Big data analytics for modeling scoring probability in basketball: The effect of shooting under high-pressure conditions

In this paper, we analyze the shooting performance of basketball players by examining the factors that may generate high-pressure game situations. Using play-by-play data from the Italian “Serie A2” Championship 2015/2016 to build the model, we validate the main results using data from the Olympic Basketball Tournament “Rio 2016” to determine whether the relationships we identified can be confirmed using data from players at a very different professional level. After a preliminary exploratory analysis, we (1) develop a multivariate model based on the Classification and Regression Tree algorithm in order to investigate how selected high-pressure situations, jointly considered, affect scoring probability and then propose new shooting performance measures; (2) investigate players’ personal reactions to selected high-pressure game situations by introducing additional new measures, improving the indices currently used to measure shooting performance. The results are interesting and easy to interpret with the aid of some insightful graphical representations. Our approach can be exploited by both scouts and coaches to understand important player characteristics and, ultimately, to measure and enhance a team’s performance.