LAB RECORD: What does sMAPE reward? An analytical and simulation study Study: smape-target · Lab record specification v0.22 · Begun 2026-11-16 · Repository: https://github.com/entrylib/study-planning Events: 190; runs: 11; results: 15,996 (kept in files, never listed here); head: seq 16186, sha256:c767565348c39f3380cd121cc7700c981775229a5ea952eae7601255500b11c8 This view is made from the log; the log is the record (lab record §11). WHO - AC1 Claude Code (agent; Anthropic Claude Opus 5.5; in Claude Code; derivations and simulations), main log; credit: conceptualization, methodology, formal_analysis - AC2 Ivan Svetunkov (human; ORCID 0000-0001-7826-0281), main log; credit: validation, supervision - AC4 Claude Code (agent; Anthropic Claude Fable 5.2; in Claude Code; proof checking), main log; credit: validation QUESTIONS AND HYPOTHESES Q1 Which point forecast minimises the expected sMAPE of a forecast, for which distributions of outcomes (positive, or also zero or negative), and why: what in the… H1 [supported] For $y > 0$ and $f > 0$, the [[TM1:sMAPE term]] $s(y,f) = |y-f|/(|y|+|f|)$ depends on $y$ and $f$ only through their ratio $r = y/f$: $s = |r - 1|/(r + 1)$. 2026-11-16 13:20 Q1 question by AC1: Which point forecast minimises the expected sMAPE of a forecast, for which distributions of outcomes (positive, or also zero or negative), and why: what in the… 2026-11-16 13:20 TM1 term by AC1: sMAPE term [variable] 2026-11-16 13:21 H1 hypothesis by AC1: For a positive outcome $y$ and a positive forecast $f$, the [[TM1:sMAPE term]] depends on $y$ and $f$ only through their ratio $r = y/f$: $s(y,f) = 200\,|r-1|/… [a_priori] 2026-11-16 14:49 PL5 plan by AC1: [confirmatory, requires approval] for H1, H2, H3, H4, H5, H6, H7, H8, H9, H10, H11, H12, H13, H14: Formal proofs (proofs.md), checked independently by AC4 (Claude Fable 5.2: a derivation from the statements alone, then… 2026-11-16 14:58 AP4 approval by AC1: PL5 approved by AC2 at 2026-11-16T14:58:18Z, recorded by agent AC1: "1. Yes 2. For this study 3. Approved 4. render proofs.md as pdf so that I can read LaTeX f… 2026-11-16 15:14 RN5 run by AC1: [confirmatory, success, under PL5, commit 1e23bdc] python3 claimset-code/smape-target/check_formulas.py --seed 6; 39 results in runs/RN5/results.json (RS14882 to RS14920) 2026-11-16 17:28 H21 hypothesis by AC1: For $y > 0$ and $f > 0$, the [[TM1:sMAPE term]] $s(y,f) = |y-f|/(|y|+|f|)$ depends on $y$ and $f$ only through their ratio $r = y/f$: $s = |r - 1|/(r + 1)$. [a_priori] 2026-11-16 17:43 CK1 check by AC4: H1: confirmed (full) 2026-11-16 17:46 ST5 status by AC1: H1 -> supported (plan PL5), evidence RN5, CK1, AR1: PL5's rule for supported is met: the proof (paper H1 in proofs.md, AR1) is complete; the independent checker AC4 confir… 2026-11-16 21:55 CK14 check by AC1: H1: confirmed (full), checked by AC2, recorded by agent AC1: "All looks fine now. NExt" 2026-11-16 22:55 PL8 plan by AC1: [confirmatory, requires approval] for H1, H2, H3, H13, H4, H5, H7, H8, H12, H14, H34: PL5 decided the hypotheses as worded before the restatement of DC5 (s without the factor 200, 'actual', merged results,… 2026-11-16 23:16 DV3 deviation by AC1: RN5 ran at commit 1e23bdc, not the plan's f865dfa. Between them only encode.py (new), targets.py and flatness.py changed (the --derived-only option of DC4); ch… 2026-11-17 09:21 PL9 plan by AC1: [confirmatory, requires approval] for H1, H2, H3, H13, H4, H5, H7, H8, H12, H14, H34: PL5 decided the hypotheses as worded before the restatement of DC5 (s without the factor 200, 'actual', merged results,… 2026-11-17 09:21 AP6 approval by AC1: PL9 approved by AC2 at 2026-11-17T09:20:47Z, recorded by agent AC1: "Record. And the plan with these corrections is approved." 2026-11-17 09:23 RN8 run by AC1: [confirmatory, success, under PL9, commit fa0ce62] python3 claimset-code/smape-target/check_formulas.py --seed 6; 39 results in runs/RN8/results.json (RS15749 to RS15787) 2026-11-17 09:23 ST21 status by AC1: H1 -> supported (plan PL8), evidence RN8, CK1, CK14, AR5: PL8's rule for supported is met for the current wording: the proof (paper H1 in proofs.md, AR1.R2) is complete; the ind… H2 [supported] For $y > 0$ and $f > 0$: (a) replacing $r = y/f$ by $1/r$ leaves the [[TM1: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 $e$ from $y$ ($0 < e < y$), the higher one scores bett… 2026-11-16 13:20 Q1 question by AC1: Which point forecast minimises the expected sMAPE of a forecast, for which distributions of outcomes (positive, or also zero or negative), and why: what in the… 2026-11-16 13:20 TM1 term by AC1: sMAPE term [variable] 2026-11-16 13:21 H2 hypothesis by AC1: For a positive outcome and a positive forecast, the [[TM1:sMAPE term]] is unchanged when the ratio $r = y/f$ is replaced by $1/r$: a forecast too high by a fac… [a_priori] 2026-11-16 14:49 PL5 plan by AC1: [confirmatory, requires approval] for H1, H2, H3, H4, H5, H6, H7, H8, H9, H10, H11, H12, H13, H14: Formal proofs (proofs.md), checked independently by AC4 (Claude Fable 5.2: a derivation from the statements alone, then… 2026-11-16 14:58 AP4 approval by AC1: PL5 approved by AC2 at 2026-11-16T14:58:18Z, recorded by agent AC1: "1. Yes 2. For this study 3. Approved 4. render proofs.md as pdf so that I can read LaTeX f… 2026-11-16 15:14 RN5 run by AC1: [confirmatory, success, under PL5, commit 1e23bdc] python3 claimset-code/smape-target/check_formulas.py --seed 6; 39 results in runs/RN5/results.json (RS14882 to RS14920) 2026-11-16 17:28 H22 hypothesis by AC1: For $y > 0$ and $f > 0$: (a) replacing $r = y/f$ by $1/r$ leaves the [[TM1:sMAPE term]] unchanged, so a forecast $k$ times too high ($f = ky$) and one $k$ time… [a_priori] 2026-11-16 17:43 CK2 check by AC4: H2: confirmed (full) 2026-11-16 17:46 ST6 status by AC1: H2 -> supported (plan PL5), evidence RN5, CK2, AR1: PL5's rule for supported is met: the proof (paper H2 in proofs.md, AR1) is complete; the independent checker AC4 confir… 2026-11-16 21:55 CK15 check by AC1: H2: confirmed (full), checked by AC2, recorded by agent AC1: "All looks fine now. NExt" 2026-11-16 22:55 PL8 plan by AC1: [confirmatory, requires approval] for H1, H2, H3, H13, H4, H5, H7, H8, H12, H14, H34: PL5 decided the hypotheses as worded before the restatement of DC5 (s without the factor 200, 'actual', merged results,… 2026-11-16 23:16 DV3 deviation by AC1: RN5 ran at commit 1e23bdc, not the plan's f865dfa. Between them only encode.py (new), targets.py and flatness.py changed (the --derived-only option of DC4); ch… 2026-11-17 09:21 PL9 plan by AC1: [confirmatory, requires approval] for H1, H2, H3, H13, H4, H5, H7, H8, H12, H14, H34: PL5 decided the hypotheses as worded before the restatement of DC5 (s without the factor 200, 'actual', merged results,… 2026-11-17 09:21 AP6 approval by AC1: PL9 approved by AC2 at 2026-11-17T09:20:47Z, recorded by agent AC1: "Record. And the plan with these corrections is approved." 2026-11-17 09:23 RN8 run by AC1: [confirmatory, success, under PL9, commit fa0ce62] python3 claimset-code/smape-target/check_formulas.py --seed 6; 39 results in runs/RN8/results.json (RS15749 to RS15787) 2026-11-17 09:23 ST22 status by AC1: H2 -> supported (plan PL8), evidence RN8, CK2, CK15, AR5: PL8's rule for supported is met for the current wording: the proof (paper H2 in proofs.md, AR1.R2) is complete; the ind… H3 [supported] For $y > 0$ and $f > 0$, the [[TM1:sMAPE term]] is $s = \tanh(|\ln y - \ln f|/2)$ ([[TM4: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|… 2026-11-16 13:20 Q1 question by AC1: Which point forecast minimises the expected sMAPE of a forecast, for which distributions of outcomes (positive, or also zero or negative), and why: what in the… 2026-11-16 13:20 TM1 term by AC1: sMAPE term [variable] 2026-11-16 13:20 TM4 term by AC1: log error [variable] 2026-11-16 13:21 H3 hypothesis by AC1: For a positive outcome and a positive forecast, the [[TM1:sMAPE term]] equals $200\tanh(|z|/2)$, where $z$ is the [[TM4:log error]]: it is 0 at $z = 0$, strict… [a_priori] 2026-11-16 14:49 PL5 plan by AC1: [confirmatory, requires approval] for H1, H2, H3, H4, H5, H6, H7, H8, H9, H10, H11, H12, H13, H14: Formal proofs (proofs.md), checked independently by AC4 (Claude Fable 5.2: a derivation from the statements alone, then… 2026-11-16 14:58 AP4 approval by AC1: PL5 approved by AC2 at 2026-11-16T14:58:18Z, recorded by agent AC1: "1. Yes 2. For this study 3. Approved 4. render proofs.md as pdf so that I can read LaTeX f… 2026-11-16 15:14 RN5 run by AC1: [confirmatory, success, under PL5, commit 1e23bdc] python3 claimset-code/smape-target/check_formulas.py --seed 6; 39 results in runs/RN5/results.json (RS14882 to RS14920) 2026-11-16 17:28 H23 hypothesis by AC1: For $y > 0$ and $f > 0$, the [[TM1:sMAPE term]] is $s = \tanh(|\ln y - \ln f|/2)$ ([[TM4:log error]] $\ln y - \ln f$). (a) It is 0 when $y = f$, strictly incre… [a_priori] 2026-11-16 17:43 CK3 check by AC4: H3: confirmed (full) 2026-11-16 17:46 ST7 status by AC1: H3 -> supported (plan PL5), evidence RN5, CK3, AR1: PL5's rule for supported is met: the proof (paper H3 in proofs.md, AR1) is complete; the independent checker AC4 confir… 2026-11-16 21:55 CK16 check by AC1: H3: confirmed (full), checked by AC2, recorded by agent AC1: "All looks fine now. NExt" 2026-11-16 22:55 PL8 plan by AC1: [confirmatory, requires approval] for H1, H2, H3, H13, H4, H5, H7, H8, H12, H14, H34: PL5 decided the hypotheses as worded before the restatement of DC5 (s without the factor 200, 'actual', merged results,… 2026-11-16 23:16 DV3 deviation by AC1: RN5 ran at commit 1e23bdc, not the plan's f865dfa. Between them only encode.py (new), targets.py and flatness.py changed (the --derived-only option of DC4); ch… 2026-11-17 09:21 PL9 plan by AC1: [confirmatory, requires approval] for H1, H2, H3, H13, H4, H5, H7, H8, H12, H14, H34: PL5 decided the hypotheses as worded before the restatement of DC5 (s without the factor 200, 'actual', merged results,… 2026-11-17 09:21 AP6 approval by AC1: PL9 approved by AC2 at 2026-11-17T09:20:47Z, recorded by agent AC1: "Record. And the plan with these corrections is approved." 