To someone watching from behind a screen, the movements of players on the field could be seen as wizardry — their strategies, the work of brilliant coaches, every leg and arm movement toned from years of harsh training. But to a computer, all that’s just numbers, and some numbers are better than others.
The recent AI boom has had a tremendous impact on all sorts of fields, and that includes sports. The use of digital tools for improving fitness is nothing new. Personal fitness trackers have been a thing for decades, but new developments in artificial intelligence mean that this fitness advice can become even more personalized and even more in-depth. The technology is still quite untested, and it’ll take a while for it to be scaled up to the level where very high-level teams can rely on it. But if it works, it could result in a revolution in how we train and practice.
First, though, there are some specific reasons why AI would be so helpful for sports. Dr. Joe Kileel, from the University of Texas at Austin, studies the applied mathematics behind artificial intelligence.
“Part of a thrust of my research pertains to what are called tensors,” Kileel said. “If you think of them as a data set, it’s like a table of numbers, but actually the table is maybe not just two-dimensional, but higher order … This is in line with part of the idea of AI of trying to find these patterns in high-dimensional data.”
Artificial Intelligence as a resource can be more efficient for certain analytical aspects of sports. Kileel emphasized how it could be beneficial specifically for statistics and simulation.
“It should have potential for improved statistical analysis, for example,” Kileel said. “There may also be possibilities to simulate sporting events using AI.”
Commercial AI startups are currently exploring how to get more and more results from artificial intelligence. Darren Montgomery works for the company Zone7, providing AI for a variety of high-level teams.
“Zone7 uses all available data sources: wearables, subjective, EMR, etc., to build a digital twin of each athlete,” Montgomery said. “Some of their algorithms simulate future load against each profile to assess potential injury risk. Some others generate optimized plans to meet particular goals for strength, endurance, or skill acquisition. All of it hyper-personalized.”
Despite the very recent adoption of this technology, there are already some major successes that have been attributed to the technology. These successes have occurred in a variety of sports.
“We’ve had teams in top-tier soccer in multiple countries, in NFL, NBA, and other sports,” Montgomery said. “All of our clients see year-on-year reduction in injuries and days lost, on average around 40%.”
The use of AI in sports is, however, not without its intricacies and issues. Matthias Jud, from the Karl-Franzens-University in Austria, is known for his PhD thesis, a systematic review of scholarly research regarding AI usage in sports.
“AI is extremely differently used in sports,” Jud said. “I would suggest a categorization like endurance sports and athletics, where technique is more important than, for example, force or endurance. These are shooting, dancing. So from my point of view, it is extremely important to make a differentiation in this case.”
Jud has studied AI programs coming from a variety of places: Britain, the U.S., Germany, and soccer matches in Austria. Additionally, he’s looked at the use of AI technology for sporting activities at a variety of different levels.
“From my point of view, AI is currently used in recreational sports in particular,” Jud said. “So you have to also distinguish between professional athletes and recreational athletes, you know … In particular, for those athletes who do not have very high goals or just doing sports for approximately, I don’t know, five or ten hours a week … Nowadays, AI tools, their functionalities, they are not designed for professionals; they are for recreational sports.”
Still, Jud’s findings make him believe that it’s important not to get too overexcited about AI’s capabilities. His findings about personal usage provide some skepticism about the growth.
“As far as I can say now, users are not really perceiving the embedded AI system,” he said. “They perceive it like a fitness tracker … they do not really perceive the AI system itself, and they do not see a benefit of this AI … AI in this field seems not to be having a huge impact. Now, I don’t know how it will be in five years.”
Whether they’re hopeful or skeptical about using artificial intelligence in sports, whether they’re excited or scared, both sides have agreed that the future is still very difficult to predict. Nonetheless, the future of AI will be dictated by the people who work on it, including at LASA. Suhan Li manages the AI Sports Lab club, meeting on Mondays in G104, where students research, write code, and discuss this topic.
“Usually it’s very hard and intimidating to write code for the AI, so we have a soft start,” Li said. “We start with research, some research papers, especially systematic review papers that highlight what work has already been done in this field. And we also look into experimental papers to see the specifics. And through this process, my co-presidents and I teach the members about certain research skills. Right now we are just starting to learn how to write the code part.”
Li provided a specific example of AI use that he’s studying, that being analysis of the saber lunge in fencing. As an important, refined move, one that can result in a point being taken if done correctly, even minuscule changes to the movement of the player. Those embodied in the aforementioned tensors that the AI models use can result in major gains. This principle, by which AI is used for even the smallest changes, results in major improvements, echoing throughout sports, even into the major leagues.
“The benefits will be pretty small, but you have to keep in mind in professional sports, a small advantage is the key to winning or not winning,” Jud said. “And so this can in fact have a high impact for athletes to improve their technique, to improve their fitness level, et cetera … I do not expect that you can run the hundred meters one second faster by AI analysis. This is not possible, and this is not really realistic.”
Artificial intelligence usage in sports still has to be refined. The current implementations are scattered, very new, being mostly only in the last decade or even the last five years, and need a lot more testing. The future, though, does look bright to some, according to Li.
“It’s a very new, tantalizing technology,” Li said. “I hope more people are interested in this technology.”
