Is 3-5-2 Working?
By James Dailey
I believe an important aspect of being an analyst is to evaluate one’s work and be honest in that assessment. This is my 10th piece for Celtic by Numbers since Alan was gracious enough to allow me to share my work on his platform (I know superior talent when I see it! – Ed). As with anyone, I am sure I’ll have some real stinkers and be horribly wrong at times. My goal is to be right more than I’m wrong, while hopefully challenging how readers think and not be too boring!
2 not 1
My first piece looked at our strikers and my basic conclusions were that Edouard plays more like a left sided forward or number 10 and would benefit from being paired with a more traditional box striker, whether that was to be Klimala or Bayo. I did not include Griffiths in my analysis for a variety of reasons, but I will admit that I was skeptical about his return to being productive at a high level.
Another one of my pieces examined how odd Lennon’s nearly exclusive use of the 4-2-3-1 prior to the winter break was out of character for him given his managerial record, and that he would probably go with two strikers with 4 fit options. Since the break, he’s done so in five out of six games.
Along with better utilizing Edouard’s talents, I also speculated in another piece that either the 3-5-2 or 4-4-2 diamond would help address what I’ve seen as the biggest problem under Lennon’s tenure, which is how he’s misused McGregor, and our midfield in general.
Striking Numbers
Now that I’ve laid out a brief summary, let’s look at some numbers:
These figures are all for SPFL league games and are broken down into pre-winter break games and then the five games league games played using 3-5-2 as the primary formation: the December Hibs game is in both samples. As one can see, performance levels are improved both offensively and defensively. The normal caution about sample size and relative strength of opposition in the samples apply.
This group of data is again all SPFL league games and compares each player’s production per 90 minutes (excluding penalties and free kicks), while again breaking them out by pre-winter break and the games using the 3-5-2 formation. I’ve also included a line at the bottom titled “Edouard Partner,” which includes when Forrest played as a striker with Edouard versus Hibs in December, and then also when Griffiths was paired with Edouard as part of the 3-5-2, but also the one Cup game versus Partick Thistle where we played 4-4-2 diamond.
The Pre-Break data segment includes the primary attacking players in the 4-2-3-1 along with McGregor. As you know, Ntcham and Christie split the number 10 position at times and Christie also played wide right frequently after Elyounoussi dropped out due to injury.
It is important to recognize that there is some reallocation of offensive performance due to the change of formation. It is only logical that Forrest would generate less xG at the wingback position versus as an attacking winger, and the data certainly reflects that. However, despite all the continued handwringing over Forrest, his output is still tremendous for a wingback – offensively. I share the concerns over Forrest’s production defensively in that role, but his xG and xA dwarf Taylor’s 0.02 and 0.16 production as a wingback in the 3-5-2. In my opinion, Bolingoli and Frimpong are better suited to offer a more balanced offensive and defensive profile at the wingback positions, and hope we get the chance to see that combination.
Ntcham’s offensive production level has dropped some, which is also to be expected due to the shift from playing an 8 versus 10 in midfield. However, look at the explosion in production from Edouard, McGregor and Griffiths! Edouard’s scoring production is slightly improved, but his xA has ramped to that of a great number 10. Edouard has produced in the 3-5-2 at the level of an elite striker AND an elite number 10 all at the same time.
To put that 0.39 xA figure into perspective, Kevin De Bruyne has an average xA per 90 minutes of 0.37 over his career in EPL league games (only Messi consistently averages over 0.4 per 90m – Ed). McGregor has similarly seen his xA explode from 0.16 to 0.29, which on an output basis, takes him into the neighborhood of the likes of Eriksen, peak David Silva, etc. As always when I make such comparisons, I am not suggesting the players or leagues are of comparable quality but offer a perspective as to the level of output at the level they are playing.
Despite all the rhetoric about not having enough fit or effective centre halfs, our xG conceded in the 3-5-2 has dropped to 0.42 per game. Brown’s defensive stats (not shown) have improved modestly with his move to a central midfield role.
The one issue where I think I’ve been objectively wrong was my skepticism towards Griffiths’ ability to return at an acceptable level. His production has been tremendous, though I will qualify it with one caveat. His direct link up play with Edouard has been very mixed, and Forrest posted an xG and xA per 90 of 0.64 and 0.40 in the game versus Hibs when he was paired with Edouard as a striker. Griffiths deserves credit for his efforts and production, but I am not yet convinced any of our natural strikers wouldn’t produce similarly playing with the brilliant Edouard and an attacking McGregor. Regardless, I am thrilled about having been wrong.
Conclusion
Lennon seems to have reverted to type by playing two strikers, which has also “fixed” our midfield and unleashed McGregor and Edouard’s creativity. We are scoring more goals, conceding less, and the underlying xG data indicate the improvement is legit. Of course, more games and expanding the sample size would help separate the signal from the noise, but so far?
All hail the 3-5-2!