Opposition analysis: Manchester City
Ahead of the Big One at the Etihad, let's try out a new method for evaluating an opponent, uncovering some vulnerabilities in their statistical footprint and the best ways to exploit them
“There is a distinct difference between “suspense” and “surprise,” and yet many pictures continually confuse the two. I’ll explain what I mean. Let us suppose that there is a bomb underneath this table between us. Nothing happens, and then all of a sudden, “Boom!” There is an explosion. The public is surprised, but prior to this surprise, it has seen an absolutely ordinary scene, of no special consequence. Now, let us take a suspense situation. The bomb is underneath the table and the public knows it […] In these conditions, this same innocuous conversation becomes fascinating because the public is participating in the scene. The audience is longing to warn the characters on the screen: “You shouldn’t be talking about such trivial matters. There is a bomb beneath you and it is about to explode!”
— Alfred Hitchcock
It was a well-timed international break, at least from a Hitchcockian perspective.
Just as things were picking up, it’s been almost three weeks of talking, talking, talking. At one point I logged on, saw a spirited discussion about whether or not Thierry Henry was overrated, and promptly logged off.
Underneath the table, there was a faint ticking sound.
Manchester City are next.
🕵🏻♂️ A sneak peek into “opposition correlation”
At some point in the last year I had an idea to build new tools for opposition analysis that could help surface key insights before matches, expanding on what I’ve one for more all-encompassing pieces like Scouting FC Porto. To oversimplify, I’d wanted to do some correlation studies to see which statistics — be they individual or team-level — had the greatest relationship with the performances of specific opponents.
For instance, if you learn that disrupting a certain kind of pass, or restricting a certain player’s carry, is disproportionately pivotal to a team’s performance, then you can use that to focus your film study — and, ultimately, your gameplan.
I’d built some straightforward measures myself, and I was interested in taking it to the next level. To help do that, I’ve been working with t — a statistician and brilliant mind on such matters. (He’s a Director at Kadre, a data science/machine learning consultancy, and is “still romantically waiting for Abou Diaby to return from injury.”)
Is it too early to show you what we’ve been working on? Perhaps. But we have several steps of refinement in the works, which I’ll cover in a bit. With such an important matchup looming, I thought it couldn’t hurt to share a sneak peek into some of the information on Man City. Hopefully, it helps us zero in on some critical areas of focus for the looming big one.
This is a work-in-progress, any errors are mine, and I may post some edits to this in the coming day or two. Don’t @ me. 😀
🔎 Looking at Man City
With that, here is a look into some Manchester City data on a team-level. The further the blue bar is to the right, the stronger the relationship the metric has to a better underlying performance (based on expected goal differential); the further the bar is to the left, the stronger the relationship it has to a worse underlying performance.
I’ll put some of the caveats and assumptions here:
As we learn more, we’re going to narrow down these reports to a more tailored list of actionable statistics. For now, you see a wider sample. Enjoy, you sicko!
Correlation is not causation. Just because something happened more frequently, it does not mean it caused the underlying xG difference. “Just stop them from getting corner kicks and prosper!” is not the takeaway here.
Not all correlations are instructive or even important. It’s not all that useful to learn that teams tend to win more games when they shoot a lot. Key passes, assists, touches in the attacking third — these are all likely to come from dominant performances.
There are sample size limitations an general noise, particularly on statistics that don’t happen at a high volume.
Next, we have some of the data on a player-level.
There’s a lot of intriguing stuff in there, which we’ll get to next. I won’t be able to get to it all. It’s also not showing all information (on KDB’s interrupted year, every bit of data for every player, etc).
⏭ What’s next for this
In the coming weeks and months, you’ll see us refine this further, including:
Cleaned up, actionable takeaways: Will add data to the original samples, include what the team’s opponents did, potentially items like tempo and challenge intensity, then simplify the final view so it can surface the key insights.
Comparison to league expectation: It only tells you so much to learn that positive stuff leads to positive outcomes and negative stuff leads to negative outcomes. The question is first about proportionality: how much does it matter, compared to other variables? Next, it’s about perspective: how much does this matter to the team compared to other, similar teams? How is it impacted by game state and the like?
Player-by-player analyses: How to impact each individual’s performance.
Statistical decision trees!
More.
Anyway.
