Every model version we have trained, with its out-of-time score against the closing market. A new version only ships if it scores at least as well as the current champion on the same out-of-time matches — more features never promote a model automatically.
| Model | Features | Trained | Holdout RPS | Market RPS | Gap vs market | Matches (n) | Calibration (ECE) |
|---|---|---|---|---|---|---|---|
| ensemble-v2lhClub | v2lh | 11 Oct 2026 | 0.2291 | — | — | 74 | 0.182 |
| ensemble-v2lhClub | v2lh | 11 Oct 2026 | 0.2094 | — | — | 554 | 0.076 |
| btts-club-championClub | v2 | 12 Oct 2026 | 0.2407 | — | — | 557 | 0.051 |
| ou25-club-championClub | v2 | 12 Oct 2026 | 0.2442 | — | — | 557 | 0.094 |
| ou25-club-championClub | v2 | 18 Oct 2026 | 0.2421 | 0.2242 | +0.0179 | 114 of 683 | 0.065 |
| ensemble-v2lhClub | v2lh | 18 Oct 2026 | 0.2106 | 0.1878 | +0.0228 | 114 of 683 | 0.043 |
| ensemble-v2lhClub | v2lh | 25 Oct 2026 | 0.2158 | 0.2005 | +0.0153 | 249 of 688 | 0.033 |
| ensemble-v2lhClub | v2lh | 26 Oct 2026 | 0.2170 | 0.2021 | +0.0149 | 283 of 696 | 0.054 |
| btts-club-championClub | v2 | 26 Oct 2026 | 0.2426 | 0.2399 | +0.0027 | 283 of 696 | 0.048 |
| ensemble-v2lhClub | v2lh | 01 Nov 2026 | 0.2110 | 0.1997 | +0.0113 | 345 of 684 | 0.024 |
| btts-club-championClub | v2 | 01 Nov 2026 | 0.2439 | 0.2430 | +0.0009 | 345 of 684 | 0.055 |
| ou25-club-championClub | v2 | 08 Nov 2026 | 0.2358 | 0.2342 | +0.0016 | 491 of 717 | 0.014 |
| ensemble-v2lhClubChampion | v2lh | 08 Nov 2026 | 0.2086 | 0.1991 | +0.0095 | 491 of 717 | 0.058 |
| btts-club-championClubChampion | v2 | 08 Nov 2026 | 0.2467 | 0.2460 | +0.0006 | 491 of 717 | 0.030 |
| ou25-club-championClubChampion | v2 | 15 Nov 2026 | 0.2372 | 0.2337 | +0.0035 | 530 of 730 | 0.026 |
| women-baselineWomen's club | baseline | 13 Oct 2026 | — | — | — | — | — |
| wc-baselineNational | baseline | 11 Oct 2026 | — | — | — | — | — |
Gap = holdout RPS − market RPS on the same matches. Negative means the model scored better than the closing line out-of-time; it turns green only when the significance test confirms it beat the market — a small gap on noise stays neutral. Beating the market is not required for promotion: a club version ships when it scores no worse than the current champion on the same out-of-time matches (a national-team version must also beat the closing line), so a champion can trail the market. That is why the forecasts we publish are anchored to the market price. A blank gap means we have no out-of-time market comparison for it yet. Matches (n) is how many out-of-time matches sit behind the row's comparison — treat a gap on a handful of matches as noise, not signal. When only some of the out-of-time matches had a recorded closing line, the RPS and gap columns use just those, and the cell shows both counts (e.g. 114 of 683); calibration (ECE) is measured on all of them.