Testing the Validity of Markov Chains in Analyzing Soccer Gameplay through Barcelona’s 2020-2021 La Liga Season

Abstract

Statistical modeling has been applied to most or all modern-day sports in order to analyze gameplay and offer teams an upperhand in games. Though games such as baseball have had mass changes due to Sabermetrics, the empirical analysis of baseball, there still remain numerous components of soccer where mathematics can be more deeply integrated. Though current literature and research like Sanyal’s “Who will receive the ball? Predicting recipients in soccer videos” do provide some form of analysis, many of these methods have limitations due to perspective or robustness. To counter these limitations, Markov Chains were implemented. As they disregard the players as a factor and can take in theoretically infinite amounts of data, Markov Chains seem to be a superior method to the previous ones suggested by current literature. By applying the data from Barcelona’s 2020-2021 La Liga Season to generate a transition matrix, it was seen that Markov Chains were effectively able to deduce patterns in soccer gameplay. However, as it only provides a transition matrix, giving the probabilities that the ball travels from one section of the field to the next, it does not provide a direction countering strategy to an opponent but rather the basis for which a strategy can be created.