How the Poisson Distribution Is Used in Football Betting

If you have ever wondered how a betting model turns “Arsenal are better than Everton” into an actual number like “62.4% chance of a home win”, the answer for football is usually the Poisson distribution. It is the standard starting point for goal-based modelling, and understanding it tells you a lot about what these models can and cannot do.

What the Poisson distribution actually is

Poisson is a probability distribution that describes how often a rare, independent event occurs in a fixed window of time. Given an average rate, it tells you the probability of seeing exactly 0, 1, 2, 3 or more occurrences.

Football goals fit this reasonably well. Goals are relatively rare (around 2.7 per match across most top leagues), they can occur at any point in the 90 minutes, and — to a first approximation — one goal does not determine the next. The formula is:

where λ (lambda) is the expected number of goals for that team, and k is the specific number you want the probability of. Every spreadsheet has this built in as POISSON.DIST.

Step 1: Calculate attack and defence strength

The model needs a λ for each team. You get there by measuring how much better or worse each team is than the league average.

Say the league averages 1.55 home goals and 1.20 away goals per match. If our home team has scored an average of 2.0 goals in home matches, their attack strength is 2.0 ÷ 1.55 = 1.29 — they score 29% more than an average home side. If the away team concedes 1.4 per away match against a league average of 1.55, their defensive strength is 1.4 ÷ 1.55 = 0.90.

Step 2: Calculate expected goals for the match

Repeat for the away side using their attack strength and the home team’s defensive record, and you might get λ away = 1.05. You now have the two numbers the whole model rests on.

Step 3: Build the scoreline matrix

Run Poisson on each λ to get the probability of each team scoring 0, 1, 2, 3... goals. Multiply the two together for every combination and you have the probability of every scoreline.

Sum all the scorelines where the home team scores more for the home win probability. Sum the equal scorelines for the draw. Sum every combination totalling 3 or more goals for Over 2.5. One matrix gives you a price for every main market, and you can then compare each to the bookmaker using the value betting formula.

Where basic Poisson breaks down

A naive Poisson model is a starting point, not a finished product. Its known weaknesses are well documented:

• Goals are not fully independent. Teams change behaviour after scoring — a side going 1–0 up often drops deeper. Poisson assumes the rate stays constant, which it does not.
• It underestimates draws and low scores. Real football produces more 0–0 and 1–1 results than plain Poisson predicts. The classic fix is the Dixon–Coles adjustment, which corrects the probabilities of low-scoring scorelines specifically.
• Raw goals are a noisy input. A team that won 3–0 on two deflections and a penalty is not as good as the scoreline says. Using expected goals (xG) instead of actual goals gives a far more stable measure of underlying performance.
• It ignores context. Injuries, red cards, fixture congestion, motivation in dead rubbers and weather all matter and appear nowhere in the base formula.
• Old matches are weighted the same as recent ones unless you apply time decay so that recent form counts for more.

From textbook model to usable model

The gap between a spreadsheet Poisson model and something that beats the closing line is mostly these refinements: xG inputs rather than raw goals, Dixon–Coles low-score correction, time decay weighting, explicit handling of team news, and machine learning layers that capture patterns a fixed formula misses.

This is roughly the pipeline we run at BetsToday — you can read a summary of the data sources and modelling approach on our research page.

This article is educational content only and is not financial or gambling advice. Betting involves risk and you can lose money. See our responsible gambling page for support resources.