Is Rodgers right about Hatate's inconsistency?

Using coefficient of variance, the midfielders of recent times are measured for this performance consistency.

Over on the sunlit uplands of Bluesky (you really must join, it’s lovely, so far) I came across this piece by the excellent Marc Lamberts. I find pieces like this on Bluesky every day now. They are lost forever in the sewer that is X.

Anyway, Marc uses the statistical technique of coefficient of variance to measure the consistency of player performance – in this case xG in strikers.

Marc explains the coefficient of variance in detail in his piece so rather than cut and paste his clear explanation, please refer to his column.

TL;DR – a low score (percentage) represents a tight clustering of values around the mean – an indicator of predictability and consistency.

A high score indicates that data values are more spread out, indicating variability or inconsistency.

I used to do this back on Celtic By Numbers some time ago. I recall showing how consistent Scott Sinclair was relative to other wingers. The new huddlebreakdown.com is searchable.

James and I have been building a profile of the central midfield over the last week or so. Mainly due to concerns over the fit of skills to requirements in that area; the ageing of Callum McGregor; and the appraisal of shiny new things - namely Arne Engels and Paolo Bernardo. Then there is the thorny question of Reo Hatate and the manager's suggestion that he needs to find greater consistency in his play.

The method used here will be to select four aggregated metrics so that we get more bang for our analysis buck. These will be:

1.        Defensive action success rate – all defensive actions with passing removed;

2.        Packing score – all packing actions – passes, receipts, dribbles, turnovers and recoveries;

3.        Expected scoring contribution – expected goals plus expected assists; and

4.        Celtic attacking threat score – a count of valuable attacking actions.

Slightly biased towards attacking actions as we will see.

I will include all the main central midfielders including the defensive and attacking ones so that we can see the profile of the different roles.

I have included all the games each player has played I have the data for where the player played at least 45 minutes. Secondly, I have adjusted each value where a player played less than 90 minutes to reflect what their score could have been had they performed at that level for the full match.

Two things to bear in mind throughout.

1.        This isn’t trying to prove who is the “best” or delivers the greatest quality, it is assessing consistency.

2.        There will be variance based on specific roles within the midfield structures.

So, coefficient of variance, meet midfielders!

Awright, let’s get fell in.