When Bastian Schweinsteiger tried two shots from distance in the Champions League game against Arsenal in March 2014 (1-1), Pep Guardiola shook his head in disdain from the dugout. The German midfielder later hinted that attempts from distance were effectively banned under the Spanish coach. Bundesliga expert and pundit Franz Beckenbauer was enraged by that revelation in the TV studio. “I disagree completely,” the “Kaiser” said, “if you have the chance to shoot from distance, especially against a well-organised defence, then you have to take it. It’s a useful method.” There was a danger, Beckenbauer added, that Bayern would end up “playing like Barcelona, where you can’t stand to watch anymore because they even pass it to each other on the goal line.”
A lot has been written about the cultural differences Guardiola has been confronted with during his reign in Bavaria. More intriguing, however, is the question as to whether Beckenbauer’s accusation of Bayern over-playing the game was factually correct. Were Schweinsteiger’s methods really as useful as the 70-year-old thought?
The answer isn’t that hard to find if, like Ted Knutson, you know exactly where to look. The 39-year-old American has until recently worked as a analyst for Championship side Brentford and Danish champions FC Midtjylland, where he was developing statistical models for player and opponent evaluation as well as free-kick routines. Knutson also writes insightful articles for the influential statsbomb.com blog.
With the help of data from thousands of games and hundreds of thousands of shots, Knutson has calculated that the probability of a shot from 22 metres (24 yards) in a central position going in is only 3 per cent. “On average, a player would need 33 attempts from that position to score one goal,” he explains. “But if he manages to play in a teammate who finds himself one-on-one with the keeper, the probability of a goal rises to 40 per cent.”
Killer balls like that don’t come too often, and they are also very hard to pull off. In the stands, the crowd get restless when the final pass goes astray and the experts bemoan the team for over-elaborating. The data shows that they’re right to take that risk, however. “The probability of a goal is 13 times higher in one-on-one positions than from a shot from distance,” Knutson says. “Therefore, passing it forward remains the right option even if only one in ten attempts bears fruit.”
The numbers are pretty unequivocal in that respect: creating a few high-probability chances (from short distance, centrally, after a pass or a dribble) are expected to yield more goals than many chances of lower quality (shots from distance, shots or headers from crosses as well as from narrow angles).
Of course there are players who score difficult goals more often due to their vast skill, but their chances from bad positions are only marginally improved in comparison to lesser skilled players. Real world-class players, Knutson says, are incredibly disciplined when it comes to shooting. “Part of the reason why Lionel Messi is the best in the world is because he rarely shoots from further out than 18 metres or from non-central positions. That’s why he scores roughly with every fifth shot.”
Clever coaches have started to listen to the numbers. Borussia Dortmund coach Thomas Tuchel sought out Brentford and Midtjylland owner Matthew Benham in London to find out more about the importance of expected goals and other analytical concepts. And he’s instructed his team to play accordingly. According to Knutson, “The data shows that Dortmund create chances that have, on average, a 50 per cent better scoring probability than they did under Jürgen Klopp last season.”
The difference is most telling when it comes to the individual shooting stats of Pierre-Emerick Aubameyang. By the end of March, the Gabonese forward had only taken 14 shots more than last season (14 goals, 84 attempts) but had scored eight goals more from open play. How? By dramatically decreasing the amount of shots from low probability spots. His expected goals tally (penalties and dead-balls excluded) has risen from 0.14 (one in every seven shots) to 0.24 (one in every fourth shots).
“Statistics can at best round off impressions,” Bayern’s technical director Michael Reschke says, “a lot of the statistical data is unclear”. The parameters of football analytics cannot and will not replace traditional scouting, but they enable the confirmation of observations and prevent mistakes in the transfer market. “There are a players who score from impossible angles in a season, but after a while the law of averages catches up with them,” Knutson says, and mentions a South American forward who moved clubs within the Bundesliga.
Knutson explains that the relatively strong importance of luck in football – one deflected shot, one refereeing mistake turning a game – can lead to distortions in the table. “There are teams that create many high quality chances and concede very few high quality chances and still lose a few games, as Juventus did at the start of the season - or the other way around, like West Ham in the Premier League. People write about teams going on great runs or having fundamental problems but the data might say something completely different. After a while, the results tend to get back in line with the probabilities but sometimes it takes more than a season.”
For fans, experts and journalists, the game remains a story that is first and foremost being told backwards, starting from the end. But top coaches such as Guardiola and Tuchel have moved a step further, towards numbers that are relevant beyond the score sheet. For them, what’s decisive is the objective performance measured by the ratio of high-class chances, because that’s what they can influence - unlike a deflected ball or a poor refereeing decision.