💭 Overall thoughts on the best approach against Manchester City
Looking through these numbers, my first thoughts are about how important it is to impact their build-up.
♦️ Arsenal out-of-possession
If there was any doubt, it is now backed up by this work: the primary enemy is Rodri. Or, rather, the enemy is Man City’s short-passing apparatus, as marshalled by Rodri.
Some key things to note:
Four of Man City’s worst short-passing percentage games (i.e., below 92.2%) have resulted in a draw (Liverpool), draw (Chelsea), loss (Arsenal), and loss (Aston Villa). Short passing success is not only important for all teams; when compared to league averages, it is disproportionately important to the success of Man City.
When they complete fewer than 250 short passes, their record is 1-4-2. When they complete more than 250 short passes, they are 18-2-1.
If they do manage to complete a lot of the short passes, but fewer than 11% of passes received are progressive, they underperform in similar ways.
The more the ball is with Rodri and Stones (who will miss this one), the better it is for Man City (as is proven to the naked eye, but especially true in these statistics). Rodri’s progression has one of their highest individual correlations with success.
If passes from Rodri (and Stones) are limited and subsumed by the likes of Dias, Aké, Ederson, and Akanji, there is an appreciable negative impact on underlying performance.
If opponents manage to block 12+ passes, they’re 3-3-2 against Man City, which is obviously worse than their performances otherwise. Tackling them on the dribble, for example, does not have this strong of a relationship.
There is a corollary to this. Pep has tried to build a juggernaut at receiving and dribbling into central areas, and will stubbornly try to do so against any opponent. It’s been working. Of note:
Man City have the highest dribbling success rate in the league (52.8%).
Man City have the most carries into the penalty area by a distance (298).
The more touches a player like Foden has, the better for Man City. Specifically, he needs to be kept out of the box (the general cutoff in the numbers is keeping his box touches to <12% of his total touches).
In short, Arsenal have to deny Rodri’s ability to be progressive, and deny central space for Man City attackers to drop in, turn, and carry into the penalty area.
This all leads to a similar gameplan: a compact, disciplined mid-block — scaling up the intensity of the press as the game wears on. Lay off, stay compact, allow touches by the back-line, block lanes (instead of directly challenge), and surround Rodri.
I’d imagine you aren’t too surprised by this being my suggested lineup.
Jorginho and Rice have both played #6 and #8 in the past couple of months. Jorginho’s role as a higher player against Liverpool was interesting, because it helped him deliver direct balls to the left side, and generally kept him in a manageable pocket of space; the bigger cavern was given to the best in the business at covering such ground (Rice).
The trade-off is one of Rice’s many other strengths: he can be incredibly disruptive as a higher presser.
This, in so many words, is the exact player you’d want following a player like Rodri. If Rodri is indeed priority #1, then you put Rice on him.
This is what was done in the Shield:
Chelsea had a similar plan for Rodri recently, with Gallagher tailing him. City did a wide, “split CBs” thing that may look familiar after playing Porto.
The Man City countermeasure is to be expected. They’ll spread out players, give the GK more responsibility, and seek to “decouple” Rice and Jorginho, trying to drag Rice toward their goal or off to the side. If you can use Rodri to pull Rice around, then you give the other DM more room to cover. This gives the DM a test of mettle: should I jump to support Rice and keep up the intensity, or hang back and be “safe”?
Here’s Rice getting mad at Partey for not supporting his jump.
These backup runs are where Manchester United have been destroyed this year.
Jorginho, despite his athletic limitations, and to his great credit, is always willing to support these jumps. He sees them a moment before they happen, which helps him compensate, but it’s still a risk against the likes of De Bruyne (who should be served a hearty helping of Gabriel) and Foden.
As such, he needs neighbors. This means bringing the block as narrow as possible, tucking the wingers in, and bringing the full-backs in — essentially enclosing Jorginho, who can also communicate so effectively in these situations. As we’ve learned, Man City are more negatively impacted by their passes being blocked than they are being tackled on the dribble. Patience is key.
For Rice, it probably just means not getting sucked too far up the pitch with Rodri (i.e., all the way into the box) unless a player like Havertz is 100% rotating behind. This is also where the Kai presence looms large. He can help deliver the swarm.
