When I presented my initial framework to analyze goalkeepers in May, there were only two goalkeepers on the list of top 50 most expensive transfers. This summer, the two on that list (Ederson and Buffon) were topped by the moves of Alisson to Liverpool and Kepa to Chelsea. Clubs could finally be realizing the undervaluation of goalkeepers. Or maybe, Liverpool just had a panic attack on their walk of shame from Kyiv and Chelsea’s insecurities exponentiated following Thibaut Courtouis’ exit for Real Madrid. Nonetheless, £65M and £71M are by no stretch of the imagination cheap rebound dates. Especially, when Kepa was a second choice for the London clubs new number 1. According to The Sunday Times report, Chelsea made a £90M bid for Atletico Madrid’s Jan Oblak before signing Kepa.
Given that I have spent so much time in May preaching the under-emphasis on quality goalkeeping, it would be easy for me to praise Liverpool and Chelsea’s moves. And in fact, I went through Kepa’s numbers hoping to cherry-pick statistics glorifying the young Spaniard. However, what I kept seeing over and over again is that - from a statistical point of view, at least- the young, Spaniard is surprisingly average. For my first written piece in football this is a bold claim to make and in a couple of years it might bite me in the ass. But at the very least, I am going to explain why I don’t think Kepa is worth £71M.
In a simplified manner, you can divide a goalkeeper’s responsibilities into three factors: Shot Stopping, Collecting Crosses and Distribution. You can rank these factors in whichever order you wish, but the bottom line is that a goalkeeper will not play if he can’t make saves, command the pitch and collect crosses, and effectively distribute the ball.
With about 95% of the 2017/2018 La Liga season available through StatsBomb data, we are able to illustrate a fairly good picture of Kepa’s performance and tendencies. With this new data, we have attackers, defenders and goalkeeper coordinates on all shots; offering us more insight into goalkeeper style and ability.
Given that raw stats like Save Percentage and Clean Sheets are quite ineffective in their ability to rank goalkeepers, we present slightly more advanced metrics:
- The first is adjusted Save Percentage (adjSV%), which represents how much better or worse a goal keeper has performed compared to a “league average” goalkeeper. adjSV% is calculated as the difference between the total expected goals against and the actual goals conceded and divided by the total number of shots against on target.
- The second is the Goals Saved Above Average (GSAA) which is an estimate for how many goals a keeper has cost their team or saved their team, compared to a league average goalkeeper. This is calculated as the difference between the sum of the expected goals against and the goals conceded.
Note: We remove all penalties against in these analyses.
Of the 126 shots on target that Kepa faced, he conceded 37 goals with a total expected goals conceded of 31.35 He ranked 16th in La Liga in terms of adjusted save percentage and can be held responsible for approximately 5 of Athletic Bilbao’s goals against last season. Five additional goals against per season offers no justification for a £71M business expense, but at 23 years old, Chelsea most likely plans to hold onto Kepa for the long haul. And if I had the data available, I would love to show you how he compares to a young David de Gea or Thibaut Courtois, but we don't have that historic data just now.
There is another fair point to be made here. Shot stopping performances are hard to replicate. And in order to really evaluate a goalkeeper, you have to look at his athleticism, positioning, and physical abilities.
Diving a little deeper into these shots against, we look at Kepa’s positioning and more specifically how often he is in the optimal position. We define the optimal position by imputing every possible keeper position inside the 18, and using the shot’s characteristics like location, defender’s positions, attacker’s positions and other common factors to calculate the goalkeeper position that minimizes the probability of scoring.
Kepa managed to crack the top ten best positioned keepers in La Liga, which might be something Chelsea sees in him. Looking at the density curves above, Kepa positions himself nearly identical to Jan Oblak. The two young goalkeepers tend to put themselves close to the best position, more often than not. This plot potentially highlights one reason why Chelsea’s eyes were set on Oblak at first and later fell to Kepa.
On crosses, Kepa again performed close to average. Using another probabilistic metric, we aggregate the number of crosses each keeper attempted to collect and the number they were expected to have collected. This is a simple way of defining how aggressive a keeper is compared to the league average.
These are the top performers in La Liga and the Premier League ranked by the difference from expected percent collected:
Kepa ranked 25th between La Liga and the Premier League. To be fair, not all keepers need to be aggressive on crosses. In fact a lot of defensive systems prefer their goalkeeper to stay home and allow their center backs to sort out the chaos in front of them. One notable goalkeeper who plays like that is the indubitably, talented David de Gea. Aggression on crosses is better thought of a stylistic attribute that can add serious value to a team depending on the team’s style.
