Applying the principles of Johan Cruyff to data science

With such an ambitious goal, the Barça Innovation Hub has presented its research at this year's world-renowned MIT Sloan Sports Analytics international conference. This research is based on a mathematical model that assesses the quality of the decisions made by players taking into account the position of their team mates and opponents every second within a match. The study is co-authored by Javier Fernández (head of Sports Analytics at FC Barcelona), Luke Bornn (former Harvard professor and current Vice-President and head of Strategy and Analysis for the Sacramento Kings) and Dan Cervone (Director of Quantitative Research for the Los Angeles Dodgers)

"Football is a sport that you play with your brain. You have to be in the right place at the right moment, not too early, not too late." This is how Johan Cruyff summarised the main goal for decision-making in football, stressing that being in the right place at the right time is the cornerstone for positional play.  On this basis, Cruyff insisted that “it is statistically proven that players actually have the ball for three minutes on average. So, the most important thing is: what do you do during those 87 minutes when you do not have the ball? That is what determines whether you’re a good player or not”. With this unequivocal view, Johan expressed how individual talent must also help to create the best possible context around the ball through the movements of these players who are not in control of the ball.

In recent years, a great deal of quantitative analysis in football has focused on applying statistical analysis to isolated actions.  Statistics such as completed passes, shots on goal or the percentage of possession, is data that refers to individual actions which do not take into account the context in which they occur.  On the other hand, the research presented seeks to reverse this fundamental aspect through the introduction of a new variable – EPV (Expected Possession Value) – which determines the quality of the players’ position every second on the pitch in terms of the possibility of the action resulting in a goal. Overtime, the evolution of the EPV allows to evaluate the influence of the players’ decisions (both those controlling the ball and without it) which results by improving or reducing this value.

To be able to build this mathematical model, the team of researchers processed the position of players in every single frame of video for the games played by FC Barcelona over the course of two seasons.  Through Machine Learning algorithms, the researchers have been able to determine during every single second, the impact of the players disposition on the probility of the possession ending in a goal.

It may seem a fortune-teller trick, but this is a cutting edge example in applying artificial intelligence to football, an area in which Barça Innovation Hub is a globally-renowned innovator. The model is still under development, but much is expected from this tool which has been made possible by the joint work of the Club's coaches, analysts, and data scientists. This can be applied to different practical cases such as the risk and benefit evaluation of the passes, the identification of positional superiorities according to the context, and the determination of alternative options for each player at each moment, among others. It is an example of how artificial intelligence is knocking the door to enter the game through a technical implementation focused on the end user, the coach. 

The complete paper can be downloaded here.