Lies, damned lies, and statistics was a phrase popularised by Mark Twain in the 19th century. The idea continues to hold a lot of 'truth' and it is why my experience has been that quality analysis is typically done with a healthy dose of scepticism and humility.

Data quality varies, models typically have positives and negatives, and the people utilising all of this have varying skillsets, training, experience, and their own natural human cognitive biases. This can all manifest in healthy scepticism about 'stats' and the risks of them being utilised in counterproductive ways. 

StatsBomb sent out the following tweet this week that gave me an idea to explore these issues:

I have tried to develop a conceptual framework for football performance attribution over the past couple of years with the “expected trophies” concept as a centrepiece.

A basic idea from the concept is that the drivers of club performance levels flow down from relative financial resources to the quality of deploying those resources (recruitment, academy, etc.), followed by coaching decisions and finally idiosyncratic player contributions combined with variance (i.e. luck). 

Much of the football analytics world is focused upon the latter and often through the lens of recruitment. However, I believe it is valuable and important to consider player-specific performance metrics within a broader context.

For example, here is a radar comparing StatsBomb On-ball Value (OBV) metrics for James Tavernier and Alistair Johnston's league games this season:

With the obvious caveat that these sample sizes are not large enough to draw enduring conclusions, it offers a starting point for discussion.

Relative to the OBV model, it is an objectively true statement to say that Tavernier has been almost twice as 'valuable' per 90 minutes as Johnston in the games covered in the report. However, value does not necessarily equate to quality.

The graphic above from StatsBomb’s tweet is indicative of the specific system and role which Tavernier plays. In contrast, under Ange Postecoglou, Celtic have largely deployed their full-backs in a different role.

Tavernier does a lot of overlapping, crossing, pressing and counter-pressing within the opposition's final third whereas Johnston is often asked to play an inverted role within Celtic’s current style of play.

Here are radars displaying attacking and defending-related metrics for the two players:

The attacking radar offers further context to the disparity in OBV between the two players, as Tavernier is obviously a cornerstone of set-pieces for Rangers while Johnston has just a fraction of involvement via some throw-ins.

It just so happens that Tavernier has been pretty good at dead-ball deliveries but that is independent of comparing the contributions of each player from open play.   

Other statistical proxies I like to monitor for additional context relate to relative efficiency and average quality.

For example, while their utilisation and roles are different, Tavernier’s average xG assisted from open play is 0.099 versus Johnston at 0.086. Given the context of Tavernier being involved with more aerial crosses, which on average are generally lower-quality chances, this is a potential signal that he may be more skilful in that phase of the game.

We can also see from the data relating to dribbling and carrying the ball in possession that Tavernier’s related OBV metric, again for this limited sample, may be a signal of some relative advantage versus Johnston.

Where the nuance of comparing the two shines through is on the defensive side. The OBV radar shows a modest advantage to Tavernier on a per 90-minute basis but this is where understanding the model and how each player is utilised is important.

Celtic Way:

Note the average distance for pressing and defensive actions for the two players - Tavernier is generally deployed further up the pitch, and combined with less controlled possession. His role while playing for Rangers results in a higher volume of actions (i.e. quantity).

Johnston, in contrast, stands out in a material way relative to quality. For example, despite him being involved in materially fewer actions, his volume of regains is comparable to Tavernier.

His efficiency in winning duels in tackles and versus opposition dribblers is also superior and benchmarks very highly amongst his peer group. The one defensive issue that stands out as a potential issue for Johnston is his fouling rate, where aggression can manifest in things like conceding dangerous free kicks.

I can imagine a reader thinking 'this rocket scientist has gone through all this to say that Tavernier is an attacking full-back and Johnston is a bit more defensive - duh!' 

The point of this exercise was to use a fairly obvious case to try to highlight some of the important variables which I believe are underappreciated. 

Neither Tavernier nor Johnston controls what system or style of play in which they play. Tavernier has provided his team with a high level of value for the role he has been asked to play but that role is very significant and quite different from Johnston’s. 

READ MORE: The ghost in Ange Postecoglou's Celtic machine

Additionally, given the way in which a model like OBV is designed, it will have a natural bias towards full-backs playing a role similar to Tavernier’s at Rangers. Expected goals-related metrics will also naturally skew towards full-backs that play an attacking role in sides that dominate their leagues, as both Rangers and Celtic do.

In many ways, a comparison of Johnston and Tavernier is really a comparison of the styles of play for Celtic and Rangers.

Rather than getting bogged down with questions such as which player is 'better', each player’s statistical profile is reflective of a broader puzzle.

Celtic’s system and style of play under the design and leadership of Postecoglou has resulted in an extreme degree of domestic dominance that is also reflected in a wide breadth of underlying statistical models and metrics.

I would argue that having players thrive in the roles they are asked to fulfil is part of what creates a dynamic where the whole is performing at a level greater than the sum of its parts. 

While stats can be used and weaponised for debates on which player is 'better' at the pub for friends on opposite sides of the Glasgow Derby, Johnston is excelling at performing what his manager demands.

With supporters celebrating 100 games into Postecoglou’s tenure, it is the commitment and flourishing of players in their respective roles which is helping the team perform at such a historic level and ahead of what the expected trophies framework suggests as a benchmark for Celtic over his reign.