Evaluating Tactical Performance in Football using Features Derived from Position Tracking Data

Over the recent years there has been an increasing interest in the use of data in professional football, fueling a growing sports analytics industry. 

This had lead to the development of features like Opta's expected goals, Impect's packing and others that try to capture different aspects of performance in easy to understand statistics. 

Data-driven features like these have gained considerable popularity over the recent years, and they are often used to evaluate performance. However, it is unclear whether or not this implied relationship with performance (match outcome) actually exists. 

To scientifically evaluate the relationship between features derived from tracking data and performance, we conducted two studies using tracking data from over 100 matches in previous Eredivisie seasons. 

In the first one we computed several features inspired on the general principles of play in football as well as on existing football analytics tools and studied the relationship between these features and match outcome. 

In the second one we used features derived from our previous research that quantify how a team disrupts the opponent through passing and again studied the relationship with match outcome. 

Next week, Matthias Kempe and I will present the results of these studies and discuss their implications at the MathSport International Conference in Athens, but you can already check out the results yourself by clicking the links in this post!

#MathSport #datascience #football #tactics #sportsanalytics #soccer