Mauro Florez
  • Home
  • Projects
  • Teaching
  • Gallery

On this page

  • Overview
  • How It Works
  • Features

epl-lpp: EPL Match Prediction

A Poisson regression model for predicting English Premier League match scores, learning team attack strength and defensive weakness from historical match data.
Published

February 1, 2026

Overview

A Poisson regression model for predicting English Premier League match scores. The model learns team attack strength and defensive weakness from historical match data.

How It Works

The model uses a Poisson GLM with the following structure:

Home Goals ~ HomeTeam + AwayTeam + Positions + Rolling Stats
Away Goals ~ AwayTeam + HomeTeam + Positions + Rolling Stats
  • HomeTeam effect → Attack strength of the home team
  • AwayTeam effect → Defensive weakness when facing that opponent
  • Positions → Current league standings
  • Rolling Stats → Goals, shots, shots on target (last 5 games)

Features

  • Rolling averages for goals scored/conceded
  • Rolling averages for shots and shots on target
  • Dynamic league position calculation
  • Trained on all current season games

View Project Website