-3",
   "id_type": "doi"
  },
  {
   "id": "B9",
   "target": "10.1016/j.ijforecast.2006.03.001",
   "id_type": "doi"
  },
  {
   "id": "B10",
   "target": "10.1007/s10618-022-00894-5",
   "id_type": "doi"
  },
  {
   "id": "B11",
   "target": "10.1371/journal.pone.0174202",
   "id_type": "doi"
  },
  {
   "id": "B12",
   "target": "10.1057/jors.2014.103",
   "id_type": "doi"
  },
  {
   "id": "B13",
   "target": "10.3390/forecast8020024",
   "id_type": "doi"
  },
  {
   "id": "B14",
   "target": "10.1016/j.ijforecast.2020.06.008",
   "id_type": "doi"
  },
  {
   "id": "B15",
   "target": "10.1017/CBO9780511802256",
   "id_type": "doi"
  },
  {
   "id": "B16",
   "target": "Smyl, S. (2025). Sparse-Proof sMAPE. Foresight: The International Journal of Applied Forecasting, 78, 29–30. https://ideas.repec.org/a/for/ijafaa/y2025i78p29-30.html",
   "id_type": "reference"
  }
 ]
}
