I came across this piece on Mustafi's tendency to rush out of defence yesterday, goes into a good amount of detail and has some really damning graphic evidence of his poor positional sense:
http://intacttactic.co.uk/?d=04-11-2016#c
When we signed him last season I was a little bit skeptical as to how he'd fit in with Koscielny, as Mustafi's high numbers for tackles and interceptions suggested a player who liked to defend on the front foot and another such defender was not what we needed (having already tried the same failed experiment with Vermaelen and Gabriel).
Having said that, he won me over with his performances over the next couple of months as Arsenal generally took a very attacking approach where conceding even 1 or 2 goals wasn't an obstacle to winning, and Mustafi helped the attacking side of the game with his ability to break the lines with his passing. Not to mention that he contributed to Arsenal being very strong at set pieces and not conceding any goals from corners until the Everton defeat (which Mustafi missed).
However, the second half of the season was a different story and Mustafi's return against Bournemouth saw the start of a dreadful period of form for him and the club in general. I think at this point the honeymoon period was over and we were starting to see the real Mustafi.
I can totally understand Wenger wanting to sell him this transfer window just gone by, but it still raises a question about our so-called 'StatDNA signings'. Does Wenger actively get involved in the decision to sign them or does he merely give the seal of approval? Should the process not be integrated with the needs of the squad rather than merely identifying good individual players and then working out how to fit them in later? And shouldn't it be reserved for attacking players, where it's much easier to quantify their contributions because their work is largely centred around recordable, on the ball actions? No amount of stats will ever tell you about a defender's ability to man-mark, or to defend space. You can have record-numbers of tackles and interceptions but if you leave (statistically-unrecorded) gaps at the back that cost you 1 or 2 goals a game, then you're a bad defender.
Article on Mustafi's impetuousness
posted on 9/9/17
You are right that defensive stats are a lot harder to interpret but still there's some good progress done during the years. Using possession adjusted stats for example, expected goals after defensive errors or misplaced passes leading to fast breaks all could be some markers for player quality. But yes - this is one of the big challenges in modern football analytics.
posted on 9/9/17
Mustafi picked up an injury and came back when the team was going through their annual shtteshow. My biggest issue is not with personnel although we have poor players. We do however seem to play at less the sum of our parts more often than we should, so the problem lies with the coaching. We have no discernible cohesion or plan other than to keep passing and hope we catch a break and score.
posted on 9/9/17
You are right that defensive stats are a lot harder to interpret but still there's some good progress done during the years. Using possession adjusted stats for example, expected goals after defensive errors or misplaced passes leading to fast breaks all could be some markers for player quality. But yes - this is one of the big challenges in modern football analytics.
---------------
I agree these sound like big leaps in our interpretation of data. Not come across anything like this in mainstream football analysis aside from a few niche blogs. But still all related to on-the-ball actions and their consequences. A defender rushing out of position or letting his man get goalside of him won't get picked up by any kind of metrics. Until such advanced tools get introduced, the best we can hope for is some sort of impact analysis i.e. how does X club's defence fare with/without Y player?
posted on 9/9/17
comment by MaineGunner (U10130)
posted 3 hours, 33 minutes ago
Mustafi picked up an injury and came back when the team was going through their annual shtteshow. My biggest issue is not with personnel although we have poor players. We do however seem to play at less the sum of our parts more often than we should, so the problem lies with the coaching. We have no discernible cohesion or plan other than to keep passing and hope we catch a break and score.
----------------
I agree but I do think regardless of gameplan and coaching that having 2 CBs who like to play on the front-foot is a bad combination. Never worked with Koscielny and Vermaelen either. 3 at the back obviously provides a little extra cover for someone of Mustafi's nature and in fact, probably requires a CB who's comfortable closing down into midfield and stepping up to build attacks. Still a place for Mustafi in our squad.
posted on 9/9/17
comment by renoog (U4449)
posted 1 hour, 8 minutes ago
I agree these sound like big leaps in our interpretation of data. Not come across anything like this in mainstream football analysis aside from a few niche blogs. But still all related to on-the-ball actions and their consequences. A defender rushing out of position or letting his man get goalside of him won't get picked up by any kind of metrics. Until such advanced tools get introduced, the best we can hope for is some sort of impact analysis i.e. how does X club's defence fare with/without Y player?
----------------------------------------------------------------------
The reason for this is that most analytics rely on Opta stats and it does not include positional off the ball data. There are proxy models for defensive pressure like speed of the attack but they don't really work for personal defensive stats. And the problem with how X club's defence fare with/without Y player is that the sample is usually so small that it's hard to be of any real significance.
I'm pretty sure that there are better stats acquired and used by at least some of the clubs for private use as I've heard Wenger mentioning those in interviews but the data is not publicly available. Not sure what data they have available for scouting as some of the leagues we buy from don't have such detailed data available at all.