Tuchel Fucked England, Right?

England reached the semifinals of the World Cup this year, and they played in that game exactly a week ago. They took the lead first, early in the second half, then conceded a goal to Argentina that tied the game in the 85th minute, then conceded the winning goal in stoppage time.

Having watched the game, I developed a lot of strong opinions for Tuchel's approach after England initially took the lead. In short, I thought he managed their lead poorly, I thought the substitutions he made were ill-advised, and I thought he was generally poorly prepared for the situation that he found himself in.

Of course, after the game, Tuchel - maybe rightly - acknowledged that it is easy to question his decisions given the outcome.

Generally speaking, he is right, and people are far more guilty of incorrectly assigning blame based on the result of a match. In this case, he was both wrong - at least in the group chat's case - about the timing of the criticism and, in my eyes, about whether it was justified.

This piece is an examination of Tuchel's approach, or England's approach, depending on what you think the manager is responsible for and when. Let's get into it.

First and foremost, we should identify what happened. These are facts, defined mostly by simple proxies for England's posture at different points in the game.

These things are relatively simply defined by:

mean-x (average field position) - On a pitch scaled 0 to 100, where 0 is your own goal line and 100 is the opponent's, this is the average location of all your touches. ~50 is midfield.

field tilt (share of the dangerous territory) - Of all the touches happening in the final third (either team's), the percentage that are yours. 50% is an even fight for the attacking zone. England's collapsed from 53% to single digits.

own-third share (how far back you're playing) - Of your own touches, the percentage that happen in your defensive third. A low number means you're getting the ball forward.

opp big-ch/10 - This is the opponent's big chance rate per 10 minute window. It is an average of a rolling 10-minute window across those minutes, so it smooths across the edges and slightly understates the raw count.

Quite obviously, at 0-0, where the game stood until the 55th minute, England's average field position was about the half-way line, they shared a higher percentage of the dangerous territory than Argentina, and only about 27% of their own touches were in their defensive third. They allowed 0 big chances to Argentina in that time frame. A fair assessment of the game prior to their goal would have been that they were slightly the better team, but the game was largely in the balance.

After taking the lead and before their first substitution, this all changed fairly drastically and relatively quickly. Over the 15 minutes or so after their goal, their average field position dropped, they gave up a considerable amount of the share of touches in dangerous territory, but they still hadn't yet given up a big chance.

Their worst phase came after their first substitution: their average field position became barely above their own defensive box and they had basically no share of the dangerous territory, numbers I only believed at first because I watched the game. Two-thirds of their touches came in their own third, and they gave up 2 big chances in that window. They also gave up the tying goal.

Honestly, watching the game was still far more alarming than these numbers appear, which says a lot about their approach.

Counter to Tuchel's opinion that it is easy to judge his performance after the result, below are some images of some smart people and me reacting to these choices in real time.

The first and most obvious question is whether sitting deep is actually bad. Before I could even answer that for myself, I had to define "deep" and separate it from other defensive postures.

So I took a big pile of games, about thirty-odd thousand of them across a bunch of leagues over the last five years or so, grabbed every stretch where a team was protecting a one-goal lead, and sorted their posture into four buckets measured against their own normal. Here is what each one actually looks like out on the pitch.

A couple of things worth saying about this before anything else. "Normal" is not a team throwing everyone forward chasing a second goal. It's a team continuing to play, keeping the ball, contesting the game. Even at normal, almost a third of your touches are in your own third. This matters because a lot of the pushback would likely stem from something along the lines of "every team sits on a lead," and that's true, but sitting on a lead the way Arsenal or Atletico do it is the mild row, not the deep one. Those two clubs get called frauds and bus-parkers constantly, and I'll come back to them, but what they do to protect a lead is nowhere near what England did.

The simple answer to the simple question is sitting right there in the last column. The deeper you sit, the more you concede. It climbs from 1.37% a minute at normal up to 1.75% at deep, and the opponent's shot rate basically doubles along the way. So on the face of it, yeah, sitting deep is bad, and the deeper you go the worse it gets.

To be upfront, though, the simple version isn't enough, largely because it's the exact mistake Tuchel was warning about. Teams don't sit deep at random. They sit deep because they're getting battered, or because they're worse than the other team, or because something in the game is already going wrong. So, the fact that deep teams concede more might just be telling us that teams already in trouble concede more, which isn't interesting and isn't the same as saying the choice to sit deep caused it. I believed the choice was causing it. Proving that is harder, and it's what the rest of this is about.

