The Premier League, by the numbers.
A projected final table and match-by-match probabilities for the 2026/27 season, generated by a Dixon-Coles Poisson model averaged with an Elo baseline. We publish the method and the input ratings up front. This is applied probability, not a tip sheet.
Man City
Title favourite · 46%
20
Teams rated
380
Games simulated
2
Models, averaged
The title race
Who wins the league?
Share of 3000 simulated seasons each team finishes 1st. The model runs every fixture home and away, samples goals from each match’s Poisson rates, and tallies the final standings.
Projected final table
Where every team lands
| # | Team | Pts | Title | Top 4 | Releg. |
|---|---|---|---|---|---|
| 1 | Manchester City | 78 | 46% | 96% | — |
| 2 | Arsenal | 75 | 28% | 92% | — |
| 3 | Liverpool | 73 | 22% | 90% | — |
| 4 | Chelsea | 60 | 1% | 28% | 1% |
| 5 | Newcastle United | 59 | 1% | 26% | — |
| 6 | Aston Villa | 57 | — | 18% | 1% |
| 7 | Tottenham Hotspur | 56 | — | 16% | 1% |
| 8 | Manchester United | 56 | — | 13% | 1% |
| 9 | Brighton & Hove Albion | 53 | — | 7% | 3% |
| 10 | Nottingham Forest | 50 | — | 4% | 6% |
| 11 | Crystal Palace | 50 | — | 3% | 8% |
| 12 | Bournemouth | 49 | — | 2% | 8% |
| 13 | Fulham | 47 | — | 2% | 12% |
| 14 | Brentford | 46 | — | 2% | 15% |
| 15 | Everton | 46 | — | 1% | 17% |
| 16 | West Ham United | 45 | — | 1% | 19% |
| 17 | Wolverhampton | 42 | — | — | 32% |
| 18 | Leeds United | 39 | — | — | 49% |
| 19 | Burnley | 37 | — | — | 60% |
| 20 | Sunderland | 35 | — | — | 67% |
Head-to-head
When the big sides meet
The model’s output for the season’s marquee fixtures — home win, draw, away win, and the most likely scoreline. The bar reads left to right: home · draw · away.
Same method, higher stakes
We point the same engine at Kenya’s 2027 election.
A football model and an election model are the same idea — quantify uncertainty from the evidence, publish every input, and score yourself afterward. The Premier League is where you can check our work weekly. Our flagship is a transparent, fundamentals-plus-polls forecast of Kenya’s August 2027 general election.
How it works
The method, in the open
1 · Ratings
Each club carries an Elo strength rating — the model’s published prior. Stronger clubs create more, concede less. These are the only hand-set inputs; every probability is derived from them.
2 · Match model
Goals follow a Poisson process with a Dixon-Coles low-score correction, averaged with an Elo baseline. Two models, shown together, so you can see where they agree.
3 · Season
We simulate all 380 games thousands of times and count how often each team wins the league, makes the top four, or goes down. The table shows the average outcome.
Elo ratings are the model's published priors (a snapshot), not the official roster. Ratings and fixtures refresh from results — a live football-data feed can auto-update them each matchweek. Nothing here is a betting recommendation — it is a published probability model you can check against results.
Get the next model
Predictions, forecasts, and the methods behind them.
New models — Premier League, elections, markets — plus the data tools and courses that teach you to build them.