Bayesian modelling of the English Premier League

Abstract

The process of modelling football data using statistical methods has become increasingly popular nowadays. The aim is to fit a Poisson log-linear and a hierarchical model using Bayesian modelling with special reference made to Markov Chain Monte Carlo methods. These models assess the teams' performance with respect to their defence and attack parameters. Furthermore, statistical tests are used to compare the efficiency of the forecasts of the proposed models against those of the bookmakers focusing, on the fixed odds betting market. Illustrations using real data from the 2011-2012 English Premiership are considered.