Soccer, the world’s most popular sport, captivates millions with its dynamic gameplay and strategic depth. Among the various metrics used to analyze performance, shot data provides valuable insights into both team efficiency and individual contributions. This talk presents spatial multidimensional item response theory (IRT) models with homologous factors to analyze ten seasons of English Premier League data. The multidimensional structure jointly models the latent shooting ability of the attacking team and the latent defensive shot-stopping ability of the opposing team, extending traditional paired-comparison frameworks. By incorporating horizontal and vertical reflectional symmetries of the soccer field, we quantify spatial variation in shot success probabilities. We assess the effects of formations, home-field advantage, seasonal trends, and match-time intervals, while distinguishing between open-play and set-piece situations as well as left- versus right-footed shots. Our findings provide deeper insights into team strategies and spatial patterns in shot success.

This is a joint work with Professor Debashis Mondal.