Wednesday 27 June 2018 3:02 pm

Deep learning: Belgium's secret weapon for England's World Cup group decider

England manager Gareth Southgate has long been be familiar with the threat posed by some of the supreme talents in the the Belgium squad such as Eden Hazard, Romelu Lukaku and Kevin De Bruyne.

Yet the Red Devils have a less well-known weapon at their disposal for tonight’s World Cup match in Kaliningrad, effectively a shoot-out for top spot in Group G: artificial intelligence technology designed to identify their opponents’ weaknesses.

Belgium, coached by Roberto Martinez, have been using the services of Dutch football intelligence company SciSports for the last five months, including at the tournament in Russia, in a bid to get an edge on their rivals for the trophy.

“We’re helping them to analyse their opponents based on data,” SciSports founder and chief executive Giels Brouwer told City A.M.

“We’re providing their technical staff with strengths, weaknesses of the teams they’re going to face.”

That might sound like standard fare in the data-drenched world of modern football, yet while England may know all about their opponent’s collection of Premier League stars, SciSports believe they will give the Belgium such an intricate level of detail about Southgate’s team that they might be aware of their weaknesses before the Three Lions coach does himself.

Most of the top nations at the World Cup, like the best European clubs, have their own performance analysts in tow pouring over the numbers from data providers such as Opta, STATS or Wyscout to help suggest way that their teams could improve.

Yet Belgium, prompted by former Everton and Wigan manager Martinez, have shirked the fallible limitations of a human mind in favour of computational deep learning provided by SciSports.

Just as machines trained in deep learning are proving to be more proficient than the greatest human minds when it comes to spotting cancer, winning the complex strategy game GO or reading comprehension, SciSports is betting they could be more astute football technicians than Pep Guardiola – never mind Southgate.

“We’ll take a look at how and where Belgium’s opponents are vulnerable,” says Brouwer, who believes Germany were the only other team at this year’s World Cup to be using AI.

“Where are the team’s weak spots? Are their backs playing too high so there’s space behind and goals come from that situation? Is there a weakness at set-pieces?

“For example, one of our models can analyse over 100m passes and automatically generate the impact of every individual pass leading to a goal within 15 seconds. We can find patterns in where the most effective passes come from.

“We can analyse formations and tactics and help them find the weak spots based on pattern recognition technology. It’s all based on fully automatic machine learning.”

Despite the end-point of such innovation potentially putting him out of a job one day, Martinez has been an influential proponent of the technology.

“The entire set-up came because of discussions we’ve had in the past,” says Brouwer.

“He’s really open to data as he showed in the Premier League. We had several conversation in the past about what he wanted to know and we built several models for him.”

It’s not just managers who have sought the services of artificial intelligence; players have too.

When Memphis Depay was struggling at Manchester United the Dutch winger and his agent sat down with SciSports to look for tactical reasons behind his patchy form and the kind of club that may better serve his talents.

SciSports drew up a model that named five clubs who would suit him – one of which was Lyon, where he is now thriving.

While almost unique now, Belgium’s usage of deep learning technology is likely to be repeated by a growing number of nations at future tournaments.

Data analysts are already set to have a more visual presence at this year’s World Cup after the game’s rulemakers IFAB ruled that small electronic devices such as tablets and laptops will be allowed on the bench for the first time “if used for coaching/tactics or player welfare”.

Coaches will be free to access live, real-time data as the matches are being played.

That’s one area where Belgium and SciSports won’t be able to hurt England just yet.

“The changes were just announced two weeks ago and our product development team was already tied up and wasn’t able to do anything in time,” he says.

Once the whistle goes, it’s still just Martinez and Southgate. Until then though, every move England have made is being watched by the machines.