The impressive Celtic data that shows how far Ange Postecoglou has taken Celtic - James Dailey

I was taught by my primary analytics mentor, who is a theoretical physicist, to start by asking a question and then align the analysis to answer the question.

Seems so simple, right?

'How did Celtic perform in Tuesday’s game versus Midtjylland?' is a simple question, but a key part of analytics is learning to ask questions which uncover valuable insight.

For the game Tuesday, here are some potential questions:

How did Celtic perform in Ange Postecoglou’s first competitive game?

How did Celtic perform for a first champions league qualification round of the season?

Both of those questions appear reasonable, but what can be learned by answering them? What are the variables and do they make for an apple to apple comparison or apples to oranges?

Here is the question I asked for today’s column:

How did Celtic perform against Midtjylland relative to performances against comparable opponents in Europe in recent years?

My primary source for football data has historically been Wyscout, though thanks to this column, I also now have access to Statsbomb for SPFL Premiership games.

Wyscout’s coverage began for the 2015-2016 season, so that is when my benchmarking sample began. I’ve included 18 games in the benchmark versus opponents such as Malmo, Hapoel Be’er Sheva, and Copenhagen.

Here are some metrics from Tuesday’s game versus the average of the 18-game benchmark I created:

These are just a few key metrics upon which to begin an analysis, but I believe offer a good starting point.

Generally speaking, Celtic were very effective in limiting the opponent's volume and quality of shots, while doing a good job of creating quality shots when attacking.

In fact, the 2.93 in xG was higher than any of the 18 games in my sample, with the next highest at 2.83 vs Copenhagen in the 3-1 loss at home back in February 2020. In addition, that xG included a penalty kick, which accounted for 0.76.

READ MORE: Why Celtic should look at the South American transfer gold mine for young, hungry players

Considering non-penalty xG, and things like xG from open play, xG from set-pieces, etc. are all varying ways to analyze performance levels.

On the defensive side, the 0.21 in xG conceded to Midtjylland was lower than all but one game in the benchmark sample. Celtic conceded just 0.03 xG to Rosenborg in the Europa League group stage game on September 20, 2018, in a 1-0 victory at Celtic Park.

Given the back four Celtic fielded in Tuesday’s game, I think many supporters would have been in disbelief in advance of the match if it were suggested that a dominant defensive performance was going to occur. For some perspective, the back four in that Rosenborg game was Lustig, Boyata, Benkovic, and Tierney. What accounted for the relative success in suppressing Midtjylland’s attack?

Here we can see the locations from which Midtjylland launched its thirteen entry balls into Celtic’s penalty area, followed by a screenshot of the only one which occurred centrally by Dreyer in the 55th minute. Celtic played a lot of 4-2-3-1 under Neil Lennon, which is a system that can be vulnerable to defensive transitions. Ange Postecoglou deployed what appeared to be more of a narrow 4-3-3 prior to Bitton’s red card.

In addition, Celtic have been deploying “inverted fullbacks” throughout the preseason, and that tactic continued on Tuesday. The result was a solid midfield structure from which to funnel Midtjylland’s attack out wide.

I believe a key to this narrow setup was deploying Soro as a single pivot, with a relatively narrow area to patrol. To highlight the nature of how he was deployed, consider his map for total actions against Midtjylland and compare it to Scott Brown’s versus Copenhagen from February 26, 2020.

Soro's is first.

With Taylor and Ralston inverting at times in addition to Soro’s relatively narrow positioning, Celtic were able to extinguish transitions before Midtjylland were able to mount serious attacking threats.

Here is a map showing the nineteen entry balls into Celtic’s penalty box, where you can see they had six which came from a more central area and accounted for two of the three goals conceded:

Of course, performance levels do not necessarily translate directly into results in football. I ran the xG sequence for both teams from Tuesday’s game through a Monte Carlo engine, and it produced a 96% probability of Celtic winning the game.

In single games and knockout football, suppressing the volume and quantity of opposition scoring chances is vital to optimizing a team’s chances of achieving results. Because I am an analyst and not an algorithm, I can use my experience and some discretion.

READ MORE: Ange Postecoglou's Celtic success hinges on how much real autonomy he’s been granted by the directors - Kevin McKenna

I thought Wyscout’s xG metrics were too high for a couple of Celtic’s chances, but even when I made some manual adjustments, that only reduced the Monte Carlo probability down to about 85%-90%.

My conclusion and answer to the question above? Celtic performed very well.

Given the context of the state of the squad and having played meaningful minutes down a player, I believe it was an excellent performance to begin the Ange Postecoglou era.

Ange’s track record and reputation for getting more out of teams than the sum of the parts was on full display Tuesday. Long may it continue.