Thomas Tuchel has a problem: he has a squad of good players but he has been criticised for not being able to knit together a coherent attacking side — but is that fair, and would moving from 3–4–3 to 4–3–3- be the answer, as so many fans seem to think?
NB: all data from fbref.com
What’s he been doing?
For the most part under Tuchel, Chelsea have played a 3–4–3 formation. In the 2021/22 season, the back three centred around the age-defying Thiago Silva (2,650 Premier League minutes) with the now-departed Andreas Christensen (1,496 mins) as his alternate. Flanked on the left by (the also departed) Toni Rudiger (3,035 mins) and on the right with a rotating combination of Cesar Azpilicueta (2,070 mins), Trevoh Chalobah (1,449 mins) and, on occasion, Reece James.
Central midfield usually saw two from N’Golo Kante (1,773 mins), Mateo Kovacic (1,558 mins) or Jorginho (2,273 mins). In the second half of the season Ruben Loftus-Cheek (1,394 mins) saw more game time whilst loanee Saul (480 mins) rarely got a look in.
Plan A was to flank the central midfielders on the right with Reece James (1,864 mins) and left with Ben Chilwell (542 mins) but injuries disrupted these plan and required an amount of improvisation.
After Chilwell was injured in the Champions League in November, Marcos Alonso resumed left wingback duties (2,167 mins). On the right, it was a mishmash of minutes for almost everyone in the squad: Azpilicueta, of course, but also Hakim Ziyech, Callum Hudson-Odoi, Christian Pulisic and Ruben Loftus-Cheek. I have probably forgotten some poor swine forced to do it.
In the forward line, Mason Mount (2,363 mins) was the mainstay. Kai Havertz (1,809 mins) and Romelu Lukaku (1,589 mins) shared the central striker duties, with the remaining minutes spread between the erstwhile wingbacks: Ziyech (1,319 mins), Pulisic (1,290 mins) and Hudson-Odoi (962 mins), plus Timo Werner (1,283 mins).
Did it work?
Defensively, this was sound. Chelsea conceded 32 goals against an Expected Goals (xG) tally of 36, equating to an xG of 0.84 per 90.
The goalkeeping contribution to this pretty neutral, with Edouard Mendy (3,060 mins) posting post-shot xG of 29.6 against a tally of 31 conceded. Excluding one own goal, this is almost bang on expectations.
Kepa (360 mins) actually performed well when deputising, conceding two goals versus post shot xG of 3.4 — a small sample, of course, but he didn’t let anyone down.
Going forward, Chelsea scored 76 goals (including one own goal) at a rate of 1.97 goals per game. This was an overperformance on an xG of 1.77 goals per game. Comparatively, this was the 3rd best xG per 90 outcome in the division, with only — you’ll never guess — Liverpool (2.35 xG per 90) and Manchester City (2.34 xG per 90) ahead of them.
Perhaps surprisingly — given the merry-go-round selection, the much-discussed issues with Lukaku and the the fact that, with just 11 goals, Mason Mount was the leading scorer — only Leicester City’s goals over xG per 90 (0.29) was superior to Chelsea’s 0.20.
What let Chelsea down last season was shot volume. At 15.4 shots per 90 — compared to 18.5 for Manchester City and 19.0 for Liverpool — Chelsea were a good three shots per game light.
All other things being equal, those extra 114 shots at the rate of 0.11 goals per shot equates to a further 12–13 goals, which would inevitably earned more points in more games.
So: can Tuchel coach an effective attack?
The numbers from 2021/22 season suggest he absolutely can, and in fact did.
Although the output wasn’t as high as the top two, and the quality of finishing was also a little lighter, there’s no doubting Chelsea have a good attack.
This should come as no surprise: whilst there is arguably little to be gained by looking at his PSG team (due to the ridiculous talent advantage PSG hold in France), his 2015–17 Borussia Dortmund side had good attacking outputs. 151 goals in 68 games. In 2016/17, his side scored at 0.12 goals per shot (City and Liverpool levels) but, interestingly generating 15.7 shots per 90 — very much in the range of 2021/22 Chelsea.
