Bayesian modelling of tactical patterns in football

Abstract

The main goal of this dissertation is to investigate the Bayesian modelling techniques applied to probabilistic graphical models in order to recognise tactical patterns in football. We also study how the said techniques relate to Markov chains and Markov process theory and how a Collapsed Gibbs Sampler can be used to estimate the required distributions for the drawn tactics. Using real spatially informative data from Manchester United games over the 2011/2012 premier league season, we apply the studied Collapsed Gibbs Sampler and discuss the results from both a statistical perspective through accuracy tests and convergence diagnostics and also the philosophical perspective where we consider a drawn tactic and discuss its footballing interpretation.