TRAIL OF H3 (claim C3) Statement: 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) $0 \le \tfrac12|\ln y - \ln f| - s \le \tfrac1{24}|\ln y - \ln f|^3$; so when all log errors are small, sMAPE is about 100 times the mean absolute error of $\ln f$ as a forecast of $\ln y$ ([[TM19: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 $\frac1n\sum_t|\ln y_t - \ln f^*| - \frac1n\sum_t|\ln y_t - \ln m| \le \frac1{12}\cdot\frac1n\sum_t|\ln y_t - \ln f^*|^3$: its 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. Status: supported In log order: the question, each version of the hypothesis, the plans that name it with their approvals, deviations and runs, its statuses, and the checks, decisions and notes that refer to them. Results are given by number and file. 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 13:44 TM19 term by AC1: log-MAE [variable] 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… 2026-11-18 11:08 H36 hypothesis by AC5: 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] 2026-11-18 12:02 PL12 plan by AC5: [confirmatory, requires approval] for H36, H37, H38, H41, H39, H40, H42: Version 3 of the paper answers a reader's comments on version 2. Proofs: in proofs.md, section H3 gains part (c); secti… 2026-11-18 12:02 AP9 approval by AC5: PL12 approved by AC2 at 2026-11-18T12:02:11Z, recorded by agent AC5: "1. Opus is fine [...] 5. (2) with 3% is fine 6. We use "m" to denote the median. Use diff… 2026-11-18 12:02 RN13 run by AC5: [confirmatory, success, under PL12, commit 2f0d2fc] python3 claimset-code/smape-target/check_v3.py --seed 8; 82 results in runs/RN13/results.json (RS16073 to RS16154) 2026-11-18 12:14 CK27 check by AC6: H36: confirmed (full) 2026-11-18 12:15 ST33 status by AC5: H3 -> supported (plan PL12), evidence RN13, CK27: PL12's rule for supported is met for H36, the current wording of H3: the proof (proofs.md, section H3, with the new par…