top of page

Football video analysis has a secret problem. We know how to solve it

There are countless studies out there on the subject of overwork. The exact statistics and results seem to vary somewhat, often based on what they are looking for. One study from South Africa found that longer hours resulted in higher productivity, and one from Boston University found the exact opposite, that workers who worked less succeeded exactly at the same rate. One thing they all seem to agree on is that after a few hours, workers will be less focused.

How are you tying this into football?

Football video analysis is focus-intensive. Event tagging, the initial phase of it, is time-consuming and can seem mindless. Over 1600 events happen in a single football match. Going over footage of the match with a comb and finding every single event is intensive enough, but some things don’t even register as events. Noticing the movement of two players while the ball is far away from them can be important, but it won’t register with most people who aren’t specifically looking for it. The second half, the analytical part, is even more difficult and requires very specialized labor. It isn’t work that a person can just mindlessly do. An unfocused analyst will miss things, even if they are the best analyst in the world.

Even a focused analyst can miss things. If five events happen at the same time, an analyst could catch one or two, but they will miss others. Alternatively, a part of the game where seemingly nothing happens could have vital information that gets missed by an analyst waiting for a big play. Nobody catches 100% of everything. An exciting event will draw anyone’s notice, and take more mental space than others.

A further problem is that the work of football video analysis needs to be done fairly quickly. Ostensibly, the solution to the previous problem is to allow someone to work at it for a while and spread the work out over a few days, but analysis needs to be done quickly. Getting insights quickly is an incredible tool, and help a club apply the lessons it learned from the game faster. This means less training time wasted on exercises that don’t help, fewer bad habits internalized, and overall, speed will lead to stronger play. Additionally, often a moment will go viral, and a good club will want to leverage that moment. There is a problem here: Giving analysts the proper amount of time to do their football video analysis is not entirely possible when getting it done quickly is so advantageous.

What do people do, then?

There is the solution the most elite clubs do. They hire many analysts, and have them work in concert. It’s financially intensive, hiring many people to do this sort of trained labor, and often they are in-house analysts on a retainer, but it works well. Other clubs without those finances often deal with work that is late, or by necessity worse than it could otherwise be.

But the best and most consistent work seems to be handled by AI. Machine learning techniques, where a program is iteratively trained on many examples until it can understand what is part of the pattern it is looking for and what isn’t, are a solid technique that is proving itself in many fields. From the obvious aspects of video game AIs to the very newsworthy self-driving cars, we see AI involved in more and more fields. AI has that secret sauce we’ve needed this whole time.

In sports analysis specifically, AI could help with all of the aspects we’ve mentioned. An AI directly connected to a camera set feed can analyze the game as it’s happening. It will spot events no matter how many of them are happening at once or where they are happening. It won’t get bored and miss things. It won’t lose focus if it runs for hours, though at the speeds AI can do analysis, it often doesn’t need to run for hours. It handles tagging and analysis quickly and efficiently.

That is why Track160 relies on AI, trained by machine learning techniques, to do football video analysis. With only a single camera set, Track160’s software can follow a match on a stadium of any size and analyze it quickly and accurately and produce the sort of data that would take a team of analysts many hours to get. We strongly believe that AI is the future of data analytics, and we are excited to be a part of pushing it forward.

Read more:


bottom of page