Poisson Distribution
A statistical model for predicting the number of goals in a football match.
The Poisson distribution models the probability of a given number of independent events occurring in a fixed interval. In football, goals are approximately Poisson-distributed — each goal is roughly independent and rare.
How OddsIntel uses it: 1. Estimate expected goals (xG) for each team using historical form, ELO, H2H, and league-specific goal rates 2. Apply the Poisson probability mass function to each possible scoreline 3. Sum scoreline probabilities to get P(home win), P(draw), P(away win), P(over 2.5), etc.
Dixon-Coles correction: Standard Poisson slightly underestimates 0-0 and 1-0/0-1 draws. Our model applies the Dixon-Coles (1997) correction to adjust low-score probabilities.
The Poisson model is blended 50/50 with an XGBoost model trained on 96,000 historical matches.
Common Questions
Is Poisson modelling accurate?
It's a strong baseline. Poisson correctly assumes goals are rare and roughly independent, but misses correlations between goals (e.g. a late goal changes game state). The Dixon-Coles correction and XGBoost blend partially address this.