Скачать или смотреть Video Poster - Conference on Statistics and Data Science (CSDS) 2022 - João Vítor Rocha da Silva

Different faces of defense: Studying the National Basketball Association’s (NBA) defensive positions.

Basketball’s classical positions have always been well defined, from an offensive standpoint. It is well known what role a point guard or a center performs, in theory, offensively. Although, many times, the offensive side of the game is neglected by teams, fans, and even players, past research by Bianchi et al. (2017) and Silva & Rodrigues (2022) show that definitions of the basketball’s classical positions do not represent the complexity and modernity of today’s gameplay style anymore.
Consequently, these definitions not only fail to correctly represent the athletes’ offensive roles in court, but also (and more importantly) defensively. Hence, this study aims to analyze the game-related defensive statistics of all active players on the 2020-21 NBA season, grouping players by their common characteristics, and creating exclusively defensive labels, which added to the classical definitions translate in a simple and objective manner the defensive role that players are performing.
To achieve this objective, an analysis on all average statistics of each active NBA player was conducted via Self-Organizing Maps and cluster analysis (K-Means Algorithm), that allowed us to identify similarities and differences between players of same and different classical positions. From that, we were able to create four new defensive-exclusive labels: (i) the rim proctectors; (ii) the perimeter defenders; (iii) the multi-position defenders; and (iv) the functional defenders. These labels represent roles that can be performed by any player on court, and combined with the classical definitions, can provide a more accurate and complete definition of a player’s role in court.