A1158
Title: Understanding football receiver routes through functional data clustering
Authors: Tianyu Guan - York University (Canada) [presenting]
Abstract: Running routes by receivers are central to offensive strategy in football because they shape how teams create space, advance the ball, and design passing plays. Functional principal component analysis (FPCA) is proposed to cluster these routes by treating each trajectory as a smooth functional curve. FPCA extends classical PCA to functional data and provides a low-dimensional representation of the trajectories. By projecting routes onto a small number of functional principal components (FPCs), the FPC scores that represent each route are obtained. Clustering methods such as K-means can then be applied to these FPC scores, and each cluster corresponds to a common route type run by receivers. These route clusters provide insight into the strategies teams use to advance the ball on each play. From a football analytics perspective, they are interpretable and useful for downstream tasks. For example, the FPC scores for each cluster can be used as predictors in a regression model to predict whether a route is successfully completed. Using a group lasso approach, with each group corresponding to a cluster, allows one to assess the contribution of different route types to completion outcomes.
