Categorization of Rugby Union Players by Performance Characteristics Using Principal Component Analysis and Cluster Analysis

Makoto Kiuchi, Hirofumi Maehana and Nobuyoshi Hirotsu

[Received October 4, 2019 ; Accepted August 4, 2020] 

The aim of this study was to categorize rugby players by performance characteristics irrespective of positions using principal component analysis (PCA) and cluster analysis (CA). Data were drawn from the Japan Rugby Top League 2015–2016 season; a sample of 231 players and 16 items were used in the analysis. We used PCA to identify the players’ ability and CA to categorize players’ performance characteristics, regardless of position. PCA reduced the 16 items to four principal components: penetration, defense and competition, ball handling and kicking, and turnovers. We categorized players into six clusters based on the four component scores: playmaker, ball carrier, tackler and supporter, competitor, passer and kicker, and spoiler. Although player categorization by CA was achieved regardless of position, the players’ performance characteristics also influenced categorization, and specific positions have performance characteristics. Unique players were identified for some positions. In addition, we clarified a difference in performance characteristics between Japanese and foreign players in the league. Since the categorization of players by performance characteristics irrespective of position enables teams to find necessary players, this study may be advantageous for team management and may contribute to more useful analysis of players and clarifying performance characteristics.

Keywords: game data, performance characteristics, player categorization, principal component analysis, cluster analysis

[Football Science Vol.17, 86-97, 2020]