Football Match Predictor
Pick a fixture, see the model's read on it — win probabilities, expected goals, goals and BTTS markets. Built on a Dixon-Coles Poisson model trained on 301 recent results.
Quick note on the maths: recent form is weighted heavily, thin samples get pulled toward the league average, opponent strength is baked in, and the whole thing lands on a Dixon-Coles Poisson grid (ρ = −0.12). Probabilities, not promises.
Get next matchweek's predictions in your inbox
One email per matchweek: the model's standout calls across the Premier League, La Liga, Serie A, Bundesliga, Ligue 1 and the Champions League. No spam, unsubscribe any time.
How it actually arrives at a number
Walk through a real one from the list above. Before the Champions League final, the model looked at PSG and Arsenal the way it looks at everyone: their last ten games, with the recent stuff turned up loud and anything older fading on a 40-day half-life — a thumping back in February barely registers, this month's results carry the weight. It keeps each side's home and away records apart, because plenty of teams are a different animal on the road, and it docks marks for soft schedules so nobody gets credit for beating up the bottom three.
Out of that comes one number per team: expected goals. The model had PSG around 1.7 and Arsenal around 1.3 — close, with PSG nudged ahead. Then it does the part people skip: it spreads those expected goals across every scoreline from 0-0 to 7-7 on an 8×8 grid, with a small Dixon-Coles correction so it doesn't fall in love with 1-0s and 0-0s the way a naive Poisson does. Tot up the cells where the home side wins and that's your home probability; tot up everything with three goals or more and that's Over 2.5. Same grid, every market. The final itself? It finished 5-4 — a reminder that the grid gives you the lean, not the lottery numbers.
Predictions by competition
World Cup 2026 Predictions
Live nowModel-driven win, draw & goals projections for every group-stage and knockout fixture across the USA, Canada & Mexico — refreshed each matchday.
See World Cup predictionsDomestic Leagues
Cups & Continental
What this model can tell you, and what it can't
Here's the honest pitch. The model is good at exactly one job: taking the goals a team has scored and shipped lately and turning that into a fair read on the next match. What it has no idea about is the stuff that decides plenty of games — a manager resting his front three before a European tie, a keeper pulling up in the warm-up, a derby that throws the form book out the window. Use the numbers as your anchor, then layer the team-news on top yourself.
The expected-goals line is the bit I'd actually look at. Say the model hands the home team 1.9 and the visitors 0.8. That is not a prediction of 2-0. It's shorthand for "play this match a hundred times and the goals roughly land here" — and most of those hundred run-throughs finish on some other scoreline entirely. That spread is why you'll rarely see a clean 70% favourite outside the odd mismatch. Football is messy, and any tool quoting you 90% certainty is flattering you.
That home-advantage slider — why it's there
Home advantage is not one fixed number. It runs higher in the Bundesliga than the Premier League, higher at some grounds than others, and across Europe it has quietly shrunk over the last ten years. The slider hands you the controls: crank it up for a genuine fortress, drag it to zero for a neutral-venue final or a one-legged tie, and watch what the probabilities do. Nine times out of ten you'll be surprised how little they budge — which tells you something useful about how much home advantage is really worth.
The off-season replay
Right now there's nothing to predict — the leagues are done until August — so the tool flips into replay mode. Take that PSG 5-4 Arsenal Champions League final sitting at the top of the list: the model only ever saw the form both sides carried into the match, made its call, and you get to see how close it landed. No quietly dropping the games it got wrong. Just the read against what actually happened, and you're free to think it called it badly.