Before I ran the numbers, I had a pretty specific idea in my head about why sitting deep hurts you. I will lay it out so you know what I was trying to prove, and so you can hold me to whether the data actually backed it up or whether I was just a guy mad at his television.

When you drop that deep, you hand the other team the ball in dangerous areas, over and over, for as long as you keep doing it. You stop having any kind of outlet, so every time you win the ball back you just give it away again. Your forward is your farthest player forward, and he's still near your box; any attempt at a counter leaves that player one-on-three against opponents who are also...still halfway into your half of the field. The opponent gets to set up camp and take run after run at you, and even if each individual moment is survivable, you're rolling the dice a lot of times, and eventually one of them comes up bad. That's the eye-test version, and it basically boils down to "if you let a good team shoot at you enough, they score."

The second thing I believed, and I was a lot less sure about this one, is that it gets worse the longer you do it. My theory was that the longer a team sits in, the more they switch off, lose their man, stop anticipating and start reacting. It feels true when you watch it. A team that has been defending its own box for twenty minutes straight looks a step slow to everything by the end. I wanted to test whether that was real or whether I was just seeing the goal coming and reading backwards from it.

So those are the two things I brought in with me. Deep is bad because the small dangers pile up, and deep gets worse the longer you commit to it. Now for the fun part.

Measuring "how deep are you right now" against "did you concede in the next fifteen minutes" answers a narrow question. It doesn't answer the one I actually care about, which is what England's decision to sit deep starting at the 54th minute and basically never stopping did to the whole rest of the game. One is a snapshot, the other is a choice England made repeatedly for ~30 minutes.

The first thing I wanted to settle was my switching-off theory, because if it was real it would change how you think about all of this. So I checked whether sitting deep for a long stretch makes each individual minute more dangerous than sitting deep for a short one, holding how deep you are constant. It doesn't.

There's basically no effect. A team that's been parked for twenty minutes is in no more danger, minute for minute, than one that just dropped in.

But in losing one theory, I sort of gained something in the other. If sitting the twentieth minute is no worse than sitting the second, then the total cost of sitting deep for a long time is just the sum of a bunch of small costs, one per minute, added up. Everything just accumulates. The opponent gets more cracks at you the longer you keep it up, which is exactly the boring eye-test version, and that turns out to be the whole mechanism.

Before I add anything up, let's look at the thing doing the adding, because the rest of this piece leans on it.

I took every minute of every one-goal lead in the data, a little over 600,000 of them across close to 20,000 games, and modeled two things: for a team sitting at a given (rolling roughly ten-minute average) depth, in a given minute, how likely they are to concede in the next minute, and how likely they are to score a second. I ran it twice, once on actual goals and once on expected goals built from a very crude expected goals model (a simple measure of chance quality, whether the looks they gave up were good ones, not just whether they happened to go in), so the two could be checked against each other.

The model lets the effect of depth bend instead of forcing it into a straight line, and it counts games rather than minutes when it decides how confident to be, because a hundred minutes out of one match aren't a hundred separate facts. And it adjusts for quality: every team carries its own pre-game win probability from the betting odds, so when I compare deep to normal I'm holding how good the team is fixed, not quietly comparing good sides to bad ones. That last part is the whole reason this isn't the confounded mess from the simple version.

Here's what it finds, per minute, around the 65th, for three depths measured against just playing normally.

Two things in that table matter to me. First, the goal columns and the expected-goals columns tell the same story, which was the whole point of doing both: about +0.2% a minute on the concede rate lines up with about +0.0017 of an expected goal a minute, because a goal is, more or less, a goal. Second, and this is the thing I kept banging on about, the damage splits cleanly into two accounts. Sitting deep raises the rate you concede and it lowers the rate you score a second, and at the 65th minute those two are nearly the same size. Deep hurts your attack about as much as it hurts your defense, minute for minute.

Run those tiny per-minute numbers across the whole half hour and here's the total, in goals and in expected goals, which land in the same place again.

So about six hundredths of a goal of extra danger invited, and about seven hundredths of your own given up. Per minute it really is almost nothing, honestly, it looks like nothing. The whole question is what all that nothing, repeated for thirty-five minutes and priced properly, actually comes to.

