Watson knows tennis. Or at least, the artificial intelligence engine run by IBM
New for the sport in 2023, IBM has upped its AI involvement in tennis, launching an AI draw analysis and generative AI commentary of match highlights, all available for the 2023 U.S. Open.
Tyler Sidell, technical program director, sports and entertainment partnerships at IBM, tells me the more than 2.7 million points of data taken before and during the tournament continually feeds updates to the company’s draw analysis, showing a player’s relative path difficulty toward advancing to future rounds.
IBM borrows from the video technology already in place at the Billie Jean King National Tennis Center in New York, whether Arthur Ashe Stadium or any of the 17 tournament courts on site, to extract 56 unique in-play physical data points from every match. IBM is tracking everything from forehand speed to serve percentages, combining the in-stadium data with player biological information to create a fan-focused element at the U.S. Open.
The AI draw analysis couples with the company’s win percentage prediction for each match. Sidell says the idea behind taking so much data to form predictions offers the USTA a way to “engage with fans and spark debate.”
By constantly monitoring in-tournament information, IBM can use current data points to keep the predictions in tune with in-form players. Sidell notes that during Wimbledon, the Watson platform predicted Carlos Alcaraz’s path to the final correctly in six out of seven opponents and also had Alcaraz winning the championship in a tight match—55% win probability—over Novak Djokovic.
Watching the draw analysis shift with new insights as the tournament went on gives fans a real-time look at the potential road to the final rounds.
Before the main draw of the 2023 U.S. Open started, IBM’s Power Index ranked American Coco Gauff as the top player in the women’s draw, five spots ahead of her No. 6 ranking and had Ons Jabeur second on the Power Index, up from her No. 5 rank.
On the men’s side, Novak Djokovic led the Power Index, jumping Carlos Alcaraz. Jannik Sinner was third, up from No. 6 in the world rankings.
As for the draw analysis, the tournament started with IBM predicting Djokovic as having the easiest path to the championship match. For the women, Jessica Pegula was predicted to have a relatively easy path to the semifinals.
Working with the U.S. Open for over 30 years, IBM has also introduced generative AI commentary for match highlights, available on USTA platforms. Using the watsonx platform, the machine learning takes the video feed from in-venue cameras and starts to understand the sport of tennis to narrate highlights.
“The goal is to create a richer community,” Sidell says. “We are turning meta data into spoken sentences.”
Using the AI system, IBM can stitch together highlights packages within minutes of a match completing, giving the operation a scalability that ensures highlight packages from all 17 match courts remain updated. And commentated.
Looking forward, IBM hopes to soon add multiple languages using AI.
The service debuted for Wimbledon and the tournament quickly made the AI commentary part of the default option when fans visited the site. “Seeing partners embrace it is encouraging,” Noah Syken, IBM vice president of sports and entertainment partnerships, tells me.
For the 2023 U.S. Open, Watson had some key prediction, all designed to get people talking.