2026-11-17 09:23 RN8 run by AC1: [confirmatory, success, under PL9, commit fa0ce62] python3 claimset-code/smape-target/check_formulas.py --seed 6; 39 results in runs/RN8/results.json (RS15749 to RS15787) 2026-11-17 09:24 ST23 status by AC1: H3 -> supported (plan PL8), evidence RN8, CK3, CK16, AR5: PL8's rule for supported is met for the current wording: the proof (paper H3 in proofs.md, AR1.R2) is complete; the ind… H4 [supported] Let $Y$ 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 [[TM5:weight]] $w(y,f) = y/(y+f)^2$: $\mathrm E[w(Y,f)\mathbf 1\{Yf\}]$. So th… 2026-11-16 13:20 Q1 question by AC1: Which point forecast minimises the expected sMAPE of a forecast, for which distributions of outcomes (positive, or also zero or negative), and why: what in the… 2026-11-16 13:20 TM5 term by AC1: pull [variable] 2026-11-16 13:20 TM6 term by AC1: sMAPE target [concept] 2026-11-16 13:21 H4 hypothesis by AC1: For positive outcomes with a continuous distribution and a positive forecast $f$, $S'(f) = \frac{200}{f}\,\mathrm{E}[w(Y,f)(\mathbf 1\{Yf\}… [a_priori] 2026-11-16 14:49 PL5 plan by AC1: [confirmatory, requires approval] for H1, H2, H3, H4, H5, H6, H7, H8, H9, H10, H11, H12, H13, H14: Formal proofs (proofs.md), checked independently by AC4 (Claude Fable 5.2: a derivation from the statements alone, then… 2026-11-16 14:58 AP4 approval by AC1: PL5 approved by AC2 at 2026-11-16T14:58:18Z, recorded by agent AC1: "1. Yes 2. For this study 3. Approved 4. render proofs.md as pdf so that I can read LaTeX f… 2026-11-16 15:14 RN5 run by AC1: [confirmatory, success, under PL5, commit 1e23bdc] python3 claimset-code/smape-target/check_formulas.py --seed 6; 39 results in runs/RN5/results.json (RS14882 to RS14920) 2026-11-16 17:28 H25 hypothesis by AC1: Let $Y$ 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 ab… [a_priori] 2026-11-16 17:43 CK5 check by AC4: H4: disputed (full) 2026-11-16 17:44 H32 hypothesis by AC1: Let $Y$ 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 ab… [a_priori] 2026-11-16 17:45 CK12 check by AC4: H4: confirmed (full) 2026-11-16 17:46 ST9 status by AC1: H4 -> supported (plan PL5), evidence RN5, CK12, AR1: PL5's rule for supported is met: the proof (paper H5 in proofs.md, AR1) is complete; the independent checker AC4 confir… 2026-11-16 21:33 H33 hypothesis by AC1: Let $Y$ 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 ab… [a_priori] 2026-11-16 21:33 CK13 check by AC4: H4: confirmed (full) 2026-11-16 21:34 ST16 status by AC1: H4 -> supported (plan PL5), evidence RN5, CK13, AR1: Verdict renewed after H4 was corrected twice: H4.R3 (the median with equal weights is a flat point, not necessarily the… 2026-11-16 21:56 CK18 check by AC1: H4: confirmed (full), checked by AC2, recorded by agent AC1: "All looks fine now. NExt" 2026-11-16 22:55 PL8 plan by AC1: [confirmatory, requires approval] for H1, H2, H3, H13, H4, H5, H7, H8, H12, H14, H34: PL5 decided the hypotheses as worded before the restatement of DC5 (s without the factor 200, 'actual', merged results,… 2026-11-16 23:16 DV3 deviation by AC1: RN5 ran at commit 1e23bdc, not the plan's f865dfa. Between them only encode.py (new), targets.py and flatness.py changed (the --derived-only option of DC4); ch… 2026-11-17 09:21 PL9 plan by AC1: [confirmatory, requires approval] for H1, H2, H3, H13, H4, H5, H7, H8, H12, H14, H34: PL5 decided the hypotheses as worded before the restatement of DC5 (s without the factor 200, 'actual', merged results,… 2026-11-17 09:21 AP6 approval by AC1: PL9 approved by AC2 at 2026-11-17T09:20:47Z, recorded by agent AC1: "Record. And the plan with these corrections is approved." 2026-11-17 09:23 RN8 run by AC1: [confirmatory, success, under PL9, commit fa0ce62] python3 claimset-code/smape-target/check_formulas.py --seed 6; 39 results in runs/RN8/results.json (RS15749 to RS15787) 2026-11-17 09:24 ST25 status by AC1: H4 -> supported (plan PL8), evidence RN8, CK13, CK18, AR5: 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 co… H5 [supported] Let $\ln Y$ have a density $g$ symmetric about $m$, $g(m+v) = g(m-v)$, and unimodal: $g(m+u) \ge g(m+v)$ whenever $|u| \le |v|$, strictly on a set of positive probability. Then the median $e^m$ of $Y$ is the only [[TM6:minimiser of expected sMAPE]]. 2026-11-16 13:20 Q1 question by AC1: Which point forecast minimises the expected sMAPE of a forecast, for which distributions of outcomes (positive, or also zero or negative), and why: what in the… 2026-11-16 13:20 TM6 term by AC1: sMAPE target [concept] 2026-11-16 13:21 H5 hypothesis by AC1: If $\ln Y$ has a density that is symmetric about $m$ and strictly unimodal, the median $e^m$ of $Y$ is the only [[TM6:minimiser of expected sMAPE]]. [a_priori] 2026-11-16 14:49 PL5 plan by AC1: [confirmatory, requires approval] for H1, H2, H3, H4, H5, H6, H7, H8, H9, H10, H11, H12, H13, H14: Formal proofs (proofs.md), checked independently by AC4 (Claude Fable 5.2: a derivation from the statements alone, then… 2026-11-16 14:58 AP4 approval by AC1: PL5 approved by AC2 at 2026-11-16T14:58:18Z, recorded by agent AC1: "1. Yes 2. For this study 3. Approved 4. render proofs.md as pdf so that I can read LaTeX f… 2026-11-16 15:14 RN5 run by AC1: [confirmatory, success, under PL5, commit 1e23bdc] python3 claimset-code/smape-target/check_formulas.py --seed 6; 39 results in runs/RN5/results.json (RS14882 to RS14920) 2026-11-16 17:28 H26 hypothesis by AC1: Let $\ln Y$ have a density $g$ symmetric about $m$, $g(m+v) = g(m-v)$, and unimodal: $g(m+u) \ge g(m+v)$ whenever $|u| \le |v|$, strictly on a set of positive… [a_priori] 2026-11-16 17:43 CK6 check by AC4: H5: confirmed (full) 2026-11-16 17:47 ST10 status by AC1: H5 -> supported (plan PL5), evidence RN5, CK6, AR1: PL5's rule for supported is met: the proof (paper H6 in proofs.md, AR1) is complete; the independent checker AC4 confir… 2026-11-16 21:56 CK19 check by AC1: H5: confirmed (full), checked by AC2, recorded by agent AC1: "All looks fine now. NExt" 2026-11-16 22:55 PL8 plan by AC1: [confirmatory, requires approval] for H1, H2, H3, H13, H4, H5, H7, H8, H12, H14, H34: PL5 decided the hypotheses as worded before the restatement of DC5 (s without the factor 200, 'actual', merged results,… 2026-11-16 23:16 DV3 deviation by AC1: RN5 ran at commit 1e23bdc, not the plan's f865dfa. Between them only encode.py (new), targets.py and flatness.py changed (the --derived-only option of DC4); ch… 2026-11-17 09:21 PL9 plan by AC1: [confirmatory, requires approval] for H1, H2, H3, H13, H4, H5, H7, H8, H12, H14, H34: PL5 decided the hypotheses as worded before the restatement of DC5 (s without the factor 200, 'actual', merged results,… 2026-11-17 09:21 AP6 approval by AC1: PL9 approved by AC2 at 2026-11-17T09:20:47Z, recorded by agent AC1: "Record. And the plan with these corrections is approved." 2026-11-17 09:23 RN8 run by AC1: [confirmatory, success, under PL9, commit fa0ce62] python3 claimset-code/smape-target/check_formulas.py --seed 6; 39 results in runs/RN8/results.json (RS15749 to RS15787) 2026-11-17 09:24 ST26 status by AC1: H5 -> supported (plan PL8), evidence RN8, CK6, CK19, AR5: PL8's rule for supported is met for the current wording: the proof (paper H6 in proofs.md, AR1.R2) is complete; the ind… H6 [inconclusive] For positive outcomes with a continuous distribution function and a continuous, strictly increasing quantile function $Q$ (no atoms, no gaps), the quantile level $\tau$ of a [[TM6:minimiser of expected sMAPE]] satisfies the balance of [[TM7:the pulls over quantile levels]]: $\int_0^{\tau} w(Q(u), Q… 2026-11-16 13:20 Q1 question by AC1: Which point forecast minimises the expected sMAPE of a forecast, for which distributions of outcomes (positive, or also zero or negative), and why: what in the… 2026-11-16 13:20 TM6 term by AC1: sMAPE target [concept] 2026-11-16 13:20 TM7 term by AC1: quantile profile [concept] 2026-11-16 13:21 H6 hypothesis by AC1: For positive outcomes with a continuous quantile function $Q$, the quantile level $\tau$ of a [[TM6:minimiser of expected sMAPE]] satisfies [[TM7:the pulls ove… [a_priori] 2026-11-16 14:32 H15 hypothesis by AC1: For positive outcomes with a continuous quantile function $Q$, the quantile level $\tau$ of a [[TM6:minimiser of expected sMAPE]] satisfies the balance of [[TM… [a_priori] 2026-11-16 14:43 H16 hypothesis by AC1: For positive outcomes with a continuous distribution function and a continuous, strictly increasing quantile function $Q$ (no atoms, no gaps), the quantile lev… [a_priori] 2026-11-16 14:49 PL5 plan by AC1: [confirmatory, requires approval] for H1, H2, H3, H4, H5, H6, H7, H8, H9, H10, H11, H12, H13, H14: Formal proofs (proofs.md), checked independently by AC4 (Claude Fable 5.2: a derivation from the statements alone, then… 2026-11-16 14:58 AP4 approval by AC1: PL5 approved by AC2 at 2026-11-16T14:58:18Z, recorded by agent AC1: "1. Yes 2. For this study 3. Approved 4. render proofs.md as pdf so that I can read LaTeX f… 2026-11-16 15:14 RN5 run by AC1: [confirmatory, success, under PL5, commit 1e23bdc] python3 claimset-code/smape-target/check_formulas.py --seed 6; 39 results in runs/RN5/results.json (RS14882 to RS14920) 2026-11-16 17:29 