Identification and Optimization of High-Performance Passing Networks in Football

TL;DR

This paper develops a method to analyze and optimize football passing networks by examining topological metrics, enabling tactical improvements and addressing the blocked player problem to enhance team performance.

Contribution

It introduces a novel framework combining temporal network analysis and logistic regression to identify, classify, and optimize passing network structures for better football team performance.

Findings

Significant topological differences between high and low performance moments.

A logistic regression model effectively classifies performance states.

Optimized passing networks improve robustness against contextual changes.

Abstract

This study explores the relationship between the performance of a football team and the topological parameters of temporal passing networks. To achieve this, we propose a method to identify moments of high and low team performance based on the analysis of match events. This approach enables the construction of sets of temporal passing networks associated with each performance context. By analyzing topological metrics such as clustering, eigenvector centrality, and betweenness across both sets, significant structural differences were identified between moments of high and low performance. These differences reflect changes in the interaction dynamics among players and, consequently, in the team's playing system. Subsequently, a logistic regression model was employed to classify high- and low-performance networks. The analysis of the model coefficients identified which metrics need to be adjusted to promote the emergence of structures associated with better performance. This framework provides quantitative tools to guide tactical decisions and optimize playing dynamics. Finally, the proposed method was applied to address the ``blocked player" problem, optimizing passing relationships to minimize the emergence of structures associated with low performance, thereby ensuring more robust dynamics against contextual changes.