Network science in sports

Abstract

Network science has become a central methodological framework for the study of complex systems, with applications across social, biological, and technological domains. Its strength lies in formalizing such systems as collections of interacting units, whose collective behavior depends not only on the properties of the individual components, but also on the structure, dynamics, and evolution of their interactions. In this Review, we analyze how this framework, rooted in statistical physics, is being applied to the study of sports systems. Sports provide a particularly suitable domain for network modeling, as performance emerges from interactions among agents, whose actions are constrained by spatial embedding and temporal ordering, and evolve under competitive, adaptive, and out-of-equilibrium conditions. We review the main network representations used in sports, from classical graph-based descriptions to more general frameworks, including spatial, temporal, multilayer, and higher-order networks. Particular attention is devoted to soccer, where network science has had its deepest methodological impact, while applications to other sports such as basketball, volleyball, American football, or badminton serve to illustrate the generality of the underlying principles. Network science provides sports analytics with a rigorous language for quantifying coordination, emergence, adaptation, and collective performance. At the same time, the availability of high-resolution event, tracking, and pose-estimation data makes sports an ideal testbed for validating and refining network models in realistic settings. As a result, the relationship between network science and sports, currently driven primarily by the application of network methods to sport, has the potential to become genuinely bidirectional. In this Review, we not only organize the literature on network methodologies and their applications to sports, but also discuss complementary concepts needed to capture the full picture, including entropy and complexity measures, as well as key ideas from statistical physics and nonlinear dynamics. Finally, despite the substantial progress made so far, the field remains under active development, and we highlight the main open questions that should be addressed in future research.