Our supercomputer played the entire Premier League — one million times
Before a single ball rolled in 2026/27, we let our supercomputer play the whole season — all 380 matches — one million times. Arsenal win every third season, the race for fourth place is wide open, and one of the promoted clubs refuses to behave like one. At the bottom it is brutally simple: the last three go down.
The title race · three clubs, three quarters of the titles
Arsenal win every third season
In 346,991 of one million simulated seasons, Arsenal finish as champions — 34.7%, clearly the most of any club. Manchester City follow at 23.2% and Liverpool at 15.4%. Together the trio take almost three titles in four. Then it thins out fast: Manchester United are fourth favourites at a modest 5.0% — closer to Aston Villa and Chelsea than to the top.
Share of 1,000,000 simulated seasons ending in the title. The top trio in red.
The forecast · all 20 clubs
The full table, position by position
| # | Team | 25/26 | Exp. pos | 80% CI | Titles | P(Title) | P(Top 4) | P(Rel) |
|---|---|---|---|---|---|---|---|---|
| 1 | Arsenal | 1 | 2.5 | 1–6 | 346,991 | 34.7% | 89.1% | 0% |
| 2 | Manchester City | 2 | 3.1 | 2–7 | 232,082 | 23.2% | 79.8% | <0.1% |
| 3 | Liverpool | 5 | 3.5 | 3–8 | 154,345 | 15.4% | 72.2% | 0% |
| 4 | Manchester United | 3 | 6.8 | 2–12 | 50,414 | 5.0% | 28.5% | 0.5% |
| 5 | Aston Villa | 4 | 8.1 | 2–14 | 41,875 | 4.2% | 22.2% | 2.5% |
| 6 | Chelsea | 10 | 7.6 | 3–12 | 34,975 | 3.5% | 21.9% | 0.9% |
| 7 | Coventry City | 18** | 9.2 | 4–16 | 28,915 | 2.9% | 16.1% | 5.1% |
| 8 | Brighton | 8 | 9.3 | 3–14 | 24,148 | 2.4% | 14.7% | 4.4% |
| 9 | Newcastle United | 12 | 10.3 | 3–16 | 23,446 | 2.3% | 13.0% | 8.8% |
| 10 | Ipswich Town | 19** | 10.0 | 4–15 | 20,985 | 2.1% | 12.4% | 6.8% |
| 11 | Fulham | 11 | 10.1 | 3–14 | 17,406 | 1.7% | 11.1% | 6.5% |
| 12 | Bournemouth | 6 | 10.1 | 5–16 | 16,573 | 1.7% | 10.8% | 6.3% |
| 13 | Brentford | 9 | 12.1 | 6–16 | 3,876 | 0.4% | 3.7% | 11.8% |
| 14 | Crystal Palace | 15 | 13.6 | 5–16 | 2,105 | 0.2% | 2.1% | 21.3% |
| 15 | Leeds United | 14 | 15.3 | 7–19 | 942 | <0.1% | 0.9% | 36.5% |
| 16 | Tottenham Hotspur | 17 | 14.5 | 10–19 | 548 | <0.1% | 0.8% | 26.1% |
| 17 | Nottingham Forest | 16 | 14.1 | 9–18 | 306 | <0.1% | 0.6% | 20.5% |
| 18 | Everton | 13 | 15.7 | 9–18 | 54 | <0.1% | 0.1% | 36.1% |
| 19 | Sunderland | 7 | 16.4 | 12–20 | 14 | <0.1% | <0.1% | 44.1% |
| 20 | Hull City | 20** | 17.7 | 13–20 | 0 | 0% | <0.1% | 61.7% |
25/26 = final league position last season. ** = clubs promoted from the Championship, which enter the simulation from positions 18–20. Exp. pos = average final position; 80% CI = in eight simulations out of ten the club finishes within that range. Titles = league titles out of 1,000,000 simulated seasons.
P(final position) per club
Where does everyone finish? The full distribution
One million final tables, compressed into a single picture. The top is dark and concentrated — Arsenal, City and Liverpool crowd the first few rows. The middle is smeared across ten positions. And at the bottom you can see why the relegation battle will run all season: only Hull have a distinctly dark right-hand edge.
| Team | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | 14 | 15 | 16 | 17 | 18 | 19 | 20 |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Arsenal | ||||||||||||||||||||
| Manchester City | ||||||||||||||||||||
| Liverpool | ||||||||||||||||||||
| Manchester United | ||||||||||||||||||||
| Aston Villa | ||||||||||||||||||||
| Chelsea | ||||||||||||||||||||
| Coventry City | ||||||||||||||||||||
| Brighton | ||||||||||||||||||||
| Newcastle United | ||||||||||||||||||||
| Ipswich Town | ||||||||||||||||||||
| Fulham | ||||||||||||||||||||
| Bournemouth | ||||||||||||||||||||
| Brentford | ||||||||||||||||||||
| Crystal Palace | ||||||||||||||||||||
| Leeds United | ||||||||||||||||||||
| Tottenham Hotspur | ||||||||||||||||||||
| Nottingham Forest | ||||||||||||||||||||
| Everton | ||||||||||||||||||||
| Sunderland | ||||||||||||||||||||
| Hull City |
Darker cell = the club finishes there more often. Positions 18–20 = relegation — the bottom three go down. Rows sorted by expected final position.
