As a data guy but, more importantly, as an analyst as much as I love to tell you all how great player X and Y is doing and here are some mind-blowing stats, I mainly see the role as that of a sceptic. Also, as a natural-born worrier, I am always looking out for stuff to be worried about.

Celtic enjoy strong dominance in most domestic Scottish matches but there is always the odd banana skin. In years gone by that was often Livingston. More recently, it is St Mirren. ‘What causes those issues and how can Celtic improve?’ is what I have in my mind.

Secondly, the gulf between Scottish league football and the Champions League is ginormous. It is a conceivable estimate individual Celtic player stats regress by about 40 per cent against this level of opposition. It is the massive difference in the quality of opponent that is causing this.

So what are the risk indicators that could cause concern at that level even if, in Scotland, it is rare teams have the ability to exploit those weaknesses?

In that spirit, then…

‘Toxic’ combinations

I introduced this phrase in the 2020-21 season to help explain how a backline with Shane Duffy, two full-backs in Jeremie Frimpong and Diego Laxalt, who often bombed up the field at the same time, and an ageing Scott Brown as its protector was all a bit… tricky.

There were certain combinations of players that simply introduced too much risk for Celtic. In this case Duffy, used to playing in a deep defence well protected by a compact midfield, was horribly exposed due to the lack of cover that the system under Neil Lennon provided.

Under Ange Postecoglou, it is remarkable that the cohesiveness of the system of play, where the positions of players are carefully rehearsed and implemented, can absorb a multitude of weaknesses.

It is noticeable, for example, that Carl Starfelt’s error rate declined markedly over the last season despite his high-risk playing style remaining extant. In short, the strength of the collective allowed for the mitigation of areas of concern. The players also, to their credit, grew more comfortable with what was being asked of them.

Finally, in the later part of the 2021-22 season, it was mainly a core of 17 players utilised so understanding grew. Nevertheless, we must remain highly sensitive to new toxic combinations.

Mitigating weakness

I wrote, on my own Celtic By Numbers site before the Kilmarnock League Cup semi-final, a warning about how unbalanced the attack had been in the Premiership encounter against the same opponent.

The thrust of the article was how Celtic recycled the ball between Carl Starfelt, Alexandro Bernabei and Reo Hatate and largely ignored the right flank despite the aggressive movement of Alistair Johnston and the normally highly active attacking outlet that is Jota.

Indeed, Tony McLaughlin (Celtic Trends) put forward that 53 per cent of Celtic’s attacks came down the left as opposed to 20 per cent down the right and the balance down the middle.

READ MORE: Don't write off 'busy' Alexandro Bernabei just yet

What transpired in that match is that lack of progress down the left was forgotten when Bernabei split the whole Kilmarnock midfield and defence with a pack pass to Daizen Maeda, who crossed for Jota to score the opener.

The second half was more cohesive and Hatate in particular took up intelligent positions and found pockets of space as well as setting up the decisive second goal by forcing Ash Taylor to put past his own goalkeeper.

This illustrates that the weaknesses Celtic have can often not manifest as glaring issues in the Scottish game due to the lack of ability of the opponent to exploit.

This is generally not true in Europe, hence worrywart me.

League Cup semi-final

The only change Celtic made for the semi-final was at right-back, where Josip Juranovic made a somewhat surprising reintroduction given the persistent transfer rumours surrounding him. The left side remained as was.

In general Celtic were far better balanced going forward than in the league game. Juranovic’s thrusts down the right, at pace, were particularly noteworthy. Yet, overall, this was a remarkable game for the champions and not in a good way.

Packing data covers a multitude of actions. It records forward passes completed that take opponents out the game as well as runs with the ball.

I also record turnovers and recoveries that take players out of the game. A recovery takes opponents out of the game. A turnover means your own players are ahead of the ball therefore out of the game and this is scored with a negative value depending on the number of players out of the game.

The sum of all these actions results in an overall packing score for each player and for the team. Packing is a great indicator of opposition quality. This season, Celtic have only had a lower packing score than their opponent on five occasions:

Celtic Way:

Guess what? All five occurred in the Champions League.

On Saturday, Kilmarnock became the first team in Scotland this season to have a higher overall packing score than Celtic – and it was not even close. Killie’s packing score was 350 to Celtic’s 267.

Only twice this season have Celtic had a packing score less than 100 more than their domestic opponent (October 22 at Tynecastle, +10, and January 2 at Ibrox, +22).

So what drove this?

Left-sided woes

Let’s go back to the combination of Starfelt, Bernabei and Hatate. If we look at the impact of turnovers across the squad, the following are the turnover pack scores for the squad so far this season:

Celtic Way:

Bernabei and Hatate are by far the most impactful in terms of turnovers damaging to the team and Starfelt is third as well as being the central defender with the most. Contrast with Cameron Carter-Vickers on 3.56 per 90 minutes as opposed to 16.3 by the Swede for what ‘circumspect’ looks like. This is the very essence of a toxic combination.

Against Kilmarnock, their most successful attacking threat was positioned on the Celtic left in Daniel Armstrong. He created three chances, had two shots and one secondary assisting pass. His total expected scoring contribution was 0.74 – the next highest in his team being 0.26.

Long balls were launched onto Celtics left side to further test the backline but it was self-inflicted in the main.

Twice in the first half Starfelt passed the ball directly to an opponent under no pressure. Both times led to shots at goal; Rory McKenzie’s shot went through Carter-Vickers’ legs, was wind-assisted and skidded off the wet turf thus forcing an exceptional Joe Hart save. Fortunately, the other effort fell to the hapless and leggy Kyle Lafferty who, unable to run much, tried to score from 45 yards.

Can you imagine if Vinicius Junior or Christopher Nkunku were the recipients of those gifts? Not a nice thought.

Starfelt completed 90 of 93 attempted passes and in all other respects was arguably Celtic’s best performer – but those two lapses were potentially game-losing.

Celtic Way:

Bernabei twice lost possession in his own defensive third and gave away three packing turnovers. In total, 10 of his 70 attempted passes went to the opponents. One turnover led to an Armstrong shot in the second half.

Hatate was responsible for five turnovers, the most on the day, and did not balance that by winning back the ball to force an opposition turnover.

Against Kilmarnock in the league, the pair had mitigated a similar pattern by providing key moments to set up goals. That did not happen here and there were no upsides to this combination of carelessness and high-risk passing. Hatate created one chance and completed only four pack passes.

The point here is, if you are going to turnover the ball in dangerous areas, there had better be some upsides in terms of positive outcomes from risky passes.

Secondly, while Kilmarnock lacked the final-third quality to exploit those moments, teams like Real Madrid and RB Leipzig did not and will not again.

At half-time in the Killie game, both Starfelt and Hatate had negative packing scores overall and Bernabei’s was only five. To put it into context, Aaron Mooy led the team on overall packing with 122.


Celtic must be wary that inherent weaknesses introduced through toxic combinations of players may go unpunished in Scotland but almost certainly will not against better opponents in Europe.

Some players are more inherently risky and/or careless in their distribution and putting three such players in the same part of the pitch is inadvisable.

The combination of Starfelt, Bernabei and Hatate breaks even Postecoglou’s well-drilled system. Some adjustments will be needed until Greg Taylor is fit enough to resume his place at left-back.