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 | 7418 | 0.680 | 56% |
| Base rate | 7418 | 0.691 | 53% |
| Recent form | 7418 | 0.690 | 54% |
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) | 26 Mar 2026 | 2469 | 0.683 | 0.680 | −0.4% |
| Polymarket | 2 Apr 2025 | 4234 | 0.680 | 0.677 | −0.3% |
| Kalshi | 16 Apr 2025 | 4039 | 0.680 | 0.678 | −0.4% |
The information from Retrosheet was obtained free of charge from and is copyrighted by Retrosheet. Interested parties may contact Retrosheet at www.retrosheet.org.
The postseason odds play the real format: three division winners and three wild cards per league, byes for the top two, and best-of-three, five and seven series.
The starter figure is runs allowed with him on the mound, which includes his defence and bullpen; it is not a pitching-only measure like FIP.
Lineups, rest and bullpen use are not in the model.
Ballparks with a roof get no weather, because whether it was closed isn't recorded.
Questions and answers
How accurate are PodiumOracle's MLB predictions?
On 7,418 games from held-back seasons (2024 to 2026), which the model never saw while it was tuned, the side it favoured won 56% of the time. Its log loss was 0.680, against 0.691 for the base rate and 0.690 for recent form. Against the betting markets on the same finished games, its log loss was 0.683 against the bookmakers' 0.680 over 2,469 games, 0.680 against Polymarket's 0.677 over 4,234 games and 0.680 against Kalshi's 0.678 over 4,039 games (lower is better).