Research

Data-driven Player Evaluation

Abstract

Analytics in sports became popular with the book “Moneyball: The Art of Winning an Unfair Game” by author Michael Lewis, published in 2003. The book demonstrates how a relatively unpopular baseball team successfully used analytics to ensure a win. Perhaps this was a major drive for baseball to be the first sport to truly embrace and believe that analytics can help improve the performance of the team and support coaches and trainers in their decision making. Other sports have also gradually introduced analytics – tennis, cricket, American football.

Football, or soccer is one of the sports that although has already embarked on the analytics journey, has not fully grasped the advantages that data has to offer to decision makers.

There are several issues with the academic field of sports analytics, including football:

  1. Research remains fragmented – most of the articles are published by authors who do not further pursue this discipline (Wright 2009). Specifically for football, it is noticeable that only few authors’ names appear constantly in publications related to analytics from a specifics group of schools – mainly in the UK and Portugal.
  2. Due to the competitive nature of sports there is a tendency to keep the results secret and therefore a large number of the efforts are not published (Coleman 2012).
  3. As an academic field SA has no natural target and thus, authors may be unsure of the most appropriate place to publish their research  (Coleman 2012).

In spite of the above-mentioned issues, research in sports science and specifically football has advanced considerably in recent years, driven by the availability of technology for notational analysis (analysis of movement patterns, strategies and tactics in team sports). The ultimate goal of such analysis is optimizing the performance of sports teams, the strategies and training programs.

Research Picture

 

Research objective

The research project at our Chair is focused on in-game player evaluation.

Football players are constantly being tracked during live games by means of automated tracking technologies. Every event that happens on the field is recorded and used in post-match analysis to improve strategies and tactics. Such data have not yet been used for in-game decision making. This project proposes a framework for enhancing the player substitution decision to support the coach and his staff during live matches. We specifically work with event data provided by OPTA Sports and we apply advanced analytic techniques, such us, network analysis, process mining, self-organizing maps. Our analysis follow the dynamic systems theory, which became popular in sports science recently, and which suggest that typical statistical techniques do not reflect the dynamic nature of football – a game in which 22 players constantly interact with each other and the ball.

Project partners