One of the main concerns for Chelsea in the next transfer window will be the goalkeeper situation.
Despite a world record fee, Kepa Arrizabalaga’s performances have been ... mixed ... with Frank Lampard briefly dropping the €80m-man to the bench, even, and starting Willy Caballero in his stead — for six straight games! Kepa was reinstated just before the suspension of the season, keeping back-to-back clean sheets against Liverpool and Everton, but his long-term future remains unclear. Neither is Caballero’s future ascertained at this point, but even if the veteran backup signs an extension to stay for next season, his role will be (and certainly should be) limited to just cup competitions, unless absolutely necessary.
While offloading Kepa would be a gargantuan task in and of itself, there’s no harm in looking at potential upgrades at the position. Chelsea have certainly been linked with a whole host of players to either replace or compete with Kepa.
I have tried to include data going back as far as 2017-18 for all the players, but that wasn’t always possible. Two players whom I really wanted to analyze were André Onana and Uğurcan Çakır, but the lack of available data for the Eredivisie and the Süper Lig meant that simply wasn’t possible.
First off, let’s look at per-90 numbers for goals conceded (GC/90; x-axis) and the goals expected to be conceded (PSxG/90; y-axis).
And here are the cumulative differentials for the PSxG numbers as a +/- metric.
Kepa does poorly here, though that should come as no surprise by now. What is actually surprising is how bad Manuel Neuer is performing in PSxG differential, with his number almost as bad as Kepa’s. It would have been really silly if our reported interest in the 34 year-old Neuer was actually true.
At the other end of the scale, Dean Henderson is performing remarkably well, easily outperforming what he is expected to save. This does explain why Manchester United are so keen on retaining him despite having a certain David De Gea amongst their ranks already. I have always had a bias towards Thomas Strakosha of Lazio, and it’s good to see him doing well in this metric.
Of course, the number of goals that a goalkeeper concedes or is expected to concede depends to a great extent on the defense in front of him and the midfield in front of them, so it’s important to look at additional metrics to try to get a better picture.
Below is a scater plot of Save% (y-axis) against PSxG/SoTA (x-axis). The latter attempts to show how many goals are expected to be conceded from each shot on target that a goalkeeper faces and is analogous to the xG/SoT that is used to evaluate attacking players. In isolation, the PSxG/SoTA only shows how good the defence is at restricting quality chances, but coupled with the Save%, it might shed some light on the goalkeeper’s role in that equation — easier shots likely lead to higher save percentages just by definition, or at least should.
For instance, while Jan Oblak does have a ridiculous Save%, his numbers are also helped by the ridiculously good defense of Diego Simeone’s Atlético Madrid. He gets to save a lot of easy shots, unlike say, Aaron Ramsdale of Bournemouth.
Dean Henderson shows well here, too, with a good Save% despite a lot of difficult shots, which again bears credence to his quality.
Next, let’s look at some passing numbers. Goalkeepers these days have so much more to do than just kick it long and far as often as possible.
While there are still teams that practice such archaic talents — and will thus affect a goalkeeper’s individual statistics in that regard (i.e. is the team set up to go long from goal kicks?) — it certainly doesn’t hurt to be able to hit an accurate long ball.
Slightly less archaic — though still greatly dependent on the defensive setup — is a goalkeeper’s ability to rule the airspace over his penalty area.
Kepa performs much better in long-range passing (as expected) and cross-stopping (not as expected) than any of the metrics, so we can surmise that the goals Chelsea have conceded from crosses and set-pieces aren’t necessarily Kepa’s fault, and can be attributed to our defensive shortcomings instead.
One note of interest is Gianluigi Donnarumma’s comfort with distribution and aerial balls, which could be a good thing in in light of Chelsea’s widely reported interest in the young AC Milan goalkeeper.
Last but not least, let’s take a look at some sweeper-keeper action numbers.
OPA is the number of times a goalkeeper ventures outside the penalty area for defensive or buildup purposes. Along with the average distance travelled from the goal, this tells us just how closely a goalkeeper sticks to the traditional ways of the position. Again, different teams will have different systems and different expectations from their goalkeepers in this regard, but it’s another data point that can be used to gain a better understanding of a player’s abilities and success rates.
Manuel Neuer unsurprisingly is at the top right corner of this chart, venturing far and often from his goal at Bayern Munich.
These charts do not paint a great picture of Kepa, though his non-shot-stopping numbers look decent. It’s also worth remembering the improvements he had shown once he was reinstated to the starting lineup, especially as we might find it practically impossible to offload the 25-year-old in coronavirus-infected football economy.
Chelsea could still choose to bring in another player to compete with Kepa, though goalkeepers generally do not like sharing their minutes, even more so than outfield players. The financial burden of adding another top-tier goalkeeper may make that unfeasible as well, which could mean that Chelsea will have to stick with Kepa, at least for the time being.
If he keeps repeating what he did against the two Merseyside teams last month, it could well be a turning point in the young man’s career. If not, he will fall well short of the legendary status he’s been aiming for since joining Chelsea.
All the player data and metrics used for this article have been taken from fbref.com.