Jürgen Klopp was in his third week as Liverpool’s manager, in November 2015, when the team’s director of research, Ian Graham, arrived at his office carrying computer printouts. Graham wanted to show Klopp, whom he hadn’t yet met, what his work could do. Then he hoped to persuade Klopp to actually use it.
Graham spread out his papers on the table in front of him. He began talking about a game that Borussia Dortmund, the German club that Klopp coached before joining Liverpool, had played the previous season. He noted that Dortmund had numerous chances against the lightly regarded Mainz, a smaller club that would end up finishing in 11th place. Yet Klopp’s team lost, 2-0. Graham was starting to explain what his printouts showed when Klopp’s face lit up. “Ah, you saw that game,” he said. “It was crazy. We killed them. You saw it!”
Graham had not seen the game. But earlier that fall, as Liverpool was deciding who should replace the manager it was about to fire, Graham fed a numerical rendering of every attempted pass, shot and tackle by Dortmund’s players during Klopp’s tenure into a mathematical model he had constructed. Then he evaluated each of Dortmund’s games based on how his calculations assessed the players’ performances that day. The difference was striking. Dortmund had finished seventh during Klopp’s last season at the club, but the model determined that it should have finished second. Graham’s conclusion was that the disappointing season had nothing to do with Klopp, though his reputation had suffered because of it. He just happened to be coaching one of the unluckiest teams in recent history.
In that game against Mainz, the charts showed, Dortmund took 19 shots compared with 10 by its opponent. It controlled play nearly two-thirds of the time. It advanced the ball into the offensive zone a total of 85 times, allowing Mainz to do the same just 55 times. It worked the ball into Mainz’s penalty area on an impressive 36 occasions; Mainz managed only 17. But Dortmund lost because of two fluky errors. In the 70th minute, Dortmund missed a penalty shot. Four minutes later, it mistakenly scored in its own goal. Dortmund had played a better game than Mainz by almost any measure — except the score.
In soccer, pure chance can influence outcomes to a much greater extent than in other sports. Goals are relatively rare, fewer than three per game in England’s Premier League. So whether a ball ricochets into the net or misses it by a few inches has, on average, far more of an effect upon the final result than whether, say, a potential home run in baseball lands fair or foul or an N.F.L. running back grinds out a first down.
Graham brought up another game to Klopp, against Hannover a month later. The statistics were weighted even more heavily in Dortmund’s favor: 18 shots to seven, 55 balls into the box compared with 13, 11 successful crosses from the wing to three. “You lost, 1-0,” he said. “But you created double the chances —”
Klopp practically shouted. “Did you see that game?”
“No, no, it’s just ...”
“We killed them! I’ve never seen anything like it. We should have won. Ah, you saw that!”
Graham had not seen that game, either. In fact, he told Klopp, he hadn’t seen any of Dortmund’s games that season, neither live nor on video. He hadn’t needed to, unless he wanted to experience one of the breathtaking acts of athleticism that can occur in soccer, or the drama of two teams fighting to assert their will upon the other — the reasons, in other words, that most fans watch sports. To understand what happened, all he needed was his data.