The last great club match of the season is in Budapest tonight: PSG vs Arsenal, the Champions League final, 16:00 UTC at the Puskás Aréna. Here's what our model sees in it — and, just as importantly, what it doesn't claim to.
The headline
On our club-football baseline — a Dixon-Coles joint-Poisson model fit over ClubElo ratings — Arsenal are clear favourites:
- Arsenal win: 54%
- Draw: 25%
- PSG win: 21%
The gap traces straight back to the ratings. ClubElo has Arsenal at 2063 and PSG at 1964 (snapshot 25 May) — a near-100-point advantage, which the model converts into an expected-goals split of 1.73 to 0.98 in Arsenal's favour.
The scoreline picture
A single-goal game is the model's base case. The most likely exact scores:
| Score (PSG–ARS) | Probability |
|---|---|
| 1–1 | 12% |
| 0–1 | 11% |
| 0–2 | 10% |
| 1–2 | 10% |
| 0–0 | 7% |
| 1–0 | 6% |
The most likely single result is a 1-1 draw, but notice how the list tilts: four of the top six involve Arsenal scoring more than PSG. Stack the margins and the tilt is unmistakable —
- Arsenal by 1+: 54%, by 2+: 30%, by 3+: 14%
- PSG by 1+: 21%, by 2+: just 7%
A heavy Arsenal win is more than four times as likely as a heavy PSG one.
Goals, both ends, and tempo
- Both teams to score: 52% — close to a coin flip.
- Clean sheets: Arsenal keep one 38% of the time; PSG only 18%.
- Total goals: over 1.5 is 76%, over 2.5 is 51%, over 3.5 is 29%.
- Tempo: the model expects an early goal — a 36% chance the opening goal arrives inside the first 15 minutes, and only a 7% chance of a goalless 90.
Put together, it's the shape of a final where Arsenal are favoured to control the chances and PSG need an early, against-the-grain goal to bend the game their way.
What this is — and isn't
Two honest caveats:
These are club numbers, not World Cup numbers. This forecast comes from our club-football fit, which is separate from the international ensemble behind our World Cup pages. ClubElo doesn't transfer cleanly to national teams, so don't read tonight's figures as a statement about any of these players' countries — that's a different model, and a different post.
The model is frozen before kickoff and stays frozen. The probabilities above do not re-price during the match. What the live page refreshes through the 90 minutes is descriptive — score, shots, expected goals, possession — not an updated forecast. The pre-match number is the model's one statement, made once.
The live match page is at /ucl-2026-final/, with the full scoreline grid, totals ladder and in-match breakdown. Methodology — the exact Dixon-Coles formulas over ClubElo — is in the UCL baseline research note. Numbers pin the snapshot at the date above.