Kepa Arrizabalaga actually stopping more shots than expected

The Premier League’s Worst Goalkeeper™ (or is it all of Europe now? I lose track) was once again up to his Worst Goalkeeper™ antics, conceding two goals on two shots last night as Chelsea sunk to even more frustrating lows with a 2-2 draw against 10-man Arsenal. David Luiz didn’t sacrifice himself on the altar of bad defending for this!

Obviously, Kepa’s critics were out en masse once again — not like they need an invitation — usually saying things like “I know he couldn’t do anything about the two goals but he has to do better anyway”. Hashtag-analysis.

And obviously, conceding twice from two shots is NOT going to help that already pretty low low lowwwww (to the window!) save percentage (to the wall!). It’s also not going to help things like xG/shot or whatever “advanced” stats we’re dabbling in these days.

Martinelli’s miscontrol and breakaway — thanks to Emerson pushing up too far and losing out at the edge of the box and Kanté slipping at midfield — was rated at around 0.6 xG, which seems low but penalties are only scored at something like 80 per cent on average as well, unless you’re Jorginho. Do we have a xG-personal stat yet that takes the player actually xG-ing into account? Probably don’t have a big enough sample size for most players for that, which is a problem with most of these things.

And of course Bellerin’s shot carries with it a miniscule xG because not only was it his (much) weaker foot, it was also from distance and two players were in the way (who just blocked the goalkeeper’s vision instead) even after he waltzed around two others (Emerson again, plus a hobbled Abraham who was filling in for Mount on the left for some reason). Not that it really matters, because Bellerin also picked out a spot where practically no goalkeeper in the history of goalkeeping, good or bad, would’ve saved it.

All that is bad and it makes everyone involved look bad. But it’s much easier to point the finger primarily at Kepa, even if he’s relatively small. He is, after all, the last line of defense, and his primary job is to save shots — though if we’re looking to blame him for anything last night, it’s for doing some of the secondary things badly, like almost giving the ball away twice at the back.

Recently, the wonderful spreadsheet nerds at FBRef (in partnership with StatsBomb) started tracking what they’re calling “post-shot expected goals” (PSxG). Unlike xG, post-shot xG only takes on-target shots into account, and thus becomes a somewhat more reasonable metric by which to (help) evaluate goalkeepers rather than just shots or shots-on-target or save percentage.

Unfortunately, I do not have a good punchline wherein Kepa would be the world’s best in this metric. (He’s the worst, duh!)

Actually, he’s the third worst among Premier League regulars, after Rui Patrício and Nick Pope*, with a whopping (-2.9) PSxG+/-. That is, he’s given up 3 goals more than PSxG would expect.

Off with his head!

Wait, here comes the punchline!

Kepa’s PSxG+/- per-90 this season is (-0.12). Guess what his PSxG+/- per-90 was last season? (-0.13)!

Hashtag-AlwaysImprove.

(These numbers, as it turns out, are a bit off as FBRef were updating their site just as I was collecting the data. Read below for details, though I think my point about the need for a broader conversation still stands.)

*BTW, Pope was one of the standouts in this metric three seasons ago, and now he’s the worst. Turns out shoulder injuries are probably not beneficial to goalkeepers. For another big fall, see also: David De Gea. Or maybe this stat is just bunk. Courtois didn’t do too well in it either in his final season at Chelsea, and there’s no data from earlier than that.

While it’s been decided that Kepa is the absolute worst and has to be replaced post-haste, it’s probably important to not let any one statistic, metric, incident, or dollar amount decide such things. That may or may not change your mind of course, but at the very least, the conversation around Kepa needs to be much, much broader than it has devolved to at the moment.

Here’s one aspect for example, just to stay with PSxG — which already may or may not be significant in any way, shape or form. Chelsea are conceding 0.38 PSxG/SoT — that is, each shot that is on target against us has a 0.38 expected goal value. You probably won’t be surprised that this is the worst in the league, slightly worse than even Norwich City and Watford, the two bottom teams in the league. You probably also won’t be surprised that Liverpool are far and away leading the league in this metric, with 0.26. The next best are surprise-package Sheffield United, at 0.30.

There’s more to conceding goals than just the goalkeeper. Chelsea were third best in this metric in each of the last two seasons (0.30 under Sarri with Kepa, 0.28 under Conte with Bad Courtois). As Lampard said after last night, Kepa’s “not the only one” who needs to improve.

(And if you think Willy Caballero is the answer, well ... let’s ask Argentina fans.)

BONUS PUNCHLINE:

Just as I was writing this, FBRef finished updating their metrics, showing that Kepa’s now “responsible” for 4.3 goals, up from 2.9 after the Arsenal match. PSxG total meanwhile only went up from 25.1 to 25.7. PSxG thus shows that Kepa gave up 2 goals on chances that some mythical average goalkeeper only gives up 0.6 goals.

Are they saying that an average goalkeeper is expected to save at least one of those shots every time and the other one nearly half the time?

Lies, damned lies, statistics!

(Kepa’s also now on (-0.18) for PSxG+/- per-90, thus turning my headline into pure clickbait, but [FUN] it. If you read this far, hope you enjoyed the journey. Here’s your refund: .)

FBRef