ST1 status by AC1: H6 -> revised: Merged into H4 by the supervisor's decision DC5: its direction result (the target lies on the side of the median that c… 2026-11-16 22:54 ST17 status by AC1: H6 -> inconclusive (plan PL5), evidence RN5: PL5's deciding run RN5 ran with H6 among its hypotheses. H6 was merged into H4 by DC5 before its verdict: its direction… 2026-11-16 23:16 DV3 deviation by AC1: RN5 ran at commit 1e23bdc, not the plan's f865dfa. Between them only encode.py (new), targets.py and flatness.py changed (the --derived-only option of DC4); ch… H7 [supported] Let $\ln Y$ have two modes, at $m - a$ and $m + a$ ($a > 0$), with the actuals concentrated near them: a mixture of a distribution near $m - a$ and its mirror image about $m$, with weights $p$ and $1 - p$. (a) With equal weights the median $e^m$ lies in the gap between the two groups; when the grou… 2026-11-16 13:20 Q1 question by AC1: Which point forecast minimises the expected sMAPE of a forecast, for which distributions of outcomes (positive, or also zero or negative), and why: what in the… 2026-11-16 13:20 TM3 term by AC1: expected sMAPE [variable] 2026-11-16 13:22 H7 hypothesis by AC1: [[TM3:Expected sMAPE]] can have more than one local minimum. For an equal-weight mixture of two distributions of $\ln Y$ symmetric about $-a$ and $+a$ whose sp… [a_priori] 2026-11-16 14:47 H18 hypothesis by AC1: [[TM3:Expected sMAPE]] can have more than one local minimum. Let $\ln Y$ have a density $g$ that is a mixture, with weights $p$ and $1-p$, of a density concent… [a_priori] 2026-11-16 14:49 PL5 plan by AC1: [confirmatory, requires approval] for H1, H2, H3, H4, H5, H6, H7, H8, H9, H10, H11, H12, H13, H14: Formal proofs (proofs.md), checked independently by AC4 (Claude Fable 5.2: a derivation from the statements alone, then… 2026-11-16 14:58 AP4 approval by AC1: PL5 approved by AC2 at 2026-11-16T14:58:18Z, recorded by agent AC1: "1. Yes 2. For this study 3. Approved 4. render proofs.md as pdf so that I can read LaTeX f… 2026-11-16 15:00 H20 hypothesis by AC1: [[TM3:Expected sMAPE]] can have more than one local minimum. Let $\ln Y$ have a density $g$ that is a mixture, with weights $p$ and $1-p$, of a density concent… [a_priori] 2026-11-16 15:14 RN5 run by AC1: [confirmatory, success, under PL5, commit 1e23bdc] python3 claimset-code/smape-target/check_formulas.py --seed 6; 39 results in runs/RN5/results.json (RS14882 to RS14920) 2026-11-16 17:28 H27 hypothesis by AC1: Let $\ln Y$ have two modes, at $m - a$ and $m + a$ ($a > 0$), with the actuals concentrated near them: a mixture of a distribution near $m - a$ and its mirror… [a_priori] 2026-11-16 17:44 CK7 check by AC4: H7: confirmed (full) 2026-11-16 17:47 ST11 status by AC1: H7 -> supported (plan PL5), evidence RN5, CK7, AR1: PL5's rule for supported is met: the proof (paper H7 in proofs.md, AR1) is complete; the independent checker AC4 confir… 2026-11-16 21:56 CK20 check by AC1: H7: confirmed (full), checked by AC2, recorded by agent AC1: "All looks fine now. NExt" 2026-11-16 22:55 PL8 plan by AC1: [confirmatory, requires approval] for H1, H2, H3, H13, H4, H5, H7, H8, H12, H14, H34: PL5 decided the hypotheses as worded before the restatement of DC5 (s without the factor 200, 'actual', merged results,… 2026-11-16 23:16 DV3 deviation by AC1: RN5 ran at commit 1e23bdc, not the plan's f865dfa. Between them only encode.py (new), targets.py and flatness.py changed (the --derived-only option of DC4); ch… 2026-11-17 09:21 PL9 plan by AC1: [confirmatory, requires approval] for H1, H2, H3, H13, H4, H5, H7, H8, H12, H14, H34: PL5 decided the hypotheses as worded before the restatement of DC5 (s without the factor 200, 'actual', merged results,… 2026-11-17 09:21 AP6 approval by AC1: PL9 approved by AC2 at 2026-11-17T09:20:47Z, recorded by agent AC1: "Record. And the plan with these corrections is approved." 2026-11-17 09:23 RN8 run by AC1: [confirmatory, success, under PL9, commit fa0ce62] python3 claimset-code/smape-target/check_formulas.py --seed 6; 39 results in runs/RN8/results.json (RS15749 to RS15787) 2026-11-17 09:24 ST27 status by AC1: H7 -> supported (plan PL8), evidence RN8, CK7, CK20, AR5: PL8's rule for supported is met for the current wording: the proof (paper H7 in proofs.md, AR1.R2) is complete; the ind… H8 [supported] 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 [[TM1:sMAPE term]] its largest value, 1. Let $Y$ be 0 with probability $p_0$ and otherwise positive, and $m_+$ the lowest expected $s$ a positive forecast reach… 2026-11-16 13:20 Q1 question by AC1: Which point forecast minimises the expected sMAPE of a forecast, for which distributions of outcomes (positive, or also zero or negative), and why: what in the… 2026-11-16 13:20 TM1 term by AC1: sMAPE term [variable] 2026-11-16 13:20 TM6 term by AC1: sMAPE target [concept] 2026-11-16 13:22 H8 hypothesis by AC1: A forecast of the opposite sign to the outcome, or zero against a non-zero outcome, or non-zero against a zero outcome, scores 200, the maximum of the sMAPE te… [a_priori] 2026-11-16 14:49 PL5 plan by AC1: [confirmatory, requires approval] for H1, H2, H3, H4, H5, H6, H7, H8, H9, H10, H11, H12, H13, H14: Formal proofs (proofs.md), checked independently by AC4 (Claude Fable 5.2: a derivation from the statements alone, then… 2026-11-16 14:58 AP4 approval by AC1: PL5 approved by AC2 at 2026-11-16T14:58:18Z, recorded by agent AC1: "1. Yes 2. For this study 3. Approved 4. render proofs.md as pdf so that I can read LaTeX f… 2026-11-16 15:14 RN5 run by AC1: [confirmatory, success, under PL5, commit 1e23bdc] python3 claimset-code/smape-target/check_formulas.py --seed 6; 39 results in runs/RN5/results.json (RS14882 to RS14920) 2026-11-16 17:28 H28 hypothesis by AC1: 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 [[TM1:sMAPE term]] i… [a_priori] 2026-11-16 17:44 CK8 check by AC4: H8: confirmed (full) 2026-11-16 17:47 ST12 status by AC1: H8 -> supported (plan PL5), evidence RN5, CK8, AR1: PL5's rule for supported is met: the proof (paper H8 in proofs.md, AR1) is complete; the independent checker AC4 confir… 2026-11-16 21:56 CK21 check by AC1: H8: confirmed (full), checked by AC2, recorded by agent AC1: "All looks fine now. NExt" 2026-11-16 22:55 PL8 plan by AC1: [confirmatory, requires approval] for H1, H2, H3, H13, H4, H5, H7, H8, H12, H14, H34: PL5 decided the hypotheses as worded before the restatement of DC5 (s without the factor 200, 'actual', merged results,… 2026-11-16 23:16 DV3 deviation by AC1: RN5 ran at commit 1e23bdc, not the plan's f865dfa. Between them only encode.py (new), targets.py and flatness.py changed (the --derived-only option of DC4); ch… 2026-11-17 09:21 PL9 plan by AC1: [confirmatory, requires approval] for H1, H2, H3, H13, H4, H5, H7, H8, H12, H14, H34: PL5 decided the hypotheses as worded before the restatement of DC5 (s without the factor 200, 'actual', merged results,… 2026-11-17 09:21 AP6 approval by AC1: PL9 approved by AC2 at 2026-11-17T09:20:47Z, recorded by agent AC1: "Record. And the plan with these corrections is approved." 2026-11-17 09:23 RN8 run by AC1: [confirmatory, success, under PL9, commit fa0ce62] python3 claimset-code/smape-target/check_formulas.py --seed 6; 39 results in runs/RN8/results.json (RS15749 to RS15787) 2026-11-17 09:24 ST28 status by AC1: H8 -> supported (plan PL8), evidence RN8, CK8, CK21, AR5: PL8's rule for supported is met for the current wording: the proof (paper H8 in proofs.md, AR1.R2) is complete; the ind… H9 [inconclusive] For positive outcomes and forecasts, $s(y,f) = 100|z| - \tfrac{25}{3}|z|^3 + O(|z|^5)$, so for small log errors [[TM2:sMAPE]] is approximately 100 times the mean absolute error of $\ln f$ as a forecast of $\ln y$, whose minimiser is the median of $Y$; sMAPE differs from it in being bounded and conc… 2026-11-16 13:20 Q1 question by AC1: Which point forecast minimises the expected sMAPE of a forecast, for which distributions of outcomes (positive, or also zero or negative), and why: what in the… 2026-11-16 13:20 TM2 term by AC1: sMAPE [variable] 2026-11-16 13:22 H9 hypothesis by AC1: For positive outcomes and forecasts, $s(y,f) = 100|z| - \tfrac{25}{3}|z|^3 + O(|z|^5)$, so for small log errors [[TM2:sMAPE]] is approximately 100 times the me… [a_priori] 2026-11-16 14:49 PL5 plan by AC1: [confirmatory, requires approval] for H1, H2, H3, H4, H5, H6, H7, H8, H9, H10, H11, H12, H13, H14: Formal proofs (proofs.md), checked independently by AC4 (Claude Fable 5.2: a derivation from the statements alone, then… 2026-11-16 14:58 AP4 approval by AC1: PL5 approved by AC2 at 2026-11-16T14:58:18Z, recorded by agent AC1: "1. Yes 2. For this study 3. Approved 4. render proofs.md as pdf so that I can read LaTeX f… 2026-11-16 15:14 RN5 run by AC1: [confirmatory, success, under PL5, commit 1e23bdc] python3 claimset-code/smape-target/check_formulas.py --seed 6; 39 results in runs/RN5/results.json (RS14882 to RS14920) 2026-11-16 17:29 ST2 status by AC1: H9 -> revised: Merged into H3 by decision DC5: the small-error approximation is part (b) of H3. 2026-11-16 22:54 ST18 status by AC1: H9 -> inconclusive (plan PL5), evidence RN5: PL5's deciding run RN5 ran with H9 among its hypotheses. H9 was merged into H3 by DC5 before its verdict: its content i… 2026-11-16 23:16 DV3 deviation by AC1: RN5 ran at commit 1e23bdc, not the plan's f865dfa. Between them only encode.py (new), targets.py and flatness.py changed (the --derived-only option of DC4); ch… H10 [inconclusive] For a positive outcome $y$ and $0 < e < y$, the forecast $y + e$ scores lower than $y - e$: $200e/(2y+e) < 200e/(2y-e)$ as [[TM1:sMAPE terms]]; forecasts $ky$ and $y/k$ ($k > 1$), equally wrong as ratios, score the same (F2). 