Artificial intelligence in sports has largely been used for fan interfaces, such as video highlights curated using crowd noises and player reactions.
Now the United States Tennis Association is using AI to assist in identifying and aiding top players. Using IBM’s Watson program, the USTA’s high performance coaching division since the late summer has been digesting thousands of hours of video footage to create customized training reports for top pros and junior players.
AI, or machine learning, can quickly digest metadata and learn how to process and analyze it. Now, video is added to the mix of content for AI to examine.
“We are treating video really like a rich data source,” said Elizabeth O’Brien, program director, IBM Sports & Entertainment Partnerships. “How can we actually see all the things that are hidden in video and turn that into data.”
The additional information and video capability allow players like Madison Keys to dive deeper on their play.IBM
Martin Blackman, the USTA’s general manager of player development, said the data has actually helped create a new statistic that measures acceleration and deceleration of players.
“We are able with Watson to look at a player’s acceleration, movement and speed over the course of the match and show them how their court position improves when they are moving at an optimal level,” he said.
Before Watson, the USTA had to manually go through player footage. For a typical 90-minute match, a “tagger” would spend two hours pulling out unforced errors, forehands, backhands and even identifying the exact time of the start and end of each point. By contrast, Watson takes two minutes to do that while analyzing trends and patterns. That means the USTA can deliver nearly instant analysis to players shortly after their matches.
The coaching supplement to AI is the brainchild of Stephen Hammer, IBM distinguished engineer, sports CTO. He helped develop the highlights program and realized it could be used to help player development.
Asked if IBM had used Watson to digest video footage elsewhere, Hammer pointed to one other example. Watson created the trailer for the horror movie “Morgan,” with AI determining the most emotional scenes that were used in that preview.
Similarly, Watson can go through a match and, instructed what to look for, quickly pull out what is required for the coaches. For example, Blackman instructed Watson to focus on points after unforced errors to see whether a player performed better, the same or worse (unforced errors are player mistakes, like a double fault).
“Those are just things we never could have done before, and it is so cutting edge because everyone talks about how important the mental part of the game is, but it is very hard to measure and quantify it and this is a real tangible way of doing so,” Blackman said.
IBM is a U.S. Open sponsor. The USTA owns and operates the Open and runs player development centers across the country. Watson is now being used at those centers.
Often, outside coaches and parents will send video footage to Blackman. Before Watson, watching all that unsolicited footage was not possible. But now Watson, which can measure racket speed, foot movement and accuracy, can alert him whether there is video worth watching.