Let’s see how Kepa stacked up against some similar goalkeepers:
Given that Chelsea’s first target was Jan Oblak, Kepa is a substantial improvement in terms of controlling the box. But still, he is performing below expectation, not even reaching the levels to replace Thibaut Courtois and pretty far from the men at the top of the table in Ederson and Nick Pope. Let’s not forget that Kepa’s transfer fee was 36.3M more Ederson’s, the former most expensive keeper of all time.
Most recently in the ever-evolving game of football, a goalkeeper’s ability to distribute has been pivotal. Although Sarri’s system at Chelsea remains, for the most part, unknown. We anticipate Chelsea to play a lot like last season’s Manchester City, a “vertical tiki-taka” as some have termed it. That means a possession based system while taking every opportunity possible to go forward, attack with pace and to press high up the field. That being said, they need a goalkeeper who can contribute to the attack. A goalkeeper who is confident on the ball, confident under pressure, but still recognizes attacking opportunities and exploits them.
Following the plot above, we anticipate Chelsea to move into the top right quadrant under Sarri. Now what an ideal keeper for that system looks like we don’t really know, but we can build a profile for that system. This quadrant as a whole, took more shots, corners and direct free kicks in the attacking half than the other three quadrants on average. They seem to be high volume teams that take advantages of potential opportunities on the attacking end, even at risk of conceding chances against.
In order to profile Kepa’s distribution at Athletic Bilbao, we define a number of different distribution metrics. Since we expect Chelsea to play quickly in the attack, we looked at the proportion of positive outcomes that occurred within 15 seconds of a goalkeeper’s pass. Positive outcomes are defined as a possession that connected through the goalkeeper and resulted in either a shot, free kick in the attacking half or a corner. About 4.2% of positive outcomes in the EPL and La Liga connected through the goalkeeper, with a maximum of 10% in Burnley’s Nick Pope. Courtois has been very effective contributing to the attack last season, contributing to nearly 5% of Chelsea’s positive outcomes. Not to mention his collection and distribution led to the goal of the tournament in this year’s World Cup. Kepa similar to Courtois contributed 4.8% of positive outcomes.
Sarri will want Kepa to be involved in general play, and he will need to be sharp at his feet amongst high attacking pressure. Intuitively more robust statistics, are calculated by looking at goalkeeper’s actions when there is pressure applied. One of the most respected, distributors from the goalkeeper position is Man City’s Ederson. When Ederson is under pressure, his completion percentage only drops 5% points. Kepa’s change under pressure is much worse, his completion percentage drops 18.7% points when under pressure. To add some context to those numbers, the average drop in completion percentage when under pressure is 10.5 %points for goalkeepers.
Kepa also plays the ball into pressure 19% points more often when under pressure 6% points greater than the league average. Among a million other distribution statistics you can come up with, on some intuitive ones, Kepa again looks very, average. And that may not fare well at Chelsea, assuming Sarri will expect Kepa to play differently and complete lots of passes, regardless of the pressure that is applied.
How well will Kepa fit in Courtois’ shoes?
In terms of shot stopping, we expect Kepa to slightly under perform Courtois. Thibaut ranked 10th in the EPL in terms of adjSV% last season, with an adjusted save percentage of -0.02. That is not to say the young keeper will not come up with a big stop now and again, but we don’t expect it to be as consistent as the Belgian.
We also expect Kepa to be more timid on crosses, but when he does win the ball, he will look to distribute quickly just like Courtois. He will look for quick counter attacking opportunities and exploit them going forward. Kepa is not terrible with the ball at his feet, but he has struggled under pressure at Athletic Bilbao and we know Sarri will expect more out of him. We think Kepa will blend into Courtois position fairly well, but we don’t think he is an upgrade regardless of his price tag. Even Sarri admitted he is not on Courtois’ level, at least not yet.
What Was Chelsea Thinking?
Kepa Arrizabalaga is only 23 years old and has logged a lot of first team caps with Athletic Bilbao, clearly showing the confidence the club had in him. His performance has attracted one of the most prominent clubs in Europe to pay the highest fee on record to acquire his talents. But, over and over again, each metric that we uncover show Kepa simply looking average.
From our analysis, we have profiled Kepa as an okay shot stopper, average cross collector, and average distributor. To be fair, a lack of temporal trends, the small sample size, and the lack of a large reference group are just some of limitations of this analytical framework. And it also could very well be that Chelsea understands keeper development a whole lot better than we can through data.
But at the end of the day, in the most expensive goalkeeper transfer of all time, we would expect to see something that makes sense. Something that makes Kepa shine. Something that justifies £71M. At that price, I don’t see the value in Kepa Arrizabalaga. We wait for years from now, to find out how wrong I am.