Using massive spatial temporal data to analyse NBA match performance under different possession outcomes

This is the authors' abstract. We don't add key points for this paper.

This study aimed to explore the ball, player, and team spatial temporal patterns that distinguish different basketball ball possession outcomes. Event, ball and player tracking data from 632 NBA games during the 2015–2016 season were gathered and provided by the league’s official SportVU tracking system. Afterwards, 4464 ball possessions meeting the following criteria were extracted from the raw data: (i) complete set play sequence without fouls or out-of-bounds events; and (ii) possessions ending with a jump shot or lay-up. A descriptive analysis of the spatial-temporal and contextual variables was carried out to identify univariate differences in variables between the made and missed shot outcomes, and a follow-up analysis was performed to assess the variations in key variables across distinct teams. The results revealed significant differences between distinct shot outcomes when the spatial temporal patterns of the shooting player, which included variables such as the shot distance, defensive angle, and defensive player distance, were considered. The spatial-temporal variables at the team level and the ball-related kinematic variables had less of an influence on successful shots. A further comparison between teams of different levels revealed that they had similar characteristics in terms of the significant variables in the above analysis, but different teams showed distinct styles considering their spatial-temporal performance. The overall approach carried out in this study may provide a novel perspective for analysing basketball performance on the basis of the integration of event and tracking data, enabling a better understanding of player and team dynamics during set play.