Statistically, the below map shows where Man City are most likely to lose the ball when playing their tougher league foes (Arsenal 2x, Liverpool 2x, and I included Chelsea 2x in this).
Pep will undoubtedly have something more creative to offer. Last time, he offset the absence of Rodri by putting most of his squad in the middle of the pitch.
Rico Lewis is so smart, mature, and skilled. That said, if he plays, he should be swarmed and leaned upon. The statistics show a relationship between him getting tackled and weaker Man City performances.
The objective in that one, again, was to try and get through the middle, and get Jorginho yanked around, or have others helping him out, and then pounce on the space that opens up.
This kind of thing may just be the price of jamming up the middle.
Before the news broke that Stones and Walker would miss this one, I considered these out-of-possession questions fairly straightforward: settle into a gross, compact mid-block with a few triggers here and there, and then rachet up the pressure has the game goes on, and especially as some super-subs arrive. This is what was opted for last time.
And that is probably what I’d do again. You can see how it worked out in the underlying statistics.
There is, however, a notion that is more tempting now that the Manchester City backline is missing two of its best players at dictating tempo: bring the full-beans press from the get-go (likely with Jorginho/Rice flipped) and try to run them off the pitch in a statement game. If they are momentarily getting too much rhythm or space, just foul them.
It’s probably too risky, and considering the location and the desired result (a draw is OK), I lean Door #1. Especially if Man City try the split-wide thing we saw at Porto, it’s probably best just to hang back a bit as things settle in.
♦️ Arsenal in-possession
I don’t see much of a compelling argument to stray from the formula that has been working so well of late: a good balance of short stuff, patience, central work, directness, and outlet balls out to Havertz and Martinelli. It’s the mixture that proves so potent, and like a racing driver turning up his fuel mix when a pass is in sight, this can get increasingly direct as the game goes on.
On that, Rúben Dias hasn’t had the easiest time with Havertz in the past. He leans and leans but has difficulty generating the leverage he achieves against so many other players. I’m not sure this is the best example, but it’s the first one I could think of.
In Rodri, Man City also have a #6 who can struggle to defend in huge spaces (though less so, and the problem may be overstated) but is excellent when there is coverage around.
Though Rodri is a solid amount faster than Jorginho, he does not have the luxury of being paired with Rice. As such, his team doesn’t fare as well when they have to intervene more directly — getting dragged, facing dribblers in space, having to tackle, and the like. The idea, then, is to “decouple” Rodri from his backline, and hit it to Havertz with runners around. Make Rodri react to some healthy chaos instead of creating it.
This will exploit one of Arsenal’s strongest strengths.
If Martinelli plays, I’d also love to see him swing centrally (as well as left) in the 4-2-2-2 thing we’ve seen of late. He can probe for his best matchup and sprint past them.
That said, the long-ball should be a weapon, but not the only one. There must be a coherent strategy for winning it.
Like Liverpool, Man City often baits the overly-hopeful long hit; they are then generally proficient at closing the area between the backline and the GK on an over-the-topper, and then launching a new round of attack from there.
The medium pass is statistically more successful. Of note:
In the five games their opponent has attempted the most medium passes, Man City haven’t won once (0-2-3)
In the eight games when their opponent attempts their most long passes, Man City have never lost (7-1-0)
From there, it’ll be about the actual Man City XI, and which matchup is the most exploitable. If Rico Lewis starts in the backline, it’s him. Otherwise, I’d probably feel best about the chances of Saka versus Gvardiol, if that takes place.
Regardless, it’s time to talk about Arsenal’s build-up.
I’d expect an all-out pressing force, but the Man City press intensity can be a little bit lower this year, though it is still about as effective as always.
Every time I look at Haaland’s defensive stats (imperfect as these metrics are at fully capturing performance), I’m surprised at how much his season totals can look like game totals. That said, I’d expect him to come out swinging for this one.
For these reasons and more, I think it makes sense for White to push up to the wing whenever possible, as was done in the first leg.
This fulfills a few objectives. First, Saka gets to tuck into the half-space, where he closer to goal and extra dangerous. Second, Ødegaard can be more active at floating and linking phases. Finally, and most importantly, this pulls the opposing left-winger down, and dulls their contributions in the attacking phase. When Man City’s wingers have to be more active defensively, their other numbers suffer.
