THE USE OF CLASSIFICATION AND REGRESSION TREE WHEN CLASSIFYING WINNING AND LOSING BASKETBALL TEAMS

The aim of the present study was to identify the best predictors when classifying winning and losing teams in basketball in consideration of situational variables by the classification and regression tree (CRT) non-parametric analysis. The sample was composed of 1,404 balanced games (score-differences: 1-14 points) from the Spanish EBA Basketball League that presented high heterogeneity and a non-parametric distribution. These games were split into faster-and slower-paced games according to ball possessions per game (using a cluster k-means). The CRT analysis was used to predict which game-related variable/ s better classified winning and losing teams during slower-and faster-paced games. In total, this approach explained 72% of the total variance in the slower-and 69.3% in the faster-paced games. The results identified importance of defensive rebounds (100%), successful free-throws (94.7%), assists (86.1%), and fouls committed (55.9%) for the classification of winning and losing teams in the fast-paced games. Conversely, in the slow-paced games the better classification of winning or losing teams was accomplished by the following variables: successful free-throws (100%), defensive rebounds (82.3%), fouls committed (68.4%), assists (66.9%), successful 2point (62.2%) and 3-point field-goals (61.6%). The influence of situational variables was identified only for team quality in the slow-paced games. The present findings allow coaches for a better control of games and competition.