Man Vs. Machine: Is Soccer Ready For Artificial Intelligence?

Man Vs. Machine: Is Soccer Ready For Artificial Intelligence?

Espanyol v Stjarnan – UEFA Europa League Second Qualifying round: 1st Leg

A head coach sees his team take the lead in a crucial match with 20 minutes remaining.

One assistant advises the coach to drive his players forward, to keep attacking in the hope of a decisive second goal. The other favors more cautious tactics, dropping players back to focus on defending the narrow lead.

The coach might draw on his past experience to make his decision or he might just go with instinct — his gut feeling in that moment of what will win the match.

Or, he might ask a machine.

Such a scenario would be unthinkable to some members of soccer’s old guard, for whom gut instinct trumps stats every time.

Soccer has been famously slow to embrace technology in decision making and, while the sport is waking up to the possibilities, clubs are still learning how to best use the data they generate.

But a growing number of consultancies are demonstrating how enhanced analytics can create marginal gains.

Olocip, a Madrid-based consultancy, uses artificial intelligence to help clubs and players “make the best decisions” in match strategy, player scouting and injury prevention.

The company is led by Esteban Granero, a 32-year-old midfielder who plays for Espanyol, in Spain’s La Liga.

“From my point of view, inside the clubs, I could see how they started to collect data but they were lost in the way of finding the utility of that data,” Granero, who began his career at Real Madrid, told me.

“The best solutions that the clubs could find in the market were tools that provided graphical and descriptive analysis of the data. However, in order to extract the most useful information from the data, it is necessary to rely on AI.

“This will allow the clubs to use predictive and prescriptive analyses to reduce uncertainty and make better decisions.”

Aiming to operate as an “external AI department” for clubs, Olocip offers three “solutions systems” that harness machine learning, a branch of AI, to predict future outcomes. Clubs can also ask the researchers to focus on other specific issues they want to address.

Olocip lead data scientist Marco Benjumeda, who has a PhD in AI, said AI had an advantage over descriptive data analysis in scouting players due to the “context” it provides.

“If you want to buy a player, you are not really interested in what the player has done in his previous club, you are interested in what will he do in your club,” he told me.

“On one hand there’s the style of play of the club, which clearly affects the performance of the players as far as we have seen. But there is also the level of the competition the player is in. Two players might produce similar variables but the level of competition they are in might be clearly different.

“We take all of this into account and the models learn from the data how to transform a player described by his past performance in the teams he played in, into a prediction for his new context.”

Olocip used its models to predict Real Madrid summer signing Eden Hazard will score more goals but assist fewer this season for his new club than he did for Chelsea, for example.

The TCT Coach application offers real-time strategy advice during a match.

“You can be at a certain moment in a game, for example you have just scored a goal. The model uses the data of what has happened in the game and we use variables that were designed by football experts to try to describe the current state of the game,” Benjumeda said.

“Then the AI model predicts a picture of what will happen in the next 15 minutes of the game. And you can interact with the variables to check how changes in some of the variables may affect the rest.

“For example, setting the pressure higher or trying to keep more possession or increasing the speed of passing or playing more in the center of the pitch.

“The model estimates how changes in these variables would modify the probability of obtaining your goals. By doing that, we can also automatically generate instructions for the coaches.”

Soccer’s reputation as an uncertain, unpredictable game is one reason the above scenario would be seen with suspicion by some stalwarts of the sport.

The head coach’s experience of being in similar scenarios or instinct for what he can see in front of him should be enough, they might argue. His “soccer intelligence” may be more valuable than that constructed by an algorithm.

 Olocip and similar agencies are not seeking to remove humans from decision making, but rather show how AI can “support” human decisions, Granero said. Less man vs. machine, more machine helping man.

What’s more, Olocip say players and clubs are ready and willing to embrace new scientific insights. It won’t say which clubs they work with specifically but “several teams from the main European leagues, several representative agencies, investment funds, and other sport projects” are using their models.

“Most of the clubs understand that science can help,” Granero said.

“AI can greatly reduce uncertainty and rejecting this technology may cause them to miss plenty of opportunities.

“Sooner or later every club in La Liga will be supported by AI technology, we are just the pioneers.”

[“source=forbes”]