Data Mining and Polar Coordinates in the Analysis by Gender of Finishing Behaviors in Professional Basketball Pick and Roll

Abstract

The open nature of basketball gives it a large uncertainty that makes hard the tactical analysis of the situations that happen in the game. Specifically, screens are one of the offensive tactical elements most used in basketball and one example of a tactical situation that needs the highest preparation level to get a good performance in the competition. The aim of this study is to differentiate these player behaviors by gender using data mining and polar coordinates analysis. Therefore, one ad hoc observational tool made by 17 criteria and 97 exhaustive and mutually exclusive (E/ME) categories has been designed and validated using the data quality analysis (correlation coefficients and concordance index 0.98) and generalizability analysis (G coefficients 0.94) to perform such a study. The observational design is nomothetic, punctual, and multidimensional. A total of 176 ball screens situations have been analyzed for the men's category and 118 for women's category, corresponding to three different teams of each gender playing in the highest competition level in Spain during the 2018/2019 season using Hoisan software tool. The analysis of the relationships among behaviors has been performed using Polar Coordinates analysis as well as data mining analysis: clustering and decision tree classifier. Results show significant relationships that allow us to tactically interpret the pick and roll situations in men's and women's professional basketball players in Spain, allowing us to develop more intervention programs which will optimize training and improve players performance.