How AI revolutionizes football video analysis for the sub-elite level
Data is now an integral part of sports. This is no longer in doubt. After all, It reached the most concrete level of evidence possible a long time ago: A movie starring Brad Pitt... It seems like the question nowadays is no longer, “Does data analytics have a place in Sports”, but, “How will data analytics evolve from here and on?”. This question is asked by everyone in the field - journalists, managers, players, owners, analysts, and so many more. We have our own view, and we will explain it through one crucial ingredient of football analytics: video analysis.
Football video analysis: the new game within the game
Seemingly, any football team can do video analysis - all it needs is a video recording of a match and a person identifying the events and analyzing them. There is a vital distinction between two terms: event-tagging and video analysis. The latter is the one you hear about everywhere; it’s the sexier term relating to end-conclusions like why Ronaldo is so dangerous inside the 16-box or why Messi should take the corners only from the right for PSG. These conclusions wouldn’t have been born without first identifying every event that happened in the match, and that means tagging each tackle, pass, throw-in, corner, shot, etc.. It’s not magic, it’s trained work that requires a professional equipped with footage of the game.
It might sound out of context after opening this article with the immense progress data has made in football, but as of today, this important work of event-tagging is done manually 99% of the time - whether by an inside-the-club persona (usually an analyst) or an outsourced vendor. How long does this work actually take? Well, it can take anywhere between 12-25 hours per match, depending on the number of analysts on it. For rich clubs like Manchester City, Liverpool, Barcelona, or PSG, that’s affordable and easy to do. They already have a crew of video and data analysts on board that can sweep through matches quickly and thoroughly.
Track160's solution uses AI to automatically tag events. Check out its video analysis features
But what about the sub-elite clubs that have limited resources? Can they tackle this issue when they have just one or zero analysts on the payroll, avoid mistakes that are bound to happen in every manual work, and still find time to enjoy the fruits of video analysis? The short answer is: they can. But how?
Let’s look at an alternative, shall we? Let’s look at AI.
Find the events hidden in the great moments
Machine learning techniques require a lot of set up. Similar to a robot looking up Sarah Connor in a phone book and getting many wrong ones before finding the correct one, any AI needs to be trained to find the correct methods. The average 1,600 events that happen in a football match provide it with perfect training to identify different types of events but also to identify what doesn’t fit them. Once it is properly trained, an AI can not just recognize and tag every event, it does so with no distractions and with full comprehensiveness.
Beyond that, though, it can find and tag events in noise. It can look at individual moments and tag multiple events, not just the kick to the goal but the dribble, the attempted poke tackle, and everything else that possibly could have happened in that moment. It can look at a sequence that will appear hum-drum to a person and find the little extra movement that an analyst could notice when focused, but maybe would miss when watching the whole game. An AI could find all of them immediately, with no wasted work, even while the match is happening. By doing so, it offers the ideal basis for any video analysis that will follow.
All the events you see below were identified and tagged using AI, with no human involvement
Every second matters
It takes a human time to do manual tagging. Beyond the effort of the tagging itself, experts work hours they’re comfortable with and take breaks, because humans need that sort of thing. Software doesn't. It is faster at tagging events due to a computer’s processing speed, and can work not just when a match is over, but during the match itself. This means immediate analysis is possible for a coach who wants to review the data right after the game, before he has time to get the team a congratulation or consolation prize, and sometimes even during half-time if he intends to make real-time changes based on objective stats.
Automated video tagging and analysis have another big advantage. When social media is functional, key moments from matches often go viral, and speed can be of the essence for a club to find the right clip for the right moment. When an event threatens to catch “digital fire” and become the center of attention for a day or two (before it’s replaced by the next boat stuck in a canal), automated videos could help a club be a discussion leader before they miss their moment.
Accurate automated event-tagging allows users to create playlists with one click
Money, dear boy
Trained video analysts don’t grow on trees. Hours of expert labor are well worth the cost to clubs that can afford them, but many clubs can not. The gap between the richest clubs and others has always been there with player selection, or as Grace Robertson put it, “Man Utd's attack is mostly a bunch of players each just doing their thing, but you get to a point where the players doing their thing are so good it doesn't usually matter”. But that gap can be minimized. When AI is available, the costs of having good football video analysis can be reduced, allowing more teams to benefit from it and more stakeholders with no initial expertise to get access to it.
Football video analysis is the next game-changer
In the effort to democratize football data analytics, video analysis AI-driven solutions are vital. They not only have the potential to be superior to the work of trained analysts, they will let less wealthy clubs get into football analytics. This doesn’t just make the clubs perform better, this improves the game of football as a whole, with sharper plays and greater moments. Automated platforms are changing the game, in more ways than one.