How IBM and the US Open are using Watsonx to create more AI-generated tennis content

Using Watsonx helps bring depth, speed and scale to USTA’s editorial coverage, said Brian Ryerson, USTA’s senior director of digital strategy. He said it also helps solve the “blank page problem” as the team keeps up with more than 200 matches. IBM also is using the tournament to showcase its Watsonx governance platform, which adds guardrails to monitor and mitigate inaccuracies from the LLM. The AI models collect more than 7 million data points during the tournament ranging from serve speed stats and win-loss ratios to analysis from thousands of tennis articles and blogs.

“These match reports will allow our editorial team within minutes of a match ending to really have a generated preview of what a match reward could be: the key moments, what happened, all of that,” Ryerson said. “And our editorial team can take that, edit it, add their own editorial flair, as I like to call it, and then publish it out to the world.”

When used with the right guardrails, AI can help editorial teams create more content and offer more personalization, said Randy Picht, executive director of the University of Missouri’s Reynolds Journalism Institute, a journalism innovation and research center. He gave the example of using AI to gather and share past and current context that would otherwise be hard to gather — something that could help writers, readers and players alike.

“The ability to go into a rich archive and pull out context and stuff is also very promising,” he said. “You can teach someone how to write a two-paragraph story about who won a tennis match, but you can’t teach them that 20 years ago, something happened with a left handed player from Ukraine.”

Assistant coach

Beyond the fans watching the broadcasts, Watsonx’s tennis know-how might also someday help players and coaches. One of the players curious about the potential of AI is Andre Agassi, one of several former and current tennis stars who are collaborating with IBM and USTA. Ahead of playing in the annual Stars of the Open event, Agassi toured IBM’s data room and shared thoughts on what might be helpful to players rising up the ranks.

According to Agassi, using AI could help analyze what or how something happened, but leaves it up to the players and coaches to figure out why something happened. Although AI might provide players with “one hell of a roadmap” to better understand themselves — and the competition — he also noted it’s important to not get too hung up on static stats.

“You never know if they played seven lefties in a row,” Agassi said, noting that AI might help discern such minutiae. “There’s certain things you can’t and you don’t just definitely run with.”

Using visual AI to analyze footage could be especially helpful when playing against someone for the first time or to think about what someone is like in various environments and surfaces. For example, Agassi said a ball has a certain bounce depending where it’s hit, what kind of spin it had before it was hit and how an arc changes after a bounce. He also noted AI might help fans better understand what it feels like to serve against someone who’s much taller or shorter than the other player.

“It’d be so important never playing somebody to understand which shots they’re actually attempting to get, either at peak, or just prior to peak or post-peak,” Agassi said. “That tells me a little bit of their mindset and aggression, and I can deduce a lot from seeing that particular point of contact.”

Taking AI to center court

Online and offline, the US Open attracts a large audience. According to IBM and USTA, nearly 1 million tennis fans attend the tournament in New York City while more than 15 million unique devices visit the US Open’s website and app during a three-week period.

Coaching Watsonx meant training it on the latest AI models on the sport’s vernacular — and how to pronounce players’ names — along with other aspects like the speed and pitch of what tennis commentary should sound like. In the future, it might look at using different voices based on each event. It’s also exploring ways to translate commentary into other languages, starting with Spanish as the new addition for 2024.

“Maybe there’s a different type of commentator — a beginning commentator, a silly commentator, or whatever would support the brand,” said Monica Ellingson, practice lead for IBM Consulting’s sports and entertainment. “We’re continuing to focus on that fan experience, making it personalized to that fan.”

IBM and USTA’s collaboration on AI-generated analysis and content creation represents just one of the ways IBM is marketing Watsonx this year. Others include during the 2024 Masters golf tournament and earlier this year during the 2024 Grammys.

The US Open is more than just a branding play, said Kristi Kolski, marketing program director, IBM Sports and Entertainment Partnerships. Enhancing fan engagement and personalization to a global tennis audience also allows IBM to market its AI capabilities amid an increasingly crowded category. She also noted IBM is sing the same hybrid cloud and other AI tech it uses for clients across a range of industries — showcased in ways that are “easily relatable and applicable.

“In the past we’ve prioritized developing unique insights from data to spark sports debates— using data like scores, stats, and serve speeds to create predictions such as ‘the likelihood to win’,” Kolski said. “Now, we’re focused on delivering personalized US Open digital content with the help of generative AI technologies — for example, different fan personas and different styles of content like short and long form.”