SoccerMix: Representing Soccer Actions with Mixture Models

Abstract

Analyzing playing style is a recurring task within soccer analytics that plays a crucial role in club activities such as player scouting and match preparation. It involves identifying and summarizing prototypical behaviors of teams and players that reoccur both within and across matches. Current techniques for analyzing playing style are often hindered by the sparsity of event stream data (i.e., the same player rarely performs the same action in the same location more than once). This paper proposes SoccerMix, a soft clustering technique based on mixture models that enables a novel probabilistic representation for soccer actions. SoccerMix overcomes the sparsity of event stream data by probabilistically grouping together similar actions in a data-driven manner. We show empirically how SoccerMix can capture the playing style of both teams and players and present an alternative view of a team's style that focuses not on the team's own actions, but rather on how the team forces its opponents to deviate from their usual playing style.

Identifying Soccer Teams’ Styles of Play: A Scoping and Critical Review

“…Narizuka and Yamazaki [ 57 ] adopted the Delaunay network to get (as adjacency matrix A(t)) the formation of a team at time t. Then, using hierarchical clustering, they obtained not only the average formation (i.e., “442”, “4141”, “433”, “541” or “343”) for each team in the match but also the positional exchange of players within the match formations. Decroos, Roy, and Davis [ 58 ] used mixture models to achieve a representation of soccer actions. In the first stage, for each action type, a mixture model was fitted to the locations (x, y).…”

“…This has practical value, as opposed to simply grouping things together. The vectors of the actions which were created in the research of Decroos, Roy, and Davis [ 58 ] gave more information than simple numbers, but they also refer to separate variables. However, the ability to examine them simultaneously can provide valuable information to the coaching staff.…”

6MapNet: Representing soccer players from tracking data by a triplet network

“…With the acquisition of large-scale sports data, several studies have tried to quantitatively characterize [5,6,9,13] or evaluate [3,7,8,11,12] soccer players using those data. Especially, some of them used locations (and directions) of actions to represent each player as a vector.…”

“…Players' feature vectors were then assembled by concatenating the weights multiplied on the components to reconstruct the original heatmaps. Decroos et al [6] also proposed a mixture model that fits the distribution of each action type as a combination of finite Von Mises distributions by the locations and directions of actions. Then, they represented each action as the vector of responsibilities for those component distributions.…”

“…Recently, there have been attempts to quantitatively grasp players' styles using video-based event stream data [5,6,9], but they have some limitations due to the properties of the data. First, since the event data were too sparse to figure out a player's style from a few matches, they used the seasonal aggregations of individual players.…”