The Lennon Sample

For those of us who write about and analyse football using advanced statistics one of the toughest challenges is making the case for looking beyond the result of a match, past how many goals have been scored and conceded, especially over a small sample of games. It is a tough sell and something counterintuitive to most football fans. Results and goals are all that really matter, right?

The main reason we issue caution about making sweeping conclusions based on results alone is that the rate of goal scoring is very unpredictable. The results over a period of, say 7 games, are in itself not a very reliable indicator of what will happen over the following 7 games or the rest of the season.

That’s why we dive in deeper; looking at the quantity of chances created and conceded (shots) and - even better - adding the quality of those chances, as measured through Expected Goals (xG). It’s statistically proven that such metrics are a better indicator of future results, than actual results themselves.

What we want to analyse is: are these current results sustainable or not?

If a team is winning games but not producing a lot of good quality chances (and conceding a fair few), that winning streak is highly likely to end soon, unless the underlying performances improve.

The same applies to a team on a bad run result wise. If the chances are there, take a deep breath, keep the faith and the results will highly likely come. If the amount of chances created and conceded are as bad as the actual results, then yeah, feel free to press that panic bottom.

Which bring us to Neil Lennon.  Yet again, he’s been one of the major stories of a Premiership season after his tumultuous exit at Hibs and then an unexpected return to Celtic Park. The current narrative seem to be that Hibs have improved and excelled under new manager Paul Heckingbottom, while Celtic under Lennon is only doing just enough to win games without convincing, relying on very late winners in between frustrating stalemates against Aberdeen and Livingston at home.

A cursory glance at the points won and goal difference under and without Lennon at the helm at both clubs seem to support this.

(Quick note:  as usual, we only look at non-penalty goals, also excluding own goals, when we analyse – these are more likely to be linked to the ability to create and concede chances, and hence the underlying performances in each game)

Hibs Goals.PNG

In the 22 league games up until he was suspended by the club on the 26th January, Hibs took 29 points, 1.3 per game, under Lennon. With Heckingbottom in charge they are undefeated in the league, gaining 20 points from those eight games, an average of 2.5 points a game.

The goals tell a similar story of a remarkable turnaround: a goal difference of 27-21 under Lennon (1.2 – 1.1 per game) compared to 12-5 (1.5 – 0.8 per game) under Heckingbottom.

While the difference is not quite as stark between Rodgers and Lennon at Celtic, the story is similar. Up until he left, Celtic had taken 2.3 points per game under Rodgers in the league, scoring 2.2 goals and conceded 0.4 goals on average.

Under Lennon Celtic are taking 0.3 points less per game so far and while they are actually conceding less (0.2), the goal-scoring rate has almost halved to 1.2 goal per league game. Again, the results and goals seems to support the narrative; Celtic grinding our results by being compact defensively but failing to produce offensively. Hardly a ringing endorsement for making Neil Lennon the permanent manager.

Celtic Goals.PNG

So with results and goals so obviously in Neil Lennon’s disfavour, it must be an open and shut case that both Celtic and Hibs have performed worse under his management this season?

Not quite….

Let’s start with Hibs under Paul Heckingbottom. They are on a great run in terms of results, the first seven matches yielding 5 wins against bottom six opposition and two draws against Rangers and Kilmarnock. It was then topped off  with Hibs’ first derby win at Tynecastle in six years last weekend. The fanbase will be delighted and a warning from a nerd with a spreadsheet that, actually,  this isn’t as rosy as it might seem, is surely likely to produce a response that could land you in trouble with the SFA compliance officer.

But a warning is what the Hibs fans are getting.

Hibs xG.PNG

While the results and goals are very encouraging, the underlying numbers under Heckingbottom are not of the same standard.  Hibs’ total expected goals under Heckingbottom is 12 which equates to 1.5 xG, equal with Lennon’s average this season. The xG against is 11.7 (also 1.5 per game) compared to 1.1 xG per game under Lennon.

What does that mean in plain English? The quantity and quality of chances created under Heckingbottom is exactly the same as with Neil Lennon in charge, but they have conceded more chances under their new manager. Simply put, Hibs underlying performances under Paul Heckingbottom are worse than under Neil Lennon.

