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 | 211 | 0.615 | 70% |
| Base rate | 211 | 0.692 | 66% |
| Recent form | 211 | 0.635 | 70% |
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 |
|---|---|---|---|---|---|
| Polymarket | 23 Jan 2026 | 58 | 0.708 | 0.765 | 7.5% |
Clubs are rated on their Premiership games and also their Champions Cup and Challenge Cup games, alongside the URC and Top 14 sides they meet there, so the European results put the leagues on one scale. Only Premiership games are scored in the track record.
The simulated table adds bonus points the league's way: four for a win, two for a draw, one for losing by seven or less, one for four tries (at each side's usual rate).
International call-ups during the Six Nations and autumn tests weaken clubs unevenly; the model doesn't know who is away.
Questions and answers
How accurate are PodiumOracle's Premiership predictions?
On 211 games from held-back seasons (2024–25 to 2026–27), which the model never saw while it was tuned, the side it favoured won 70% of the time. Its log loss was 0.615, against 0.692 for the base rate and 0.635 for recent form. Against the betting markets on the same finished games, its log loss was 0.708 against Polymarket's 0.765 over 58 games (lower is better).