Premier League simulator 26/27
We take the real table after 10 rounds, play the 280 matches that are left 10,000 times, and count how each season ended. What comes back is not one prediction but a range: Arsenal are champions in 79% of those runs, and 4 clubs win the league in at least one season in a hundred.
Every club across 10,000 simulated seasons
The bar is the 80% band: the points range the club finished in on eight runs out of ten. The dot is its average.
| Club | Now | Points range (80% of runs) | Finish | Title | Top 4 | Down |
|---|---|---|---|---|---|---|
| Arsenal | 27 | 76–92 pts · avg 84 | 1–2 | 79% | 100% | – |
| Man City | 21 | 64–82 pts · avg 73 | 1–4 | 10% | 91% | – |
| Brighton | 22 | 64–81 pts · avg 73 | 2–4 | 10% | 91% | – |
| Liverpool | 18 | 56–73 pts · avg 64 | 3–8 | 1% | 53% | – |
| Chelsea | 16 | 49–67 pts · avg 58 | 4–11 | – | 17% | – |
| Bournemouth | 15 | 48–66 pts · avg 57 | 4–12 | – | 15% | – |
| Brentford | 14 | 48–65 pts · avg 57 | 4–12 | – | 14% | – |
| Aston Villa | 13 | 46–63 pts · avg 54 | 5–13 | – | 8% | 1% |
| Everton | 16 | 44–61 pts · avg 53 | 5–14 | – | 5% | 1% |
| Leeds | 14 | 43–60 pts · avg 52 | 6–15 | – | 4% | 2% |
| Newcastle | 12 | 39–56 pts · avg 48 | 7–17 | – | 1% | 6% |
| Man United | 8 | 39–56 pts · avg 47 | 7–17 | – | 1% | 8% |
| Hull | 11 | 38–56 pts · avg 47 | 8–17 | – | 1% | 7% |
| Nott'm Forest | 9 | 38–55 pts · avg 46 | 8–17 | – | – | 9% |
| Tottenham | 12 | 36–52 pts · avg 44 | 10–18 | – | – | 16% |
| Fulham | 10 | 35–52 pts · avg 43 | 10–18 | – | – | 19% |
| Sunderland | 8 | 31–48 pts · avg 39 | 12–20 | – | – | 41% |
| Ipswich | 11 | 29–45 pts · avg 37 | 14–20 | – | – | 57% |
| Coventry | 9 | 28–44 pts · avg 36 | 14–20 | – | – | 64% |
| Crystal Palace | 7 | 27–43 pts · avg 35 | 15–20 | – | – | 68% |
Sorted by average finishing position. “Finish” is the 80% band of final places. Ties are split by goal difference inside every simulated season. Rebuilt several times a day; results to 2026-11-08.
How the Premier League simulator works
1. Rate every club
A Dixon-Coles goal model is fitted to two years of results. Each club gets an attack and a defence rating, and the league gets a home advantage. Nothing is hand-tuned: the ratings are whatever fits the results best.
2. Play the match
For a fixture those ratings give an expected goals figure for each side, and the actual score is drawn at random around it. The same fixture can finish 0-0 once and 4-1 the next time, which is the point.
3. Play the season 10,000 times
All 280 remaining matches are played out, the points are added to the real table, and the final standings are recorded. Then it starts again, 10,000 times over.
4. Count, don't guess
A club's title chance is simply the share of those seasons in which it finished first. The points range is where it landed in the middle 80% of them. No opinion enters at any step.
What it does not know: injuries, suspensions, a new manager, a January window or a cup run. It sees results and nothing else, which is why the honest output is a range rather than a scoreline — and why the bands above are as wide as they are in November.
Premier League simulator: quick answers
How does the Premier League simulator work?+
It fits a Dixon-Coles goal model to two years of results, which gives every club an attack and a defence rating plus a league-wide home advantage. Each of the 280 remaining matches is then played out by drawing a score at random around the expected goals those ratings imply, the points are added to the real table, and the whole season is replayed 10,000 times. Every number on this page is a count of what happened across those runs.
Who does the simulator say will win the Premier League?+
Arsenal came out on top in 79% of the 10,000 simulated seasons, finishing on about 84 points (76–92 in the middle 80% of runs). 4 clubs win the league in at least 1% of them, so it is a long way from settled.
How accurate is a football season simulation?+
The useful question is whether the percentages are calibrated - whether the things given a 30% chance happen about 30% of the time. We publish our model's calibration from a walk-forward test over 10,930 matches rather than asking you to take the simulator on trust. A simulation is not a forecast of one season; it is a map of how many different ways the season can still go.
What is the points range?+
It is the 80% band: the club finished inside that points range in eight of every ten simulated seasons, with one in ten above it and one in ten below. A wide band means the season is still open for that club; a narrow one means most paths end in the same place.
Can I make my own version?+
Yes. The table predictor lets you pick the winner of every remaining match yourself and watch the table rebuild around your results, which is the hand-made version of what the simulator does 10,000 times.
Is it free?+
Yes. The simulator, the tables and the match predictions are free and need no account.
