Analysis of Long-Term Player Action Prediction Performance Based on Causal Modelling in Rugby League

Jiaxuan Wang, Ruigeng Wang, Shahrokh Heidari, David Arturo Soriano Valdez, Mitchell Rogers, Gaël Gendron, Yani He, Nicolas Mir, Yalu Zou, Riki Mitchel, Alfonso Gastelum Strozzi, Marcel Noronha, Michael Witbrock, Patrice Delmas

Abstract: Predicting athletes’ behavior is important in designing game strategies and enhancing audience engagement in sports analytics. While machine learning has enabled advancements in action prediction across various sports, most existing models primarily capture correlations rather than underlying causal relationships. Here, we study long-term player action prediction in team sports (e.g., Rugby League) by integrating causal behavior modeling with Graph Neural Networks (GNNs). We assessed the model’s performance over extended video sequences, evaluating its resilience to varying observation window sizes. Our results demonstrate that incorporating causal structures significantly improves long-range action prediction accuracy, with the transformer-based graph neural network (TransformerConv) exhibiting superior robustness.

External IDs:dblp:conf/acivs/WangWHVRGHMZMSNWD25