Event Based Coupled Dynamics in Soccer: Attack-defense Interaction Networks

In Soccer two teams are constantly interacting between each other, changing and adapting continuously accordingly to the rival’s actions and environmental changes. As such, soccer is an excellent example of a complex adaptative system which, thanks to the recent growth of available data, allows to make novel approaches to study and describe its dynamics.

One of such approaches is that of complex networks, which has already been used mainly to study passing networks and to characterize player and team-centered dynamics. Nonetheless, in most of the literature, the analysis and description takes into account each team separately and the intertwined interaction dynamics of the two teams has just started to be explored recently. In this work we attempt to include the inherent coupled dynamics of the two teams involved in a soccer match, and to describe it on the basis of the notion of events.

Unlike other team sports as Football and Baseball in which the game develops in discrete sequences of events or turns, in Soccer the game develops in a continuous fashion and is only interrupted by certain events. When these particular events occur, the possession of the ball switches from team to team and a new sequence of events is started. Nonetheless, changes of ball possession are not necessarily triggered by this specific set of events and events such as interceptions or unsuccessful passing attempts can switch the team in possession of the ball without interrupting the flow of the game. We will refer to the set of events that can trigger a change of possession of the ball as inter-team events and to those which don’t involve the loss of control of the ball as intra-team events.
During inter-team events, teams adapt and modify the players’ interactions to respond to the current situation at each moment of the game, constantly switching from an attacking to a defensive behavior and vice versa. We consider that these events are the binding elements between team dynamics and the building blocks of the description of the game we propose.

Based on these observations, using a publicly available dataset from different european leagues and tournaments, we build mono and multi-layer networks of events (Fig. 1) which summarize the teams dynamics and behavior occurring during a match. By a statistical analysis of their topology we identify recurring patterns of sequences of events which more often lead to desired outcomes such as the occurrence of goals. This approach can also be used as a coarse grain assessing tool of team performance, as it is able to identify sequences of actions which lead to both desired and undesired game configurations.

We consider that event-based descriptions of the coupled dynamics of soccer matches like the one we propose, can shed light for the development of a holistic description of the game which takes into account its collective nature.

Συνεδρία: 
Authors: 
Martin Zumaya, Nelson Fernández and Carlos Gershenson
Room: 
3
Date: 
Thursday, December 10, 2020 - 18:00 to 18:15

Partners

Twitter

Facebook

Contact

For information please contact :
ccs2020conf@gmail.com