How much is a great goalkeeper worth?
It's an article of faith in the analytics community that GKs are undervalued, with the occasional exception of a recruiting blunder like Chelsea's €80m signing of Kepa from Athletic Bilbao. Great GKs just aren't as salient as star strikers; when GKs excel, they get a low-key high-five from a CB rather than a pile-on celebration at the corner flag. But how can we put numbers to this notion?
By taking a GK's game apart piece by piece, we can start to calculate how much he or she is really worth. Here we'll look at shotstopping. As our FAQ explains, we use a metric based on the chance that a GK will concede from a batch of shots where a generic GK would concede 50% of the time. We generate this metric for five different categories of shots, handicapping the shots using expected goals and also by simultaneously rating the finishing skill of the strikers.
To figure out how these metrics translate into extra goals prevented, we have to do a sort of thought experiment. Let's imagine that a GK faces a typical batch of shots – by what percentage would his chance of stopping each shot have to rise or fall in order to match his metric? In other words, if the metric says he'd have a 45% chance of conceding from the entire batch, what would that mean for each shot? It might mean that you'd have to multiply the chance of stopping each shot by, say, 102% so that the chance of conceding from the whole batch was 45% instead of 50%.
Once we figure out this change in the chance of stopping each shot, we can apply it to all the shots that the GK would be expected to face during the season – and their expected goals, since different clubs allow shots of different quality. As an example, let's take the shotstopping phenomenon that is Jan Oblak:
We can see that Oblak is well above average as a shotstopper except when it comes to headers from dead balls. In particular, he is simply freakish at stopping non-headers in open play, the biggest category of shots.
First we looked at the number of shots Oblak had faced so far in each category and scaled them up to a 38-game season. Then we tried to figure out how much his skill would affect his opponents' chances of scoring. Here's how we thought he'd fare by the end of the campaign, versus a generic GK in La Liga:
(Note: Products may not be exact because of rounding for this table.)
So Oblak may be worth a total of 13 goals to Atleti from his shotstopping alone. His enormous skill against non-headers in open play accounts for most of the difference. And it's interesting to note that his performance against direct free kicks hardly affects his goals conceded, which could be true because he's expected to face so few of them and/or because GK skill against these shots doesn't vary too much.
Before we go on, let's allow that number to sink in. If a single player is worth 13 goals more than a generic player, surely he's one of the two or three most valuable at the club. Are there many strikers who would score 13 more goals than a generic player from the very same chances?
At the top of La Liga last season, 13 goals were the difference between fourth and seventh. It was easily the difference between the Champions League and the Europa League, a gap of tens of millions of euros. In domestic leagues with merit payments for finishing higher up the table, the value would go even higher.
Or just imagine the value of an Oblak to a club facing relegation. That's something Sam Johnstone may know a little about. If West Bromwich Albion are relegated this season, it won't be because of his shotstopping:
West Brom signed Johnstone, a perennial Manchester United loanee, for about £6.5m in 2018. That's looking like a steal today.
We'll look at the value of GKs as outfield players in a future article. For now, the takeaway from this analysis is to look for the undervalued assets in the market. Smarter scouting means identifying GKs who are under-the-radar shotstopping experts and pick them up on the cheap. You might not notice the difference they make on a day-to-day basis, but you will when you look at the final table.
[Photo: Анна Нэсси]
