Dr. John Harrison digs into the data to investigate how Chelsea’s goalkeepers have performed recently, and how suitable they are to the new manager’s style…
On the 7th of September, after a prolonged average run of form, Chelsea decided to part company with manager Thomas Tuchel. The Blues moved moved quickly and brought in Brighton and Hove Albion manager Graham Potter to replace Tuchel just one day later.
Potter is known for his attractive but solid brand of football, liking his teams to be malleable and able to change formation and tactical approach fluidly depending on the opposition team’s set up.
When Potter arrived at Chelsea, he brought with him a host of backroom staff including Brighton’s goalkeeper coach Ben Roberts. Many have suggested that the change in manager and backroom staff may bring £72m goalkeeper Kepa Arrizabalaga back into the picture at Chelsea.
In order to investigate whether there is a case for Kepa to be Potter’s number one, lets dig into both Kepa and Mendy’s data.
Kepa’s last season as number one (2019/2020) was not a successful one. His poor shot stopping record was apparent to all and the statistics painted a similar picture, Kepa conceded 0.32 more goals per match than would be expected if an average Premier League goalkeeper had faced the shots Kepa did. It was the worst shot stopping performance of any Premier League goalkeeper in the last 5 years. Additionally, Kepa’s shot prevention was similarly poor as he allowed Chelsea to face an additional 0.11 expected goals per match more than would be expected if an average Premier League goalkeeper had faced the crosses and through balls that Kepa did.
These poor performance levels forced Chelsea back into the goalkeeping transfer market and before the start of the 2020/2021 season they spent £22m to secure Stade Rennais goalkeeper Edouard Mendy. Last season, Mendy’s form for Chelsea fluctuated, however overall he put up some solid shot stopping and shot preventing numbers. Mendy saved 0.06 more goals per match than would be expected if an average Premier League goalkeeper had faced the shots he did. Mendy’s shot prevention was similarly solid as he allowed Chelsea to face 0.10 expected goals per match less than would be expected if an average Premier League goalkeeper had faced the crosses and through balls that Mendy did.
So when we compare Mendy last season to Kepa’s last season as number one we find Mendy was worth a huge additional 0.59 goals per match! The perceived problem with Mendy, though. is his distribution (and I assume people believe Kepa could only improve his output from the 2019/2020 season). High profile mistakes against Real Madrid, West Ham United, and more recently Leeds United have put his ability on the ball under the spotlight.
On the other hand, while Kepa is not renowned for his shot stopping or shot preventing skills (as the statistics highlighted earlier) Kepa is thought of positively for his ball playing ability and given Mendy’s recent issues questions have been asked of whether Potter and his staff would prefer a ball playing goalkeeper.
Below is Mendy’s distribution under pressure data over his last 15 Premier League matches in comparison with Kepa’s, supplied by Goalkeeper-XG.
The first thing we can see is, under pressure, both Chelsea goalkeepers have been below average in their last 15 games! Mendy has cost Chelsea 0.52 expected goals more than the average Premier League goalkeeper would be expected to if they received the ball in the same situations as Mendy, although this number is heavily influenced by his mistake vs Leeds. Kepa has been even worse and cost Chelsea 0.59 expected goals more than the average Premier League goalkeeper which is mainly driven by his mistake vs Liverpool which Mane capitalised on.
So, it looks like Kepa’s ball playing reputation while under pressure could be unfounded, at least in the recent past.
One interesting difference between Mendy and Kepa when they are pressured is Kepa seems to have a more varied range as he tries to find teammates relatively frequently regardless of their location in his half, whereas Mendy rarely looks for a teammate who is not either in the mid right or mid left sections of his own half. Kepa’s completion rate when playing through pressure into the opposition half is also better than Mendy’s, and the Spaniard's poor overall outcome is driven by lapses in concentration inside his own 18 yard box.
Potentially, Kepa’s inflated reputation on the ball derives from these factors, but one thing is for certain: in a Chelsea shirt, Kepa has been just as error prone on the ball as Mendy in the recent past.
It isn't just distribution under pressure that will be important to Potter. Below, we have graphics for Mendy’s distribution-under-no-pressure data over his last 15 Premier League matches. In comparison with Kepa’s, we find again that Mendy comes out on top.
This time, both Chelsea goalkeepers have been above the Premier League average in their last 15 games. Mendy’s excellent distribution has added 0.14 expected goals more than the average Premier League goalkeeper would be expected to if they received the ball in the same situations as Mendy. Kepa does again have worse numbers, but his distribution has still added 0.03 expected goals more than the average Premier League goalkeeper would be expected to.
This time, both Mendy and Kepa have similar looking distribution graphics with a large fraction of the attempted passes being into the sides of the 18 yard box and the middle of their own half. The reason Mendy has a higher goal value is twofold. Firstly, Mendy has only given the ball away in his own third twice while Kepa has given the ball away three times, and secondly, Mendy’s long passes into the opposition half are successful far more often than Kepa’s (37.1% success vs 30.6% success).
So it turns out that statistically, overall, Mendy has been a below average distributor for Chelsea. He has cost them 0.38 expected goals more than an average Premier League goalkeeper due to his passing. But the claim that Kepa could solve this issue seems unfounded given his recent Chelsea performances have cost Chelsea even more expected goals due to poor distribution (0.56). Additionally, Kepa’s poor shot stopping and shot prevention record would certainly be a cause for concern if he was reinstated if they remained at the same level as they were in 2019/2020.
If Potter wants to displace Mendy due to his recent distribution levels and seeks an upgrade, it is likely that Chelsea will have to dip into the transfer market for a goalkeeper once more. Below are Robert Sanchez’s (Potter’s goalkeeper at Brighton) distribution graphics over the past 15 Premier League games.
The first thing that becomes apparent is Potter likes his goalkeeper to mix things up far more than Mendy and Kepa have become used to at Chelsea. Far more of his passes (under pressure and under no pressure) are aimed deep into the opposition half. Crucially, Sanchez has been worth 0.34 expected goals above average to Brighton due to the fact he has seldom given the ball away in his own third while also consistently producing accurate passes deep into the opposition half.
If Potter feels neither of his current goalkeepers can improve when it comes to distribution, nor can they mix things up more (i.e. go longer more frequently), then I imagine when Chelsea do go into the transfer market for a goalkeeper Sanchez will be one of the first goalkeepers they contact. Not only is his distribution top quality, but his consistently solid shot stopping and shot preventing performances for Brighton make him far less of a risk then reinstalling Kepa who at least recently hasn’t even been a superior distributor to Mendy.
Could we see Chelsea break the bank once again for a goalkeeper, and Brighton receive another big money fee for a key Potter player? It's definitely not impossible. The pockets of the Stamford Bridge elite are deep, and given the reported targeting of Potter as the ‘chosen one’ for the Chelsea hot seat, it's hard to imagine they'll refuse to back him in the transfer market.