Because it just adds up, you can actually add it up.

Here's the concession side of that on its own. Two teams, same one-goal lead at the 54th minute, one sits deep the rest of the way and one keeps playing normally.

The two lines start together and never touch again. By full time the team that sat deep is about 3.7 points more likely to have coughed up the equaliser, and the shape of it tells the story: the gap widens a sliver at a time, every minute, for half an hour. You let a good team shoot at you a few more times than you had to, over and over, and eventually the dice come up bad.

If England had held a genuinely deep posture from the 54th minute to the end, and you price everything through what a goal is really worth at each minute of a one-goal game, sitting deep like that cost them about 4.7 percentage points of win probability. Because the model holds quality fixed, that figure is for an evenly matched game, which is basically what this was, two live sides with the result in the balance. About 3.3 of the 4.7 is the defensive side, the extra chance of conceding. The other 1.4 is the offensive side, the goals you stop yourself from scoring by never going forward.

In one picture:

Most of it, about 3.3 of the 4.7 points, is defense, the extra chance of letting the equaliser in, and that much I expected. The part I almost argued myself out of is the orange band, the roughly 1.4 points you lose on offense by never going forward and never threatening the second goal that would have put the game to bed. That orange band is those seven hundredths of a goal you never scored, turned into win probability.

Now, nearly five points of win probability off a fairly small change in expected goals made me suspicious too, so I looked at it from a few different angles. Basically, a goal in a one-goal game late on is worth an enormous amount, and it isn't symmetric.

Looking at the 65th minute, you're up a goal and you win about 75% of the time. If you concede the equaliser you drop to roughly 25% win probability, so that one goal is worth about half a win. A second goal only takes you from 75% up into the high nineties, so it's worth a lot less. That's why the defensive side dominates. When you're protecting a lead, giving up a goal is a catastrophe and scoring another is a luxury, and sitting deep trades a tiny bit more of the catastrophe for a tiny bit less of the luxury. Do that for half an hour and it adds up to almost five points of win probability.

I still didn't fully trust my own number here, so I checked it against the betting market, which prices this exact stuff for a living.

Here are the numbers under that line, both the straight win probability and the two-way version that throws out draws.

The gap between an even team and one favored by a quarter of a goal is worth about 6.9 points of win probability before a game even kicks off. England's shell swung the game by a little over half of that quarter-goal, worth roughly 3.5 points pre-game, and in play it cost them 4.7. So the same swing was worth about a third more once the game was late and there was a lead on the line, which is what you'd expect, because a goal matters more when there's less time left to answer it.

One more thing before I move on, because it's the part that actually answers "did England blow it, or just start it too early." The cost is front-loaded but it never turns into a benefit. If England had climbed out of the shell at the 65th minute instead of staying in it, they'd have saved about 2.2 of those 4.7 points. So more than half the damage came not from dropping deep, but from refusing to come back out. The insistence was worse than the instinct.

A few quick caveats here. This is the cost of a fixed, genuinely deep posture held the whole way, and England's actual game ramped into it over about ten minutes rather than flipping a switch at the 54th, so treat 4.7 as the price of the posture, not a minute-by-minute replay of that specific night. And all of this is measured. Teams choose how deep to sit, so there's always some risk that deep teams are just worse teams in a way I haven't fully scrubbed out. I've done what I can to compare like with like.

This is the part I was less angry about while it was happening while admittedly still fairly angry. Tuchel, reportedly, was already unhappy that his players had dropped so deep, and his answer was to bring on defenders and go to a back five. My gut reaction was that this was stupid, that you don't respond to your team sitting too deep by subbing on more defenders so they can sit even deeper. I'd add that in a knockout stage game, this also leaves you with a less-than-ideal side to manage the next 30 minutes after you give up the goal you are practically begging to give up, but that's a bit harder to test.

Here's what actually happened with the shape:

The first thing this shows actually cuts against my complaint some. The collapse in England's posture, that seventeen-point drop in field position, happened while they were still in a back four, before the first substitution. So the shell was either happenstance or an instruction, not a personnel move. Tuchel didn't create the deep block by subbing. His players were already doing it, and the subs came after.

Then the question becomes whether the subs made it worse, and here I have to eat a bit of my own take. I looked at every game in the data where a leading team changed shape, and specifically the ones where they added a defender and went to a back five, like England did. The comparison I care about isn't "sub versus no sub," because fresh legs help on their own and that isn't the interesting question. The interesting question is whether going to the back five specifically was the wrong call. Here's what those changes do.

