Next Big Thing: ‘Watson’ the AI tech that puts you on the pitch and predicted Mahrez’s move to Man City
The Athletic, 9th February, 2020 - The Leicester City right-back has taken up a decent defensive position alongside Jonny Evans when he suddenly dashes past me and out towards the touchline to close down Andrew Robertson on Liverpool’s left flank. A huge hole is created in Leicester’s defence and with Demarai Gray trying to cover Sadio Mane, Liverpool break through as Evans and Roberto Firmino both miss Robertson’s pass.
As I watch, marvelling at the speed and movement around me, Mane is in on goal and before I can blink his shot is past Kasper Schmeichel and he is racing towards the far corner to celebrate with his team-mates.
This is the work of IBM’s master inventor Joe Pavitt of the emerging technologies department. His virtual reality (VR) technology will have many uses — I also had a brief stint in the International Space Station in my new world, which will be used to help train astronauts to fix problems when on board in real life — but it could have a huge impact on how football clubs analyse their opponents, and their own performances. It is as addictive as any video game, but it has a more serious and practical purpose.
It is technology Leicester are considering installing in their new, state-of-the-art training ground just north of the city when it opens next summer, although Pavitt admits there is work to be done before it is ready for full use. This is just a demonstration of its potential, after all, and what I have experienced through the headset and handheld remotes took just six days to programme using all the technical data and GPS data available, but its potential is vast.
Not only could I stand on the pitch and relive the key moments from the 1-1 draw between Liverpool and Leicester, one point of a laser towards an executive box in the stand and a click later I am looking down on the pitch from inside a room. A huge table-top pitch is behind me with the game playing out in graphics on it, while the actual television footage is shown on a screen on the wall. To the right are large menus, hanging like paintings on the wall, detailing all the technical analysis from the game that I can access just by asking for it.
“Show me all the Liverpool set plays!” “Show me Jamie Vardy’s key moments!” This is the natural evolution from the technical analysis that has made such a huge impact in football over the past few years.
“The idea of doing this in virtual reality is you get a clearer view of what is happening than on a 2D screen,” Pavitt, a West Ham United supporter, says. “You get more context of the players’ positions, and you get the idea of standing in the players’ shoes.
“The graphics aren’t exactly FIFA [the video game] standard yet, there is still work to be done, but this demonstrates what is possible with this technology. You get the sense of how fast the players and the ball move in real time.
“Players may review their performances by watching videos but this takes that analysis to another level as they can revisit moments in games so they can learn from mistakes and experience what they can do differently next time through VR. They can also look at their future opponents. It is a richer experience.”
The virtual reality is part of IBM’s Artificial Intelligence programme, “Watson”, which is designed to take performance data analysis to the next level. Traditional computer programmes are just that, programmed. AI is taught, meaning it has the ability to learn and adapt.
“The problem with elite sports is there is too much data,” said Pavitt. “With Watson, analysts can access relevant data much faster, which is handy when you are preparing for games and only have a couple of days to do so between fixtures.
“Rather than analyse three games manually, AI can help analyse 10 games more efficiently.”
As an experiment, IBM used Watson to help Isthmian Premier League side Leatherhead last season — and the transformation in their fortunes was remarkable. They went from losing six of their first nine games to finishing eighth in the table, narrowly missing out on the play-offs.
The key to Watson is to access the most relevant data quickly using “conversational interface”. Simply put, that means asking for it with Watson learning football terminology to produce the data quickly.
“We are in conversation with Premier League clubs about conversational interfacing and it is one of the key functionalities of the virtual reality,” Pavitt explains.
“Traditionally, you have all this data to look through using drop-down menus, sliders and filters. With AI, it can understand natural language so just ask it a question. For example, you could ask, ‘Show me the goals of Jamie Vardy’ and it will jump straight to the footage of his goals.
“You don’t have to have people sitting through 90 minutes after 90 minutes of analysis. You can just ask it for the information you care about.”
As part of opposition scouting, Watson can also read online match reports and even Twitter to understand more about each player and team, depending on how often each player is mentioned and can even distinguish between a positive and negative comment on that player’s performance.
The system is also being used in recruitment and can match suitable candidates for identified positions using its vast database. More impressively, it can also predict likely future moves.
Alongside the virtual reality platform at IBM’s Innovation Centre just outside Winchester in Hampshire, Pavitt demonstrates how. Using data available in 2016-17 it tries to match players to certain clubs depending on an array of criteria, including the amount of games played, length of contracts, player strengths and weaknesses. It correctly predicts Riyad Mahrez’s move to Manchester City, although it doesn’t see Danny Drinkwater’s transfer to Chelsea.
Watson has also been used to analyse tennis player performances at Wimbledon, for the England rugby team and by Fox Sports during last summer’s women’s World Cup. Though not everyone has been so quick to make use of it, with Pavitt saying only three Premier League clubs are using data properly.
“The perception of AI at elite level is interesting,” said Pavitt. “We are seeing differing opinions across different clubs.
“There seems to be two different views. We spoke to a Champions League side in Spain about this and they said they knew that AI would be massive in the sport in the next five to 10 years, but they didn’t know what it was or how it worked so they didn’t trust it.
“Then we have had teams who say they don’t understand it but just want to give it a go. The fear factor is interesting so we are going slowly in introducing the technology. There aren’t many clubs using AI or Watson at the moment.
“What surprised me when I started looking into elite sports a few years ago was how few teams in the Premier League have analysts. I don’t mean someone just watching games. They aren’t data scientists crunching numbers like this. There are just three clubs doing it properly that I am aware of: Liverpool, Manchester City and Leicester. It is no coincidence they are the top three in the Premier League.
“Watson has learned over time to distinguish where a play originates and, using data from the Premier League since 2017-18, it analysed that 20 per cent of shots on goal in the Premier League came from plays which began with a throw-in. Liverpool manager Jurgen Klopp recently appointed a throw-in coach. There was an uproar when he did that at the time but look where they are now! Maybe they have a point.”
Leicester have been using analytics since Nigel Pearson joined the club in 2008. Arriving with him then was Paul Balsom, the head of performance innovation, and the club has embraced sports science and data analytics for more than a decade. Last April, they appointed Mladen Sormaz as their first head of football analytics, responsible for providing data-driven insights into performance, recruitment and sports science.
“There is a lot going off for the new training ground,” Balsom, who is also performance manager for the Sweden national team and head of innovation for Belgian side OH Leuven, told the Training Ground Guru podcast.
“In terms of performance analysis, we have a lot of ideas which are quite futuristic. We have talked to companies, and IBM have been one of them, and the types of things they are telling us may be possible in the not too distant future — the whole virtual reality scenario where a player can walk into the analysis room with glasses on and position himself anywhere on the field in any situation.
“What we think will also be interesting with virtual reality is to put our younger players in those positions as well. So it could be a good feedback tool for our own senior players but certainly for a young player, we could use this technology to upscale them quickly.”
While Leicester are ready to embrace the possibilities of virtual reality analysis, the key remains to combine technological advances with practical improvements on the pitch.
“You can show a player something but it still, in my opinion, has to be translated to work on the field, work on the grass,” Balsom said. “That is where I think the successful coaches of the future will be able to understand the advantage of technology and ask questions of the technology and the data, get the information back but then will still be able to get out on the grass and do the job on the grass.
“Certainly not in the time I have left in football, that will not have disappeared by then.”
Read the full article here.
UKI External Communications