How Ange Postecoglou 'dark mutterings' gave way amid Celtic line-up volatility and absorbed variance - Alan Morrison

“Buckle up,” we were told. “It’ll be a rollercoaster.”

The ‘Age of Ange’ would see all-action, high-intensity attacking football.

Defence? “That’s for drongos, mate.”

The early part of the season seemed to bare out seasoned Ange Postecoglou watchers’ predictions.

There were extreme sugar-rush performances: 6-0 wins against Dundee and St Mirren; a 3-2 League Cup win against Hearts that should have been 10-1; four goals away in Europe at FK Jablonec.

There were cold-turkey downers: defeat in Denmark; three away losses in the league including to arch-pragmatists Rangers; eight goals shipped against Betis and Leverkusen; a seat-of-the-pants excursion to Alkmaar.

This was just how it was going to be. Binge, fast, sleep. Repeat.

The nadir of defeat at Livingston on September 19 saw Celtic “toiling” (so said the BBC match report) in sixth position and four points off leaders Rangers. That gap would grow to six points. The report called the Hoops “beleaguered”.

The 4-0 reverse to Bayer Leverkusen 11 days later saw a thrilling football match but the nagging doubt that Celtic were playing a naïve style, in stark contrast to Steven Gerrard’s rigid but efficient Rangers.

The data hinted that the underlying performances were an improvement on the disastrous 2019-20 campaign. There was hope – but the results were suffering due to adverse variance.

Variance is spoken about a lot in analytical chat. What this means is, for example, injuries, opposition goalkeepers playing a blinder, underperforming your expected goals by missing good chances consistently or the opposition are scoring 30-yard worldies.

Dark mutterings came from those that track results but not performances. “He has no Plan B,” they said. “He has to find a way to win”. Those were common refrains. For some, the knives were being sharpened.

Fast-forward 31 unbeaten league games. As we sit here in May with the title secured following Celtic’s 1-1 draw at Tannadice, it’s clear the ‘adrenaline junkies’ did not burn out by Christmas as many sage observers predicted but instead found a way to win. The title has been certain for many weeks. It has been almost serene.

Thirty-one league games unbeaten with only six draws. 70 goals have been scored and 18 conceded over that run. Surely that’s the very definition of (looks up thesaurus): consistency.

So what happened, mate?

Ange knows his xG

Let me bring you down to earth with reference to our staple analytical measure: expected goals.

Under Neil Lennon last season, the difference between expected goals for and expected goals against measured on a six-game rolling basis, was often between 0 and +1 with the occasional run of games where it nearly got to +2.

What does this mean? And why is it important?

Expected goals measures are more predictive than real goals (I know, sit down and have a cup of tea). Over the long term, if you are generating and conceding around the same level of xG at each end, then the likelihood is you will have mixed results. Some wins, some draws, some defeats.

In Lennon’s case, the xG differential was close enough (usually less than +1 xG per game difference between xG for and against) that the normal variances of football could often impact the result. The keeper saves eight shots, your striker hits the post when a goal is more likely, that controversial refereeing call… Simply, Lennon’s team did not generate enough quality chances versus those conceded to remove many of the variances at play.

Back to the expected ‘Ange-ball’ rollercoaster. How has his Celtic team performed?

As you can see, there was significant variability around September and October. This period was also the only time the six-match rolling average xG differential (great band name) fell below +1.5. Remember the personnel challenges and injuries at this time? More on this later.

Focus on the orange trend line. You will see that the trend for xG differential has slowly improved over the season but always been just either side of ‘+2’. This means Celtic have been generating 2 or more xG for more than the xG the opposition are generating from their chances.

Very simply, this is how you generate consistency.

If you are regularly posting two or more xG more than your opponent, there is significant margin for variance to be absorbed. Meaning you can suffer that bad penalty call, score a freak own goal or concede from 40 yards and still be generating enough chances the other end to not suffer.

The question of ‘how’ Postecoglou achieved this is an article for another day involving discussion on tactical system consistency, recruitment and organisational culture amongst other things. But ‘what’ he has achieved is something that could not have been foreseen when ‘Angeball’ first came into view.

Let’s compare with Celtic’s closest rivals, Rangers, and their own six-match rolling xG differential.

This has been trending downwards across the SPFL season and around an average of under +1 xG difference.

There is also greater variability since Giovanni van Bronckhorst took charge. What is a strength in Europe – adapting to the opponent, tactical flexibility depending on game-state – has not been a blessing in the league. A team top-heavy in athletic, workmanlike midfielders, suffering injuries to key attackers and whose most potent attacking weapon is a penalty-scoring right-back, has struggled to break down low blocks in the SPFL on key occasions.

There is insufficient of an xG differential being generated to a mitigate for goalkeeper Allan McGregor dropping in two goals at Ross County, for instance, or the side scoring one from 24 shots at home to Motherwell.

Celtic are champions because they have consistently outperformed their opponents in creating and stopping chances to a degree that has largely removed the impact ‘luck’ plays in a low-scoring sport like football.

Line-up volatility

What makes Postecoglou’s achievement more admirable is that consistency of result has been achieved despite inconsistency of selection.

This graph shows the average number of chances to line up per match and the total number of players deployed within SPFL matches.

Celtic have used 36 players in the SPFL and average 2.5 line-up changes per match. You can see those with high volatility like Hibernian and St Johnstone have had very poor seasons. Dundee United are an outlier and it is remarkable, given the line-up instability they have had, that European qualification has been achieved.

Those in the bottom left segment have seen the lowest levels of line up volatility. Ross County, Hearts and Motherwell have all had very successful seasons by their expectations while St Mirren are probably performing about as expected.

Rangers’ line-up changes are also high, reflecting the latter part of the season when European adventures midweek has led to greater rotation of resources.

Conclusion

It started off with hope but also with a realisation that, such was the size of the rebuild and the relative stability of the major opponent, there would be progress without expectation of triumph.

Early-season indicators bore this out. There were great performances and solid underlying metrics but patchy outcomes and results.

Postecoglou’s playing style and coherent recruitment then generated sufficient xG differential on a consistent basis, meaning Celtic have been able to win the league with unforeseen ease. That consistency of performance outcome is surprising.

Wondering what can be achieved with a more stable playing squad and decreased line-up volatility makes this a summer to savour.