2026-11-16 13:20 Q1 question by AC1: Which point forecast minimises the expected sMAPE of a forecast, for which distributions of outcomes (positive, or also zero or negative), and why: what in the… 2026-11-16 13:20 TM1 term by AC1: sMAPE term [variable] 2026-11-16 13:22 H10 hypothesis by AC1: For a positive outcome $y$ and $0 < e < y$, the forecast $y + e$ scores lower than $y - e$: $200e/(2y+e) < 200e/(2y-e)$ as [[TM1:sMAPE terms]]; forecasts $ky$… [a_priori] 2026-11-16 14:49 PL5 plan by AC1: [confirmatory, requires approval] for H1, H2, H3, H4, H5, H6, H7, H8, H9, H10, H11, H12, H13, H14: Formal proofs (proofs.md), checked independently by AC4 (Claude Fable 5.2: a derivation from the statements alone, then… 2026-11-16 14:58 AP4 approval by AC1: PL5 approved by AC2 at 2026-11-16T14:58:18Z, recorded by agent AC1: "1. Yes 2. For this study 3. Approved 4. render proofs.md as pdf so that I can read LaTeX f… 2026-11-16 15:14 RN5 run by AC1: [confirmatory, success, under PL5, commit 1e23bdc] python3 claimset-code/smape-target/check_formulas.py --seed 6; 39 results in runs/RN5/results.json (RS14882 to RS14920) 2026-11-16 17:29 ST3 status by AC1: H10 -> revised: Merged into H2 by decision DC5: the comparison of high and low forecasts with equal absolute errors is part (b) of H2. 2026-11-16 22:54 ST19 status by AC1: H10 -> inconclusive (plan PL5), evidence RN5: PL5's deciding run RN5 ran with H10 among its hypotheses. H10 was merged into H2 by DC5 before its verdict: its content… 2026-11-16 23:16 DV3 deviation by AC1: RN5 ran at commit 1e23bdc, not the plan's f865dfa. Between them only encode.py (new), targets.py and flatness.py changed (the --derived-only option of DC4); ch… H12 [supported] For a positive actual $Y$: (a) [[TM10: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 $\mathrm E[(Y-f)^2 - (Y-g)^2] = (f-g)(f+g-2\,\mathrm E[Y])$ is finite and the mean is preferred to every other… 2026-11-16 13:20 Q1 question by AC1: Which point forecast minimises the expected sMAPE of a forecast, for which distributions of outcomes (positive, or also zero or negative), and why: what in the… 2026-11-16 13:20 TM10 term by AC1: RMSE [variable] 2026-11-16 13:20 TM11 term by AC1: MAE [variable] 2026-11-16 13:20 TM12 term by AC1: MAPE [variable] 2026-11-16 13:28 H12 hypothesis by AC1: For positive outcomes $Y$ and the other three measures: (a) [[TM10:RMSE]]: $\mathrm{E}(Y-f)^2$ is finite exactly when $\mathrm{E}Y^2 < \infty$; when $\mathrm{E… [a_priori] 2026-11-16 14:49 PL5 plan by AC1: [confirmatory, requires approval] for H1, H2, H3, H4, H5, H6, H7, H8, H9, H10, H11, H12, H13, H14: Formal proofs (proofs.md), checked independently by AC4 (Claude Fable 5.2: a derivation from the statements alone, then… 2026-11-16 14:58 AP4 approval by AC1: PL5 approved by AC2 at 2026-11-16T14:58:18Z, recorded by agent AC1: "1. Yes 2. For this study 3. Approved 4. render proofs.md as pdf so that I can read LaTeX f… 2026-11-16 15:14 RN5 run by AC1: [confirmatory, success, under PL5, commit 1e23bdc] python3 claimset-code/smape-target/check_formulas.py --seed 6; 39 results in runs/RN5/results.json (RS14882 to RS14920) 2026-11-16 17:29 H29 hypothesis by AC1: For a positive actual $Y$: (a) [[TM10: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… [a_priori] 2026-11-16 17:44 CK10 check by AC4: H12: confirmed (full) 2026-11-16 17:47 ST13 status by AC1: H12 -> supported (plan PL5), evidence RN5, CK10, AR1: PL5's rule for supported is met: the proof (paper H10 in proofs.md, AR1) is complete; the independent checker AC4 confi… 2026-11-16 21:56 CK23 check by AC1: H12: confirmed (full), checked by AC2, recorded by agent AC1: "All looks fine now. NExt" 2026-11-16 22:55 PL8 plan by AC1: [confirmatory, requires approval] for H1, H2, H3, H13, H4, H5, H7, H8, H12, H14, H34: PL5 decided the hypotheses as worded before the restatement of DC5 (s without the factor 200, 'actual', merged results,… 2026-11-16 23:16 DV3 deviation by AC1: RN5 ran at commit 1e23bdc, not the plan's f865dfa. Between them only encode.py (new), targets.py and flatness.py changed (the --derived-only option of DC4); ch… 2026-11-17 09:21 PL9 plan by AC1: [confirmatory, requires approval] for H1, H2, H3, H13, H4, H5, H7, H8, H12, H14, H34: PL5 decided the hypotheses as worded before the restatement of DC5 (s without the factor 200, 'actual', merged results,… 2026-11-17 09:21 AP6 approval by AC1: PL9 approved by AC2 at 2026-11-17T09:20:47Z, recorded by agent AC1: "Record. And the plan with these corrections is approved." 2026-11-17 09:23 RN8 run by AC1: [confirmatory, success, under PL9, commit fa0ce62] python3 claimset-code/smape-target/check_formulas.py --seed 6; 39 results in runs/RN8/results.json (RS15749 to RS15787) 2026-11-17 09:24 ST29 status by AC1: H12 -> supported (plan PL8), evidence RN8, CK10, CK23, AR5: PL8's rule for supported is met for the current wording: the proof (paper H10 in proofs.md, AR1.R2) is complete; the in… H13 [supported] For actuals and forecasts of any sign, $0 \le s \le 1$. So the [[TM3: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, H12). For a positive random actual $Y$, at least one f… 2026-11-16 13:20 Q1 question by AC1: Which point forecast minimises the expected sMAPE of a forecast, for which distributions of outcomes (positive, or also zero or negative), and why: what in the… 2026-11-16 13:20 TM3 term by AC1: expected sMAPE [variable] 2026-11-16 13:28 H13 hypothesis by AC1: The [[TM1:sMAPE term]] lies in $[0, 200]$ for outcomes and forecasts of any sign and is undefined only at $y = f = 0$, where it is set to 0, so expected sMAPE… [a_priori] 2026-11-16 14:49 PL5 plan by AC1: [confirmatory, requires approval] for H1, H2, H3, H4, H5, H6, H7, H8, H9, H10, H11, H12, H13, H14: Formal proofs (proofs.md), checked independently by AC4 (Claude Fable 5.2: a derivation from the statements alone, then… 2026-11-16 14:58 AP4 approval by AC1: PL5 approved by AC2 at 2026-11-16T14:58:18Z, recorded by agent AC1: "1. Yes 2. For this study 3. Approved 4. render proofs.md as pdf so that I can read LaTeX f… 2026-11-16 15:14 RN5 run by AC1: [confirmatory, success, under PL5, commit 1e23bdc] python3 claimset-code/smape-target/check_formulas.py --seed 6; 39 results in runs/RN5/results.json (RS14882 to RS14920) 2026-11-16 17:28 H24 hypothesis by AC1: For actuals and forecasts of any sign, $0 \le s \le 1$. So the [[TM3:expected sMAPE]] of every forecast is finite, between 0 and 200, whatever the distribution… [a_priori] 2026-11-16 17:43 CK4 check by AC4: H13: confirmed (full) 2026-11-16 17:46 ST8 status by AC1: H13 -> supported (plan PL5), evidence RN5, CK4, AR1: PL5's rule for supported is met: the proof (paper H4 in proofs.md, AR1) is complete; the independent checker AC4 confir… 2026-11-16 21:55 CK17 check by AC1: H13: confirmed (full), checked by AC2, recorded by agent AC1: "All looks fine now. NExt" 2026-11-16 22:55 PL8 plan by AC1: [confirmatory, requires approval] for H1, H2, H3, H13, H4, H5, H7, H8, H12, H14, H34: PL5 decided the hypotheses as worded before the restatement of DC5 (s without the factor 200, 'actual', merged results,… 2026-11-16 23:16 DV3 deviation by AC1: RN5 ran at commit 1e23bdc, not the plan's f865dfa. Between them only encode.py (new), targets.py and flatness.py changed (the --derived-only option of DC4); ch… 2026-11-17 09:21 PL9 plan by AC1: [confirmatory, requires approval] for H1, H2, H3, H13, H4, H5, H7, H8, H12, H14, H34: PL5 decided the hypotheses as worded before the restatement of DC5 (s without the factor 200, 'actual', merged results,… 2026-11-17 09:21 AP6 approval by AC1: PL9 approved by AC2 at 2026-11-17T09:20:47Z, recorded by agent AC1: "Record. And the plan with these corrections is approved." 2026-11-17 09:23 RN8 run by AC1: [confirmatory, success, under PL9, commit fa0ce62] python3 claimset-code/smape-target/check_formulas.py --seed 6; 39 results in runs/RN8/results.json (RS15749 to RS15787) 2026-11-17 09:24 ST24 status by AC1: H13 -> supported (plan PL8), evidence RN8, CK4, CK17, AR5: PL8's rule for supported is met for the current wording: the proof (paper H4 in proofs.md, AR1.R2) is complete; the ind… H14 [supported] Let $\ln Y = \mu + \sigma\varepsilon$, where $\varepsilon$ has a fixed density $g$ 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, $… 2026-11-16 13:20 Q1 question by AC1: Which point forecast minimises the expected sMAPE of a forecast, for which distributions of outcomes (positive, or also zero or negative), and why: what in the… 2026-11-16 13:20 TM3 term by AC1: expected sMAPE [variable] 2026-11-16 13:20 TM6 term by AC1: sMAPE target [concept] 2026-11-16 13:28 H14 hypothesis by AC1: Let $\ln Y = \mu + \sigma\varepsilon$, where $\varepsilon$ has a fixed bounded density $g$ with bounded continuous first and second derivatives, and write a fo… [a_priori] 2026-11-16 14:43 H17 hypothesis by AC1: Let $\ln Y = \mu + \sigma\varepsilon$, where $\varepsilon$ has a fixed bounded density $g$ with bounded continuous first and second derivatives, and write a fo… [a_priori] 2026-11-16 14:47 H19 hypothesis by AC1: Let $\ln Y = \mu + \sigma\varepsilon$, where $\varepsilon$ has a fixed bounded density $g$ with bounded continuous first and second derivatives and distributio… [a_priori] 2026-11-16 14:49 PL5 plan by AC1: [confirmatory, requires approval] for H1, H2, H3, H4, H5, H6, H7, H8, H9, H10, H11, H12, H13, H14: Formal proofs (proofs.md), checked independently by AC4 (Claude Fable 5.2: a derivation from the statements alone, then… 2026-11-16 14:58 AP4 approval by AC1: PL5 approved by AC2 at 2026-11-16T14:58:18Z, recorded by agent AC1: "1. Yes 2. For this study 