Modelling home advantage in sport: A new approach

Abstract

Home advantage (HA) is well documented in a wide range of team sports, and the numerous factors that may contribute to it necessitate robust multivariate modelling techniques in order to establish their independent effects. As performance measures such as goals scored by each team in a single sporting match are not independent, bivariate distributions are increasingly being used to model HA. An alternative method is proposed whereby repeated measures regression analysis using Generalised Estimating Equations is used to account for within-match correlation. Using European Europa League football as an example, repeated measures analysis is used to estimate HA and its variance in terms of the home/away ratio for goals scored, and a simple formula provided for expressing this in terms of the percentage of goals scored by home teams, with standard errors. Methods of estimating covariate adjusted HA at different levels of a predictor variable, such as crowd size, are also described. All STATA (2009) commands used to run the analyses are provided. Although the proposed modelling strategy is most appropriate for sports where match outcomes are determined by discreet events of equal value (e.g. goals scored in football), generalisations of this approach to continuous and binary measures of team performance are also described.