Poisson Distribution in Football Predictions

When you see a probability like "Over 2.5 Goals: 62%" on ProSoccer, it didn't come from gut feeling or historical win/loss counts alone. It came from the Poisson distribution — a mathematical model that describes how often a random event (like a goal) occurs in a fixed time window.

What Is the Poisson Distribution?

The Poisson distribution answers one question: if an event happens on average λ times, what is the probability it happens exactly k times?

Where e ≈ 2.718, λ is the expected number of events (expected goals), and k is the number of events we're predicting (actual goals scored).

Step 1 — Estimate Expected Goals (λ) for Each Team

Before applying Poisson, we need the expected number of goals each team will score. We start with league averages:

Then we calculate each team's attack and defence strength:

• Man City Attack Strength (home) = City home goals scored per game ÷ league home avg = 2.1 ÷ 1.51 = 1.39
• Arsenal Defence Strength (away) = Arsenal away goals conceded per game ÷ league away avg = 0.9 ÷ 1.23 = 0.73

Expected goals for Man City at home vs Arsenal away:

Step 2 — Build a Scoreline Probability Matrix

With λ_home = 1.53 and λ_away = 1.10, we calculate P(k) for each team scoring 0, 1, 2, 3… goals:

Each scoreline probability = P(City scores X) × P(Arsenal scores Y). For example:

Step 3 — Derive Market Probabilities

By summing scoreline probabilities we get market odds:

• Home Win (Man City): sum of all P(City > Arsenal) = ~49%
• Draw: sum of all P(City = Arsenal) = ~24%
• Away Win (Arsenal): sum of all P(Arsenal > City) = ~27%
• Over 2.5 Goals: sum of all P(total goals ≥ 3) = ~58%
• BTTS Yes: sum of all P(City ≥ 1 AND Arsenal ≥ 1) = ~52%

Why Poisson Works for Football

Goals in football are rare, independent events — a goal at minute 10 doesn't make a goal at minute 60 more or less likely. This independence is the core assumption of the Poisson model, and football satisfies it reasonably well. Studies using decades of match data from the Premier League, La Liga, Bundesliga and Serie A show Poisson models produce well-calibrated probabilities, especially for Over/Under and correct score markets.

Limitations & How We Improve It

Pure Poisson ignores in-match context: red cards, injuries, weather, tactical shifts. At ProSoccer, our model layers xG data and live Pressure Index on top of the Poisson base — adjusting λ in real time as the match progresses. This is why our live predictions update every 15 seconds rather than being static pre-match numbers.