Seattle Sports Center – The latest chatter on Seattle sports teams.
Seeing in to the future: modelling football player movements

Seeing in to the future: modelling football player movements

May 29, 2020



David Sumpter seminar for Linköping University 15:15 Wednesday 22nd April.

Based on article from Sloan sports analytics conference 2020.

Full paper here: http://www.sloansportsconference.com/wp-content/uploads/2020/02/SLOAN-Peralta-Final-submission.pdf

Abstract: Soccer has some of the most complex team movement patterns of any team sport. Recently, several measurements have been proposed for evaluating state of play and for identifying the expected value of dribbles, passes or shots [1-6]. The next step is to automatically identify the alternative actions available to players both on and off the ball. We address this challenge by defining three optimization criteria that drives the movement of players during attack. (1) Pass probability: A player moves to maximize the probability of pass success to either himself or to another player, e. g. by opening up a passing lane. (2) Pitch Impact: Occupy point on the field which is maximally dangerous. For example, a striker moves to a point directly in front of goal. (3) Pitch Control: Maximize the amount of space controlled by the team. Soccer players often rate their teammates in terms of their ability to anticipate the movement of the other players on the pitch a few seconds in to the future. To account for this, and building on studies of pedestrian movement, we assume players maximize their future value position on a weighted combination of these three criteria.

We then built a ‘self-propelled player’ model, simulating attacking roles by maximizing a weighted combination of pass probability, impact and control. We compared the simulations to player decisions during matches by top-flight men’s teams of Hammarby IF and FC Barcelona. In simulations, we found that the two or three players nearest to the ball tended to optimize the product of pass probability and pitch impact. We found that simulations in which players optimized pitch control did not reliably capture the movement of players.

In a first-team coaching intervention at Hammarby, players re-watched attacking situations in which they had been involved in the form of pass probabilities, pitch control visualisations and comparisons to the simulation model. The players often agreed that the model captured complex game patterns, including attacking runs to displace defenders and pressing that narrows down the opponent’s passing opportunities. The model also recommended runs that the players hadn’t taken, which the players also found realistic and aided discussions. Despite the fact that discussion of models with professional players is rare, the players showed a high willingness to engage with them. We further explored how these techniques can be used to provide automated feedback to players within the match cycle.

source

Leave a Reply