Role revolution: towards a new meaning of positions in basketball

Abstract

Most team sports are characterized by positions to which the players in a team are assigned. The goal of this classification is to attribute specific responsibilities during a game. Moreover, the same classification drives the buying and selling of players according to team managers and coaches strategies. The existing positions - often defined a long time ago - tend to reflect traditional points of view about the game and sometimes they are no longer well-suited to the new concepts arisen with the evolution of the way of playing.

This paper focuses on basketball and aims at describing new roles of players during the game, by means of the analysis of players' performance statistics with data mining and machine learning tools. In detail, self-organizing maps and fuzzy clustering procedures are adopted in tandem to define groups of players with similar way of playing. The results show that, when considering the modern basketball players' statistics, classical positions are not able to fully represent their way of playing, and a new set of 5 roles emerges as a meaningful classification of players' characteristics.