CMPN: Modeling and analysis of soccer teams using Complex Multiplex Passing Network

Abstract

Nowadays, coaches exploit data analysis in soccer (football) matches to plan their strategies against opponents. Network science, a subdomain of data analytics, is widely used to analyze soccer matches by treating players as nodes and passes between them as edges. However, single-layer methods for analyzing games overlook critical information by aggregating different types of passes into one layer. This paper introduces a new model called the Complex Multiplex Passing Network (CMPN) for analyzing team sports performance, with a focus on soccer matches. We utilized a real-world dataset to construct the multilayer structure of the CMPN. Each layer represents a specific type of pass between players. Using the CMPN, we conducted various analysis tasks at different topological scales. Firstly, we identified the core players of teams by calculating the PageRank versatility of each player. Next, we discovered the types of passes between trios of players based on multilayer motifs. Additionally, we measured similarities between passing tactics using the Pearson inter-layer assortativity measure. Finally, we employed a long short-term memory network to predict the outcomes of attacking plays using the CMPN model. The predictions achieved over 90% accuracy and approximately 70% F-measure. These findings offer practical value to coaches and performance analysts, as they enable appropriate planning by predicting playing styles in different competitions and neutralizing the strategies of opposing teams.