3. Approved 4. render proofs.md as pdf so that I can read LaTeX f… 2026-11-16 15:14 RN5 run by AC1: [confirmatory, success, under PL5, commit 1e23bdc] python3 claimset-code/smape-target/check_formulas.py --seed 6; 39 results in runs/RN5/results.json (RS14882 to RS14920) 2026-11-16 17:29 H30 hypothesis by AC1: Let $\ln Y = \mu + \sigma\varepsilon$, where $\varepsilon$ has a fixed density $g$ with bounded continuous first and second derivatives, and write a forecast a… [a_priori] 2026-11-16 17:44 CK11 check by AC4: H14: confirmed (full) 2026-11-16 17:47 ST14 status by AC1: H14 -> supported (plan PL5), evidence RN5, CK11, AR1: PL5's rule for supported is met: the proof (paper H11 in proofs.md, AR1) is complete; the independent checker AC4 confi… 2026-11-16 21:56 CK24 check by AC1: H14: confirmed (full), checked by AC2, recorded by agent AC1: "All looks fine now. NExt" 2026-11-16 22:55 PL8 plan by AC1: [confirmatory, requires approval] for H1, H2, H3, H13, H4, H5, H7, H8, H12, H14, H34: PL5 decided the hypotheses as worded before the restatement of DC5 (s without the factor 200, 'actual', merged results,… 2026-11-16 23:16 DV3 deviation by AC1: RN5 ran at commit 1e23bdc, not the plan's f865dfa. Between them only encode.py (new), targets.py and flatness.py changed (the --derived-only option of DC4); ch… 2026-11-17 09:21 PL9 plan by AC1: [confirmatory, requires approval] for H1, H2, H3, H13, H4, H5, H7, H8, H12, H14, H34: PL5 decided the hypotheses as worded before the restatement of DC5 (s without the factor 200, 'actual', merged results,… 2026-11-17 09:21 AP6 approval by AC1: PL9 approved by AC2 at 2026-11-17T09:20:47Z, recorded by agent AC1: "Record. And the plan with these corrections is approved." 2026-11-17 09:23 RN8 run by AC1: [confirmatory, success, under PL9, commit fa0ce62] python3 claimset-code/smape-target/check_formulas.py --seed 6; 39 results in runs/RN8/results.json (RS15749 to RS15787) 2026-11-17 09:24 ST30 status by AC1: H14 -> supported (plan PL8), evidence RN8, CK11, CK24, AR5: PL8's rule for supported is met for the current wording: the proof (paper H11 in proofs.md, AR1.R2) is complete; the in… H31 [supported] 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 $\mu$ appro… 2026-11-16 13:20 Q1 question by AC1: Which point forecast minimises the expected sMAPE of a forecast, for which distributions of outcomes (positive, or also zero or negative), and why: what in the… 2026-11-16 13:20 TM6 term by AC1: sMAPE target [concept] 2026-11-16 14:36 RN1 run by AC1: [confirmatory, success, under PL1, commit 3cb6c7c] python3 claimset-code/smape-target/targets.py --seed 3 --reps 1000; 11,503 results in runs/RN1/results.json (RS1 to RS13195) 2026-11-16 17:29 H31 hypothesis by AC1: Let the actuals be $\mu + \varepsilon$, where $\varepsilon$ is symmetric about 0 and its density decreases away from 0 (for example, normal). If $\mu > 0$, eve… [from_exploration] 2026-11-16 17:35 PL6 plan by AC1: [confirmatory, requires approval] for H31: Proof (proofs.md, H9 in the paper), checked independently by AC4 and by the supervisor, and one deciding run of python3… 2026-11-16 17:35 PL7 plan by AC1: [confirmatory, requires approval] for H31: Proof (proofs.md, H9 in the paper), checked independently by AC4 and by the supervisor, and one deciding run of python3… 2026-11-16 17:38 AP5 approval by AC1: PL6 approved by AC2 at 2026-11-16T17:37:45Z, recorded by agent AC1: "1. Approved. 2. Done 3. Confirmed" 2026-11-16 17:40 RN6 run by AC1: [confirmatory, success, under PL7, commit 2da728a] python3 claimset-code/smape-target/check_jump.py; 56 results in runs/RN6/results.json (RS14921 to RS14976) 2026-11-16 17:44 CK9 check by AC4: H31: confirmed (full) 2026-11-16 17:47 ST15 status by AC1: H31 -> supported (plan PL7), evidence RN6, CK9, AR1: PL6's rule for supported is met: the proof (paper H9 in proofs.md, AR1) is complete; AC4 confirmed it (CK9, full); and… 2026-11-16 21:56 CK22 check by AC1: H31: confirmed (full), checked by AC2, recorded by agent AC1: "All looks fine now. NExt" 2026-11-16 23:16 DV2 deviation by AC1: 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 mea… H34 [supported] The forecast with the lowest expected sMAPE is the median $m$ of the actuals in each of two cases. (a) $\ln Y$ has a density symmetric about $\ln m$ and unimodal. (b) More generally, $Y$ is positive, takes no single value with positive probability, and its expected sMAPE has a single local minimum;… 2026-11-16 13:20 Q1 question by AC1: Which point forecast minimises the expected sMAPE of a forecast, for which distributions of outcomes (positive, or also zero or negative), and why: what in the… 2026-11-16 13:20 TM6 term by AC1: sMAPE target [concept] 2026-11-16 22:55 H34 hypothesis by AC1: The forecast with the lowest expected sMAPE is the median $m$ of the actuals in each of two cases. (a) $\ln Y$ has a density symmetric about $\ln m$ and unimod… [a_priori] 2026-11-16 22:55 PL8 plan by AC1: [confirmatory, requires approval] for H1, H2, H3, H13, H4, H5, H7, H8, H12, H14, H34: PL5 decided the hypotheses as worded before the restatement of DC5 (s without the factor 200, 'actual', merged results,… 2026-11-16 22:56 CK25 check by AC4: H34: confirmed (full) 2026-11-17 09:21 PL9 plan by AC1: [confirmatory, requires approval] for H1, H2, H3, H13, H4, H5, H7, H8, H12, H14, H34: PL5 decided the hypotheses as worded before the restatement of DC5 (s without the factor 200, 'actual', merged results,… 2026-11-17 09:21 AP6 approval by AC1: PL9 approved by AC2 at 2026-11-17T09:20:47Z, recorded by agent AC1: "Record. And the plan with these corrections is approved." 2026-11-17 09:23 RN8 run by AC1: [confirmatory, success, under PL9, commit fa0ce62] python3 claimset-code/smape-target/check_formulas.py --seed 6; 39 results in runs/RN8/results.json (RS15749 to RS15787) 2026-11-17 09:24 ST31 status by AC1: H34 -> supported (plan PL8), evidence RN8, CK25, CK18, CK19, AR5: 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 proo… H35 [supported] The [[TM32:msMAPE term]] has the properties of the [[TM1:sMAPE term]] except near zero. (a) Where $|y|+|f| \ge 0.5$, $m(y,f) = s(y,f)\,\frac{|y|+|f|}{|y|+|f|+\epsilon}$: the 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 comp… 2026-11-16 13:20 Q1 question by AC1: Which point forecast minimises the expected sMAPE of a forecast, for which distributions of outcomes (positive, or also zero or negative), and why: what in the… 2026-11-16 13:20 TM1 term by AC1: sMAPE term [variable] 2026-11-17 11:22 TM32 term by AC1: msMAPE term [variable] 2026-11-17 11:22 H35 hypothesis by AC1: The [[TM32:msMAPE term]] has the properties of the [[TM1:sMAPE term]] except near zero. (a) Where $|y|+|f| \ge 0.5$, $m(y,f) = s(y,f)\,\frac{|y|+|f|}{|y|+|f|+\… [a_priori] 2026-11-17 11:22 PL10 plan by AC1: [confirmatory, requires approval] for H35: Proof of H35 from the definition of the msMAPE term (a short section added to the proof document), checked independentl… 2026-11-17 11:22 AP7 approval by AC1: PL10 approved by AC2 at 2026-11-17T11:19:45Z, recorded by agent AC1: "1. Estimation. 2. Distributions are fine, we don't need all of them. Approved. Go" 2026-11-17 11:22 RN9 run by AC1: [confirmatory, success, under PL10, commit 838c642] python3 claimset-code/smape-target/check_msmape.py --seed 7; 8 results in runs/RN9/results.json (RS15788 to RS15795) 2026-11-17 11:23 CK26 check by AC4: H35: confirmed (full) 2026-11-17 11:26 ST32 status by AC1: H35 -> supported (plan PL10), evidence RN9, CK26, AR8: PL10's rule for supported is met: the proof of H35 (proof-msmape.md, AR8) is complete, every step naming what it rests… Q2 Where sMAPE and MAE both target the median of the outcomes, which measure's minimiser over a test set of n outcomes varies less from one test set to another, a… H11 [inconclusive] For independent outcomes from a continuous positive distribution whose expected sMAPE has a single minimiser $\theta$ with $S''(\theta) > 0$, the [[TM8:sMAPE minimiser of a test set]] of $n$ outcomes is consistent for $\theta$ and asymptotically normal with variance $\mathrm{E}[\psi(Y,\theta)^2]/(n… 2026-11-16 13:20 Q2 question by AC1: Where sMAPE and MAE both target the median of the outcomes, which measure's minimiser over a test set of n outcomes varies less from one test set to another, a… 2026-11-16 13:20 TM8 term by AC1: sMAPE minimiser of a test set [variable] 2026-11-16 13:20 TM9 term by AC1: relative efficiency [variable] 2026-11-16 13:22 H11 hypothesis by AC1: For independent outcomes from a continuous positive distribution whose expected sMAPE has a single minimiser $\theta$ with $S''(\theta) > 0$, the [[TM8:sMAPE m… [a_priori] 2026-11-16 14:49 PL5 plan by AC1: [confirmatory, requires approval] for H1, H2, H3, H4, H5, H6, H7, H8, H9, H10, H11, H12, H13, H14: Formal proofs (proofs.md), checked independently by AC4 (Claude Fable 5.2: a derivation from the statements alone, then… 2026-11-16 14:58 AP4 approval by AC1: PL5 approved by AC2 at 2026-11-16T14:58:18Z, recorded by agent AC1: "1. Yes 2. For this study 3. Approved 4. render proofs.md as pdf so that I can read LaTeX f… 2026-11-16 15:14 RN5 run by AC1: [confirmatory, success, under PL5, commit 1e23bdc] python3 claimset-code/smape-target/check_formulas.py --seed 6; 39 results in runs/RN5/results.json (RS14882 to RS14920) 2026-11-16 17:29 ST4 status by AC1: H11 -> abandoned: Dropped by decision DC5: it applied a standard large-sample theorem; the efficiency of the sMAPE minimiser against the… 2026-11-16 22:54 ST20 status by AC1: H11 -> inconclusive (plan PL5), evidence RN5: PL5's deciding run RN5 ran with H11 among its hypotheses. H11 was dropped by DC5 before its