In terms of goals per shot on target, 16/17 Dortmund and 21/22 Chelsea were very similar, at 0.35 and 0.34 respectively.
So, in terms of output, Tuchel has somewhat recreated his Dortmund side at Stamford Bridge — and no-one thought Dortmund were lacking in attack. Any claim Tuchel has a defensive set up or negative approach is wrong.
So: what’s the problem?
Despite all of the above, there’s no doubt watching Chelsea attack can be a frustrating experience. The numbers in 2021/22 — and a look back at Tuchel’s Dortmund — show he can create a team that is decent going forward.
The Chelsea and Dortmund data suggests Tuchel is satisfied with 15 shots per 90. However, without either exceptional finishing quality or an incredible defence this will always lag the top two when it comes to turning performance into goals and points.
Tuchel himself is evidently often frustrated with how the team plays and how often match superiority fails to translate into scoreboard advantage — and we all know how the football Gods love to punish that.
In short: Chelsea need to deliver more. But is a switch in formation the way to do it?
3–4–3 is not a defensive system as, in possession, the wingbacks join the front three to make an attacking unit of five to keep the pitch wide and asks questions of a back four and shielding midfield. Nonetheless, that reduces the number of attack-first players on the pitch compared to other formations, especially 4–3–3.
A move to 4–3–3 would allow Chelsea to swap a defender for a more attacking player and that might generate the extra shots per 90 that are missing. However, this becomes a question of personnel.
Switching to a back four from three-plus-wingbacks is tricky decision on who to leave out. Assuming the first choice wingbacks remain Reece James and Ben Chilwell then they stay in the team at full back. So we’re talking two from three centre backs.
Kalidou Koulibaly has just joined to be a straight replacement for the departed Rudiger — so surely makes up one half of the centre back pairing. He has proven in Italy he has the tools to succeed in a back four.
Based on 2021/22 minutes, you would naturally choose Thiago Silva as the other main centre back — but therein lies a risk. Thiago’s numbers for defensive actions still look good; there’s no obvious decline that pops in the figures (though perhaps a successful tackles when facing a dribbler of 44% — same as Marcos Alonso — is a point to watch). Nonetheless, at 38 in September, taking him out of the relative security of the middle of a back three, where the space he is asked to defend is smaller, and into a centre back pairing feels like a risky approach against a significant portion of the Premier League.
On the other hand, are Chelsea really going to go into big games leaving a fit and ready Thiago Silva on the bench? Preferring a Chalobah, Levi Colwill or even a Jules Kounde (if that saga ever concludes)? That too feels risky.
So for Tuchel, this is a tricky choice: it’s not just the addition of the forward-minded player, it’s the rejigging of the backline and the relative risk and reward of playing a four with Thiago in it (or not).
Further forward, how does Tuchel composite the midfield? Assuming a 4–3–3 that has a ‘six’ behind two ‘eights’, then Tuchel is spoiled for choice with the eights. Two from Mason Mount, Conor Gallagher, Mateo Kovacic and N’Golo Kante: a set that any team in world football would be happy to choose from. Hell, even Ross Barkley would be a fine fifth option in that set up. Loftus-Cheek would do too.
Indeed, the clarion call to switch to 4–3–3 from Chelsea fans is, in no small part, down to the allure of the promised land of these players working as a group that gets about 1,800 odd minutes each across a season.
Lovely. But what about the six?
Jorginho is, of course, machine tooled to play at the base of a midfield three. But we’ve been here before: whilst Jorginho can and does control the tempo of a game he doesn’t really impact in the final third, as his 21/22 shots assisted rate of one every nine games attests. His defensive activity per game in 21/22 stacked up reasonably well against those generally considered the best in the league. Jorgi’s 5.06 tackles and interceptions per 90 compares well to Kalvin Philips’ 5.59, Fabinho’s 4.40, Rodri’s 3.28 and Declan Rice’s 4.87.
