Neil Lennon's role in Celtic's failures: Three key decisions that show former manager must shoulder responsibility - James Dailey

While there has to be a degree of empathy for all who were involved - as they were asked to perform under the challenging circumstances of serial lockdowns and isolation - Lennon's apparent insinuations that he played little to no active role in the team's fate has left many fans incredulous.

Last season there were clear risk factors with regards to some of Lennon's decision-making and it seemed equally obvious that failure to qualify for the Champions League once again could trigger an exodus of important players.

Over a year later there is now extensive performance data with which to conduct a managerial decision-making autopsy.

To perform this procedure, three active decisions which Lennon took will be highlighted. These have been chosen to go under the knife as they appear to have had a material impact on Celtic's 2020-21 failures.

1. Scott Brown as first-choice

First, and arguably the most important, was beginning the season with Scott Brown as a first-choice starter in defensive midfield. The confluence of the then-captain's style of play and his normal ageing curve made this decision a predictably poor one. 

Brown had already begun to show, as is normal for players in his position and age, a material decline in performance metrics over the preceding seasons.

His increasing vulnerabilities were amplified by Lennon’s decision to use Brown within a system that relied upon intense pressing. Here were various team-level defensive metrics since the 2018-2019 season, which correlated with the higher average defensive and pressing distances for Brown displayed above.

From these figures, it is clear last season's side did not stand out as lacking in pressing intensity. In fact, in some ways, it was more intense than the preceding 2019-2020 season with lower passes per defensive action (PPDA) and a higher defensive distance, which was the average distance from Celtic’s goal-line in which a defensive action took place. 

Those metrics offer evidence that the team was actively pressing and doing so further up the pitch. Here was a defensive radar comparing Ismaila Soro with Brown last season.

Generally speaking, they are not significantly different. But there are a couple of vital outliers which may relate directly to Brown’s predictable age-based physical decline.

Brown was in the 14th percentile in getting dribbled past and the 31st percentile in the rate at which opposing players were successful at beating him via dribble or in a tackle. Those were at an SPFL Premiership level for a midfielder who was asked to play an aggressive pressing role in a system that was inherently vulnerable to defensive transitions.

READ MORE: The three reasons Celtic are leaking goals and why it's not the fault of Ange Postecoglou's system - Ross Goodwin

Is it a surprise that a 22-year-old may have an athletic advantage relative to a 35-year-old? Celtic’s xG conceded was significantly lower in games in which Soro started: 0.60 vs 0.80. Soro’s sample included the two post-Dubai games where he played in patchwork sides. Excluding those two games, the average was just 0.47.

2. Diego Laxalt over Greg Taylor

Next up is Lennon's decision to select Uruguay international Diego Laxalt, who joined on loan from AC Milan, ahead of Greg Taylor during what was the most material drop in performance levels last term.

This run stretched from the first Glasgow derby in early October through to early winter. Here were Laxalt and Taylor's attacking and defending radars in league games last season:

The 4-2-3-1 system Lennon deployed over that stretch required overlapping and attacking full-backs capable of creating chances. From an attacking perspective, Laxalt was below average at SPFL level for xG Assisted from open play, and was in the 17th percentile in crossing accuracy. Taylor, in contrast, was one of the highest in the league.

While Laxalt was more intense in his defensive actions and pressing, he was not as effective as Taylor Like Brown, he was quite poor at defending against dribblers. Given Celtic’s left-sided bias, Laxalt’s poor creative output and struggles against dribblers compounded some of the side's midfield issues in dealing with defensive transitions and often resulted in leaving centre-backs - such as Shane Duffy - exposed while playing a high line.

3. Indecision between the sticks

Last to be dissected is Lennon's carousel selection decisions when it came to goalkeepers.

Celtic developed a severe issue defending set-plays and crossed balls as the season progressed, and those issues coincided with Vasilis Barkas’ removal as the starting keeper. Here was how it broke down last season:

The numbers show Barkas’ relatively poor shot-stopping manifested in more goals being scored than what the xG value suggested should have occurred. However, the team still conceded fewer goals overall in those phases of play when he started. So what was the missing link?

It seems Barkas was far better at commanding his box against balls which were available to be claimed. The graphic below compares all keepers who played 900 or more minutes in the league by how aggressive their average distance was from their goal-line when they made a defensive action (Y axis) and by how successful they were at claiming balls (X axis).

Here we can see that Scott Bain was similar in his aggressive positioning, but was 15th of 17 in his effectiveness at claiming balls.

This speaks to the argument that goalkeeper play, and related performance value, is far more nuanced than simply shot-stopping. Celtic conceded just 0.45 goals with Barkas playing in league games last season, which compared to 0.85 with Bain, even though Barkas was worse when it came to shot-stopping metrics.

Conclusions

Starting the season with Brown; selecting Laxalt over Taylor; dropping Barkas. All of these decisions had materially negative measurable impacts on Celtic's season.

Barkas, Soro and Taylor combined to start together in just one out of 38 league matches. That was in the 2-2 draw at Livingston in the first game following the return of players from the Dubai-related isolation.

Despite the extenuating circumstances, including Mohamed Elyounoussi and Leigh Griffiths selected to start together at striker for the first and only time, and a portion of the game being played on a snowy pitch, the xG was 2.10 vs 0.34.

That remains the best performance by xG at Livingston since their return to the Premiership, and not by a small margin. The 1.76 in xG differential compares to the next highest of 0.93.

This takes us all the way back to the start and the influence of Lennon on last season. If starting Barkas, Taylor and Soro together in just one of 38 games was not the manager’s responsibility, then whose was it?