Ioannis Ntzoufras

Publications Page

On modelling Soccer Data 

Karlis, D. and Ntzoufras, I. (2000).  Student, 3, 229-244.

Abstract

        This paper examines the plausibility of Poisson regression models for the interpretation and prediction of association football (soccer) scores. The paper is divided into two parts; the first part consists of an explanatory analysis  on the two basic assumptions of modelling soccer data:  Poissonity and independence, while the second presents a practical implementation using data from 1997-98 season league results. In the second part of the paper, a simple Poisson model formulation is presented and an application on data from English Premier Division  and `Italian Serie A' of 1997-98 season is provided. The proposed model setting allows for  comparison of different models with different natural interpretation.  The model with constant home effect over all the teams  is selected by the AIC criterion. Negative binomial models are also considered but  the improvement is minor.

Keywords: negative binomial distribution; English Premier Division; log-linear models; model selection; overdispersion.

[Download scanned version of the paper, 1,75MB]

All Contents Copyright.
Last revised: 15-11-2001