Premier League predictions
England
10 upcoming matches. Here the market has been at least as accurate as the model, so the site shows probabilities but no value picks.
Saturday, 21 November 2026
- Man CityFulhamO2.5 57%BTTS 51%xG 2.13–0.85
- Aston VillaSunderlandO2.5 48%BTTS 50%xG 1.62–0.96
- BournemouthNott'm ForestO2.5 51%BTTS 55%xG 1.55–1.15
- ChelseaLeedsO2.5 52%BTTS 55%xG 1.63–1.12
- CoventryCrystal PalaceO2.5 53%BTTS 57%xG 1.32–1.47
- TottenhamIpswichO2.5 67%BTTS 62%xG 2.33–1.13
- NewcastleArsenalO2.5 50%BTTS 51%xG 0.95–1.73
Sunday, 22 November 2026
Monday, 23 November 2026
Recent results
Saturday, 12 September 2026
- FTAston VillaNott'm Forest1-2O2.5 50%BTTS 56%xG 1.52–1.21backfilled
- FTBournemouthBrentford2-2O2.5 59%BTTS 61%xG 1.67–1.33backfilled
- FTChelseaHull2-2O2.5 53%BTTS 44%xG 1.32–0.88backfilled
- FTCrystal PalaceIpswich2-3O2.5 56%BTTS 56%xG 2.05–0.99backfilled
- FTLiverpoolFulham0-0O2.5 63%BTTS 59%xG 1.99–1.12backfilled
- FTTottenhamEverton0-0O2.5 49%BTTS 54%xG 1.36–1.22backfilled
Sunday, 13 September 2026
Monday, 19 October 2026
Monday, 2 November 2026
Team ratings
Attack and defence on a log scale against the league average (0). Higher attack scores more; lower defence concedes less. Fitted on goals and expected goals, recent matches weighted more.
| # | Team | Attack | Defence | Strength |
|---|---|---|---|---|
| 1 | Arsenal | 0.33 | -0.66 | 1.00 |
| 2 | Man City | 0.42 | -0.34 | 0.77 |
| 3 | Liverpool | 0.35 | -0.20 | 0.55 |
| 4 | Brighton | 0.21 | -0.13 | 0.33 |
| 5 | Chelsea | 0.25 | -0.05 | 0.30 |
| 6 | Aston Villa | 0.18 | -0.05 | 0.23 |
| 7 | Newcastle | 0.22 | -0.01 | 0.23 |
| 8 | Brentford | 0.15 | -0.04 | 0.19 |
| 9 | Bournemouth | 0.13 | -0.06 | 0.18 |
| 10 | Man United | 0.13 | -0.01 | 0.14 |
| 11 | Leeds | -0.06 | -0.16 | 0.10 |
| 12 | Hull | -0.20 | -0.27 | 0.06 |
| 13 | Nott'm Forest | -0.02 | -0.09 | 0.06 |
| 14 | Everton | -0.15 | -0.20 | 0.05 |
| 15 | Tottenham | 0.08 | 0.04 | 0.05 |
| 16 | Fulham | -0.04 | -0.07 | 0.03 |
| 17 | Crystal Palace | -0.07 | 0.02 | -0.09 |
| 18 | Sunderland | -0.22 | -0.10 | -0.13 |
| 19 | Coventry | -0.14 | 0.23 | -0.37 |
| 20 | Ipswich | -0.14 | 0.36 | -0.50 |
How the model does in this league
Out of sample on 1231 matches, the blend scored a log loss of 0.9689 against the closing market's 0.9648 (lower is better). The difference is within noise, so the site shows probabilities but no value picks. What this means.