And yet. Like Thiago in a back four, Jorginho as the six feel risky. Even bolstered with the perpetual motion of a Kante or a Gallagher slightly ahead of him, this feels like an Achilles heel in waiting.
This was the issue Frank Lampard faced in his (difficult) second season. He tried to counterbalance the risk of Jorginho at six with a hyper conservative approach to possession, where coughing the ball up was a mortal sin. It appeared to be the tactic that ate itself as the players became so risk averse and with a commitment to recycling that Greta Thunberg herself would admire. They lost their way as they lost their way to goal. The football became almost sad as a result and Frank didn’t last much longer.
Tuchel must be aware of this, so the pull factors of deploying the phalanx of eights are counterweighted by the push factors of death by possession.
Much would be fixed by a more dynamic presence at six, but that would require an addition — an expensive addition — to the squad as neither Loftus-Cheek nor Billy Gilmour really fit the bill. As lone sixes they offer much the same puzzle as Jorginho. Ethan Ampadu is more like it stylistically, as is Trevoh Chalobah, but it would be a big ask to expect them to anchor a midfield three in the Premier League in 22/23.
So Chelsea need a six. Let’s face it: it’s Declan Rice, isn’t it? I’m sure he’ll come eventually. I’m sure in time he’ll slot right in and solve a lot of issues. But eventually and in time do not look like the 22/23 season. And that causes a squad building headache for Chelsea. Given the range of might-work-for-a-bit-after-a-fashion options already on the books, a mid-price stop gap — a Remo Freuler, Djibril Sow or Chieck Doucoure — doesn’t look smart business if, in 12–18 months, serious cash will be spent to bring Rice over from West Ham.
This is a definite quandary for Tuchel and the new executives. There are no great options — or at least not great, inexpensive options.
In an immediate upgrade, Chelsea have swapped out the disappointing Lukaku for Raheem Sterling. With a career xG per 90 of 0.54 and goals per 90 of 0.41 (at a conversion rate of 0.18 goals per shot) Sterling can help Chelsea’s attacking numbers increase; he should help get closer to 18 shots per game and nudge the finishing quality up to 0.12 goals per shot. Excellent. No further questions.
Mount can, of course, play in the front three — but Chelsea have a bunch of players who offer much and deliver not quite enough. The poster boy here is Timo Werner.
Werner’s xG per 90 last season was 0.53 and his xG per shot was 0.17. These are numbers to like very much. Keep talking.
But he underperformed them. His goals per 90 were 0.28, so he frittered away half of his own xG and his goals per shot was 0.09 — so he all but halved that too.
What’s going on? I think there is a clue in his possession stats where the percentage of time he successfully received a pass was 63% — only Lukaku’s frankly dire 47% was worse at Chelsea last year. Now, attackers can have low figures here — Mohammed Salah’s was 66% last year , for example — but Werner’s skews towards the low end. The only Manchester City player with any meaningful game time to have a comparable figure last year (62%) was Ferran Torres, which may explain why City took Barcelona’s money.
But if you couple that with Werner’s dribbling stats — in 21/22 38% successful - and progressive carries per 90 of 4.55, neither of which compare well to other wide forwards, then you see a player who is not making the kind of contribution required from a key component of a side challenging for the league. You don’t have to be top of the tree in all of them, but you have to be do at least one of them well.
I don’t mean to bury Timo Werner. As a fan I like him: he never hides and clearly wants to do well. He’s also been subject to some downright rudeness from some fans and no-one deserves that. In truth, if the rest of the attackers had offered more it would matter less. But no-one really stood up, Mount as the exception.
