Match Forecast
Probabilities built on evidence,
not guesswork.
Every fixture priced by our own model — a machine-learning ensemble over team strength, form, xG, availability, match context and the market. Win, draw and away, plus Over 2.5 and both-teams-to-score. Our number, our call, scored in public against what actually happened.
257
Fixtures ahead
257
Priced · rolling window
1X2 · O2.5 · BTTS
Markets per priced fixture
Probabilities from our own model — estimates, not certainties, and sometimes wrong.
Sunday, 20 September 2026
17 matches
Probabilistic projection, not advice. Forecasts are informational only and can be wrong. Football outcomes are inherently uncertain. Terms
How it works
Each fixture is priced by a weighted ensemble led by our neural network (trained on 40,000+ matches), blended with a Bayesian bivariate-Poisson goals model, falling back to a Dixon-Coles team-strength model where a fixture lacks the inputs. Features span team strength and Elo, recent form and xG, match facts and pressure, squad availability, referee, travel and rest, weather and team news. Where market prices exist we fold them in as a 15% prior, then calibrate the final probabilities against our own settled results. Football is high-variance: these are probabilities, not certainties, and plenty of them will miss. Read the full methodology →
