This paper explores the application of the Temporal Graph Network (TGN) model to predict the receiver and outcome of a pass in soccer. We construct two TGN models that estimate receiver selection probabilities (RSP) and receiver prediction probabilities (RPP) to predict the intended and actual receivers of a given pass attempt, respectively. Then, based on these RSP and RPP, we compute the success probability (CPSP) of each passing option that the pass is successfully sent to the intended receiver as well as the overall pass success probability (OPSP) of a given situation. The proposed framework provides deeper insights into the context around passes in soccer by quantifying the tendency of passers` choice of passing options, difficulties of the options, and the overall difficulty of a given passing situation at once.
© Copyright 2024 Machine Learning and Data Mining for Sports Analytics. MLSA 2023. Communications in Computer and Information Science. Published by Springer. All rights reserved.