Stuart Armstrong: The End of the Affair?
Armstrong’s form continues to concern Celtic supporters. The nadir was a first half against Anderlecht that saw a mid-game substitution, and a noticeably more effective performance from the replacement, Ntcham. Armstrong undoubtedly suffered early season until his contract extension was sorted out in August. But now, is it the same Armstrong?
Your magnificently maned Glossary can be a faithful companion when terms get nerdy.
With Thanks
This article is intended as a companion piece to one from the excellent The Backpass Rule which I recommend and can be found here What’s Wrong With Stuart Armstrong?. In addition, the equally excellent The Rangers Report tweeted a number of statistical indicators illustrating well the worth of Armstrong. I will reproduce them throughout with thanks to @TheGersReport.
I am not going to repeat the detail of The Backpass Rule article as he covers it well and would summarise the following:
Armstrong’s early season form suffered probably due to uncertainty over contract negotiations and a possible transfer.
Mainly through the purchase of the expensive Ntcham plus the development of McGregor, Armstrong has played less minutes than last season.
Armstrong’s goal scoring, especially from range, was statistically unlikely to continue, with a regression of performance to the mean likely.
His role in the team this season has changed subtly, and he is more an orthodox central midfielder, getting forward beyond the strikers slightly less.
What I will do is share the data I have for Armstrong this season to supplement the above. My data covers Celtic in all competitions, whilst the fellows mentioned above have data for Scottish league games only.
I include full season data for 2016/17 and season to date for 2017/18 EXCEPT the 5-1 home win versus Motherwell and the 3-1 home win versus Hamilton Academical that I have not captured yet.
Where you see “NA” it means I was not collecting that data until this season.
Appearances
More pertinently, Armstrong’s utilisation is 55% - that is he has appeared in 55% of all available game time. Last season he was utilised 62% of the time. Armstrong has completed 90 minutes only 8 times this season. Across all competitions, he is being used slightly less than last season.
Defending
Brown unsurprisingly leads the Celtic midfield defensive stats with 64% and, as a side bar, it is perhaps a concern no other central midfielder comes close to this. However, Ntcham and Armstrong are jointly on 52% defensive action success rate. This is a 5% uptick for Armstrong on last seasons performance. His more orthodox midfield role demands this, and this at least shows development.
Note that Armstrong is dribbled past on average 1.12 times per 90m which is the most in the squad. In terms of body shape, positioning and strength in those scenarios, room for improvement.
Possession
Armstrong’s in game involvement is almost identical to last season (79 possessions per 90m). But he is completing around 7 more passes per 90m, and 3% more accurately. He has also improved in terms of the instances of possession won versus lost.
Overall his possession effectiveness (% of time he keeps possession from any possession event) is up, although I have tweaked the way this is recorded so that may be playing a part here.
Overall, steady improvement commensurate with a more disciplined central midfield role.
Creativity
Headlines are that Armstrong’s assist rate has gone up very slightly by 0.03 per 90m. He is creating 2.49 chances per 90m – only Rogic and Griffiths exceed this.
Getting beyond the jargon of Pack Pass data: essentially it puts values on forward passes in terms of how many opposition players are bypassed. It scores opposition members bypassed by position, the higher the score the more opposition bypassed, and the more defenders bypassed the higher the Impect score. Armstrong WAS the Packing King. New kid on the block Ntcham leads Pass Impect with 58.73, Armstrong is second.
19% of Armstrong’s passes take out at least one opponent, the highest in the squad. This is Armstrong’s special skill – the ability to play passes through defensive lines. It is an invaluable skill.
Note his xA per 90m matches precisely his actual assist rate as noted in The Backpass Rule article.
I reproduce here three tables from The Rangers Report to illustrate further Armstrong’s worth. They all relate to SPFL matches only.
Establishing Passes
I will quote @TheGersReport. “Establishing Passes per 90 leaders in @spfl - basically the 'third shot assist'. These players are consistently making the right pass prior to the build-up that leads to shots.”
Sidebar comment – Boyata??
2. xG Chain Leaders
Again: “xG Chain per 90 leaders - gives a value to players ability to be involved in passing network (up to 3 passes) that lead to shots. Yes, I have no problem with @celticfc leaving Stuart Armstrong on the bench moving forward”.
3. Scoring Chances
Same as Creativity above but SPFL only – “Scoring Chance Key Passes per 90 alongside what % of their Key Passes set up Scoring Chances (kicked shots from heart of box & headed or kicked shots from within 6-yard box).”.
These are all different ways to illustrate Armstrong’s creative talents. League leading in many instances.
Goal Scoring
Armstrong’s 17 goals last season was always going to be a stretch to maintain. He had scored 6 by mid-December last season and has 3 this, so don’t discount a post New Year scoring streak!
He is managing more shots on target this season but is taking more shots from outside the box. Overall, he is maintaining 40% shot accuracy, but perhaps influenced by last seasons results is taking more long shots. His xG per 90m is very close to his actual of 0.18 goals per 90m.
It would be great to see Armstrong making more of those late runs into the box beyond the strikers, and to taking less shots from distance.
Overall Attacking Threat
Armstrong’s overall Impect score is second only to Rogic. He is an all-round midfielder capable of making passes, finding space to receive in advanced position, and running with the ball.
His overall Scoring Contribution (goals plus assists) is down from 0.66 to 0.414 and he ranks 9th in the squad, slightly disappointing. His Scoring Contribution per minute and per possession are down significantly.
Overall, his CAT score is 10th highest in the squad.
Conclusion
Football statistics is pretty good at describing goals and chances. It is less well developed describing general midfield play but The Rangers Reports data plus the Packing concepts illustrate Armstrong’s special skill set. He leads in many creative categories. He has improved in many aspects of general central midfield play. His goal scoring contribution is significantly lower than last season’s exceptional return.
We need to relearn to love Armstrong and I hope the various data points introduced here may assist in this regard.
