Chelsea FC & Artificial Intelligence: A reality check
Chelsea FC are now adopting a new player data tracking system that measures and analyses player performance and decision making using an Artificial Intelligence analysing system developed by students at Loughborough University, London.
The goal is to help provide coaches, technical staff and, strength and conditioning staff to help players reflect on their actions after games and slowly improve their decision-making skills.
How the System works:
Currently used by the Chelsea FC academy the system analyses numerous seasons of data that tracks players and the ball for 90 minutes and develops a computer model of various playing positions poses and physical exertion. The computer data provides a benchmark to compare the performance of different players. This helps to measure the performance of individual players independent of the actions of teammates or opponents.
The analysed data can then be visualised to show what would have happened if the players with or without the ball made a different decision in any case. Although measuring is difficult as it is not humanly possible to track all events during a game. It is also difficult to isolate actions of one player from another. For example, if a player makes a pass and seconds later his team loses possession, did the player attempt the pass at the wrong time to the wrong player or was it someone else’s fault?
Imitation learning a particular branch of AI solves this problem of decision-making. It understands the fundamental decision-making policy by looking at how an expert performs that same task and mimics the expert.
For the model to be realistic, tons of historical data must be analysed and the analysed historical data will try to reflect the real world as close as possible. Apart from showing player movements with and without the ball, physical exertion, position, poses, and current game situation will also be captured and will help show if players want to park the bus or keep the attacking pressure, a deciding factor in winning or losing a game.
Fans and commentators constantly call into question players actions during the game saying they should have chosen a different pass or movement without any real way of testing their theory until now. This AI model will show how accurate these suggestions might be.
When the football critics say a player should have passed the ball instead of an attempted shot, the system will look at the alternative outcome, factors such as physical exertion, position at that point in the game will also be factored into the data and helps to show in visual what would have happened if the player chose to pass rather than shoot during that point in the game.