How Match Predictions Work: The Maths Behind the Numbers

Every probability you see on Football Predictor comes from a model. It is not a gut feeling or a pundit’s hunch — it is a repeatable calculation. Here is how it works, explained without assuming you love statistics.

Step 1: Estimate expected goals

The model begins by asking a single question for each team: how many goals would we expect them to score in this match? This number — the expected goals, or xG — is built from three ingredients:

1. Attacking strength — how prolific the team is relative to an average side.
2. Opposition defensive strength — how mean the other team is at the back.
3. Home advantage — a boost for the home side, a small discount for the visitor.

Multiply a league-average goal rate by these factors and you get an expected-goals figure for each team. For example, a strong attack facing a leaky defence at home might be expected to score 2.1 goals; the visitors might be expected to manage 0.9.

Step 2: Turn expected goals into a range of scorelines

A team expected to score 2.1 goals will not score exactly 2.1 — goals are whole numbers. To handle this, we use the Poisson distribution, a well-established way of modelling how often discrete events (like goals) occur when you know the average rate.

The Poisson distribution takes the expected-goals figure and tells us the probability of 0 goals, 1 goal, 2 goals, 3 goals and so on:

If a team is expected to score 2.1 goals, the most likely single outcome is actually 2 goals (~27%), closely followed by 1 and 3 — and there is still a meaningful chance of 0 or 4+.

Step 3: Combine both teams into a score matrix

We calculate this distribution for both teams, then combine them into a grid of every plausible scoreline — 0-0, 1-0, 2-1, and so on — with a probability attached to each. Summing the right cells gives us:

• Home win probability — every cell where the home team scores more.
• Draw probability — the diagonal where scores are equal.
• Away win probability — every cell where the away team scores more.

The single cell with the highest probability becomes the predicted score.

Step 4: Express confidence honestly

Finally, the model reports a confidence score. This reflects how dominant the favoured outcome is. A 75%-vs-15%-vs-10% match is high confidence; a 38%-vs-33%-vs-29% match is a genuine coin-toss and is reported as such. Confidence is never 100% — football does not work that way.

Why this approach?

The Poisson goal model is the backbone of countless professional football forecasts because it is transparent, fast and well-calibrated. It will not catch every red card or moment of magic, but across hundreds of matches its probabilities hold up remarkably well.

Want to see it in action? Head to the Match Predictor and try any fixture, or read Understanding Expected Goals to go deeper on the key input.