Spatio-Temporal Cohesive Networks for Evaluating Team Behavior in Soccer

Thesis Type: Postgraduate

Institution Of The Thesis: Middle East Technical University, Graduate School of Natural and Applied Sciences, Turkey

Approval Date: 2019

Thesis Language: English


Consultant: Hande Alemdar


In this thesis, we analyze positional organizations of soccer teams during game transitions which end with an important result and individual sprinting performances of soccer players. Social networks and distance matrices of teams are used to obtain organizations of teams. Spatial features of pitch such as pitch value and pass probability value are used to evaluate sprint performances. Social networks that we call cohesion matrices are used as weights in both attacking and defending transitions. The norm of the weighted distance matrices forms team spread values. Cohesion matrices show player to player interactions and connections between clusters of teams on the pitch. The team spread values are used to characterize the behaviors of teams in a transition. The average team spread values show that top teams are more expansed while attacking and more tighter while defending. Moreover, the average team spread values confirms that teams are wider while attacking except when a transition ends with losing the possession of the ball. These results characterize organizations of teams in the Turkish Super League and the effects of individual players on those organizations. Sprint analysis results show that full-back and winger players have higher sprint value averages while midfielders have less. Also, teams that are focused on having the possession of the ball have less average sprint value than teams playing in counter-attack style.