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 | 928 | 0.596 | 69% |
| Base rate | 928 | 0.688 | 55% |
| Recent form | 928 | 0.627 | 67% |
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) | 8 May 2026 | 352 | 0.592 | 0.582 | −1.8% |
| Polymarket | 16 May 2025 | 638 | 0.601 | 0.594 | −1.1% |
| Kalshi | 22 May 2025 | 643 | 0.600 | 0.596 | −0.8% |
Expansion teams start at an average rating and find their level from their results.
The playoff odds play the real format: the top eight by record, a fixed bracket, best of three, best of five and a best-of-seven final. A series already under way continues from its current score.
Rest and travel are not in the model.
Who is playing is judged from box scores and the injury report, not from lineups announced before tip-off, so a late scratch can be missed.
Questions and answers
How accurate are PodiumOracle's WNBA predictions?
On 928 games from held-back seasons (2024 to 2026), which the model never saw while it was tuned, the side it favoured won 69% of the time. Its log loss was 0.596, against 0.688 for the base rate and 0.627 for recent form. Against the betting markets on the same finished games, its log loss was 0.592 against the bookmakers' 0.582 over 352 games, 0.601 against Polymarket's 0.594 over 638 games and 0.600 against Kalshi's 0.596 over 643 games (lower is better).