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 | 93 | 0.411 | 85% |
| Base rate | 93 | 0.662 | 78% |
| Recent form | 93 | 0.573 | 72% |
Ratings use every test in the related competitions (see Sources); only Rugby World Cup games are scored in the track record.
The 2027 tournament is played in full in every simulation: the pool games with bonus points (four for a win, two for a draw, one for losing by seven or less, one for four tries), then the top two in each pool and the four best third-placed sides into the round of 16 and World Rugby's fixed bracket. Every ground is neutral except that Australia is at home.
The best third-placed sides are placed in the round of 16 by the pools each slot allows; where World Rugby's own table differs for a combination of pools, the first valid placement is used.
Injuries and selection are not in the model.
Questions and answers
How accurate are PodiumOracle's Rugby World Cup predictions?
On 93 games from held-back seasons (2019 to 2023), which the model never saw while it was tuned, the side it favoured won 85% of the time. Its log loss was 0.411, against 0.662 for the base rate and 0.573 for recent form.