How good are these forecasts?
Every forecast scored on seasons the model never saw while its settings were chosen.
Held-back seasons
| Forecast | Forecasts | Log loss | Favourite won |
|---|---|---|---|
| The model | 133 | 0.667 | 59% |
| Base rate | 133 | 0.681 | 58% |
| Recent form | 133 | 0.673 | 58% |
Against the betting markets
Each market's price before the start, scored on the same finished games as the model. A negative figure means the market was more accurate.
| Market | Since | Priced | Model log loss | Market log loss | Model vs market |
|---|---|---|---|---|---|
| Bookmakers (Pinnacle, DraftKings, bet365) | 20 Jan 2026 | 83 | 0.656 | 0.670 | 2.0% |
The table is the PWHL's: three points for a regulation win, two for an overtime or shootout win, one for losing in one. The simulation gives overtime at the league's rate.
With only a few seasons played, the rating's speed, margin of victory and home ice come from the other professional hockey leagues rather than a fit to one short season, and the track record scores the seasons since the first; treat it as indicative.
Goaltending and the players on the ice are not in the rating.
The first seed chooses its semi-final opponent; the simulation assumes it takes the fourth.
Questions and answers
How accurate are PodiumOracle's PWHL predictions?
On 133 games from held-back seasons (2025–26), which the model never saw while it was tuned, the side it favoured won 59% of the time. Its log loss was 0.667, against 0.681 for the base rate and 0.673 for recent form. Against the betting markets on the same finished games, its log loss was 0.656 against the bookmakers' 0.670 over 83 games (lower is better).