verdict (abandoned, ST4): t… 2026-11-16 23:16 DV3 deviation by AC1: RN5 ran at commit 1e23bdc, not the plan's f865dfa. Between them only encode.py (new), targets.py and flatness.py changed (the --derived-only option of DC4); ch… TIMELINE (every event but results, in log order: seq, time, ID, type, writer) 1 2026-11-16 12:57 INIT init by -: began "What does sMAPE reward? Derivations of the forecast that minimises expected sMAPE, checked by simulation" under lab record v0.22 2 2026-11-16 12:57 AC1 actor by AC1: registers Claude Code (agent) 3 2026-11-16 13:20 AC2 actor by AC1: registers Ivan Svetunkov (human, ORCID 0000-0001-7826-0281) 4 2026-11-16 13:20 DC1 decision by AC1: The supervisor (AC2), 2026-11-16: the study asks which forecast minimises expected sMAPE and why (Q1), and, as an extension, whether sMAPE's minimiser varies l… 5 2026-11-16 13:20 Q1 question by AC1: Which point forecast minimises the expected sMAPE of a forecast, for which distributions of outcomes (positive, or also zero or negative), and why: what in the… 6 2026-11-16 13:20 Q2 question by AC1: Where sMAPE and MAE both target the median of the outcomes, which measure's minimiser over a test set of n outcomes varies less from one test set to another, a… 7 2026-11-16 13:20 TM1 term by AC1: sMAPE term [variable] 8 2026-11-16 13:20 TM2 term by AC1: sMAPE [variable] 9 2026-11-16 13:20 TM3 term by AC1: expected sMAPE [variable] 10 2026-11-16 13:20 TM4 term by AC1: log error [variable] 11 2026-11-16 13:20 TM5 term by AC1: pull [variable] 12 2026-11-16 13:20 TM6 term by AC1: sMAPE target [concept] 13 2026-11-16 13:20 TM7 term by AC1: quantile profile [concept] 14 2026-11-16 13:20 TM8 term by AC1: sMAPE minimiser of a test set [variable] 15 2026-11-16 13:20 TM9 term by AC1: relative efficiency [variable] 16 2026-11-16 13:20 TM10 term by AC1: RMSE [variable] 17 2026-11-16 13:20 TM11 term by AC1: MAE [variable] 18 2026-11-16 13:20 TM12 term by AC1: MAPE [variable] 19 2026-11-16 13:21 AC3 actor by AC1: registers Ivan Svetunkov (human, ORCID 0000-0001-7826-0281) 20 2026-11-16 13:21 H1 hypothesis by AC1: For a positive outcome $y$ and a positive forecast $f$, the [[TM1:sMAPE term]] depends on $y$ and $f$ only through their ratio $r = y/f$: $s(y,f) = 200\,|r-1|/… [a_priori] 21 2026-11-16 13:21 H2 hypothesis by AC1: For a positive outcome and a positive forecast, the [[TM1:sMAPE term]] is unchanged when the ratio $r = y/f$ is replaced by $1/r$: a forecast too high by a fac… [a_priori] 22 2026-11-16 13:21 H3 hypothesis by AC1: For a positive outcome and a positive forecast, the [[TM1:sMAPE term]] equals $200\tanh(|z|/2)$, where $z$ is the [[TM4:log error]]: it is 0 at $z = 0$, strict… [a_priori] 23 2026-11-16 13:21 H4 hypothesis by AC1: For positive outcomes with a continuous distribution and a positive forecast $f$, $S'(f) = \frac{200}{f}\,\mathrm{E}[w(Y,f)(\mathbf 1\{Yf\}… [a_priori] 24 2026-11-16 13:21 H5 hypothesis by AC1: If $\ln Y$ has a density that is symmetric about $m$ and strictly unimodal, the median $e^m$ of $Y$ is the only [[TM6:minimiser of expected sMAPE]]. [a_priori] 25 2026-11-16 13:21 H6 hypothesis by AC1: For positive outcomes with a continuous quantile function $Q$, the quantile level $\tau$ of a [[TM6:minimiser of expected sMAPE]] satisfies [[TM7:the pulls ove… [a_priori] 26 2026-11-16 13:22 H7 hypothesis by AC1: [[TM3:Expected sMAPE]] can have more than one local minimum. For an equal-weight mixture of two distributions of $\ln Y$ symmetric about $-a$ and $+a$ whose sp… [a_priori] 27 2026-11-16 13:22 H8 hypothesis by AC1: A forecast of the opposite sign to the outcome, or zero against a non-zero outcome, or non-zero against a zero outcome, scores 200, the maximum of the sMAPE te… [a_priori] 28 2026-11-16 13:22 H9 hypothesis by AC1: For positive outcomes and forecasts, $s(y,f) = 100|z| - \tfrac{25}{3}|z|^3 + O(|z|^5)$, so for small log errors [[TM2:sMAPE]] is approximately 100 times the me… [a_priori] 29 2026-11-16 13:22 H10 hypothesis by AC1: For a positive outcome $y$ and $0 < e < y$, the forecast $y + e$ scores lower than $y - e$: $200e/(2y+e) < 200e/(2y-e)$ as [[TM1:sMAPE terms]]; forecasts $ky$… [a_priori] 30 2026-11-16 13:22 H11 hypothesis by AC1: For independent outcomes from a continuous positive distribution whose expected sMAPE has a single minimiser $\theta$ with $S''(\theta) > 0$, the [[TM8:sMAPE m… [a_priori] 31 2026-11-16 13:22 NT1 note by AC1: Origin of F1 to F11: they were derived during the study measure-rankings (EntryLib submission pending; record on branch claimset/measure-rankings), after its r… 32 2026-11-16 13:28 DC2 decision by AC1: The supervisor (AC2), 2026-11-16: two further formal hypotheses, on where the measures fail: one for RMSE, MAE and MAPE (infinite moments and what their target… 33 2026-11-16 13:28 H12 hypothesis by AC1: For positive outcomes $Y$ and the other three measures: (a) [[TM10:RMSE]]: $\mathrm{E}(Y-f)^2$ is finite exactly when $\mathrm{E}Y^2 < \infty$; when $\mathrm{E… [a_priori] 34 2026-11-16 13:28 H13 hypothesis by AC1: The [[TM1:sMAPE term]] lies in $[0, 200]$ for outcomes and forecasts of any sign and is undefined only at $y = f = 0$, where it is set to 0, so expected sMAPE… [a_priori] 35 2026-11-16 13:28 H14 hypothesis by AC1: Let $\ln Y = \mu + \sigma\varepsilon$, where $\varepsilon$ has a fixed bounded density $g$ with bounded continuous first and second derivatives, and write a fo… [a_priori] 36 2026-11-16 13:44 TM13 term by AC1: candidate forecast [concept] 37 2026-11-16 13:44 TM14 term by AC1: preferred candidate [concept] 38 2026-11-16 13:44 TM15 term by AC1: derived value [concept] 39 2026-11-16 13:44 TM16 term by AC1: selection design [method] 40 2026-11-16 13:44 TM17 term by AC1: efficiency design [method] 41 2026-11-16 13:44 TM18 term by AC1: flatness design [method] 42 2026-11-16 13:44 TM19 term by AC1: log-MAE [variable] 43 2026-11-16 13:44 TM20 term by AC1: ranking agreement [variable] 44 2026-11-16 13:44 TM21 term by AC1: inverse Gaussian outcomes [population] 45 2026-11-16 13:44 TM22 term by AC1: log-skewed outcomes [population] 46 2026-11-16 13:44 TM23 term by AC1: bimodal outcomes [population] 47 2026-11-16 13:44 TM24 term by AC1: outcomes straddling zero [population] 48 2026-11-16 13:44 TM25 term by AC1: zero-inflated outcomes [population] 49 2026-11-16 13:44 TM26 term by AC1: log-symmetric outcomes [population] 50 2026-11-16 13:44 TM27 term by AC1: widening outcomes [population] 51 2026-11-16 13:46 PL1 plan by AC1: [estimation, requires approval] for Q1: Selection design (TM16) on 33 outcome distributions: inverse Gaussian (6), log-skewed (12) and lognormal (2), bimodal (… 52 2026-11-16 13:46 PL2 plan by AC1: [estimation, requires approval] for Q2: Efficiency design (TM17) on 16 log-symmetric distributions (TM26): lognormal sigma and log-Laplace b, each 0.25, 0.5, 1… 53 2026-11-16 13:46 PL3 plan by AC1: [estimation, requires approval] for Q1: Flatness design (TM18) on 30 distributions (TM27): lognormal, log-Gumbel right and log-Gumbel left, with sigma 0.1, 0.2… 54 2026-11-16 14:20 DC3 decision by AC1: The supervisor (AC2), 2026-11-16: the flatness plan compares only the conventional measures RMSE, MAE, MAPE and sMAPE; log-MAE is dropped. Each measure's sensi… 55 2026-11-16 14:20 PL4 plan by AC1: [estimation, requires approval] for Q1: Flatness design (TM18) on 30 distributions (TM27): lognormal, log-Gumbel right and log-Gumbel left, with sigma 0.1, 0.2… 56 2026-11-16 14:28 AP1 approval by AC1: PL1 approved by AC2 at 2026-11-16T14:28:16Z, recorded by agent AC1: "Approved. Go" 57 2026-11-16 14:28 AP2 approval by AC1: PL2 approved by AC2 at 2026-11-16T14:28:16Z, recorded by agent AC1: "Approved. Go" 58 2026-11-16 14:28 AP3 approval by AC1: PL4 approved by AC2 at 2026-11-16T14:28:16Z, recorded by agent AC1: "Approved. Go" 59 2026-11-16 14:32 H15 hypothesis by AC1: For positive outcomes with a continuous quantile function $Q$, the quantile level $\tau$ of a [[TM6:minimiser of expected sMAPE]] satisfies the balance of [[TM… [a_priori] 60 2026-11-16 14:36 RN1 run by AC1: [confirmatory, success, under PL1, commit 3cb6c7c] python3 claimset-code/smape-target/targets.py --seed 3 --reps 1000; 11,503 results in runs/RN1/results.json (RS1 to RS13195) 6652 2026-11-16 14:43 H16 hypothesis by AC1: For positive outcomes with a continuous distribution function and a continuous, strictly increasing quantile function $Q$ (no atoms, no gaps), the quantile lev… [a_priori] 6714 2026-11-16 14:43 H17 hypothesis by AC1: Let $\ln Y = \mu + \sigma\varepsilon$, where $\varepsilon$ has a fixed bounded density $g$ with bounded continuous first and second derivatives, and write a fo… [a_priori] 6751 2026-11-16 14:43 AC4 actor by AC1: registers Claude Code (agent) 7268 2026-11-16 14:44 RN2 run by AC1: [confirmatory, success, under PL4, commit 3cb6c7c] python3 claimset-code/smape-target/flatness.py --seed 5 --reps 10000; 1,692 results in runs/RN2/results.json (RS7205 to RS10003) 9548 2026-11-16 14:47 H18 hypothesis by AC1: [[TM3:Expected sMAPE]] can have more than one local minimum. Let $\ln Y$ have a density $g$ that is a mixture, with weights $p$ and $1-p$, of a density concent… [a_priori] 9604 2026-11-16 14:47 H19 hypothesis by AC1: Let $\ln Y = \mu + \sigma\varepsilon$, where $\varepsilon$ has a fixed bounded density $g$ with bounded continuous first and second derivatives and distributio… [a_priori] 10378 2026-11-16 14:49 PL5 plan by AC1: [confirmatory, requires approval] for H1, H2, H3, H4, H5, H6, H7, H8, H9, H10, H11, H12, H13, H14: Formal proofs (proofs.md), checked independently by AC4 (Claude Fable 5.2: a derivation from the statements alone, then… 13263 2026-11-16 