Chris Middleton was court-side at this year’s Wimbledon tennis championships to see how matches are as much an AI-enhanced big data battle for the players themselves as they are a showcase of sporting excellence.
The Wimbledon lawn tennis championships – home of strawberries, Pimm’s, and, this year, a fortnight of unbroken sunshine – drew to a close on Sunday 15 July, overshadowed by the World Cup Final.
Internet of Business went backstage in the final week of the championships with IBM, during what was the 150th anniversary of the All England Lawn Tennis Club (AELTC). IBM has enjoyed a 28-year relationship with Wimbledon as its official technology partner, and has been providing cognitive services to the tournament since 2015.
Buried deep in Wimbledon’s broadcast centre are two studios occupied by IBM’s AI and digital media staff, where banks of monitors alert them to everything from an overhead smash from Nadal to an overseas hack attempt from Natal. All flash onscreen to be logged and analysed by the IBM team – smash and hack alike, complete with the direction of travel.
While the BBC has long been synonymous with UK coverage of the championships, since 2017 all on-court footage from the event has been provided by Wimbledon itself, which is now a fully fledged broadcasting and rich-media organisation, powered by artificial intelligence and big data analytics.
Like an on-court umpire, IBM’s Watson AI sits at the heart of this process, helping Wimbledon to grab highlights and action shots from its 18 tournament grass courts and make them available to global media channels in near real time.
As previously reported by Internet of Business, Watson has been taught to recognise players’ and fans’ emotions at the 2018 championships, and this new ability helps it to generate video highlights from key matches for Wimbledon’s digital and content teams.
With an average of three matches on each court every day, hundreds of hours of video footage soon pile up, which could take many more hours to pull together into highlights packages for fans, partners, and broadcasters.
Watson auto-curates highlights based on its analysis of crowd noise, players’ movements and emotions, and score data, to help fans and commentators alike focus on key moments in a match.
So if Novak Djokovic punches the air or Angelique Kerber puts away a winning backhand, Watson will see it, grab it, and rush it to viewers’ screens.
The players’ angle
But while all this may put Wimbledon at the cutting edge of AI, big data analytics, and rich media management for fans and broadcasters, Watson and the IBM team are also transforming the championships for another equally important group: the players themselves.
Indeed, Wimbledon is now as much a data analytics tournament as it is a test of sporting prowess. “We’re capturing a robust set of information and statistics from all the courts, and this is made available immediately to the commentators and to the players in the players’ lounge,” says Sam Seddon, Wimbledon client executive at IBM.
“And from this they can get information, say if they’ve come in with a particular coaching plan for that day, they can get all that from this system. And commentators will sound informed and intelligent, because they have this system in their commentary booth.”
So everything that happens on court from each player’s perspective is captured by the system: every shot, every winner, every position on court throughout the match, all become part of a big data audit trail that can be sliced and diced, not only by fans, but also by the players themselves to help them prepare for their next match.
All of this is captured on a point-by-point basis, says Seddon, linked to video files of the winning – or losing – shots, helping players to put together detailed analyses of their rivals’ strengths and weaknesses from what he calls “layers and layers of stats”.
And of course, players can see how they’re measuring up to their own high standards – whether they followed their own game plan, or didn’t put away enough shots at the net.
“We’ve used machine learning to profile different playing types,” Seddon continues, “and from this we can create head-to-head match-ups.”
So how was the system trained originally?
At the outset, the SPSS batched data and statistical analysis system acquired by IBM in 2009 would have been core to the process, explains Seddon. This was taught to recognise patterns and profiles in the data via machine learning. In this way, Watson came to understand the game of tennis itself – as a battle of competing stats and playing styles.
Watson was also trained with reams of historical data, including books on tennis – both structured and unstructured data, which was fielded by a new component, called Hudson, he says.
But where does the match data itself come from? The answer to that question is logical, surprising, and yet somehow ‘very Wimbledon’: up-and-coming tennis players sit in boxes at the court side – three on each show court, and one on each of the outside courts – and key in the data as each shot happens, using their expertise to recognise each stroke and on-court position.
So remember: behind every good AI system are reams of expert human beings. Not so much a black box solution, then, as a glass box one.
• Editor’s note: We apologise for publishing the incorrect image of the women’s championship winner earlier: an old image from our database was substituted by mistake.
Internet of Business says
Wimbledon has long been a show court for new technologies.
For example, IBM’s SlamTracker tool was redesigned for this year’s tournament. The aim is to provide tennis fans with new levels of analysis, insight, and engagement as each match unfolds, with mobile devices in mind.
Real-time data is integrated from multiple sources, including those court-side statisticians, together with chair umpires, radar guns, ball positions, player locations, and Twitter for social sentiment.
IBM’s SlamTracker works by analysing each player’s cognitive ‘keys’, to help spectators understand what tactics to look for in a match, revealing the hidden patterns in player and match dynamics, such as the pace of play or serve placement.
The tool uses Watson APIs to refine and update each player’s style, based on real-time match data, tracking how the momentum of play has shifted over the course of a match.