Arsenal v Chelsea
Sun, Sep 6, 11:30 AM ET · Matchweek 3 · Premier League
Not validated out of sample. 26 Premier League matches have been graded under this model — far too few to support any accuracy claim. No win rate or accuracy figure is quoted, and this model has not been validated out of sample. These are the model's own probability distributions, published so you can see what it says — not picks, not advice, and not compared against any price.
Full time, 90 minutes
The draw is a real outcome and is never folded away.
Two of three outcomes
Each is a sum of the three probabilities above, so it cannot disagree with them.
The ten likeliest scorelines
These ten account for 69.4% of the distribution — the remaining 30.6% is spread across every other scoreline.
| Score (Arsenal–Chelsea) | Probability | |
|---|---|---|
| 2–1 | 9.8% | |
| 2–0 | 9.8% | |
| 1–1 | 8.8% | |
| 1–0 | 8.8% | |
| 3–1 | 7.3% | |
| 3–0 | 7.3% | |
| 2–2 | 4.9% | |
| 1–2 | 4.4% | |
| 4–1 | 4.1% | |
| 4–0 | 4.1% |
Total goals in the match
Expected 3.23. The middle half of the distribution falls between 2 and 4 goals.
Probability of each total (goals)
Over / under ladder
| Line | Over | Under |
|---|---|---|
| 0.5 | 96.1% | 3.9% |
| 1.5 | 83.3% | 16.7% |
| 2.5 | 62.7% | 37.3% |
| 3.5 | 40.5% | 59.5% |
| 4.5 | 22.5% | 77.5% |
Half lines only — a match cannot land exactly on one, so over and under sum to 100% with no push.
How many each side scores
The marginal of the same grid — not a separate fit, so these add back to the total above.
Goalscoring shape and margin
Every figure is an exact sum over the same score grid.
Winning margin (Arsenal goals − Chelsea goals)
Positive means Arsenal win by that many, 0 is the draw, negative means Chelsea. Margins beyond ±4 carry the remaining mass and are not drawn.
Likeliest scorers
Read off the posted lineup: each figure is for the state the player is actually in, starting or off the bench.
| Player | Club | Appearances | To score | A shot on target |
|---|---|---|---|---|
| Kai Havertz | Arsenal | 88 | 30.5% | 56.6% |
| Bukayo Saka | Arsenal | 117 | 28.8% | 59.2% |
| Martin Ødegaard | Arsenal | 114 | 20.7% | 46.1% |
| Cole Palmer | Chelsea | 91 | 18.9% | 65% |
| João Pedro | Chelsea | 73 | 16.9% | 57.8% |
| Declan Rice | Arsenal | 141 | 13.3% | 31.4% |
| Christos Tzolis | Arsenal | none | 13% | 32.4% |
| Viktor Gyökeres | Arsenal | 10 | 11.4% | 18.3% |
| Morgan Rogers | Chelsea | 82 | 10.7% | 46.5% |
| Mikel Merino | Arsenal | 23 | 10.7% | 23.1% |
| Myles Lewis-Skelly | Arsenal | 20 | 10.2% | 18.3% |
| Gabriel Magalhães | Arsenal | 130 | 9.5% | 20.4% |
Showing the 12 likeliest of 40. No injury or suspension feed exists here, so a player who is unavailable can still appear in this list. Shots on target is a separate model from the scorer figure beside it, fitted and tested on its own — the two are not derived from one another and can disagree.
Where these numbers come from
Per-club attack and defence rates are fitted to Premier League results only, then combined into an exact Poisson score matrix for this fixture. Every probability on this page is a closed-form sum over that one grid — there is no sampling here, so there is no run count to quote and the figures are identical for every visitor, every time.
The fit takes an explicit cutoff at the run clock, so it cannot see a result from a match it is forecasting. Team-level only: no lineup, injury or team-news input exists in this model, by design.
- Model
- epl-model-v1-split-poisson
- Scoring rates (λ)
- Arsenal 2.23 · Chelsea 1.00
- Generated
- 2026-09-06T22:51:07Z
- Validation
- NOT_VALIDATED_OUT_OF_SAMPLE
Paper-only and educational. Not betting advice. All Premier League fixtures