Every settled matchday is disclosed after the evaluation worker finishes: shortlist wins, misses, abstentions, league splits, and every modelled market. No outcome is hidden by a subscription.
This report is rebuilt from the evaluation worker's settled postmortems. It measures the exact market shortlisted for each match and separately audits every other modelled angle.
The page never substitutes demo results. A report appears after completed fixtures have been processed by the evaluation worker.
This section remains separate from the reliability shortlist. It evaluates the original EPL 1X2 probability distribution, scoring quality, and publication-time odds.
Accuracy is wins divided by settled wins plus losses. Pending and void results are excluded. Brier score and log loss evaluate the original probability distribution; lower is better. The calibration gap compares average recorded confidence with observed accuracy. ROI and unit profit/loss are calculated only when every eligible settled record has immutable publication-time decimal odds; later prices are never substituted.
ROI eligibility: unavailable because at least one record lacks publication-time odds · window: all time · sample 22 Aug 2026, 12:21 to 28 Aug 2026, 23:40 · models Apex Hybrid@1.1.0
Changed: system reforecast — Superseded before kickoff by a materially changed model output. · 28 Aug 2026, 06:51
Changed: system reforecast — Superseded before kickoff by a materially changed model output. · 27 Aug 2026, 14:02
Changed: system reforecast — Superseded before kickoff by a materially changed model output. · 28 Aug 2026, 14:38
Changed: other — Reason was not captured by the legacy publication workflow. · 27 Aug 2026, 10:24
Changed: other — Reason was not captured by the legacy publication workflow. · 26 Aug 2026, 20:52
Changed: system reforecast — Superseded before kickoff by a materially changed model output. · 27 Aug 2026, 17:18
Changed: other — Reason was not captured by the legacy publication workflow. · 27 Aug 2026, 02:41
Changed: system reforecast — Superseded before kickoff by a materially changed model output. · 28 Aug 2026, 10:12
Changed: system reforecast — Superseded before kickoff by a materially changed model output. · 28 Aug 2026, 02:24
Changed: other — Reason was not captured by the legacy publication workflow. · 26 Aug 2026, 23:14
Changed: system reforecast — Superseded before kickoff by a materially changed model output. · 28 Aug 2026, 15:45
Changed: system reforecast — Superseded before kickoff by a materially changed model output. · 28 Aug 2026, 19:14
Changed: other — Reason was not captured by the legacy publication workflow. · 27 Aug 2026, 05:58
Changed: system reforecast — Superseded before kickoff by a materially changed model output. · 28 Aug 2026, 04:38
Changed: other — Reason was not captured by the legacy publication workflow. · 27 Aug 2026, 12:47
Changed: system reforecast — Superseded before kickoff by a materially changed model output. · 27 Aug 2026, 19:41
Changed: system reforecast — Superseded before kickoff by a materially changed model output. · 28 Aug 2026, 05:44
Changed: other — Reason was not captured by the legacy publication workflow. · 26 Aug 2026, 19:38
Changed: other — Reason was not captured by the legacy publication workflow. · 27 Aug 2026, 01:33
Changed: other — Reason was not captured by the legacy publication workflow. · 27 Aug 2026, 07:04