Premier League 2026-27
LIVE20,000 simulated seasonsfrom a model that blends this site's own season data with market odds, replays the real results as they land, and predicts every fixture - then keeps score on itself all season.
poisson-v2 · updated 2026-08-02 · preseason
The title race
Share of simulated seasons each club finishes first.
The next fixtures, called
Win-draw-win probabilities for the upcoming round. Market odds join each column once the books post them. Every prediction is frozen when first published and graded against the real result below.
| Date | Fixture | Home | Draw | Away | Pick |
|---|---|---|---|---|---|
| Fri 21 Aug | Arsenal v Coventry City | 89% | 8% | 3% | Arsenal |
| Sat 22 Aug | Brentford v Tottenham Hotspur | 47% | 23% | 30% | Brentford |
| Sat 22 Aug | Everton v Crystal Palace | 41% | 29% | 30% | Everton |
| Sat 22 Aug | Hull City v Manchester United | 8% | 14% | 78% | Manchester United |
| Sat 22 Aug | Ipswich Town v Sunderland | 45% | 26% | 28% | Ipswich Town |
| Sat 22 Aug | Nottingham Forest v Leeds United | 47% | 26% | 27% | Nottingham Forest |
| Sun 23 Aug | Brighton & Hove Albion v Aston Villa | 44% | 25% | 32% | Brighton & Hove Albion |
| Sun 23 Aug | Manchester City v AFC Bournemouth | 76% | 15% | 9% | Manchester City |
| Sun 23 Aug | Newcastle United v Liverpool | 32% | 23% | 45% | Liverpool |
| Mon 24 Aug | Fulham v Chelsea | 28% | 25% | 47% | Chelsea |
How the model is doing
Every fixture prediction above is frozen the moment it is published. Once the season kicks off, this section keeps the running score: how many picks landed, and the model's Brier score against the betting market's - in public, win or lose.
Every club, every outcome
Expected points, finishing ranges and the odds of each landing spot: the title, the top five (Champions League), the top seven (Europe) and the bottom three. “Finish” is the median simulated position with the 5th-95th percentile range.
| Club | xPts | Finish | Title | Top 5 | Top 7 | Relegated |
|---|---|---|---|---|---|---|
| Arsenal | 86.1 | 2 (1-4) | 47% | 98.6% | 99.7% | 0% |
| Manchester City | 84.3 | 2 (1-5) | 38.6% | 97.4% | 99.4% | 0% |
| Liverpool | 74.7 | 3 (1-8) | 10% | 83% | 93.2% | <0.1% |
| Chelsea | 66.0 | 5 (2-12) | 2% | 54.1% | 74.8% | 0.2% |
| Manchester United | 61.7 | 7 (3-14) | 0.9% | 37.3% | 59.9% | 0.5% |
| Newcastle United | 60.5 | 7 (3-14) | 0.7% | 33.5% | 55.3% | 0.7% |
| Aston Villa | 57.7 | 8 (3-15) | 0.3% | 23.5% | 44.5% | 1.1% |
| Brighton & Hove Albion | 56.4 | 9 (4-16) | 0.2% | 19.9% | 39.7% | 1.8% |
| Brentford | 51.7 | 11 (4-17) | 0.1% | 10.6% | 24% | 4.1% |
| AFC Bournemouth | 52.6 | 10 (4-17) | 0.1% | 11.9% | 26.8% | 3.5% |
| Tottenham Hotspur | 50.6 | 11 (5-18) | <0.1% | 8.6% | 20.5% | 5.3% |
| Fulham | 47.8 | 12 (5-18) | <0.1% | 5% | 13.7% | 8.5% |
| Crystal Palace | 47.0 | 13 (6-18) | <0.1% | 4.1% | 11.6% | 9% |
| Everton | 46.8 | 13 (6-18) | <0.1% | 3.8% | 10.8% | 9.1% |
| Nottingham Forest | 48.3 | 12 (5-18) | <0.1% | 5.3% | 14.6% | 7.4% |
| Leeds United | 45.0 | 14 (7-19) | <0.1% | 2.6% | 8.3% | 12.6% |
| Ipswich Town | 36.0 | 17 (10-19) | 0% | 0.3% | 1.4% | 43.4% |
| Coventry City | 35.0 | 17 (11-20) | 0% | 0.3% | 1.2% | 47.8% |
| Sunderland | 34.8 | 17 (11-19) | 0% | 0.2% | 0.7% | 47.1% |
| Hull City | 16.9 | 20 (18-20) | 0% | 0% | 0% | 98% |
Beat the Model
Back the calls the simulator is least sure about - your champion, a dark horse into the top five, and a favourite to fall - and see whether you out-predict it across the season.
Play Beat the Model: Premier LeagueHow the model works
Each club's attack and defence are goal rates per game from its last three league seasons (2025-26, 2024-25, 2023-24), recency-weighted, with the current campaign's real goals folded in as it plays out. The promoted clubs' Championship rates are translated with factors calibrated on every promoted side in this site's hub archive (75 club-seasons: attack ×0.637, goals conceded ×1.839). On top of the site data sits a market signal: de-vigged match odds (football-data.co.uk closing/posted odds (E0/E1)) fitted into team ratings and blended in at weight 0.45. Played matches count as real results; every remaining fixture is simulated with Poisson goals (league scoring rate 1.4423per team-game, home advantage ×1.11), each simulated season drawing every club's strength from a distribution (σ 0.15) rather than a fixed number. Fixture picks blend model and market 50/50 and are graded against results above. Site-data goal rates blended with market-implied team ratings; actual results folded in and only the remaining fixtures simulated.