Top 4 · three places booked, one wide open
The race for fourth place
Arsenal (89%), City (80%) and Liverpool (72%) are close to locks for the top 4. The fourth spot is wide open: United at 28.5%, Villa at 22.2% and Chelsea at 21.9% — with Coventry and Brighton lurking behind them. Six clubs with a double-digit shot at a single place.
Near locks: Arsenal 89% · Manchester City 80% · Liverpool 72%
Top-4 probability for the chasing pack. Six clubs are effectively fighting over one place.
The scout report · a newcomer that refuses to act like one
Coventry are no ordinary promoted side
Promotion usually means survival mode. Not for Coventry: the supercomputer places them 7th on average, gives them a 16.1% shot at the top 4 — and only a 5.1% risk of going straight back down. Their Championship season was that dominant: 95 points and an attack in a class of its own. Even with a hefty margin for the step up, they profile as a top-half side. Ipswich look steady too, 10th on average. Hull are the classic newcomer story: the biggest relegation candidates of all.
Positions 18–20 · the bottom three go down
The relegation battle
Hull are relegated in 61.7% of the simulations — in a class of their own. Then comes the interesting group: Sunderland finished 7th last season but created the second-fewest chances in the league, and the supercomputer trusts the underlying performances more than the table — 44.1% relegation risk. Tottenham are punished for the same reason (26.1%). In between sit Leeds (36.5%) and Everton (36.1%), and not even Crystal Palace or Forest (both ~21%) can relax.
Track record · tested against the world's best
Should you trust a supercomputer?
Fair question. The model behind the simulation has competed several times in the international forecasting contests at Futbolmetrix, where forecasters from around the world are measured against each other — with strong results. In the World Cup contest it finished 5th, well ahead of the bookmaker consensus. But a million simulations still cannot see the future: injuries, late transfers and managerial changes can move the numbers. What the simulation gives you is a fair set of odds to measure the market against.
Further reading
Want to bet on the Premier League?
The simulation is a data-driven starting point for season-long markets — then find the best price and a welcome bonus at a licensed bookmaker.
18+ · Gamble responsibly · stodlinjen.se · Affiliate links · The forecast is probabilities, not a guarantee.
Frequently asked questions
Who wins the Premier League 2026/27?+
Arsenal are the favourites: they win the league in 34.7% of one million simulated seasons. Manchester City follow at 23.2% and Liverpool at 15.4%. Together the trio account for almost three quarters of all simulated titles — then there is a big gap down to Manchester United at 5.0%.
Who gets relegated from the Premier League 2026/27?+
The bottom three go down. Newly promoted Hull City are the biggest relegation candidates at 61.7%, followed by Sunderland (44.1%), Leeds (36.5%) and Everton (36.1%). Crystal Palace and Nottingham Forest (both around 21%) are not safe either.
Can Coventry reach the top 4 as a newly promoted side?+
In 16.1% of the simulations, yes — and they finish 7th on average, by far the best of the promoted clubs. Their Championship campaign was so dominant (95 points) that the supercomputer rates them as a top-half Premier League side. Their relegation risk is just 5.1%.
Why is the simulation so down on Sunderland?+
Sunderland finished 7th last season, but created the second-fewest chances in the entire league. The supercomputer looks at how teams actually play rather than where they happened to finish — and by that measure last season looks like an overperformance. Relegation risk: 44.1%. Tottenham are punished the same way at 26.1%.
How reliable is the model behind the simulation?+
The model has competed several times in the international forecasting contests at Futbolmetrix with strong results — most recently finishing 5th in the World Cup contest, well ahead of the bookmaker consensus. But a million simulations still cannot see the future: injuries, transfers and managerial changes can move the numbers.
Method: the supercomputer played the entire 2026/27 Premier League season — all 380 matches — one million times on 17 August 2026, before the opening round. Each club's strength is based on how the team actually played last season (the promoted clubs' numbers are adjusted for the step up from the Championship) plus squad value and age going into the new season. The simulation does not see individual transfers, injuries or managerial changes. Position 1 = the title, 1–4 = Champions League, 18–20 = relegation; the bottom three go down. The figures are exact shares of 1,000,000 simulations. Source: PlaymakerAI.