Artificial Intelligence in Sports
Large sporting events often are a perfect stage to showcase innovation, and this year’s Olympics were no different. In two separate articles I read about technology at the Olympics, I came across one common denominator: Artificial Intelligence.
However, both articles approach the solution of AI a little differently; talent development and talent identification, or better yet computer vision and machine learning. Both topics lay dear to my heart and therefore I want to explore the current state of AI in sports in this column, and how they play a role in developing and identifying talented athletes.
Talent and machine learning
There are many companies focused on talent identification since it is big business, and clubs will pay a small fortune to be able to recognize talent before anyone else does. This conservative happens by analyzing large data sets, collected from statistics, video, performance data, and anything else that matters. Machine learning brings a new dimension to the mix since it is self-learning and can recognize patterns way before any other analytics, it has a lot of potential all it needs is data to work with, lots of data. So, who are the players in the realm of machine learning today?
Well, there aren’t many, at least not any I know about using it specifically for talent ID, since for that you need loads of data, over multiple years (including early youth days), from multiple players, and we just haven’t got there yet. So for now, most companies are still old skool data analytics companies.
Like Scisports, or Catapult Sports, they take video, GPS, statistics, and combine all of this to make recruitment solutions for their customers, but this data is all matters after the fact, rather than looking at a future perspective. That is what machine learning can do, not just take data and predict a potential outcome, but create a more realistic model regarding the future by combining past outcomes with the current.
Insightful computer vision
But enough about what could be, let’s talk about computer vision and its influence on talent development. Seeing yourself perform will always help you improve. By using this technology to measure performance, athletes can get in-depth insights into their movements, their positioning, accuracy, and any other metrics recognizable to the bare eye. The big difference between the bare eye and computer vision is that the latter can capture a lot more and can analyze performance way faster than we humans could ever do, helping us do what we all aim for, to improve faster.
In this field of technology, many companies use computer vision to help athletes improve; VEO, Homecourt, SwingVision, to name a few. Homecourt caught my eye a few months ago, with their designs and UX, however from a technical standpoint, not as impressive. The app is more focused on making basketball exercises more fun and engaging with the use of gamification, which is admirable, but it does miss the actual sports science behind it.
SwingVision is more my cup of tea when it comes to actual talent development. The tennis app doesn’t just measure your count, but takes a deep dive into your movements, giving you real-time feedback on how to improve your swing. This hits home for me since JOGO and sport science go hand in hand.
The missing link
But talent identification and development consists of more than just physical constraints. The one thing I generally feel is missing in most of the currently used technologies are the mental and cognitive aspects. The will, the ability to learn and process things is crucial to grow and advance in football, or any sport for that matter. Athletes who rely solely on their talent but don’t have the intrinsic motivation and processing power to learn will never make it to the highest level.
VEO looks to solve a piece of the puzzle here, by using video to track training sessions and matches for tactical analysis making it possible to see some mental abilities. However, their computer vision is not as advanced as one would like it to be, for the technology is not at a point yet where it can recognize what a player does individually. So for now the software still uses manual coding to get snippets from players, and can only detect a goal, corner, or any other situation the ball leaves the field. And we all know that is not enough for a proper analysis.
All in all, we still have loads to discover but the way we handle and find our most promising athletes is changing, for the better. Moving us towards a future with more reliable insights, and hopefully, fairer opportunities for athletes to develop and to get seen.
Feel free to reach out for any question, discussion, or collaboration at email@example.com