baseline (33.3%)
evaluated
| # | Team | Games | Pre-Match | In-Game | Correct |
|---|---|---|---|---|---|
| Loading… | |||||
The model performs best above 60% confidence — 60.6% accuracy. Predictions below 40% are essentially coin flips. Always cross-check with form and head to head data.
Uses the pre-match model (… accuracy) trained on 8 seasons of EPL data (2017–2025). Features include last-5 form, home/away specific form, current league position and points.
A 1-0 lead is never safe. Our model predicts the final result with % accuracy using the half-time score, form, and league position.
Enter the teams and half time score to predict the final result using the in-game model (… accuracy)
The in-game model uses the half time score alongside team form and league position, trained on 8 seasons of EPL data (2017–2025, COVID excluded). It accounts for the high rate of second-half turnarounds to predict the final result from any half-time scoreline.
When the model predicts each outcome, how often is it right?
Do higher-confidence predictions win more often?
Does the model do better when there's a clear favourite?
| # | Team | P | W | D | L | GF | GA | GD | Pts |
|---|---|---|---|---|---|---|---|---|---|
| Loading… | |||||||||