Predicting Team Performance with Spatial Temporal Graph Convolutional Networks

TL;DR

This paper introduces Spatial Temporal Graph Convolutional Networks (ST-GCN) for predicting team performance from agent movement data, demonstrating superior accuracy in sports analytics applications.

Contribution

The paper proposes a novel ST-GCN architecture combining graph convolutions and Gated Recurrent Units for spatiotemporal team performance prediction.

Findings

ST-GCN outperforms other classification methods in game score prediction

The architecture effectively captures spatial relationships among team members

Ablative analysis highlights the importance of each architectural component

Abstract

This paper presents a new approach for predicting team performance from the behavioral traces of a set of agents. This spatiotemporal forecasting problem is very relevant to sports analytics challenges such as coaching and opponent modeling. We demonstrate that our proposed model, Spatial Temporal Graph Convolutional Networks (ST-GCN), outperforms other classification techniques at predicting game score from a short segment of player movement and game features. Our proposed architecture uses a graph convolutional network to capture the spatial relationships between team members and Gated Recurrent Units to analyze dynamic motion information. An ablative evaluation was performed to demonstrate the contributions of different aspects of our architecture.