Play-by-play data analysis for team managing in basketball

The sports analytics literature regarding basketball is vast but the analyses based on disaggregated data, such as the play-by-play match data, are not very common. The analysis of the whole sequence of play-by-play match events has an undeveloped potential, yet most of the available methods focus on the final match results. The present work illustrates a model-based strategy for the analysis of the match progress, built upon the literature of Adjusted Plus Minus for the evaluation of player efficiency. This approach is extended in two main directions. The first extension consists in the adoption of a response variable which considers the most relevant events in the game, and not only the number of scored points. This offers some useful advantages, including the possibility of obtaining separate estimates about different complementary aspects. Further, next to player efficiency effects, the efficiency of five-man lineups is estimated. The model fitting procedure follows an empirical Bayes approach, which provides a suitable regularization. For the empirical analysis, we consider a dataset regarding the Italian Basketball League (Serie A1), focusing on the matches of the first round of the current championship 2018/2019. The dataset collects the play-by-play information along with the matches box scores, which are made available by the league website (www.legabasket.it). The results of the analysis could support the decision-making process of team management, and some illustrations on this point are provided.