As a group, the rest were about OK. Havertz underscored his xG a little, Pulisic overperformed his xG a little but they ended up in the 0.40 to 0.45 goals per 90 range (as did Mount). OK. Lukaku also got to 0.45 goals per 90, outperforming his xG by a little bit. Not quite OK for a £97m striker.
So, irrespective of formation, Chelsea need a bit more from their forward line. You can understand the interest in Raphina, with better metrics than Werner all round. It would be no surprise to see Chelsea look to make an additional attacking signing before the transfer window closes. But that probably means moving some of the well-paid players out. Easier said than done.
Ziyech looks like a possibility to go to AC Milan. Was there a smidge of interest in Werner from Juventus? Maybe. Either way, it will be hard to keep so many players around and a bit of squad pruning is required. Pulisic would probably find a suitor, though with him and Hudson-Odoi I think injuries robbed them of good seasons in 21/22 so would prefer them to stay around. Benefit of the doubt.
Chelsea could also do with an alternative option to Havertz through the middle, though — like with the midfield six — there are plenty of folks about who could do it ok-ish. Chelsea obviously have the option of keeping Armando Broja, though his stats at Southampton show he is more Lukaku-like, profile wise: a goal every three games and sub-50% possession retention doesn’t scream Tuchel-ball unfortunately, though to be fair to Broja his pressures per 90 (17) is much better than Lukaku (10) and in line with most of Chelsea’s 21/22 forwards. So you never know.
Where should Tuchel go for 22/23?
For Thomas Tuchel, trying to fit the pieces together means trade offs and risks. The groundswell of support for a move to 4–3–3 risks exposing Thiago Silva and highlights more clearly the Declan Rice shaped hole at the base of midfield. And Tuchel has the ghost of tactics past on his shoulder, sending a shiver down the spine at the thought of reliving the sterility of late Lampard football.
Sticking to 3–4–3 allows the best of Thiago Silva and probably allows Levi Colwill to see a path to first team minutes. However, the issues of last season with all the combinations in central midfield being good-but-not-great remain, with the added challenge of getting Gallagher into the side often enough to keep him happy and deny the media a narrative.
In either formation, Sterling should help improve the attack, but it looks a central striker light and 22/23 will need to see one or both of Pulisic and Hudson-Odoi live up to previous billing. They might do yet.
Despite everything, I think Tuchel should continue with the 3–4–3 formation for another season. It keeps Thiago in the side, it allows Colwill (and Chalobah) to get minutes and it pushes James and Chilwell forward. All good things.
In midfield, Gallagher would make an interesting partner for Jorginho — we know the pros and cons of Kante and Kovacic in there already. By managing minutes there should be time enough for them all. Mount could again spend a season primarily in the front line. He and Sterling could keep Havertz company.
This is not a bad side at all. It would probably lead to similar outcomes as 21/22, with a chunk of an uptick if James and Chilwell stay fit and Sterling’s output replaces Werner’s.
That would be unlikely to bridge the gap to the top two. OK. Not ideal but let’s be realistic.
The risk, though, is it isn’t enough to hold Arsenal and Spurs at bay. Spurs are likely to have their ‘Conte Supercharged’ season. We know what they are like. Arsenal have recruited well and seem settled. They probably have one more year of Saka before he heads to Manchester City too (that’s going to happen, right?).
So a 3–4–3 does risk top four football. But then so does 4–3–3. A move to the higher risk/higher reward formation seems like it would need a lot to go right for it to be a better bet for consistent success than 3–4–3.
A shift to 4–3–3 might mean we go to town more often on weaker sides — and who doesn’t love a goal fest — but I think, on balance, it would make Chelsea more vulnerable to more sides in the Premier League, and in such a tight division we want to reduce variance not embrace it.
Overall, we should not lose sight of the fact that this is a very good squad competing with two all time great sides for the league. The 3–4–3 has been good to Chelsea these last one-and-a-bit seasons. Let’s keep it for one more year before some targeted recruitment, and possibly some fond farewells, mean it is time to shift to 4–3–3.