Visual analysis of pressure in football

• Published in: Data mining and knowledge discovery Vol. 31; no. 6; pp. 1793 - 1839
• Main Authors: Andrienko, Gennady, Andrienko, Natalia, Budziak, Guido, Dykes, Jason, Fuchs, Georg, von Landesberger, Tatiana, Weber, Hendrik
• Format: Journal Article
• Language: English
• Published: New York Springer US 01.11.2017; Springer Nature B.V
• Subjects: Artificial Intelligence; Chemistry and Earth Sciences; Computation; Computer Science; Data Mining and Knowledge Discovery; Football; Information Storage and Retrieval; Physics; Players; Pressing; Sports Analytics; Statistics for Engineering; Tactics; Football data analysis; Movement data analysis; Visual analytics; Collective movement patterns
• ISSN: 1384-5810; 1573-756X
• DOI: 10.1007/s10618-017-0513-2

Modern movement tracking technologies enable acquisition of high quality data about movements of the players and the ball in the course of a football match. However, there is a big difference between the raw data and the insights into team behaviors that analysts would like to gain. To enable such insights, it is necessary first to establish relationships between the concepts characterizing behaviors and what can be extracted from data. This task is challenging since the concepts are not strictly defined. We propose a computational approach to detecting and quantifying the relationships of pressure emerging during a game. Pressure is exerted by defending players upon the ball and the opponents. Pressing behavior of a team consists of multiple instances of pressure exerted by the team members. The extracted pressure relationships can be analyzed in detailed and summarized forms with the use of static and dynamic visualizations and interactive query tools. To support examination of team tactics in different situations, we have designed and implemented a novel interactive visual tool “time mask”. It enables selection of multiple disjoint time intervals in which given conditions are fulfilled. Thus, it is possible to select game situations according to ball possession, ball distance to the goal, time that has passed since the last ball possession change or remaining time before the next change, density of players’ positions, or various other conditions. In response to a query, the analyst receives visual and statistical summaries of the set of selected situations and can thus perform joint analysis of these situations. We give examples of applying the proposed combination of computational, visual, and interactive techniques to real data from games in the German Bundesliga, where the teams actively used pressing in their defense tactics.