In the first leg of their Round of 16 Champions League tie at Sevilla, Manchester United came away with a 0-0 draw. On the surface, this may seem like a characteristic José Mourinho performance, shutting down the opposition with defensively minded, pragmatic tactics, and setting the stage for a grind-it-out victory at home.
While Mourinho may well have intended for his team to do exactly that, they did not earn a draw by shutting down Sevilla’s attack. The Spanish side had 25 shots, eight of which were on target, and the hosts also possessed the ball 57 percent of the time. United, meanwhile, had just six attempts, and only one of them was on target.
Nonetheless, they will head into the second leg at Old Trafford knowing that a win by any scoreline will take them through because of one man, David de Gea, and, in particular, one spectacular save.
The Sevilla match is an extreme example of the story of Manchester United’s season. Despite conceding the fewest goals in the Premier League, Mourinho’s defense has been uncharacteristically bad when it comes to conceding chances. Ahead of Sunday’s huge match against Chelsea, they are fifth-best in the league when it comes to expected goals conceded, which uses a model that assesses the quality of a chance based on its location, the defensive pressure on the shooter, and a host of other factors. Their roughly 32 xG conceded is almost double Manchester City’s 19. The only top-four contender to have gifted their opposition better chances is Arsenal (around 37 xG against), while Liverpool’s defense has kept pace with Mauricio Pochettino’s steely Tottenham, as both teams have conceded roughly 26 xG thus far.
For my last piece for The Ringer, I created a second xG model (xG2) that included the quality of the shot taken in addition to the quality of the chance itself. If a player gets a good chance but misses the target, it will get 0 xG2. If he puts it into the top corner with pace, it might get 0.8 or higher depending on the situation. In that context, it was used from the point of view of the shooter, but it can easily be inverted to assess the impact of a goalkeeper. If a team is conceding fewer open-play shots on target than the shot quality model expects, their goalkeeper is saving shots that otherwise might have gone in. As it turns out, United have been relying heavily on their Spanish savior. They have conceded 10 fewer open-play goals than the shot-quality model would expect, given the shots on target that de Gea has faced.
In absolute terms, de Gea, who has played every single minute available to him in the Premier League this season, has saved Manchester United roughly 10 open-play goals. Lukasz Fabianski and Nick Pope—of Swansea City and Burnley, respectively—follow the Spaniard in terms of xG saved this season, both saving their clubs around seven open-play goals. Importantly, though, de Gea has faced fewer shots than Fabianski and Pope; if anything, absolute totals underplay his performances. Although the United defense is bad relative to its immediate competitors in the top six, it is still good in the context of the league as a whole. To see the proportional impact that goalkeepers are making, it makes more sense to divide their open-play goals conceded by the expected goals from the shot-quality model.
Spectacularly, de Gea has been conceding just 60 percent of the amount of goals that might be expected given the quality of the shots on target he has faced. The average at the team level has been 99 percent this season, and West Ham United have conceded a league-worst 115 percent of what might be expected.
In the Arsenal game in December alone, de Gea “saved” United roughly three goals (compared to what the average keeper would allow), conceding a lone goal instead of the four that might have been expected given the shots on target he faced. The Spaniard effectively scored the goalkeeping equivalent of a hat trick—except, this feat is much rarer than the outfield version. In the league this season, only Nick Pope’s performance in Burnley’s 3-0 loss to Manchester City saved his team more goals.
In the same way that most goal scorers’ shot conversion fluctuates from season to season, goalkeeper shot-stopping is notoriously prone to regressing to average conversion. Nonetheless, there are elite players who manage to over-achieve for extended periods, and these particularly talented shot-stopping goalkeepers tend to show up after a couple of seasons of data.
For a presentation at this year’s OptaPro Analytics Forum in London, where the cutting edge in soccer analytics is on show, Marek Kwiatkowski developed an augmented expected goals model that includes the identities of the players involved in a shot. The model is aware of who takes the shot, what goalkeeper is facing it, and who is managing both teams involved.
Strikers who are particularly adept finishers—such as Lionel Messi or Harry Kane—increase the chance that the shot is a goal by taking it, whereas bad finishers, such as Sevilla’s Jesús Navas, detract from a chance by being worse than the average. By the same token, goalkeepers who have historically been elite shot-stoppers over the sample (since 2011-12) reduce the chance that a shot goes in just by facing it, and vice versa.
De Gea is one such goalkeeper. The model predicts that a shot with a 25 percent chance of going in against an average goalkeeper would only go in 20 percent of the time against the Manchester United man. In effect, a 1-in-4 chance becomes a 1-in-5 if it is taken against the Spaniard.
“David de Gea is undoubtedly one of the stand-out shot-stoppers of the current decade,” Kwiatkowski said. “Based on my analysis, since the start of 2011-12, he has saved Manchester United 36 goals compared to an average keeper, which puts him behind only Hugo Lloris of Tottenham, Samir Handanović of Inter Milan, and Manuel Neuer of Bayern Munich in terms of total goals during that time frame.”
Goalkeepers are notoriously undervalued in soccer, so de Gea’s achievement this season of almost single-handedly dragging José Mourinho’s side into the Premier League’s Champions League spots is likely to be underappreciated. After all, this is a sport where popular coverage still judges the men between the posts in terms of clean sheets which, generally, is more a function of a defense’s ability to prevent shots than it is a goalkeeper’s to stop goals.
In a recent piece on the value that goalkeepers might provide teams in the transfer market, The Economist used analysis by an analytics consultancy, 21st Club, that claims that the best goalkeepers in the world are worth just four or five points a season to their team, whereas strikers can be worth more than twice as much. However, unlike sports like baseball with discrete events that comprise a full game, the performance of players and their subsequent impacts in soccer are too interdependent for this to be a convincing claim. De Gea’s potential impact is invariably tied to his defense’s performance: If they concede a lot of chances, as they have this season, he gets more chances to save goals.
Only when the defense fails at its job can goalkeepers in soccer do theirs. It is often said that star athletes raise the level of their teammates, but those in gloves cannot do that. Manchester United have conceded the fewest goals in the league in spite of their defense, not because of it. De Gea’s jaw-dropping saves this season do not make Phil Jones and Chris Smalling any better at defending. If anything, their misgivings are the only reason we get to witness his brilliance.
Bobby Gardiner is a product manager at Football Whispers. He writes about football through an analytical lens, with a particular interest in team tactics.
This article was written with the aid of StrataData, which powers the StrataBet Sports Trading platform.
A previous version of this article incorrectly stated that De Gea saved an outsize number of shots while Manchester United were either drawing or winning.