Every summer the same question arrives before a ball is kicked: who actually got better? Not who spent the most, not who made the most noise — who genuinely improved the eleven that walks out in August. We built a model to answer it, and then ran the season twenty thousand times.
How the Supercomputer thinks
It starts with last season's final table, because the Premier League is stubborn: roughly seventy per cent of a club's points total carries into the following year. Then it adjusts for the window — but not by transfer fee. Fees measure ambition, not quality.
Instead every permanent signing and departure is valued by what that player actually produced in the Premier League last season: minutes on the pitch multiplied by their match rating above a replacement-level baseline. A defender who played thirty-four games at 7.2 is a real, measurable asset moving between two clubs in the same table. A name on a shirt is not.
From there, twenty thousand seasons are simulated with the year-to-year variance the league genuinely produces. The table below is the average of all of them.
The predicted table
| # | Club | Projected | 25/26 | Move | Key probability |
|---|---|---|---|---|---|
| 1 | Arsenal | 80 | 85 | — | 63.1% title |
| 2 | Manchester City | 74 | 78 | — | 24.4% title |
| 3 | Manchester United | 70 | 71 | — | 11.0% title |
| 4 | Aston Villa | 58 | 65 | — | 0.5% title |
| 5 | Tottenham | 57 | 41 | ▲ 12 | 0.4% title |
| 6 | Liverpool | 55 | 60 | ▼ 1 | 0.2% title |
| 7 | Brentford | 54 | 53 | ▲ 2 | 2.2% drop risk |
| 8 | Chelsea | 54 | 52 | ▲ 2 | 2.9% drop risk |
| 9 | Sunderland | 53 | 54 | ▼ 2 | 3.1% drop risk |
| 10 | Bournemouth | 53 | 57 | ▼ 4 | 2.9% drop risk |
| 11 | Brighton | 50 | 53 | ▼ 3 | 6.8% drop risk |
| 12 | Everton | 50 | 49 | ▲ 1 | 7.6% drop risk |
| 13 | Fulham | 49 | 52 | ▼ 2 | 8.2% drop risk |
| 14 | Leeds | 48 | 47 | — | 10.5% drop risk |
| 15 | Crystal Palace | 45 | 45 | — | 18.9% drop risk |
| 16 | Newcastle | 43 | 49 | ▼ 4 | 27.7% drop risk |
| 17 | Nottingham Forest | 42 | 44 | ▼ 1 | 33.7% drop risk |
| 18 | Ipswich | 38 | — | new | 55.7% drop risk |
| 19 | Coventry | 38 | — | new | 58.2% drop risk |
| 20 | Hull City | 38 | — | new | 58.4% drop risk |
Arsenal retain, but not comfortably
Arsenal come out champions in 63.1% of simulated seasons. That is a clear favourite, not a coronation — Manchester City take the title in 24.4%, and across twenty thousand runs the two are separated by a handful of results more often than not. The model has Arsenal on 80 points, down five on last season's 85, because regression pulls every outlier back toward the pack.
Tottenham are the story of the window
No club in the division gained more proven Premier League quality this summer. Spurs finished seventeenth. They have since added three established starters — Jan Paul van Hecke from Brighton, Marcos Senesi from Bournemouth and Sandro Tonali from Newcastle — and the model moves them seventeenth to fifth, the largest projected climb in the league.
The three biggest net gains: Tottenham +24.3 · Manchester United +6.9 · Arsenal +5.8.
Newcastle paid for it
The same window that rebuilt Tottenham took Newcastle apart. Bruno Guimarães left for Arsenal and Tonali for Spurs, and the model registers the largest quality loss in the division. From twelfth last season, Newcastle project to sixteenth, with a 27.7% chance of going down.
The three heaviest losses: Newcastle -11.9 · Nottingham Forest -7.3 · Aston Villa -5.6.
Three promoted clubs, three identical numbers
Ipswich, Coventry and Hull City all project to 38 points and all carry a relegation probability in the high fifties. That is not the model being lazy — it is the model being honest. None of the three has a Premier League record to separate them from the others, so it declines to invent one.
What the Supercomputer cannot see
Worth saying plainly, because a projection that hides its limits is worth less than one that names them. Of 105 permanent arrivals across the division, only 37 came from another Premier League club. The other sixty-five per cent arrived from abroad or the Championship with no Premier League minutes to measure, so the model cannot value them at all.
Chelsea, Newcastle and Liverpool signed the most players in that unmeasured group. Read their projections as a floor rather than a forecast — the model knows what they lost far better than it knows what they bought.
It also knows nothing about managers, injuries or fixture runs. A season is not the sum of its transfers. But it is a great deal more than a hunch, and every number above comes from the record rather than an opinion.