Machine Learning Modeling to Provide Assistance to Basketball Coaches

This study proposes the development of decision-support systems, based on neural networks and k-Nearest Neighbors (kNNs) techniques. The goal is to evaluate, for each substitution during a match, which group of players on the court, known as lineup, has a higher probability of outperforming their opponent. The developed model was tested using a database obtained from Brazilian Basketball League (NBB) matches, involving players statistics, match details and different contexts. The model achieved an accuracy of 76,99% in projections of superiority between the playing lineups, demonstrating the potential of using computational intelligence methods in decision-making applied to professional sports. Finally, the study highlights the importance of using such tools in conjunction with human experience, encouraging future research for the development of even more sophisticated and effective models for decision-making in the sports field.