The back five does two things at once, and they pull in opposite directions. It genuinely shores up the box. Teams that added a defender conceded less over the next fifteen minutes, not more, and that holds even once you account for those teams already sitting a bit deeper to begin with. The extra body works. So no, I can't tell you Tuchel's subs made England concede, because on average they don't.

But the back five also makes you noticeably deeper, and it locks you there. And that's the whole problem, because everything above says the deep posture is what was bleeding the game away, minute by minute, and refusing to climb out was the expensive part. Tuchel solved a small problem while reinforcing the bigger one. He made them better at sitting deep at the exact moment the right move was to stop sitting deep. The subs weren't the blunder so much as the doubling down. And unlike World Cups or knockout competitions in the past, this one featured a hydration break in every single match. He certainly had the opportunity to adapt, but he just chose to dig in.

England conceded anyway, of course, back five and all, but that has more to do with who they were playing and the prolonged sitting in than with the shape, which is the next thing.

A fair objection to everything so far is that England didn't sit deep against a park bench. They sat deep against Argentina, who were the 2022 World Cup champions and one of the "favorites" in this iteration, and maybe the real story is just that Argentina were very good and would have beaten anyone (well...). So I wanted to check how much of England's collapse belongs to England's decision and how much belongs to Argentina being Argentina.

In the window where England were deepest, from the 72nd to the 84th minute, Argentina produced about 0.75 expected goals off six shots. A typical team chasing a one-goal deficit in a game between evenly matched sides produces something like a quarter of that over the same span. So Argentina generated roughly three times the danger you'd expect from an average side in that spot. And this wasn't a one-off. Across their entire knockout run, whenever they went behind, they forced the game at a level almost no other team in the data got near.

Every number is a percentile against teams in the same scoreline and the same quality bracket, so a favorite trailing a weaker side is judged against other favorites in the same spot, not against the whole pile. The last column is the honest one: it strips out what any opponent's deepness would have gifted anyway and asks how much of the forcing was Argentina's own.

So Argentina were genuinely, unusually good at this, and some of England's punishment is simply credit to them. But I want to be careful about where that credit stops, because it's easy to hand it all over and quietly let England off. Two things keep it honest.

First, part of Argentina's raw output was England's gift. If you sit that deep, any decent team is going to pile up territory and shots, and Argentina did, more than any comparable side. But when you strip out how much of their forcing was just England handing them the space, and ask how much was Argentina being better than the average team would have been in the same situation, they still grade out around the 84th percentile. So the majority of their threat was their own doing rather than a product of England's retreat, but a real chunk of it was the retreat.

Second, the 4.7 points I put on England's shell earlier was measured against a like opponent, not against the above-charted version of Argentina. Sitting deep was a bad decision against anybody. But if you buy that Argentina's small sample here means something, that Messi is uniquely equipped to punish exactly this, and that they carry more attacking threat than a typical side in that spot, then you might still be comfortable assigning more of that blame to England. I certainly am.

That said, I wouldn't outright dismiss an argument that this all happened in reverse of the way I am alleging. Argentina ramped up, England sat deep in response, they got pushed deeper, then they went to survival mode. To me, that is what makes soccer such a beautiful game. Some of this is conjecture by nature, mostly because it's not entirely possible to prove which in-match decisions precipitated others, and both teams get to make them.

The benchmark I'm comparing Argentina to is huge and solid, thousands of team-games. But Argentina's own phases are small, a handful of games in one tournament, so "Argentina are an elite forcing side" is a strong read on this run.

If you've read this far you're probably where I was, which is fine, sitting deep is bad, but surely there's some point late in a game where you're supposed to shut up shop and see it out. Everybody believes this. I believed it. So I went looking for the moment in a game after which sitting deep stops costing you and starts protecting you.

No existe.

When you price the cost of sitting deep properly, through what a goal is worth at each minute, sitting deep costs you win probability at essentially every minute of the game, no matter how good the team doing it is. There's no break-even point. Here's the per-minute cost of sitting deep, split by the quality of the team.

