Let’s go back in time, back to the 2015/16 season. I know that’s a traumatic season for many of us but it’s been a few years and I actually have fond memories; because for most of the season I thought we (Arsenal) were going to win the League.
My belief that we were going to win the League was based entirely on the expected goals stats that I saw Arsenal putting up each week. And what I saw from the teams around Arsenal that season. Leicester were the feel good story and I would never take that away from them. They overperformed by about 12 points but what kind of scrooge would I have to be to say that they didn’t deserve the title? Sometimes underlying metrics like xG are used to say that a team “doesn’t deserve to win” but I think of it the other way round: Leicester outperformed their expected defensive metrics, they worked harder than anyone else, they deserved the title. Most teams who win the league overperform in some area so, it’s normal and to be expected.
Arsenal’s problem that season was, as you all remember, we couldn’t put the ball in the damn net. That season we had an expected goals of 74 and an actual goals scored of 68. Many people blamed Alexis and his goals haul of 13 wasn’t great but looking at the expected goals stats for that season, he was expected to score 13 off the shots he took. Could he have taken more shots? Probably. Could he have taken more shots in better positions? Probably. But he didn’t.
The six missing goals were actually all down to Aaron Ramsey (-3.6), Theo Walcott (-3.4), and to a lesser extent Mesut Ozil (-1.9). Players who overperformed were Giroud (+1.5) and Welbeast (+1). So, what Arsenal got, actually, was average performances from a lot of players, some slightly sub par performances from a few key players, and bang, we lost the title to a team who performed better.
One of the metrics that we didn’t have back then is expected goal chain and expected goal buildup (xGC and xGB). Both of these metrics exist in order to measure how much players are involved in the eventual shot (in the case of xG Chains) or just how much they are involved in the play that doesn’t create a shot (xG Buildup). Just to give you an example: John Stones passes to Fernandino who passes to Walker, who passes to Aguero, who passes to Silva, who passes to Jesus and Jesus takes a big chance shot (0.45xG), all of the players in that chain get the (0.45).
So, that season Mesut Ozil led all players with an expected goals chain of 32.79. It’s important to put that number into some perspective, though. You can have a high xGC and not be a creative player. For example, Aaron Ramsey was 2nd on Arsenal in xGC with 23. Meaning that Ramsey was involved in a lot of the passes which led to a shot. But Ozil also had an expected assists of 18 (17.6) and expected goals of 8 (7.9). Meaning that while he was involved in the touches which led to a shot, he was also the one creating and taking those shots. Ramsey’s expected assists were just 4 that season.
Expected goals buildup is all of the passing work that doesn’t lead to a shot. You can still calculate expected goals without a shot. Have you ever seen a ball crossed in the six yard box and wondered why there wasn’t a forward on the end of it? Or maybe it’s a blocked cross or an Alexis turnover in the final third. Whatever the action was, there was still a lot of buildup that happened before the turnover. That’s xGB.
Mesut Ozil led the League in this metric as well with 18.4. But perhaps a bit of a surprise was the fact that Nacho Monreal was 2nd in the League in this metric with 15.3 and Hectare Bellerin was 4th with 14.4. Eriksen was 3rd, Fabregas was 5th and Ramsey was 6th. These are the sum total of the expected goals from these player’s buildup play over the whole season and I think that they show that Arsenal’s buildup play was the best in the League.
But if you remember that was a strange season. It started with Coquelin and Cazorla in midfield and Cazorla suffered an horrific injury which would keep him off the pitch for a long time. After Cazorla went down, Coquelin had to be dropped and Arsenal never really recovered.
If we prorate the stats on a per90 basis, Santi Cazorla led the League in xG Buildup with 0.73. He was also 5th in the League in xG Chain per90 with 0.83. Cazorla’s expected assists per90 was only 0.3 that season (Ozil was 0.52) but the stat shows how important he was to Arsenal’s buildup and shots chains. Surprisingly, Coquelin was 2nd in the League that season in xGB per90 with 0.56 and had a decent xG Chain of 0.58!
Those numbers dropped off the next season for Coquelin. His xG Buildup dropped to 0.42. I think there’s a decent case to be made that he was helped out enormously by the pairing with Cazorla.
Now let’s travel forward in time to last season. Expected goals buildups and chains are almost certainly going to favor the Man City players. They scored 100 goals and Guardiola’s philosophy is that the team has to make a certain number of passes before they can take a shot. And sure enough, the top 6 xGBp90 players (I limited to 1500+ minutes played) are all on City: Delph, Silva, Fernandinho, Walker, de Bruyne, and Gundogan. But there in 7th and 8th place are Granit Xhaka (0.63) and Mesut Ozil (0.61).
And in expected goals chains per 90 it turns out that Aaron Ramsey makes a surprise appearance in 7th place (again I limited this to just players who played more than 1500 minutes) with a whopping 0.93 per90. That’s more expected goal involvement than Mesut Ozil.
Granit Xhaka’s numbers last season were also pretty surprising. He was 2nd in the League in total xGB with 22.6 and he had an xGB per 90 of 0.63. Wenger deployed Granit Xhaka as the lone central midfielder, giving him the keys to the house. These numbers show that Xhaka was involved in almost every play last season. Though, it’s just as a shuttler. His job is just to get the ball to other players who will do the work. His expected assists last season per90 were just 0.08. Francis Coquelin had a similar xAper90 during his halcyon days, 0.02.
I wish we had more data from the seasons when Mikel Arteta played with Arsenal. But I do know that in his last season he had an xGB of 0.84 p90 an xGC of 1.14 per 90 and an xA of 0.27 per90. I wish I had his numbers from 2013/14 and 2012/13 but I don’t. I suspect that we would see insanely high buildup and chain numbers and also Cazorla-like expected assists.
It’s wild to think that Arsenal went from Fabregas (who still appears on these tables as one of the best buildup/chain players) to Arteta (who was an amazing force of nature in Arsenal’s midfield) to Cazorla (who was another absolutely amazing CM).
What’s interesting about this season is that Emery is pairing Xhaka with a second midfielder and Xhaka’s xGB and xGC numbers have taken a hit, though he’s still 2nd at Arsenal in xGB behind Guendouzi.
And he’s also taken a hit in xGC (again, below Guendouzi):
The one name on that list that probably surprises you is Torreira. It’s not a surprise to me to see him so far down the list in both buildup and chain. These metrics heavily favor players who get a lot of touches and Torreira has (so far) demurred to Granit Xhaka when both are on the pitch.
This season Granit is also suffering from a change in formation, playing style, and from Emery’s indecisiveness in terms of lineup. So, his buildup numbers and chain numbers will have to be something to watch as the season goes on.
Anyway, I was very excited to find a source for expected goals. And this one, understat.com, is fantastic. I don’t think it will settle any debates about any players. Especially not Granit Xhaka who it seems is either untouchable from some fan’s perspectives or unredeemable. I think the truth is somewhere in between. I don’t see anything that suggests he’s at the level of midfielders we had in the past (Fab, Arteta, Cazorla) but he’s useful in terms of keeping the ball moving and he’s almost never injured so he gets a lot of touches and plays a lot.
I also think the future is bright. Guendouzi’s numbers early on this season show a player who can (probably) pick up for Xhaka if he were to get an injury.
Source: understat.com
You must be logged in to post a comment.