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 | 725 | 1.007 | 50% |
| Base rate | 725 | 1.075 | 44% |
| Recent form | 725 | 1.044 | 46% |
| Betting market | 722 | 0.991 | 52% |
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) | 23 Jan 2026 | 214 | 0.958 | 0.957 | −0.1% |
| Polymarket | 7 Aug 2026 | 87 | 0.893 | 0.894 | 0.2% |
| Kalshi | 31 Oct 2025 | 243 | 0.993 | 0.983 | −1.0% |
A draw is its own outcome: the model has a band around an even contest in which a draw is the likeliest single result, fit on the tuning seasons.
The market comparison uses the bookmakers' closing odds with their margin taken out.
The regular season is simulated; the play-off groups that follow it (points halved) are not, so the odds are for how the regular season ends.
Team news, injuries and rotation for European games are not in the model.
Expected goals are not in the rating yet; it learns only from results.
Questions and answers
How accurate are PodiumOracle's Belgian Pro League predictions?
On 725 games from held-back seasons (2024–25 to 2026–27), which the model never saw while it was tuned, the side it favoured won 50% of the time. Its log loss was 1.007, against 1.075 for the base rate and 1.044 for recent form, and 0.991 for the closing betting odds. Against the betting markets on the same finished games, its log loss was 0.958 against the bookmakers' 0.957 over 214 games, 0.893 against Polymarket's 0.894 over 87 games and 0.993 against Kalshi's 0.983 over 243 games (lower is better).