Who should be Celtic's first-choice midfield three?

How long can Callum McGregor be relied upon? Has Aaron Mooy’s post-World Cup form cemented his place in the team? Can Matt O’Riley recapture the form he showed prior to deputising for the captain?

How many scoring cameos does David Turnbull need before he gets more minutes? How do you explain an enigma like Reo Hatate?

There is an even bigger question beyond the scope of this piece: is the midfield athletic enough to compete in the Champions League?

For now, we will take a deep dive into the central midfield three’s performances, to try to answer a more basic question – who should be the first-choice three for the rest of the season?

Context

Context is everything in this game, and it is important to make a few things clear before consuming the appropriate data.

The Scottish champions play a variation of a 4-3-3 and the scope is the middle three positions. One of those areas is more of a pivot or number six role, normally played by McGregor. It is reasonable to expect Tomoki Iwata to get some minutes here during the remainder of the season.

Ahead of the pivot are two nominal eights who have a more attacking brief. At the moment there are four players vying for these two slots – Hatate, Mooy, O’Riley and Turnbull. Sead Haksabanovic has played minutes as a more orthodox number ten at times but is not included here.

It must be remembered that for 12 matches and 966 minutes, O’Riley performed in McGregor’s pivot role and his data will therefore reflect that. This means his defensive numbers will be higher than normal and his attacking data lower.

Below is the number of 90 minutes worth of game-time each player has completed and the average number of minutes of each appearance.

Turnbull and Mooy are just over the 10-match threshold I consider meaningful for the data. Note also that Turnbull only averages 36 minutes per appearance and Mooy 45. McGregor’s game time averages out at 85 minutes per appearance. These are all factors to bear in mind.

Defending

The framework for defensive performance analysis compares defensive action success rate and possessions won from defensive actions.

For the midfielder cadre, this shows:

McGregor is the clear defensive leader.  As set out above, O’Riley is next most effective as might be expected given his 12-game stint in the pivot role.

READ MORE: Will Celtic actually miss Moritz Jenz after loan transfer exit?

Mooy is not far behind O’Riley, with Hatate delivering much less defensively and Turnbull trailing. The former Motherwell man's numbers are more similar to Tom Rogic who tended to play as a traditional number 10 in a more advanced attacking midfielder role.

The next graph shows the average packing score per 90 minutes from recoveries and turnovers. A pack recovery is when you win the ball back from the opposition thus leading to some of their players being on the wrong side of the ball and out of the game. A pack turnover is where you give the ball away leading to some of your own team-mates being out of the game.

O’Riley has a distinct skill set here and averages a pack score of 18.6 per game recovering the ball and -13.1 losing it. He has the best differential between the two. This means, on average, O’Riley will win the ball back more than he will cough it up in transition.

Mooy also has a positive differential and McGregor is neutral.

Both Hatate and Turnbull recover the ball to the same extent but the Japanese player's propensity to turnover means he has an overall -10.2 packing score differential between recovering and turning it over. 

Hatate and Bernabei would have the worst profiles in this regard across the squad. Which is why playing them together is problematic.

Ball Progression

Moving on metaphorically and in actual ball forward momentum then. 

The below charts ball carries which are where the player progresses the ball at their feet in the opposition half with pack passes – forward passes that take opponents out of the game.

Mooy and McGregor have the highest numbers on pack passes with a score of 54 each per 90 minutes.

Turnbull and O’Riley are both on 42, Turnbull plays slightly further forward which will reduce the pack pass score and O’Riley’s low number is a bit of a surprise.

Hatate is in the middle, pack passing-wise, but has the most ball carries per 90 minutes with 2.64. 32-year-old Mooy with 1.02 has the least.

Chance Creation

We are getting to the exciting part of the midfield play now and it going to be broken down into three parts.

Firstly, a deeper build-up is reflected by plotting the number of secondary assisting passes per 90 minutes (these are passes completed before the pass that sets up the chance) and chances created from pack passes. The latter is a completed pack pass (a forward pass that takes opponents out the game) that also results in a chance.

Four of the midfielders are in a cluster – almost like a central midfield formation. Turnbull is the outlier with 2.56 secondary assisting passes and 1.18 pack passes resulting in chances per 90 minutes. Some of this may be due to set piece taking but this is a nice surprise as in the past Turnbull has not usually come out on top for creative deep passing.

Secondly, let’s look at getting the ball in the 'mixer'. This time successful crosses have been plotted with the number of completed passes into the danger zone (the area within the box and central to goal) per 90 minutes:

Given these are creative actions much closer to goal it is no surprise McGregor is lower placed within this view.

