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 | 257 | 0.660 | 58% |
| Base rate | 257 | 0.681 | 58% |
| Recent form | 257 | 0.680 | 58% |
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) | 4 Jun 2026 | 83 | 0.609 | 0.573 | −6.2% |
| Polymarket | 4 Jun 2026 | 81 | 0.613 | 0.576 | −6.5% |
| Kalshi | 4 Jun 2026 | 80 | 0.594 | 0.575 | −3.3% |
The playoffs are simulated the league's way: three teams from each division, the crossover when a fourth-placed team has more points than the other division's third, division semi-finals and finals at the higher seed, and the Grey Cup on neutral ground.
CFL.ca's pages go back to 2016 (there was no 2020 season), and they list older playoff games without scores, so those few games are left out of the ratings.
Live scores come from CFL.ca's scoreboard; the in-game chance moves with the score and the clock, not with field position.
Quarterbacks and injuries are not in the model.
Questions and answers
How accurate are PodiumOracle's CFL predictions?
On 257 games from held-back seasons (2024 to 2026), which the model never saw while it was tuned, the side it favoured won 58% of the time. Its log loss was 0.660, against 0.681 for the base rate and 0.680 for recent form. Against the betting markets on the same finished games, its log loss was 0.609 against the bookmakers' 0.573 over 83 games, 0.613 against Polymarket's 0.576 over 81 games and 0.594 against Kalshi's 0.575 over 80 games (lower is better).