Every one of those numbers is negative. The reason the "shut up shop at the 70th" thing feels right is that the raw danger of sitting deep does fade as the clock runs down. But it fades for a reason that doesn't actually help you, which is that a late concession is so much more expensive than an early one. The two effects almost exactly cancel. Your posture penalty gets smaller and the price of each mistake gets bigger, and you end up costing yourself about the same the whole way, and I can't show you it ever crosses over into being worth doing.

So the honest answer to "when should you sit deep" is closer to "basically never, and the popular idea that there's a safe hour to start is wrong."

Being abundantly clear, we are talking about deep, not necessarily England deep, but deep nonetheless. The math is not as clear when you work your way back to 'flat' through moderate and mild, but it definitely doesn't ever tell you to go England deep.

That's per minute, in an even game, for holding each posture instead of just playing. Mild is mild, the bleed roughly doubles from there to deep, and nothing on the scale rewards going as far as England did. Take the deep row, multiply by the half hour they sat down there, and you're back at the kind of figure from earlier.

Let me bring in two teams everybody has an opinion about. Arsenal and Atletico Madrid get mocked constantly for being boring, defensive, bus-parking sides. So I pulled what they actually do when they're protecting a one-goal lead over the last three seasons.

The two clubs with the ~worst reputations for parking the bus defend a lead at about a third of their touches in their own third, still contesting half the pitch. That's the mild row from way back at the start. It's nowhere near deep. And they hold about 87% of the one-goal leads they take into the final half hour. Their whole identity is control, not retreat. England, the supposed victims of circumstance here, sat about twice as deep as the teams the internet calls frauds. So good teams do sit on a lead, but again, they just do it better.

What about a genuinely worse team, level, late, just trying to hang on for a draw or a point? Surely that team gains something by defending, by turning some of the other team's win into a draw. This is the "it's easier to defend than to score" idea, and it's the most sympathetic case there is for sitting deep.

It doesn't really hold up either. When an underdog sits deeper in a level game, their own shots collapse and the opponent's shots climb, which you'd expect. But the opponent's shots don't get any less dangerous per shot.

Sitting deep doesn't compress the game toward a draw the way the theory wants it to. It just quiets your half of the game and hands the other team more good looks. The outcome barely moves. Defending deeper mostly just surrenders your own share of the game for nothing.

I'll be honest that this last part is where my evidence is weakest, because it's exactly here that the chicken-and-egg problem is worst. Teams that are level, late, and sitting deep are very often teams that are getting battered, and I can't fully separate "they sat deep and it hurt them" from "they were getting hurt so they sat deep." I'm measuring teams going deeper than they normally play, not teams that sit deep as an identity. This isn't "Burnley should come out and trade with City." A side that already lives in a low block never shows you the version of itself that doesn't, so whether that permanent setup is right is a question I can't really get at here. What I can say is that taking an already-defensive team and pushing it deeper still to hold a point doesn't buy the point.

So I wouldn't die on the hill that defending is actively bad for a weaker team. That said, there's no evidence it's the best state possible that people - or at least teams - tend to think it is, and a fair amount that it doesn't help.

Definitely.

Sitting as deep as England sat, that early, and staying there was a genuine, measurable mistake. It cost England something like five points of win probability against a like opponent, most of it on the defensive side, and more than half of that came not from dropping deep but from refusing to come back out once they had. The subs didn't cause that, but they cemented it.

My switching-off theory was wrong. The danger doesn't compound the longer you sit. It just piles up, one ordinary risk at a time, until one of them goes in. And the comforting idea that there's a safe point late in a game to shut up shop, the thing managers and pundits lean on, I couldn't find any evidence for at all.

So, he fucked 'em slowly, by allowing the most costly possible defense of a lead and then insisting on it while a very good team took the game off him. Easy to see after the result, sure. But it was there in the moment too, which is the whole reason the group chat was losing its mind while it was happening.

Are we happy to just blame Tuchel for this? I don't see why not. He manages the team, and though I'm not certain he has a complete ability to react to what was happening immediately, it is his job to find a way to communicate to the players that they need to get unstuck, or to bring on players who he can communicate that to on their way into the game, who too can communicate that to their teammates. Or to do so during the hydration break, a rare opportunity in this World Cup for managers to have more influence on tactics than usual.

Looks like they signed him through the Euros in two years. I, for one, look forward to another tournament where the greatest footballing nation of all time that somehow hasn't won anything since before the Berlin Wall fell disappoints their fans again.

Notes