Skip to Main Content

ESPN Previews Champions League Final With STATS’ AI-Based Goalkeeper Metric

By: Stats Perform

There are any number of places to look up the number of goals per match conceded by Alisson or Hugo Lloris. It’s 0.59 for the former and 1.0 for the latter in Premier League play. It’s about as easy to dig up their save percentages for the season – 76.3 percent for Alisson and 74.4 for Lloris.

That only gets us so far in comparing their ability because they face so few shots in the average match and they of course don’t face the same shots. Football isn’t going to drastically change to all of the sudden make it easier on those of us evaluating. And keepers of course will never face the same shots, but STATS has worked toward solving this problem of evaluating performance though an equal sample, as ESPN explained leading up to the Champions League final a season after goalkeeping played an infamous role in Liverpool falling just short of winning Europe’s most coveted club trophy.

“On top of the scarcity of events, there’s an issue of context,” Ryan O’Hanlon wrote in the ESPN article, ‘Alisson is the real reason Liverpool are in the Champions League final,’ featuring STATS AI Scientist Paul Power’s work. “The style of defense a team plays will affect the type and total of shots they allow. Stopping shots for Burnley – a side that concedes a ton of chances but tries to keep as many men behind the ball as possible – is a vastly different exercise than stopping shots for Liverpool, a team that doesn’t allow many shots but has fewer bodies back in defense whenever they do. Traditional numbers like ‘clean sheets’ or ‘save percentage’ don’t account for these effects, and they also don’t account for the quality of the finish.”

Artificial intelligence, and more specifically machine learning, is unlocking methods to get beyond the basic.

“[Paul] Power, however, is working to change how keepers are assessed and, in turn, how they’re valued,” O’Hanlon wrote. “At STATS, he helped create a model that uses artificial intelligence to determine what each keeper’s specific skills actually are.”

Since the beginning of April, STATS has ranked every keeper in various leagues. With this method, we can conclude Alisson was well worth the then-record £56.25 million Liverpool paid. He’s been the best shot stopper in the Premier League, preventing an average of 0.31 goals per match more than the average keeper.

“In the 2017-18 season, when Liverpool conceded shots, the shots were very dangerous,” Power said in the piece. “So Liverpool needed a goalkeeper who was able to cope in one-on-one situations where the defense just completely collapsed and the keeper had to do something amazing. Alisson’s true strengths were that he was able to make these kind of superhuman saves. He would have saved at least seven goals that [Simon] Mignolet or [Loris] Karius would have conceded.”

Spurs keeper Lloris has also been significantly above average at 0.21 goals prevented per 90 minutes, ranking fouth behind Alisson, David De Gea (0.27) and Ederson (0.22).

“In terms of actually picking out shots into the corners, Lloris is exceptional,” Power said. “Where he sometimes gets caught are shots that are actually straight at him. He compensates for shots that go through him by being able to reach these really high-probability shots that should be goals.”

Power’s method, “Trading Places – Simulating Goalkeeper Performance using Spatial & Body-Pose Data,” has also appeared in FourFourTwo.