CMStatistics 2021: Start Registration
View Submission - CMStatistics
B1543
Title: Football strategies via the effective playing space and tracking data Authors:  Dimitris Karlis - RC Athens University of Economics and Business (Greece) [presenting]
Marius Oetting - TU Clausthal (Germany)
Abstract: Modern analysis in football makes use of tracking data, i.e. we know exactly the position of the players and the ball with a high frequency. Such data can help coaches and scouts in several aspects, including game strategy and tactics, player evaluation, goal analysis, judging referee decisions, and talent identification, to name a few. We consider tracking data and focus on a metric that measures the position of the team on the field aiming at examining several tactical aspects. The metric is the convex hull created by the players of a team excluding the goalkeeper. It is also referred to as Effective Playing Space (EPS), calculated as the surface area (in square meters) of the convex hull of all players (excluding goalkeepers). For our analysis, we consider a novel hidden Markov model (HMMs) for modelling the EPS time series data jointly for the two teams, as they naturally accommodate the idea of a match progressing through different phases, with potentially changing tactics. The unobserved states in our HMM serve for the underlying tactics of a team (e.g. defensive vs. offensive style of play). The model enables to deepen the insights into the tactics of the teams, as interactions between them can be modelled additionally.