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 | 11338 | 0.602 | 67% |
| Base rate | 11338 | 0.671 | 61% |
| Recent form | 11338 | 0.630 | 64% |
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) | 13 Feb 2026 | 2239 | 0.639 | 0.660 | 3.3% |
| Kalshi | 13 Feb 2026 | 2433 | 0.627 | 0.642 | 2.4% |
Only games ESPN carries are in the file: nearly every Division I game in recent seasons, fewer in the early ones.
The title odds use an approximate 64-team tournament of conference leaders and the best of the rest, seeded by simulated wins and rating. Regionals and super regionals are double-elimination and best-of-three in reality; here each round is one game.
Starting pitchers are not in the model yet.
The season runs from February to June; outside it the pages show how the last one finished.
Questions and answers
How accurate are PodiumOracle's NCAA baseball predictions?
On 11,338 games from held-back seasons (2024 to 2026), which the model never saw while it was tuned, the side it favoured won 67% of the time. Its log loss was 0.602, against 0.671 for the base rate and 0.630 for recent form. Against the betting markets on the same finished games, its log loss was 0.639 against the bookmakers' 0.660 over 2,239 games and 0.627 against Kalshi's 0.642 over 2,433 games (lower is better).