O’Riley has the most successful crosses with 0.58 while Turnbull has the most danger zone passes completed with 1.87 and Mooy is not far behind.

READ MORE: How squad rotation has helped Celtic avoid an injury pile-up

The outlier is probably Hatate with significantly lower numbers than the other number eights.

Finally, in this section the more traditional view of expected assists (xA) and number of chances created per 90 minutes. Essentially chance creation quality and quantity.

With this view, it is O’Riley that is the real positive outlier with 0.54 xA (anything over 0.4 is very good) despite those 12 matches spent in a deeper position. He also creates 3.84 chances per game, 1.16 more than his nearest peer.

It is remarkable McGregor rates so highly given his deeper role in midfield. 

Mooy appears to be more creative in the build-up as opposed to that final killer pass.

Goal Threat

Using another traditional framework of expected goals (xG) and volume of shots attempted gives us the overall goal threat for each player:

McGregor is a predictable outlier due to his role.

Mooy perhaps surprisingly has the highest overall expected goal value at 0.35. A small sample warning plus he did some admirable stat padding in the recent 5-0 drubbing of Greenock Morton. Mooy had seven shots and 1.86 xG – his total xG for the season is 3.2 so more than 50 per cent of his season’s total occurred in that one game against lower-league opposition.

O’Riley is just about the only outfield player without a goal this season despite averaging 0.28 xG per match. 

Turnbull has scored a couple recently and getting goals late in games when coming on as a substitute is a great skill but also context is important. The starters soften up the opposition for 60 minutes before the finishers come off the bench.

Hatate is probably the outlier of the eights with the lowest xG of 0.26 and attempts only 1.79 shots per game.

Summary Stats

Summarising the overall attacking threat of the players can be achieved by plotting expected scoring contribution (xSC – which is expected goals plus expected assists) with my proprietary attacking threat metric that aggregates a host of volume metrics such as shots, touches in the box and chances created.

There is a slight surprise that Turnbull and Mooy carry a higher overall attacking threat than the first picks at the beginning of the season in O’Riley and Hatate.

Two important pieces of context again are the role O’Riley played for 12 matches (i.e. deeper) and that Turnbull and Mooy have made more substitute appearances and the data explains that the average xSC coming off the bench is higher than if you start due to the depth of Celtic’s squad against domestic opposition.

Finally, the StatsBomb On Ball Value (OBV) calculates the positive and negative impact of each action a player makes in relation to the team’s ability to score a goal.

They have OBV’s for passing, defensive actions, dribbling & carrying the ball and shooting resulting in an overall aggregated OBV.

In overall OBV reverse order (right to left):

O’Riley’s passing OBV is the highest single value at 0.15. No player scores positive in all values but neither Mooy nor McGregor score negative in any.

McGregor has the highest overall OBV with 0.29 with a defensive OBV of 0. The model does not like Mooy’s shooting or dribbling which meets the eye test.

O’Riley is surprisingly negative for defensive actions which I would question given the data above. Hatate is generally low on all values with a negative shot OBV.

StatsBomb hates Turnbull’s shooting and while he has converted two recent long-range strikes into the bottom corner, the model remembers all the many failed efforts from similar distance.

Conclusion

What does this all mean? Unsurprisingly, McGregor is unique in his ability to move the team forward from the deeper pivot six position. The more contentious question is who partners him?

Here is the summary per discipline:

A complex picture to be fair. If Celtic relied upon the data there is a very strong case for Mooy to be the first choice number eight looking across a range of attributes. The caveat would be that he did not play many minutes in the Champions League (152) therefore most of his minutes have been against inferior opposition. 

Turnbull and O’Riley are very close and the context for them is O’Riley’s data has been suppressed by playing 12 matches as a pivot whilst Turnbull has benefitted from the substitutes dividend – coming on late against tiring teams to rack up good data. 

So, it could be argued O’Riley is second and Turnbull is third based on this season’s data. 

Hatate is, as signposted, the enigma. Adored by much of the support he is the creator of magical moments – the goal against Hearts at Tynecastle last season, two goals in the 2022's midweek derby, the volleyed cross-field pass against Real Madrid. But across an enormous range of metrics covering defending, deep build-up, creativity and goal threat, he is consistently behind his peers on overall volume and impact.

That is not to say, as a young player, there is not plenty of room for growth and development and hopefully that comes to fruition.

Back to the start, while the midfield three of McGregor, Mooy and O’Riley is the one the data would point to as first pick, there would still be a concern about the speed and athleticism and physicality needed to be a Champions League midfield capable of upsetting a pot two team.