As you’ll see from a sample of Man City matchups against Liverpool/Arsenal/Chelsea, the primary direct 1v1 threat is on the left. Jack Grealish hasn’t played much, but all these correlations seem to show how impactful his presence has been. With him out, Doku has offered a totally different threat — he is probably the sport’s most active dribbler at the moment. If these happen in more settled situations, Arsenal have a better chance of dealing.
He is especially dangerous on the cut-back cross, as you’ll see below. Many of the crosses from the other side happen closer to the corner of the box.
By pulling him down more often to support his left-back, these opportunities can happen with lower frequency.
➕ Some other notes
Timing: Here’s a look (from Footystats) at when goals tend to happen in Man City games. I’ve got my eyes on that 31’-45’ zone, when the game has settled down, and perhaps some impatience starts to creep in.
Set plays: Next, here’s a brief look at their output on set plays. Despite the huge amount of corners they generate (they now lead the league), that hasn’t accumulated into a particularly high level of xG. Based on my own experience watching them, though, their routines have gotten increasingly sophisticated throughout the year, and they look a genuine threat of late.
Oh yeah: I shied away from talking too much about player-to-player matchups, because of uncertainty on the injury front, but it occurs to me that we haven’t talked about Phil Foden or Kevin De Bruyne quite enough. That’s because I don’t have many bright ideas there: it’s all about shadow-marking them from the front (with the DMs), and then following up with pressure from behind (Gabriel and the FBs) as quickly as possible. I’m hopeful that by tucking the wingers in more, it’ll feel more claustrophobic, but KDB will likely be going all the way to the touchline and causing havoc. It’s hard to feel blasé about this all when you look at something like his.
🔥 Parting words
I never intended this to be a full opposition report; the idea was to show more of a preview into some future work in opposition analysis. But I just kept typing, as I often do. I hope you don’t mind.
Because Manchester City are a team that has been under the microscope for so long, some of the takeaways may come across as expected; others, perhaps not so much.
But after diving into all the subtleties of the data, I’d surface five important factors to the success of Arsenal on Sunday.
1. Disrupt the short-passing game, particularly the link between Rodri and the central playmakers, at all costs
Suggestion: Have Rice essentially man-marking Rodri, but with conditions, and with Havertz close by; provoke more touches by the back-line
2. Constrict the middle of the pitch for their carriers
Suggestion: Run a rugged, compact mid-block that increases intensity as the game goes on (depending on game state)
3. Push a full-back up to drag their winger down
Suggestion: Have White hold width at right-wing, pulling opposing left-winger down, dulling their attacking threat
4. Make the second ball a primary feature of play in their half
Suggestion: Bounce it up to Havertz, but only with purpose
5. Achieve directness through medium passes and advanced dribbles
Suggestion: Go long, but not indiscriminately. Wait for moments to hit channel balls. Drag players around on the carry, even originating them from deep.
The biggest factor may just be personnel: who is available to start, and how impactful the subs are. With so many questions, there could be some real surprises in the offing. With Arsenal returning players to fitness, there could be some refreshingly pivotal contributions.
There are plenty of other little novel insights from the numbers. Man City aren’t as sensitive to their shots being on target as others; their ability to switch and cross effectively are important to their success; fouls and yellow cards are effective at interrupting their flow.
I want to thank t for working through this with me. Again, I promise that any errors in framing it up are mine alone, and think there are likely to be some fun (and simplified) permutations in the future.
But enough with our shit. The players take the stage from here, and it’s difficult to overstate the level of difficulty for Arsenal. As Sky pointed out, Guardiola’s side are unbeaten in their last 38 home games, and Arsenal haven’t won there in almost a decade. There are injuries, sure, but I still see a starting XI that is likely to include Kevin De Bruyne, Rodri, and Erling Haaland — as well as eight other mostly-really-expensive world-beaters.
This rendition of Arsenal feel a little different, though. Can they use Sunday to prove it?
Tick, tick…
Shout out to “t”. I really like the direction you are taking here. A winning combination: a data analytics guy and a scribe. What more could I ask for?
Perfect prep for the big one tomorrow! Almost split equally between being excited and terrified. Thanks for another insightful piece Billy. Hope someone at Arsenal is taking notes on how to disrupt this City machine 😄