Cooperative passing network features are associated with successful match outcomes in the Australian Football League

Abstract

Collective behaviour is an important component of team performance in team sports. This study used a binomial generalised linear mixed effects regression model to investigate the relationship between cooperative passing network characteristics and match outcomes of professional Australian Football League competition games across four seasons between 2016 and 2019. It divided a sample of 1629 observations into a training and testing partition used to develop and assess the validity of the model used in this study, respectively. The results of this study reveal that a team's connectedness is associated with the probability of winning Australian Football League games (Akaike Information Criterion = 1637.3, residual df= 1297, deviance = 1625.3). When most players within a team are involved in the team's passing network bidirectionally (i.e. a well-connected network; odds ratio = 1.053; 95% confidence interval: 4.2–6.5%, p < 0.001), teams have a higher probability of winning. The centralisation of a team's passing network was not significantly related to match outcomes. The classification accuracy for the model associating network characteristics with match outcomes was 69%. Collectively, these findings suggest that Australian Football League-specific network features should be incorporated within existing performance analysis methods and can provide a useful, practical tool for coaches to measure collective performance during team practice.