UTABIRI
Search teams/ 18+
Back to Match Analysis Model ยท 8 min read

How our predictor model matches the Poisson model

The mathematical foundations of our prediction engine and how it builds on the traditional Poisson approach.

The Foundation of Football Prediction

The Poisson distribution has long been a standard tool for predicting football matches. Football is a low-scoring game where goals arrive fairly randomly over 90 minutes, which fits the mathematical definition of a Poisson process well.

Moving Beyond Basic Poisson

A basic Poisson model uses season-long average goals scored and conceded. UTABIRI's model goes further: it rates each team's attack and defence across whole competitions, adjusts for the strength of every opponent, and gives recent matches more weight. If a team has too little history to rate, it falls back to form averages or an AI-generated expected-goals estimate.

Bivariate Adjustments

One known flaw of the standard Poisson model is that it under-predicts low-scoring draws, specifically 0-0 and 1-1. Our model applies the Dixon-Coles adjustment to raise the probability of those low-scoring results, which brings it closer to how football actually behaves.

Responsible useNo statistical model can predict sports outcomes with certainty. Individual match variance remains high, so treat every number here as information, not a promise.