An Analysis of Formation Disruption in Soccer

Abstract

Modern soccer tactics are largely designed with the objective of maintaining control of space on the pitch. In order to maximize their command of space, a team will deploy their players in a specific arrangement known as a formation. However, certain situations may cause a defending team to temporarily shift away from their defensive formation, and in this disrupted state they may be more vulnerable to attack. In this paper, we use tracking and event data from matches in an elite European soccer league to develop an analysis of formation changes as well as of defensive disruption. We begin by creating our own implementation of a technique developed by Bialkowski et al. (2014) to discover formations from tracking data. Based on the results from this implementation, we develop a new method of identifying statistically significant changes in a team’s formation using comparisons to a bootstrapped distribution of a 2-Wasserstein distance metric. In addition, we also compute a chi-squared statistic that quantifies the defensive disruption of a team at any time in a match, and we characterize possible implications of trends in this statistic. Our work in this paper can serve as a first step towards a better understanding of when formations change as well as when teams become defensively disrupted.