Ball possession dynamics in the game of Football: From data analysis to modeling

Abstract

Collective dynamics studies of team sports such as football is a hard task involving large amounts of data and analysis. In this contribution we aim to address a particular subject of this broad field such as the dynamics of the ball possession intervals. To this end we have analyzed a novel football database comprising one season of the five major leagues of Europe. First we obtained the key elements of ball possession dynamics from a statistical analysis of the database. Using this input we developed a simple stochastic model based on three agents, two teammates, and one defender. This model includes four parameters and can capture the main emergent statistical observables of the ball possession intervals in the database: the distribution of (i) possession time, (ii) the number of passes performed, and (iii) the ball distance traveled on passes. In the last part, we show that the dynamics of the model, can be mapped into a Wiener process with a drift and an absorbing barrier.