Advanced football data science
- Reference number
- SM18-0026
- Start and end dates
- 190101-210331
- Amount granted
- 1 445 864 SEK
- Administrative organization
- Uppsala University
- Research area
- Computational Sciences and Applied Mathematics
Summary
We will build modern data science tools for the analysis of player performance and tactics. We aim to develop a world-leading competence in understanding football using detailed data of player and ball positions during the match. We will develop data-driven models to evaluate player contributions (PC), study team synchronisation (TS) and understand decision-making and spatial positioning of players (DM). In each of these examples we take mathematical methods, many of which have been used in Sumpter's earlier academic research and create novelty by applying them to football. By working together with analysts, the technical director (Ola Larsson) and coaching staff at Hammarby we will iteratively improve models of the game. This will form a basis for a deeper understanding of the game and have an immediate impact in scouting, opposition analysis and development of training exercises. We expect to (1) develop an automated way to prepare coaching staff for opposition style of play based on on-the-ball data; (2) develop improved training methods for team work, based on finding movement patterns that are more likely to produce goals; (3) improve the basic research understanding of football and, through interaction with other clubs, build a basis for an analytical approach to football. The project will transfer research from academia to industry and strengthen a growing sector in Swedish industry based around sports analysis tools that goes beyond simply data collection.
Popular science description
Football is the most mathematical of sports. From the triangles of 'wall' passes and the symmetry of formations to the game theory of management and the physics of shooting, all aspects of football can be described and better understood using maths. And, in the era of data science, where computers are used in every aspect of sport, it is more important than ever to have novel methods for analysing the vast amounts of data clubs collect on their players and the opposition. Yet, mathematics is only now entering clubs. In this SSF funded project, mathematics professor David Sumpter will work for two years at one of Sweden's leading football clubs, Hammarby. He will look at the geometry of the game, analysing player performance, how the team moves together and how players make decisions of the field. This will give new insights in to the game and make use of all of the data now collected. David will work together both with Hammarby and with other leading international clubs, to produce a better model of football. Most people have either played or seen the console game FIFA. While these games are realistic in many aspects, they do not truly capture the tactical intricacies of the game. We will build mathematical models of player interactions that do explain and allow us to better predict how football is played.