At this year’s US Open tennis tournament, IBM’s AI-powered features bring fans even closer to the action. It turns out that sports and AI make for a pretty good match—and AI may even prove to be a game-changer for the sports industry.

For the US Open, IBM delivers AI-generated summaries for every men’s and women’s singles match. And for the second year, IBM is providing an enhanced version of AI commentaries with automated English-language audio and subtitles using watsonx.

The goal? To deliver a more informative and engaging experience for millions of tennis fans.

When AI meets the audience

Generative AI seems everywhere in sports events this summer. The recent Paris Olympics set an AI agenda through its partnership with Intel, offering a chat service to athletes built on French start-up Mistral AI’s tech.

Earlier this summer, IBM launched a new feature, “Catch Me Up,” based on watsonx and the IBM Granite model at Wimbledon.

“Catch Me Up gave fans relevant content in the form of short-form stories, before a match and after a match for the Gentlemen’s, Ladies and Wheelchair singles draws,” explains Kevin Farrar, IBM’s Head of Sport Partnerships UK and Wimbledon Partnership Executive. “The stories shown were based on fans’ favourite players, their location and/or trending stories, as well as an AI-generated end of day summary for each of the days of the tournament.”

This feature was a great opportunity to offer personalized content, notes Farrar. “Gen AI not only gives you this ability to create personalized content, but also to do so at scale.”

IBM also brought back its Match Insights including “Likelihood to Win,” which predicted a 61% chance of victory for Carlos Alcaraz in the final against Novak Djokovic, the second-year running Likelihood to Win called this correctly.

“The third set really validated what we had predicted,” recalls Aaron Baughman, IBM Fellow and Master Inventor. “What I found interesting is that the technique I used was a classical machine learning algorithm, a logistic regression probability distribution function that we trained. This is typically a building block for neural network activation and/or loss functions.”

“Wimbledon gives us the opportunity to showcase IBM’s Technology and Consulting on a global stage while the tournament is in progress, which is really powerful,” believes Farrar.

How sports can provide insightful data

Beyond the tournaments, certain teams and leagues have recently announced the integration of AI into their own practice.

Among them is the Boston Celtics, which uses AWS and Mission Cloud to leverage their data and analytics and improve their performance and operational efficiency. The NBA also launched an AI voice assistant for its last All-Star game earlier this year. In European soccer, Sevilla FC partners with IBM to enhance player recruitment.

This mix of sports, innovation and data is logical, believes Brian Hall, adjunct instructor at NYU and founder and head of AI of AlphaPlay AI, a start-up that aims to provide AI solutions to professional sports teams.

Hall, a chess player, credits notably IBM’s Deep Blue experience in 1997, for inspiring his interest in sports and data.

“I realized that a sporting match is very similar to a chess game,” Hall says. “They have a set number of players with a set number of rules and they can do certain things in a certain amount of time. Why shouldn’t AI assist a team in professional sports, look at the data of players and how they play at events and matches to discover new things that the human mind hasn’t seen before?”

AlphaPlay AI provides insights to teams and coaches that would help them uncover patterns. “We see ourselves as trying to enhance the intuition and the experience of players, scouts, coaches, managers and owners,” says Alice Wang, CEO of AlphaPlay AI.

“AI is very scalable. One of our clients recently said: ‘No matter how successful a sports team is, it can never hire enough analysts.’ It stayed with me for such a long time because it’s so true. You think you’ve got this army of humans, but ultimately we only all have 24 hours in a day. A human’s ability to crunch numbers is just incomparable to the computational power of AI,” she adds.

The ambition to democratize sports analytics is also an aspiration of the Norwegian B2B startup SportAI.

Founded in December 2023, SportAI uses AI and machine learning to provide insights directly to users from any uploaded video.

“Technique coaching analysis and commentary for sports is still very subjective, expensive and unscalable,” says Lauren Pedersen, CEO and founder of SportAI.

As a former college tennis player herself, she explains that she took many coaching classes during her youth.

“The coaches would all give me varying opinions on my technique, based on their training, background and what they’re seeing with my technique. It wouldn’t be backed by any data,” she observes. “We realized that we could change this with AI. We could actually put data behind technique analysis and really democratize access to great sports technique coaching analysis and commentary.”

SportAI recently announced a new round of funding. It targets coaches, broadcasters and equipment brands and will expand to more sports.

What’s next for AI in sports

While startups and big tech players alike are already embracing AI-powered tools in sports, the analytics revolution hasn’t fully permeated the sports domain yet, believes Sam Robertson, a researcher and consultant specializing in innovation and sports decision-making.

“The biggest example is injuries: injury rates have not moved in most sports,” he notes.

Robertson, however, thinks that computer vision has the potential to open up new possibilities, notably to bring a new experience to the audience. AI models can also help decision-making.

“How can a human and an AI model work together? How can we use AI to refine human judgment?” Robertson asks. “That’s particularly important in cases where people have to make decisions over and over again, like in football refereeing. It’s very unreliable and subjective, and it’s not a criticism: it’s a normal part of human decision-making. For me, it’s an obvious example of where AI could be used not only to improve accuracy but also consistency.”