U.S. Soccer brings analytics into picking players for national teams

Every time a national soccer team announces a roster for a game or tournament, some fans complain that some players shouldn’t be there and some others should.

There has always been an element of coaches' personal preferences in those decisions, and there probably always will be. But within the U.S. national team program, data analysis has become an increasingly important element.

At last month’s United Soccer Coaches convention, U.S. Soccer Federation data analysts Tyler Heaps and Joris Bekkers gave a presentation on how the men’s and women’s national teams have incorporated advanced statistical measures into the roster selection process. The event was hosted by Opta, the sports stats behemoth that tracks data on U.S. Soccer’s behalf.

Heaps said that national team coaches and Federation executives have embraced bringing more data into their decision-making process. It’s especially true for men’s national team general manager Earnie Stewart and new head coach Gregg Berhalter.

“They’ve both been great, and they’ve both been forward-thinking, and they want this stuff, and there’s an appetite,” Heaps said. “Working with them on a day-to-day basis, and figuring out what they value so that we can fit different models and different things into their style of play and what they want to do, will be huge for us."

The U.S. women’s program is just as interested. Heaps said the staff has been using analytics “to help in the decision-making process on player selection for certain camps. Opta has also tracked data for games worldwide that U.S. coaches want to analyze.

“We made team profiles for every single team that is in the World Cup,” Heaps said. “As they play more and more games this year, we can start building out player profiles for those teams in our group ... looking at style of play and how we think they’re going to set up to defend us and attack us.”

In the presentation, Bekkers highlighted a metric he helped develop to measure the impact of a player’s passing skills.

“We came up with an idea to measure whether a pass kept possession or not, by counting if the pass was still in possession of the team after 10 seconds or within three passes after the pass was made,” he said. “This will give us a very easy way to track if a player is helping with the buildup of the game, and all the actions that kind of fall out of that scope but still are valuable."

Heaps spoke about how the Federation can measure data on MLS players who’ve been tracked since their time in U.S. Soccer Development Academy youth clubs. The names were left out, though that didn’t stop some attendees from guessing.

For example, there was a player who has "done really well in the USL, and there’s a lot of buzz about him in the Twitter world, which I’m sure some of you have written about ... [but] hasn’t gotten his chance in MLS yet and is actually making a move overseas.”

There was also a centerback who “has made a very good young career in MLS and was just recently called up to the national team. ... His game has translated very well into the next step.”

“The biggest question for us is: How can we help influence technical staffs?” Heaps said. "How can we use this objective information to ask questions? ... How can we teach with what we have, and how can we translate it to those that are making decisions?”

He later added that analytical tools are especially handy when a top-tier player gets injured and national team coaches want to find a replacement with similar traits.

“Trying to find ways to use data to answer their questions, but also to fit their philosophy and fit their style, I think, is a huge thing,” Heaps said. “Doing that allows us to objectively look at the entire player pool in a very short amount of time, as opposed to looking at individual games.”

Heaps wants to help increase the use of analytics in women’s soccer. That includes testing whether data sets derived exclusively from women’s games will produce different statistical models from what’s been seen in men’s games. He cited expected goals, a widely used measure of attacking prowess, as an example.

“We get asked all the time from our women’s staffs [when] we give them an xG number ... ‘Is this men or is this women?’ ” Heaps said. “The games are different. I think having that model will open up some really cool things in terms of what’s out there and what is different. How does a shot from the top of the 18 [yard box] differ, etc.?”

Unfortunately, there aren’t enough data points in women’s soccer yet to create the quality of model that’s desired.

“I think we’re still far away,” Bekkers said. “Right now, just to give you an idea, we’ve trained our expected goals model on some, I think, 200,000 goal attempts for the men’s game. I don’t see a data set being close [to that] for the women’s game in the near future.”

He is at least sure of this: America’s soccer culture has been much more open to analytics than the European clubs he used to work with, including England’s Bournemouth. Bekkers said some at Bournemouth weren’t too welcoming to his ideas when he got there, but the club has changed their ways since then.

“Some clubs are looking into it more, because it becomes a trend and they see that they need to do something with it,” the Netherlands native said. “But they might not even have the financial means or the capabilities to figure out what they actually need to do with the data that they then get."