Computer Vision's Growing Role in Sports Data
Computer vision is playing an increasing role in sports data—and by extension, sports gaming.
The use of computer vision to analyze where players are on the court or field holds major potential to improve sports data.
Second Spectrum popularized visual data to help teams understand things like which exact locations on the floor a player shoots best, or which types of plays most result in points, or which players perform best in pick-and-rolls, and so on.
But new computer vision technology is taking this data to a new level by analyzing players locations on the court and their actions in real-time. This opens up whole new benefits for sports leagues, teams and sports gaming.
Sportradar, for example, said in its earnings report this week that it has launched computer vision technology for table tennis that can capture stats and data for real time betting. This could enable consumers to “place bets within the table tennis volley, for example, on a number of bounces or the last player [to] hit the ball, with rapid playing cycles of less than one minute,” CEO Carsten Koerl said.
This computer vision technology can capture a wealth of data on individual players and result “directly into probabilities of the two balls that are played and make some predictive models, which are working pretty accurate on who might win this ball or who might lose it.”
In other words, companies could offer odds and lines thats adjusts based on real-time data that the computer vision is capturing during game play.
In terms of other sports, Koerl said team sports with more players—and a larger space with more data to capture—like the NBA are harder. “You have more players on the pitch. But having that detailed information, of course, will drive better predictive models,” he said.
Sportradar and the NBA last year announced a data deal for the 2023-24 season which included an expansion of distribution rights for player tracking data. “This will include utilizing data to create a deeper understanding and appreciation of the game, developing new data products to enrich fan engagement, and revolutionizing how sports betting data is utilized by betting operators and media partners,” the release said.
While the exact kind of NBA player tracking data to be used was not defined in the release, some believe it is optical tracking data, which sounds similar to the table tennis computer vision data Sportradar described in its earnings call.
Furthermore, Sportradar could use this tracking or computer vision data of where players are on the court to inform lines and odds. Other companies are also surely interested as well. This could enable data on things like: Is a player moving faster or slower than he normally does—which could affect his stats? Is he positioned near the basket more than usual, which could result in more rebounds? Is he standing outside the three point line, which could result in more three pointers? This could all be fed into lines and odds in real time.
Meanwhile, the NBA last week announced a deal with Sony's Hawk Eye, a company providing 3-D visual tracking technology, which will replace Second Spectrum, which previously provided the NBA with data that located players’ locations in a single point.
The Hawk Eye technology will be used to help referees make calls or to help teams understand athletes’ movements. And Hawk-Eye will also work with Sportradar to generate tracking data, ESPN reports. Does this mean this data will eventually also end up in player props in some way?
More broadly, this type of visual data could be combined with other non-visual data—i.e. existing data on player statistics—to provide powerful data on player performance.
It seems inevitable that companies will eventually use computer vision to analyze game play and offer micro market odds and lines in real-time for major sports like basketball, baseball and football. It should provide much more data for daily fantasy companies and sportsbooks to have a more precise real-time understanding of players’ performance. And it should also increase the number of markets for consumers.