Dunkerque v Metz — Prediction, Odds & Stats
Friday 23 October 2026 · FRA-Ligue 2 · 2627
AI prediction
ai-blend v1.0 dailyCalibrated: the model's number corrected by how often calls at that confidence have landed, out of sample (5,276 settled matches in the thinnest band), then normalised. The smaller figure is the raw model output.
Top 5 scorelines
- 1-1 12.0%
- 2-1 8.5%
- 1-2 8.3%
- 1-0 7.7%
- 0-1 7.5%
Goals
| Over 2.5 | 55.1% | Under 2.5 | 44.9% |
| BTTS yes | 58.8% | BTTS no | 41.2% |
| Dunkerque 1+ | 76.8% | Dunkerque 2+ | 42.9% |
| Metz 1+ | 76.0% | Metz 2+ | 41.7% |
Expected goals 1.46 – 1.43. Every market above is read off one scoreline distribution.
Footing
Positive factors
- Dunkerque carry the stronger rating, 1434 Elo against 1384.
Risk factors
- No specific risk flags beyond the usual variance of a football match.
Why the model thinks this
The price
Dunkerque have an estimated 36.7% win probability, against 36.7% for Metz and 26.7% for the draw. Expected goals from the ensemble are 1.46 for Dunkerque and 1.43 for Metz.
Why Dunkerque
Dunkerque carry the stronger rating, 1434 Elo against 1384.
Why the draw is live
The draw sits at 26.7%. The sides are closely matched on expected goals, 1.46 to 1.43. Draws run at 28% in this competition.
Why Metz can win
Nothing in the data singles out Metz.
Key players
Metz's leading contributors by expected goal involvements a match: Gauthier Hein (0.40), Habib Diallo (0.26), Giorgi Kvilitaia (0.25).
Availability
No lineup on record for Dunkerque, so availability is unknown. 1 of Metz's top three contributors did not start their last match. Metz changed goalkeeper for their last match.
The matchup, as far as the data reaches
Dunkerque: 56% possession, 14.8 shots a match, 13.4 faced, 0.30 xG a match from set pieces. Metz: 49% possession, 14.2 shots a match, 13.8 faced, 0.36 xG a match from set pieces.
Inputs last refreshed: matches 2026-09-05 22:20 · team_stats 2026-09-05 09:41 · availability 2026-09-05 21:03 · prediction 2026-09-05 18:39 ·
Dunkerque are the model's most likely outcome at 36.7%. The likeliest scoreline is 1-1 at 12.0%. Confidence 73/100, data quality 65/100. These are probabilities, not certainties: the less likely outcome happens as often as its number says.
Written from the numbers the model saw, not by a language model; nothing here is a statistic the snapshot does not hold. How predictions work.
Was this pick made in advance?
-oos is out of sample: it was fitted only on matches
played before this one, so its pick is a genuine forecast rather than hindsight.
Anything without that suffix was fitted on all available data.What the chances say
Metz are conceding sharply more than the chances against them merit — 4.4 goals above expected, which tends to correct.
Finishing swings far more than chance creation, so a gap this size usually narrows. Over ten matches it is suggestive, not settled — a genuinely elite finisher beats expected goals for years.
What the model said
| Model | Home / Draw / Away | Expected goals | Likeliest score | O2.5 | BTTS |
|---|---|---|---|---|---|
| dc-xg | 1.27–1.66 | 1-1 11% | 56% | 59% |
A model whose name ends -oos was produced by the walk-forward backtest:
it was fitted only on matches played before this one, so the pick is genuinely out of sample.