Automatically Recognizing On-Ball Screens

Download the
Full Paper Here
Authors

Armand McQueen
Jenna  Wiens
John Guttag

Abstract

Abstract: The pick  and roll is a powerful tool; as former coach Stan Van Gundy once said of his  Magic team, "[The pick and roll is] what we're going to be in when the  game's on the line. [...] I don't care how good you are, you can't take away  everything" [1]. In today's perimeter oriented NBA, the pick and roll is  more important than ever before. The player tracking data that is now being  collected across all arenas in the NBA holds out the promise of deepening our  understanding of offensive strategies. In this paper we approach part of that  problem by introducing a pattern recognition framework for identifying on-  ball screens. We use a machine learning classifier on top of a rule-based  algorithm to recognize on-ball screens. Tested on 21 quarters from 14 NBA  games from last season our algorithm achieved a sensitivity of 82% and  positive predictive value of 80%