End of game shot selection for individual players in the NBA

Abstract

In the game of basketball, the state of the game is affected only when a field goal attempt succeeds. Most NBA athletes ritualize shooting practice in their training. The circumstances that a shot is attempted within affects its outcome. The aim of the work is to (i) create a classifier that will classify the outcome of a shot attempt given the circumstances it was attempted in, (ii) utilize the classifier to successfully infer the end of game shot selection parameters, (iii) deduction of a human-level performance for determining the quality of results and (iv) determination of whether the inclusion of additional training data will improve classification performance. The classifier-built results in 62.479% accuracy and 64.5% AUC. The end of game shot selection parameters such as the location to select and action to adopt for the shot attempt execution for a successful field goal are inferred.