This study examined the fourth quarters in the close games in the regular NBA games in the last decade, ranging from the 2013-14 season to the 2022-2023 season. A close game is categorically defined by a scenario where the point differential is confined within a 10-point margin at the onset of the fourth quarter and narrows further to a 5-point disparity by the end of the game. In total, 2,295 close games were identified in this study. Advanced game statistics, including offensive rate, defensive rate, assistance ratio, pace of game, and true shooting percentage, etc., are obtained from the NBA box scores using a python script. Understanding key factors that determine the outcome of the basketball games is critical, as such can be used to develop predictive models for coaches to design game strategies. This study developed a Bayesian Logistic Modeling approach to estimate the winning probability of a basketball team in the fourth quarter, using the pace of the last quarter and a team's shooting percentage. The accuracy of the model is used to evaluate if the model can correctly classify game outcome based on the identified game statistics in the fourth quarter of a game. The binary outcome of the close game is modeled as a Bernoulli distribution. Results reveal that the True Positive Rate and False Positive Rate is 0.93 and 0.07, respectively. Insights from this study can be used to help design coaching strategies in basketball games, illuminating potential tactical pivots that could tilt the game in their favor.
Keywords: Bayesian logistic regression; basketball games; game performance; pace of basketball games; statistical methods.
Copyright © 2024 Wang, Zheng and Li.