14:58 AP4 approval by AC1: PL5 approved by AC2 at 2026-11-16T14:58:18Z, recorded by agent AC1: "1. Yes 2. For this study 3. Approved 4. render proofs.md as pdf so that I can read LaTeX f… 13264 2026-11-16 14:58 DC4 decision by AC1: The supervisor (AC2), 2026-11-16: (1) the independent checker AC4 records its own checks of H1 to H14 in the record; (2) the non-numeric derived values that th… 13265 2026-11-16 15:00 H20 hypothesis by AC1: [[TM3:Expected sMAPE]] can have more than one local minimum. Let $\ln Y$ have a density $g$ that is a mixture, with weights $p$ and $1-p$, of a density concent… [a_priori] 13266 2026-11-16 15:01 NT2 note by AC1: The first attempt of the deciding run of PL2 (started 2026-11-16 14:29 UTC, the same command and seed as the plan) was killed at about 14:57 UTC when the conta… 13267 2026-11-16 15:01 NT3 note by AC1: The first attempt of the deciding run of PL2 (started 2026-11-16 14:29 UTC, the same command and seed as the plan) was killed at about 14:57 UTC when the conta… 13268 2026-11-16 15:05 RN3 run by AC1: [exploratory, success, under PL1, commit 1e23bdc] python3 claimset-code/smape-target/targets.py --derived-only; 906 results in runs/RN3/results.json (RS13196 to RS14101) 14175 2026-11-16 15:09 RN4 run by AC1: [exploratory, success, under PL4, commit 1e23bdc] python3 claimset-code/smape-target/flatness.py --derived-only; 780 results in runs/RN4/results.json (RS14102 to RS14881) 14956 2026-11-16 15:14 RN5 run by AC1: [confirmatory, success, under PL5, commit 1e23bdc] python3 claimset-code/smape-target/check_formulas.py --seed 6; 39 results in runs/RN5/results.json (RS14882 to RS14920) 14996 2026-11-16 17:27 DC5 decision by AC1: The supervisor (AC2), 2026-11-16, after reading the proofs: the results are restated and reduced. The contribution s of one actual is defined without the facto… 14997 2026-11-16 17:27 TM28 term by AC1: sMAPE term [variable] 14998 2026-11-16 17:28 TM29 term by AC1: sMAPE [variable] 14999 2026-11-16 17:28 TM30 term by AC1: expected sMAPE [variable] 15000 2026-11-16 17:28 TM31 term by AC1: weight [variable] 15001 2026-11-16 17:28 H21 hypothesis by AC1: For $y > 0$ and $f > 0$, the [[TM1:sMAPE term]] $s(y,f) = |y-f|/(|y|+|f|)$ depends on $y$ and $f$ only through their ratio $r = y/f$: $s = |r - 1|/(r + 1)$. [a_priori] 15002 2026-11-16 17:28 H22 hypothesis by AC1: For $y > 0$ and $f > 0$: (a) replacing $r = y/f$ by $1/r$ leaves the [[TM1:sMAPE term]] unchanged, so a forecast $k$ times too high ($f = ky$) and one $k$ time… [a_priori] 15003 2026-11-16 17:28 H23 hypothesis by AC1: For $y > 0$ and $f > 0$, the [[TM1:sMAPE term]] is $s = \tanh(|\ln y - \ln f|/2)$ ([[TM4:log error]] $\ln y - \ln f$). (a) It is 0 when $y = f$, strictly incre… [a_priori] 15004 2026-11-16 17:28 H24 hypothesis by AC1: For actuals and forecasts of any sign, $0 \le s \le 1$. So the [[TM3:expected sMAPE]] of every forecast is finite, between 0 and 200, whatever the distribution… [a_priori] 15005 2026-11-16 17:28 H25 hypothesis by AC1: Let $Y$ 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 ab… [a_priori] 15006 2026-11-16 17:28 H26 hypothesis by AC1: Let $\ln Y$ have a density $g$ symmetric about $m$, $g(m+v) = g(m-v)$, and unimodal: $g(m+u) \ge g(m+v)$ whenever $|u| \le |v|$, strictly on a set of positive… [a_priori] 15007 2026-11-16 17:28 H27 hypothesis by AC1: Let $\ln Y$ have two modes, at $m - a$ and $m + a$ ($a > 0$), with the actuals concentrated near them: a mixture of a distribution near $m - a$ and its mirror… [a_priori] 15008 2026-11-16 17:28 H28 hypothesis by AC1: 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 [[TM1:sMAPE term]] i… [a_priori] 15009 2026-11-16 17:29 H29 hypothesis by AC1: For a positive actual $Y$: (a) [[TM10: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… [a_priori] 15010 2026-11-16 17:29 H30 hypothesis by AC1: Let $\ln Y = \mu + \sigma\varepsilon$, where $\varepsilon$ has a fixed density $g$ with bounded continuous first and second derivatives, and write a forecast a… [a_priori] 15011 2026-11-16 17:29 H31 hypothesis by AC1: Let the actuals be $\mu + \varepsilon$, where $\varepsilon$ is symmetric about 0 and its density decreases away from 0 (for example, normal). If $\mu > 0$, eve… [from_exploration] 15012 2026-11-16 17:29 ST1 status by AC1: H6 -> revised: Merged into H4 by the supervisor's decision DC5: its direction result (the target lies on the side of the median that c… 15013 2026-11-16 17:29 ST2 status by AC1: H9 -> revised: Merged into H3 by decision DC5: the small-error approximation is part (b) of H3. 15014 2026-11-16 17:29 ST3 status by AC1: H10 -> revised: Merged into H2 by decision DC5: the comparison of high and low forecasts with equal absolute errors is part (b) of H2. 15015 2026-11-16 17:29 ST4 status by AC1: H11 -> abandoned: Dropped by decision DC5: it applied a standard large-sample theorem; the efficiency of the sMAPE minimiser against the… 15016 2026-11-16 17:35 PL6 plan by AC1: [confirmatory, requires approval] for H31: Proof (proofs.md, H9 in the paper), checked independently by AC4 and by the supervisor, and one deciding run of python3… 15017 2026-11-16 17:35 PL7 plan by AC1: [confirmatory, requires approval] for H31: Proof (proofs.md, H9 in the paper), checked independently by AC4 and by the supervisor, and one deciding run of python3… 15018 2026-11-16 17:35 NT4 note by AC1: Before PL6 was written, its script check_jump.py was run once outside the record, as a trial of the code, on the same grid of means; it is deterministic (quadr… 15019 2026-11-16 17:38 AP5 approval by AC1: PL6 approved by AC2 at 2026-11-16T17:37:45Z, recorded by agent AC1: "1. Approved. 2. Done 3. Confirmed" 15020 2026-11-16 17:38 DC6 decision by AC1: The supervisor (AC2), 2026-11-16, has read the proofs (proofs.md, the version of the restructured results H1 to H11 of the paper: record H1, H2, H3, H13, H4, H… 15021 2026-11-16 17:40 RN6 run by AC1: [confirmatory, success, under PL7, commit 2da728a] python3 claimset-code/smape-target/check_jump.py; 56 results in runs/RN6/results.json (RS14921 to RS14976) 15078 2026-11-16 17:43 CK1 check by AC4: H1: confirmed (full) 15079 2026-11-16 17:43 CK2 check by AC4: H2: confirmed (full) 15080 2026-11-16 17:43 CK3 check by AC4: H3: confirmed (full) 15081 2026-11-16 17:43 CK4 check by AC4: H13: confirmed (full) 15082 2026-11-16 17:43 CK5 check by AC4: H4: disputed (full) 15083 2026-11-16 17:43 CK6 check by AC4: H5: confirmed (full) 15084 2026-11-16 17:44 CK7 check by AC4: H7: confirmed (full) 15085 2026-11-16 17:44 CK8 check by AC4: H8: confirmed (full) 15086 2026-11-16 17:44 CK9 check by AC4: H31: confirmed (full) 15087 2026-11-16 17:44 CK10 check by AC4: H12: confirmed (full) 15088 2026-11-16 17:44 CK11 check by AC4: H14: confirmed (full) 15089 2026-11-16 17:44 H32 hypothesis by AC1: Let $Y$ 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 ab… [a_priori] 15090 2026-11-16 17:45 CK12 check by AC4: H4: confirmed (full) 15091 2026-11-16 17:46 AR1 artefact by AC1: files/proofs.md 15092 2026-11-16 17:46 AR2 artefact by AC1: files/proofs.pdf 15093 2026-11-16 17:46 AR3 artefact by AC1: files/smape_weights.png 15094 2026-11-16 17:46 AR4 artefact by AC1: files/proof-check-report.md 15095 2026-11-16 17:46 ST5 status by AC1: H1 -> supported (plan PL5), evidence RN5, CK1, AR1: PL5's rule for supported is met: the proof (paper H1 in proofs.md, AR1) is complete; the independent checker AC4 confir… 15096 2026-11-16 17:46 ST6 status by AC1: H2 -> supported (plan PL5), evidence RN5, CK2, AR1: PL5's rule for supported is met: the proof (paper H2 in proofs.md, AR1) is complete; the independent checker AC4 confir… 15097 2026-11-16 17:46 ST7 status by AC1: H3 -> supported (plan PL5), evidence RN5, CK3, AR1: PL5's rule for supported is met: the proof (paper H3 in proofs.md, AR1) is complete; the independent checker AC4 confir… 15098 2026-11-16 17:46 ST8 status by AC1: H13 -> supported (plan PL5), evidence RN5, CK4, AR1: PL5's rule for supported is met: the proof (paper H4 in proofs.md, AR1) is complete; the independent checker AC4 confir… 15099 2026-11-16 17:46 ST9 status by AC1: H4 -> supported (plan PL5), evidence RN5, CK12, AR1: PL5's rule for supported is met: the proof (paper H5 in proofs.md, AR1) is complete; the independent checker AC4 confir… 15100 2026-11-16 17:47 ST10 status by AC1: H5 -> supported (plan PL5), evidence RN5, CK6, AR1: PL5's rule for supported is met: the proof (paper H6 in proofs.md, AR1) is complete; the independent checker AC4 confir… 15101 2026-11-16 17:47 ST11 status by AC1: H7 -> supported (plan PL5), evidence RN5, CK7, AR1: PL5's rule for supported is met: the proof (paper H7 in proofs.md, AR1) is complete; the independent checker AC4 confir… 15102 2026-11-16 17:47 ST12 status by AC1: H8 -> supported (plan PL5), evidence RN5, CK8, AR1: PL5's rule for supported is met: the proof (paper H8 in proofs.md, AR1) is complete; the independent checker AC4 confir… 15103 2026-11-16 17:47 ST13 status by AC1: H12 -> supported (plan PL5), evidence RN5, CK10, AR1: PL5's rule for supported is met: the proof (paper H10 in proofs.md, AR1) is complete; the independent checker AC4 confi… 15104 2026-11-16 17:47 ST14 status by AC1: H14 -> supported (plan PL5), evidence RN5, CK11, AR1: PL5's rule for supported is met: the proof (paper H11 in proofs.md, AR1) is complete; the independent checker AC4 confi… 15105 2026-11-16 17:47 ST15 status by AC1: H31 -> supported (plan PL7), evidence RN6, CK9, AR1: PL6's rule for supported is met: the proof (paper H9 in proofs.md, AR1) is complete; AC4 confirmed it (CK9, full); and… 15106 2026-11-16 21:33 H33 hypothesis by AC1: Let $Y$ 