The organizers of Wimbledon, the world’s oldest tennis grand slam and most prestigious grass court tournament, think some of its followers don’t know all that much about the modern game.
It’s not intended as a slight; not everyone can be an armchair pro.
“We did some research a few years ago, which demonstrated to us that most people who engage with Wimbledon are actually not year-round tennis fans,” says Alexandra Willis, director of marketing and communications at the All England Club, which hosts the tournament.
“What we heard anecdotally was, ‘I’ve heard of a few top players, but I actually haven’t heard of many others’ and ‘this all feels a bit confusing and bamboozling,’” she adds.
It’s understandable. Tennis is experiencing an era in which the men’s game and to a degree the women’s have been defined by a small quota of dominant players with astonishing career longevity.
To fill the knowledge gap, the All England Club has teamed up with IBM to use artificial intelligence (AI) and big data to boost fan engagement – and try to predict every match winner in the process.
Think Moneyball, only aimed at the fans.
As part of the “Match Insights with Watson” feature on the Wimbledon app and Wimbeldon.com, an ever-shifting “IBM Power Index” ranking has been assigned to each player, courtesy of IBM Watson, the company’s AI for business.
The ranking is generated by analyzing athletes’ form, performance and momentum, explains Kevin Farrar, sports partnership leader at IBM UK & Ireland. “Because it’s updated daily … you can see (players) to watch, (and) it can start to identify potential upset alerts – all interesting to the fans,” he explains.
The idea is to help less-initiated fans to find players to follow, “developing their own fandom,” says Willis. Users can choose to track players and are served up personalized highlights as the tournament progresses.
Watson’s party piece is using data to predict every match winner. Displayed as a simple percentage likelihood, the AI makes the call by drawing on millions of data points recorded before and during the tournament. Factors include previous results between the athletes, current form, and more granular details like first serve win percentage, ace frequency and percentage of points won returning first serve.
Farrar explains that tournament data is compiled by a team of “very good tennis players” – usually county level and above – watching each match at Wimbledon, with three statisticians on show courts and one on the outside courts. Hawk-Eye ball and player tracking is also used.
Not all data fed into the predictor is based on hard stats, however. Intriguingly, positive or negative media sentiment is also taken into account, scanning thousands of news articles about players.
“One of the markers of ‘who’s interesting?’ is ‘who is the media excited about?’” says Willis. “Many members of the media, particularly in a sport like tennis, where they’re with the players week in, week out, have a sense and an understanding of how well people are playing – those sort of soft factors that don’t necessarily show up in (structured data points).”
Farrar reported that Watson predicted results with “pretty much 100% accuracy” on day one of the tournament, but day three provided its first big upset when women’s number 2 seed and 66% match favorite Anett Kontaveit was beaten by unseeded Jule Niemeier in straight sets.
Despite employing one of the world’s most famous AIs, Willis insists “this is not intended to be exact or an exact science.”
And even if Watson loses, it’s still a win-win, insists Farrar. “That’s an interesting talking point, and it’s engaging with fans, which is the key goal.”
“Sports fans love debate. So we’re giving them something to debate about.”