Integration of Machine Learning and Nomogram Models for Predicting Game Outcomes in the Chinese Basketball Association

The research focused on the technical and tactics aspects linked with team success in professional basketball, examining data across 596 matches in the 2019-2020 season of the Chinese Basketball Association (CBA). Key performance indicators (KPIs) were identified through lasso and logistic regression, which revealed that steals (STL), defensive rebounds (DREB), true shooting percentage (TS%), offensive rebounding percentage (OREB%), and screen-off made (SOM) were positively correlated with winning outcomes, while turnovers (TOV) had a negative correlation. Consequently, a logistic regression model was developed to predict the outcomes of the game, with an overall prediction accuracy of 85.6%. Hence, interpretation of the odds ratios showed that with increased steals, there was a 41.7% (±28.1%) increase in the chances of winning; with every added defensive rebound, there was an increase in points of 36.5% (±18.6%); and with every block pulled off, 30.3% (±48.4%) more winning chances. To further clarify the presentation of results, a nomogram graph has been constructed. This graph has taken a column-line shape to clearly reveal the mixed effect of the selected KPIs on the likelihood of winning. These results underline how important actions for retrieving the ball are and the successful use of the off-ball screen offense in relation to the shaping of game results. This study shows ways to enhance team efficiency and introduces one broadly applicable method to understand how technical and tactical KPIs interact in achieving success in professional basketball.