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 ab… [a_priori] 15107 2026-11-16 21:33 CK13 check by AC4: H4: confirmed (full) 15108 2026-11-16 21:34 ST16 status by AC1: H4 -> supported (plan PL5), evidence RN5, CK13, AR1: Verdict renewed after H4 was corrected twice: H4.R3 (the median with equal weights is a flat point, not necessarily the… 15109 2026-11-16 21:35 NT5 note by AC1: The second attempt of the deciding run of PL2 (started 2026-11-16 about 15:00 UTC, unchanged) was killed when the container restarted again, after lognormal si… 15110 2026-11-16 21:55 AR5 artefact by AC1: files/proofs.md 15111 2026-11-16 21:55 AR6 artefact by AC1: files/proofs.pdf 15112 2026-11-16 21:55 AR7 artefact by AC1: files/proof-check-report.md 15113 2026-11-16 21:55 CK14 check by AC1: H1: confirmed (full), checked by AC2, recorded by agent AC1: "All looks fine now. NExt" 15114 2026-11-16 21:55 CK15 check by AC1: H2: confirmed (full), checked by AC2, recorded by agent AC1: "All looks fine now. NExt" 15115 2026-11-16 21:55 CK16 check by AC1: H3: confirmed (full), checked by AC2, recorded by agent AC1: "All looks fine now. NExt" 15116 2026-11-16 21:55 CK17 check by AC1: H13: confirmed (full), checked by AC2, recorded by agent AC1: "All looks fine now. NExt" 15117 2026-11-16 21:56 CK18 check by AC1: H4: confirmed (full), checked by AC2, recorded by agent AC1: "All looks fine now. NExt" 15118 2026-11-16 21:56 CK19 check by AC1: H5: confirmed (full), checked by AC2, recorded by agent AC1: "All looks fine now. NExt" 15119 2026-11-16 21:56 CK20 check by AC1: H7: confirmed (full), checked by AC2, recorded by agent AC1: "All looks fine now. NExt" 15120 2026-11-16 21:56 CK21 check by AC1: H8: confirmed (full), checked by AC2, recorded by agent AC1: "All looks fine now. NExt" 15121 2026-11-16 21:56 CK22 check by AC1: H31: confirmed (full), checked by AC2, recorded by agent AC1: "All looks fine now. NExt" 15122 2026-11-16 21:56 CK23 check by AC1: H12: confirmed (full), checked by AC2, recorded by agent AC1: "All looks fine now. NExt" 15123 2026-11-16 21:56 CK24 check by AC1: H14: confirmed (full), checked by AC2, recorded by agent AC1: "All looks fine now. NExt" 15124 2026-11-16 22:53 NT6 note by AC1: The third attempt of the deciding run of PL2 (started 2026-11-16 21:36 UTC with the parallel code, unchanged results) was killed when the container was reclaim… 15125 2026-11-16 22:54 ST17 status by AC1: H6 -> inconclusive (plan PL5), evidence RN5: PL5's deciding run RN5 ran with H6 among its hypotheses. H6 was merged into H4 by DC5 before its verdict: its direction… 15126 2026-11-16 22:54 ST18 status by AC1: H9 -> inconclusive (plan PL5), evidence RN5: PL5's deciding run RN5 ran with H9 among its hypotheses. H9 was merged into H3 by DC5 before its verdict: its content i… 15127 2026-11-16 22:54 ST19 status by AC1: H10 -> inconclusive (plan PL5), evidence RN5: PL5's deciding run RN5 ran with H10 among its hypotheses. H10 was merged into H2 by DC5 before its verdict: its content… 15128 2026-11-16 22:54 ST20 status by AC1: H11 -> inconclusive (plan PL5), evidence RN5: PL5's deciding run RN5 ran with H11 among its hypotheses. H11 was dropped by DC5 before its verdict (abandoned, ST4): t… 15129 2026-11-16 22:55 H34 hypothesis by AC1: The forecast with the lowest expected sMAPE is the median $m$ of the actuals in each of two cases. (a) $\ln Y$ has a density symmetric about $\ln m$ and unimod… [a_priori] 15130 2026-11-16 22:55 PL8 plan by AC1: [confirmatory, requires approval] for H1, H2, H3, H13, H4, H5, H7, H8, H12, H14, H34: PL5 decided the hypotheses as worded before the restatement of DC5 (s without the factor 200, 'actual', merged results,… 15131 2026-11-16 22:56 CK25 check by AC4: H34: confirmed (full) 15132 2026-11-16 23:13 RN7 run by AC1: [confirmatory, success, under PL2, commit fa0ce62] python3 claimset-code/smape-target/efficiency.py --seed 4 --reps 10000; 772 results in runs/RN7/results.json (RS14977 to RS15748) 15905 2026-11-16 23:16 DV1 deviation by AC1: 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 anot… 15906 2026-11-16 23:16 DV2 deviation by AC1: 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 mea… 15907 2026-11-16 23:16 DV3 deviation by AC1: RN5 ran at commit 1e23bdc, not the plan's f865dfa. Between them only encode.py (new), targets.py and flatness.py changed (the --derived-only option of DC4); ch… 15908 2026-11-17 09:21 PL9 plan by AC1: [confirmatory, requires approval] for H1, H2, H3, H13, H4, H5, H7, H8, H12, H14, H34: PL5 decided the hypotheses as worded before the restatement of DC5 (s without the factor 200, 'actual', merged results,… 15909 2026-11-17 09:21 AP6 approval by AC1: PL9 approved by AC2 at 2026-11-17T09:20:47Z, recorded by agent AC1: "Record. And the plan with these corrections is approved." 15910 2026-11-17 09:23 RN8 run by AC1: [confirmatory, success, under PL9, commit fa0ce62] python3 claimset-code/smape-target/check_formulas.py --seed 6; 39 results in runs/RN8/results.json (RS15749 to RS15787) 15950 2026-11-17 09:23 ST21 status by AC1: H1 -> supported (plan PL8), evidence RN8, CK1, CK14, AR5: PL8's rule for supported is met for the current wording: the proof (paper H1 in proofs.md, AR1.R2) is complete; the ind… 15951 2026-11-17 09:23 ST22 status by AC1: H2 -> supported (plan PL8), evidence RN8, CK2, CK15, AR5: PL8's rule for supported is met for the current wording: the proof (paper H2 in proofs.md, AR1.R2) is complete; the ind… 15952 2026-11-17 09:24 ST23 status by AC1: H3 -> supported (plan PL8), evidence RN8, CK3, CK16, AR5: PL8's rule for supported is met for the current wording: the proof (paper H3 in proofs.md, AR1.R2) is complete; the ind… 15953 2026-11-17 09:24 ST24 status by AC1: H13 -> supported (plan PL8), evidence RN8, CK4, CK17, AR5: PL8's rule for supported is met for the current wording: the proof (paper H4 in proofs.md, AR1.R2) is complete; the ind… 15954 2026-11-17 09:24 ST25 status by AC1: H4 -> supported (plan PL8), evidence RN8, CK13, CK18, AR5: 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 co… 15955 2026-11-17 09:24 ST26 status by AC1: H5 -> supported (plan PL8), evidence RN8, CK6, CK19, AR5: PL8's rule for supported is met for the current wording: the proof (paper H6 in proofs.md, AR1.R2) is complete; the ind… 15956 2026-11-17 09:24 ST27 status by AC1: H7 -> supported (plan PL8), evidence RN8, CK7, CK20, AR5: PL8's rule for supported is met for the current wording: the proof (paper H7 in proofs.md, AR1.R2) is complete; the ind… 15957 2026-11-17 09:24 ST28 status by AC1: H8 -> supported (plan PL8), evidence RN8, CK8, CK21, AR5: PL8's rule for supported is met for the current wording: the proof (paper H8 in proofs.md, AR1.R2) is complete; the ind… 15958 2026-11-17 09:24 ST29 status by AC1: H12 -> supported (plan PL8), evidence RN8, CK10, CK23, AR5: PL8's rule for supported is met for the current wording: the proof (paper H10 in proofs.md, AR1.R2) is complete; the in… 15959 2026-11-17 09:24 ST30 status by AC1: H14 -> supported (plan PL8), evidence RN8, CK11, CK24, AR5: PL8's rule for supported is met for the current wording: the proof (paper H11 in proofs.md, AR1.R2) is complete; the in… 15960 2026-11-17 09:24 ST31 status by AC1: H34 -> supported (plan PL8), evidence RN8, CK25, CK18, CK19, AR5: 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 proo… 15961 2026-11-17 11:22 TM32 term by AC1: msMAPE term [variable] 15962 2026-11-17 11:22 H35 hypothesis by AC1: The [[TM32:msMAPE term]] has the properties of the [[TM1:sMAPE term]] except near zero. (a) Where $|y|+|f| \ge 0.5$, $m(y,f) = s(y,f)\,\frac{|y|+|f|}{|y|+|f|+\… [a_priori] 15963 2026-11-17 11:22 TM33 term by AC1: msMAPE target [concept] 15964 2026-11-17 11:22 TM34 term by AC1: Poisson counts [population] 15965 2026-11-17 11:22 PL10 plan by AC1: [confirmatory, requires approval] for H35: Proof of H35 from the definition of the msMAPE term (a short section added to the proof document), checked independentl… 15966 2026-11-17 11:22 PL11 plan by AC1: [estimation, requires approval] for Q1: Exact computation, with no sampling, of the forecasts that minimise the expected msMAPE term and the expected sMAPE ter… 15967 2026-11-17 11:22 AP7 approval by AC1: PL10 approved by AC2 at 2026-11-17T11:19:45Z, recorded by agent AC1: "1. Estimation. 2. Distributions are fine, we don't need all of them. Approved. Go" 15968 2026-11-17 11:22 AP8 approval by AC1: PL11 approved by AC2 at 2026-11-17T11:19:45Z, recorded by agent AC1: "1. Estimation. 2. Distributions are fine, we don't need all of them. Approved. Go" 15969 2026-11-17 11:22 RN9 run by AC1: [confirmatory, success, under PL10, commit 838c642] python3 claimset-code/smape-target/check_msmape.py --seed 7; 8 results in runs/RN9/results.json (RS15788 to RS15795) 15978 2026-11-17 11:23 CK26 check by AC4: H35: confirmed (full) 15979 2026-11-17 11:25 RN10 run by AC1: [confirmatory, success, under PL11, commit 838c642] python3 claimset-code/smape-target/msmape_targets.py; 201 results in runs/RN10/results.json (RS15796 to RS15996) 16181 2026-11-17 11:26 AR8 artefact by AC1: files/proof-msmape.md 16182 2026-11-17 11:26 ST32 status by AC1: H35 -> supported (plan PL10), evidence RN9, CK26, AR8: PL10's rule for supported is met: the proof of H35 (proof-msmape.md, AR8) is complete, every step naming what it rests… 16183 2026-11-17 11:45 RN11 run by AC1: [exploratory, success, commit e294cdf] python3 claimset-code/smape-target/plot_msmape.py claimset/smape-target/lab/runs/RN10/results.json 16184 2026-11-17 14:47 AR9 artefact by AC1: files/proofs.md 16185 2026-11-17 14:47 AR10 artefact by AC1: files/proofs.pdf 16186 2026-11-17 14:47 AR11 artefact by AC1: files/proof-msmape.md