Identification of Soccer Defensive Transition Play Using Latent Class Analysis

Hiroki Matsuoka, Kozue Ando and Takahiko Nishijima

[Received February 2, 2021; Accepted July 14, 2021] 

The purpose of this study was to identify latent subgroups with respect to defensive transition plays from an achievement pattern data of the defensive transition in the professional soccer game. The research procedure comprised two steps, 1) an analysis of achievement criteria of the defensive transition items by a decision tree analysis, 2) an analysis of latent classes of defensive transitions by latent class analysis (LCA). The 158 defensive transitions in two games of the J-league in 2016 were used. An achievement pattern data of nine defensive transition items consisted of approach, press, and DFL control were used for LCA. Three latent classes, class 1: High achieved transition play (HA), class 3: Moderate achieved transition play (MA), and class 2: Low achieved transition play (LA) were identified. HA (54.25±5.35) showed significantly higher defensive transition skill (DTS) score compared with MA (42.61±5.10) and LA (27.85±3.24), and MA showed significantly higher defensive transition score than LA (p < 0.05). In addition, the relationship between LA and the opponent PA entry was found. These results indicated that LA allowed the opponent counter attack leading to failure of defensive transition play. In conclusion, this study successfully identifies latent classes of defensive transition plays by LCA.

Keywords: tracking data, classification of play, criterion-referenced evaluation

[Football Science Vol.18, 60-71, 2021]