League · England
Premier League 2026-27 statistics
The season has just started: 40 matches played so far. The percentages and standings below cannot settle on so few matches — use the season picker above for last season's full table.
2026-27 season
- Matches played
- 40
- Goals per match
- 2.85
- Home win
- 33%
- Draw
- 35%
- Away win
- 33%
- 3+ goals
- 53%
- Both teams scored
- 53%
Table
| # | Team | P | W | D | L | GF | GA | GD | Pts |
|---|---|---|---|---|---|---|---|---|---|
| 1 | 4 | 4 | 0 | 0 | 8 | 1 | +7 | 12 | |
| 2 | 4 | 4 | 0 | 0 | 8 | 2 | +6 | 12 | |
| 3 | 4 | 2 | 2 | 0 | 7 | 3 | +4 | 8 | |
| 4 | 4 | 2 | 2 | 0 | 5 | 2 | +3 | 8 | |
| 5 | 4 | 2 | 1 | 1 | 13 | 5 | +8 | 7 | |
| 6 | 4 | 2 | 1 | 1 | 10 | 9 | +1 | 7 | |
| 7 | 4 | 1 | 3 | 0 | 7 | 4 | +3 | 6 | |
| 8 | 4 | 1 | 3 | 0 | 6 | 4 | +2 | 6 | |
| 9 | 4 | 1 | 3 | 0 | 5 | 3 | +2 | 6 | |
| 10 | 4 | 2 | 0 | 2 | 7 | 10 | -3 | 6 | |
| 11 | 4 | 1 | 2 | 1 | 4 | 4 | 0 | 5 | |
| 12 | 4 | 1 | 2 | 1 | 7 | 8 | -1 | 5 | |
| 13 | 4 | 1 | 1 | 2 | 7 | 7 | 0 | 4 | |
| 14 | 4 | 1 | 1 | 2 | 3 | 5 | -2 | 4 | |
| 15 | 4 | 0 | 3 | 1 | 6 | 7 | -1 | 3 | |
| 16 | 4 | 1 | 0 | 3 | 6 | 11 | -5 | 3 | |
| 17 | 4 | 0 | 2 | 2 | 0 | 5 | -5 | 2 | |
| 18 | 4 | 0 | 1 | 3 | 4 | 7 | -3 | 1 | |
| 19 | 4 | 0 | 1 | 3 | 1 | 7 | -6 | 1 | |
| 20 | 4 | 0 | 0 | 4 | 0 | 10 | -10 | 0 |
The table is computed from match results; matches decided by forfeit count with the score declared by the federation. Points deducted or awarded separately are not reflected. Only regular-season matches count — play-offs and relegation matches are excluded.
Teams in this league
- Arsenal
- Aston Villa
- Bournemouth
- Brentford
- Brighton
- Chelsea
- Coventry
- Crystal Palace
- Everton
- Fulham
- Hull City
- Ipswich
- Leeds
- Liverpool
- Manchester City
- Manchester United
- Newcastle
- Nottingham Forest
- Sunderland
- Tottenham
Every team in this season's fixture list. Unlike the standings table, teams that have not played yet also appear here.
Season simulation
The remaining fixtures were played out 10,000 times using the model's current probabilities. Percentages below are how often each team finished in that position.
| Team | Title | Top 4 | Bottom 3 | Pts (avg) |
|---|---|---|---|---|
| Arsenal | 61.6% | 98.4% | 0.0% | 78.3 |
| Manchester City | 33.8% | 95.4% | 0.0% | 74.6 |
| Liverpool | 1.5% | 39.1% | 0.4% | 59.7 |
| Hull City | 0.7% | 29.1% | 0.7% | 57.8 |
| Manchester United | 0.6% | 22.8% | 1.4% | 56.3 |
| Brentford | 0.4% | 21.1% | 1.6% | 55.7 |
| Bournemouth | 0.3% | 17.1% | 2.2% | 54.7 |
| Aston Villa | 0.3% | 17.4% | 2.4% | 54.6 |
| Brighton | 0.2% | 12.8% | 3.8% | 53.1 |
| Newcastle | 0.1% | 10.9% | 4.6% | 52.4 |
| Chelsea | 0.2% | 10.2% | 4.7% | 52.0 |
| Leeds | 0.1% | 8.2% | 5.1% | 51.5 |
| Everton | 0.1% | 6.5% | 6.4% | 50.6 |
| Nottingham Forest | 0.1% | 5.8% | 8.3% | 49.6 |
| Fulham | 0.0% | 3.2% | 13.0% | 47.6 |
| Sunderland | 0.0% | 0.7% | 32.1% | 42.6 |
| Coventry | 0.0% | 0.5% | 39.0% | 41.3 |
| Crystal Palace | 0.0% | 0.4% | 40.5% | 40.9 |
| Tottenham | 0.0% | 0.3% | 47.7% | 39.7 |
| Ipswich | 0.0% | 0.0% | 86.0% | 32.0 |
Method: ratings are frozen as of the run date; ties break by points-GD-goals, country-specific rules not applied. Pairings below the publishing threshold use the model's prior rating. The table refreshes whenever the script is re-run.
Generated: 2026-08-23 · 3 · 10,000
The model's record in this league
The model's claim is not “I know every match” — it is “I know where I'm confident”. Measured in this league:
| Model at least | Predictions | Realised |
|---|---|---|
| ≥ 50% | 820 | 64% |
| ≥ 60% | 456 | 70% |
| ≥ 70% | 217 | 81% |
| Model at least | Predictions | Realised |
|---|---|---|
| ≥ 75% | 1609 | 82% |
| ≥ 80% | 891 | 86% |
| ≥ 85% | 450 | 90% |
| Model at least | Predictions | Realised |
|---|---|---|
| ≥ 50% | 2447 | 59% |
| ≥ 55% | 1772 | 61% |
| ≥ 60% | 972 | 65% |
And if it had to answer every match?
- Model
- 55%
- Always pick home
- 44%
- Picking at random
- 33%
- Reachable ceiling
- 76%
Why that ceiling is not 100%: 24% of matches in this league end in a draw, and a draw is almost never the single most likely outcome on its own. So one “accuracy” figure cannot be read on its own — it is read next to these anchors.
Matches measured: 1440. The thresholds were not cherry-picked — all of them are shown. Incorrect predictions are included and none are hidden.
Full track record →