Using AI in football | Artificial intelligence


Goal-line technology and VAR have woven their way into the football world, and although not without their teething problems, most of us are now fully adjusted to their presence come match day. Despite Football’s anti-technology tendencies, there’s a new kid on the block who’s looking to make a name for themselves amongst players and coaches alike – the name? Artificial intelligence. Many of us are familiar with the term ‘AI’ but only those behind the scenes seem to fully grasp the science and purpose behind it. The phrase tends to conjure up ideas of a robot-run dystopia, or a Terminator 2 / Skynet type world, but thankfully the reality is far less sinister. AI in Football is aimed at creating value for coaches, managers and tacticians, allowing for a much more in-depth player analysis than what is currently possible with today’s technology. SAG, Acronis and Innovate are all leading the way with ‘deep learning-based technology’ which focuses on fixing the cumbersome methods currently used by teams.

 

Which problems will artifical technology solve?

 

The problem that AI intends to solve is the time-consuming nature of player performance analysis. The current approach relies on the team’s analyst manually reviewing training or match day footage and inputting data to create statistics and reports.

Teams have to sift through hours of video and log individual actions such as passes, shots or tackles, account for where each action took place on the pitch, and whether or not the players were successful in their attempts.

This method, although the norm for most clubs, is both error-prone (due to human input) and time intensive. Even the best analysts in the country can fall victim to their biases, lapses in judgement or fatigue, all of which can ultimately result in inaccuracies in the data.

The amount of time required to carry out the analysis also means that there are limitations in the amount of statistics that can be presented for each player. Technologies do exist already which allow player tracking, but they lack the level of sophistication and detail that AI in football intends to bring to the game.

 

How can AI in football help?

 

The primary goal of artificial intelligence in football is to fully automate the analysis process. Instead of an analyst having to watch fragments of a game, focusing on one player at a time and then present their report, AI will do this automatically. The system intends to track the players, the ball, analyse all of the important actions of the game and present statistics based on its findings.

The process relies on the AI reviewing match-day footage and creating statistics from the actions it has been programmed to track. The technology will be built to track on-pitch actions at three levels.

SAG has broken this down into low, mid and high. Low level involves ball detection, player detection and kit number recognition. Mid level involves 3D ball tracking, player identification and tracking. High level involves recognizing specific game aspects, events and calculating statistics.

When the technology becomes more widely adopted, the aim is that it will help to produce personalised training plans for individual players, make adjustments to match strategies, streamline the scouting of new talent and even analyse fans attention to determine the best way to integrate adverts and sponsorships.

 

Out with the old?

 

On paper, the upside of AI in football seems limitless, but it isn’t without its challenges. For AI to work, it has to be trained. This involves inputting the system with thousands of hours of game footage from different leagues all over the world in order for it to correctly identify different kits, positions, angles, backgrounds, stances, poses, movements and actions.

There have been claims that AI will be able to create models, based on pattern recognition, which predict a new signings impact on the club they are moving to, i.e this player will create less assists but score more goals than last season.

Football is, by its very nature an incredibly unpredictable sport, with an innumerable amount of outside factors which can influence a players performance from game to game. The idea that a system can predict such a feat would be too hard to swallow for many fans.

It seems that the perfect pairing would be a combination of AI and football analysts working together to create reports, strategies and performance plans. Artificial intelligence may be able to track events and create greater statistics, but knowledge of where players should be moving to, positioning, and situational awareness would almost certainly be best left to scouts, coaches and tacticians.

 


 

You could earn up to £100 (or currency equivalent) in bonus funds by joining Colossus with our New Player Bonus. Click here to join the action.