How Poisson Models Predict Soccer Scores
Goals in soccer are rare, roughly independent events — which makes them a natural fit for the Poisson distribution. With just two numbers, each team's expected goals, a Poisson model produces a probability for every possible scoreline, and from those, win/draw/loss and over/under figures.
Step 1 — estimate expected goals (λ)
A simple approach uses attack and defence strength relative to league average:
- Home attack strength = home team's home goals per game ÷ league average home goals.
- Away defence strength = away team's goals conceded away ÷ league average away goals conceded.
- Home λ = home attack × away defence × league average home goals.
Repeat in reverse for the away side. Modern analysts often replace raw goals with xG (expected goals from shot quality), which is more stable over small samples.
Step 2 — scoreline probabilities
The probability a team scores exactly k goals is e−λ λk ÷ k!. Multiply the home and away probabilities for each pair to get a grid of scorelines — 0–0, 1–0, 1–1 and so on.
Step 3 — aggregate
Sum cells where home goals exceed away goals for a home win, the diagonal for a draw, and the rest for an away win. Sum cells where total goals exceed 2 for "over 2.5". Our Poisson predictor does all of this live.
Known weaknesses
- Draws are underestimated. Low scores are slightly correlated (game states, tactics). The Dixon–Coles adjustment (1997) corrects low-score cells.
- Team strength changes. Weighting recent matches more heavily (time decay) helps.
- Context is missing. Injuries, rotation, motivation and weather are not in the raw numbers.
Why it is still useful
Poisson models are transparent, fast and surprisingly competitive as a baseline. They are a superb learning tool: every assumption is visible, so you can see exactly why a model disagrees with consensus. Our daily soccer predictions on the Predictions page use a Poisson-style approach built from current league tables.
Keep learning
Next: Elo Ratings Explained: From Chess to the NBA
Get one 3-minute lesson every Sunday in the free Edge Report.
Get the Edge Report