Behind the Scenes at Wimbeldon

IBM Watson might have had the busiest summer of anyone apart from Ben Stokes and Spider-Man, providing bespoke information and technology services to both Wimbledon and the U.S. Open on an unprecedented scale. Since its doubles partnership with Wimbledon began in 1990, the company has collected a net amount of 62,814,711 data points, which are spun up into insights, highlights and analysis.

IBM Watson’s data scientists and digital  editors at work at Wimbledon

IBM Watson’s data scientists and digital  editors at work at Wimbledon

“Wimbledon described the championships as like putting on a fantastic wedding,” says Sam Seddon, Wimbledon client and program executive at IBM. “So you’ve got to manage and micromanage the detail to make it the most wonderful day and experience possible. But they’re inviting 500,000 guests, they’re streaming it to 20 million guests, they’re doing it for 13 days and they need 24/7 security.”


IBM Watson supplies the technology elements needed to ace this objective and to enable Wimbledon’s emerging media enterprise. “They’re starting to set up a completely integrated broadcast and digital capability,” explains Seddon, citing the increasingly disruptive role sport is playing in the media business. ”The FIFA Women’s World Cup, British Grand Prix, Tour de France—they’re all happening at the same time so there’s a huge amount of focus on getting eyeballs. So the way that they’re doing that is by essentially turning themselves into a media business and thinking and acting like a media business.” 

At the same time, social media organizations like Facebook and telco giants such as AT&T are branching  into sport broadcasting and streaming, hence Wimbledon’s attempts to smash into the broadcast arena. “What we help them with is all of the enablement layers that allow them to put on the show,” says Seddon. “To a large extent that starts with data—whether that’s photography as data or stats as data, or whether that’s video.” 

IBM Hybrid Cloud is used to disseminate that data to 200 territories around the world, distributed from IBM Watson’s on-premise Cloud to mobile phones, tablets and desktops; 49.5 million users visited the website in 2018 over the course of the two-week tournament. The site is home to digital experiences designed and built by IBM Watson, including live scores and the MyWimbledon guest experience program where members can personalize their web portal. Users can also vote on the content that’s available on the large-screen TV on Henman Hill. “So whether you’re a debenture holder, or whether you’re a hospitality user, or whether you’re a player or whether you’re a fan, you’ll all come into a single access point,,” says Seddon.


Another major role the company plays is in security, having helped to prevent more than 200 million cyber threats during the 2018 championships. Seddon uses an example of a security event from March 2019 where QRadar identified 663 different IP addresses associated with the attack. “IBM Watson for Cyber Security is able to distill that down to just 18 IP addresses in this set of particular events, which were then deemed to be a coordinated attack.

“What that allows us to do is be a lot more focused in terms of how we’re going about analyzing and tracking the source of this and being able to then put further measures in place if needed to ensure that nothing comes through,” he continues. “Since we put this in we’re about 60 times quicker on being able to respond to cyber security events, which from Wimbledon’s point of view is something they’re really, really focused on because their whole business model is built on quality.”


Perhaps the most advanced aspect of IBM Watson’s solution is its AI deployment, capable of generating a postmatch highlights package in just two minutes. During the 2018 tournament, the company analysed over 2,000 hours of video content to generate hundreds of these packages for the website and other broadcasters. “We want to create highlights packages faster than a global news organization, we want them to be high quality, we want them to be Wimbledon quality, and we want them to be exciting,” Seddon remarks. “So how do you define excitement?”

Seddon describes the three sources of data used to make this judgement: sound coming from the umpire’s chair, video of the players themselves and match data.

“By looking at those three things together we’re able to define per points, how exciting is it by: are the crowd really animated? Is it a key point in the match? Are the players gesticulating, are they excited and fist pumping?” Using these metrics, IBM Watson ranks every single point by excitement, and packages the highest scoring into highlights along with automated graphics.

The 2019 championships saw the introduction of two new AI components. The first is Watson Acoustics, a machine learning tool trained to listen for the sound of the tennis ball on the racket. “That helps us from a clipping point of view,” explains Seddon. “Once the ball hits the racket you know the point’s started, and from an editing point of view that allows us to be very tight on the clipping, which is important to Wimbledon because they have only a certain amount of rights footage themselves, because they give that out to the BBC and worldwide broadcast partners.

“The other thing that we’ve introduced is an integrity solution,” he continues. “At the end of a match the crowd will disperse, they’ll go and get a drink, top up on Pimms, whatever it is they need to do. And it doesn’t matter that we’ve got another great quarter final coming on straight after it, the crowd noise in the stadium will be very, very different in the first period of that match. So what this allows us to do is moderate that, ensure there’s complete fairness for the next people that are coming on the court, and that their highlights are as high quality as the match that’s playing out. So just because there’s not much crowd in there, it won’t have an impact on that.”


IBM Watson captured approximately 4.5 million points of tennis data during the 2019 tournament. These are input by trained tennis players, whose knowledge of the game allows them to see ahead of the point. “If there are any issues around the grounds then what they can do is call through a radio comms panel, talk to the operators that are watching the game, say ‘I’ve got an issue with 592 on whatever court’—so set five, game nine, point two.' The operators can then replay said point and correct the system where necessary. 

“It’s also very valuable to the players,” adds Seddon. “They get a match report within 20 minutes of the match finishing, but they also get a video file with the match analytics embedded into that file. So they can then jump through that and look at all of their break points or their winners or whatever it may be. By the time they’ve got out the shower they’ve got access to it on a mobile phone or tablet.”

The press also make use of this data via IBM Watson’s bespoke reporting service, as Seddon explains: “We’ve got results back to 1877 and all of our statistical data back to 1990. We’ve got a screen called the ‘Did you know?’ screen in the digital teams room that’s on a rolling ‘we found a new thing’, just helping seed ideas for content and prompt ideas.”


How much divergence is there between the solution deployed for Wimbledon and that of the U.S. Open? “They’re both different unique client requirements,” says Seddon. “Some of the digital convergence is new here, that’s very much a Wimbledon thing; the Acoustics and integrity solution are new here. But the core part obviously like scoring systems are common between the two environments—but Flushing Meadows is a very different beast altogether as a venue and as an organization. 

“It’s a year round process so we’re constantly innovating, so we actually have a slightly offset delivery program; it’s about 18 months,” continues Seddon. “We do a kind of random ideation session that if we could dream up how to fly to the moon what would we do? And then we’ll come out of that with some proof of concept we want to do, give ourselves time to test those, maybe develop some prototypes in the six-month window, and then we’ve got a nine-month window to get production ready for it,” he concludes.