Short Course – Statistical Modelling of Football Data

The course is aimed at giving the background of modern statistical football modelling, starting from the seminal works focused on the double Poisson distributions for the number of the goals until the most recent research lines oriented towards the use of additional covariates, tracking data and machine learning methods. 

The course covers many statistical modelling strategies and requires a satisfactory knowledge of R.

The instructors will provide both theoretical material and R code to reproduce some practical examples. Although the flavor of the course is mainly Bayesian, ‘classical’ statisticians can anyway enjoy the course’s contents in a perfect way.

Registration fee for the short course: May 31, 202 – 120 euro

Lecture 1

  • Current topics in Football Analytics  
  • Double Poisson Models, Prediction, and League re-generation
  • Using covariates and additional info
  • Application to English Premier League data

Lecture 2

  • Bayesian Models for Prediction Using the footbayes R package
    Model checking via PP checks
  • Dynamic models, attacking/defence abilities
  • Prediction probabilities and model comparisons

Lecture 3

  • Prediction with Advanced Models
  • Introduction to in-play analytics
  • Bradley-Terry and ordered probit models
  • Overdispersion and correlation in Poisson-based models
  • Draw inflation
  • Copula models
  • Tracking data

NOTE: All participants should have R installed in their computers. Some implementations will be possibly in OpenBUGS, Jags or STAN and footbayes package in R (Discover more here)