Hibs’ great results under Heckingbottom have come from their opponents failing to take their chances. In these eight league games Hibs have conceded only 5 non-penalty goals but have given up chances totalling 11.7 expected goals. Non-penalty goals scored (12) perfectly matches the 12 expected goals from their chances created, with Hibs getting another 3 goals from penalties.

Should Hibs fans care? Well, they should be aware that these set of results are unsuitable if their team under Heckingbottom continue to concede chances at this current rate. Over a full season, their current expected goals difference per game (0.3) would likely not achieve any better results than what was the case under Neil Lennon.

This graph shows the relation between the quality (measured in xG) of chances created and conceded in each league game this season. If the graph goes above the line it means Hibs created chances of higher quality than opponent, if it dips below it means the opposition had the advantage

This graph shows the relation between the quality (measured in xG) of chances created and conceded in each league game this season. If the graph goes above the line it means Hibs created chances of higher quality than opponent, if it dips below it means the opposition had the advantage

But these results are a great platform for Heckingbottom to improve from. Good results inevitably breeds confidence within a squad and the belief and commitment to the manager’s methods are likely to cement and improve. With even more time to implement his ideas, Heckingbottom might well be successful in improving the team’s underlying performances, making his great start more sustainable. But the warning is clear: if Hibs don’t improve quickly, this rate of results will dry up just as fast.

It does put Neil Lennon’s tenure at Hibs this season in a slightly different light, although it’s important to note that the performances under him were almost equal to the results: a goal difference of 6 compared to an Expected Goals difference of 6.8. Hibs got pretty much ‘what they deserved’ point wise based on their performances under Lennon.

An inability to produce goals has been the main worry for Celtic fans with Lennon in charge, the team scoring at just only half the rate they did under Brendan Rodgers this season.

 Is this reflected in the actual chances created? Absolutely not.

Celtic xG.PNG

Lennon’s Celtic have created chances worth 2.6 expected goals per game, compared to 2.3 under Rodgers. That’s correct; Celtic’s chance creation has actually improved under Lennon, even though the goals have dried up.

Such goal-scoring droughts happening at the same time as a team are producing a lot of chances is not uncommon. In fact, Celtic had a similar spell at the start of this season, when they only scored 6 non-penalty goals in their first 7 games, even if their expected goals created was 13.3.

In the next four games? 18 goals from chances worth 13.5 expected goals.

Combined, Celtic scored 25 non-penalty goals from chances worth 26.8 expected goals over these first 11 league game. It’s a classic example of what we like to label ‘regression to the mean’: a team or player might go on a hot/dry streak in terms of goal scoring over a shorter sample of games, but sooner or later they will return to a level that is close to their expected goals total. 

This is exactly why expected goals are a much better predictor of future performances than goals and results: in the vast majority of cases the scoring rate will eventually align with the expected goal rate.

I fully expect the same to happen to this Celtic team; if they keep creating this amount of chances under Lennon an increase in goals will come very soon.

Celtic have also conceded chances worth slightly less expected goals under Lennon (0.6 per game) compared to Rodgers (0.7). So both when it comes to chances created and chances conceded, Lennon’s 6 league games in charge have seen a slight improvement in the underlying performances compared to the league season under Rodgers.

This graph shows the quality (measured in xG) of chances created and conceded in each league game this season.

This graph shows the quality (measured in xG) of chances created and conceded in each league game this season.

Both the sample of games under Lennon at Celtic and Heckingbottom at Hibs is on the small side, especially for anyone wanting to make sweeping conclusions from them. But as mentioned, expected goals have proven to be a much better indicator of future results over such samples than the actual results and goals achieved within in. So if any conclusion can be made at this point, it is that the underlying performances do not match up to the results for either of the new managers and if the level of chances created and conceded continues at this current rate, results will soon align to them, meaning improvement for Celtic and a decline for Hibs.

Whether Neil Lennon should become the next permanent Celtic manager should depend on a lot more than just the results or expected goals performance under him this spring. Stats like these are one puzzle of a much bigger piece the Celtic hierarchy need to put together to come to their decision. But for anyone pointing to Hibernian’s results of improvement under Heckingbottom or the lack of goals for Celtic in recent games as points against Lennon, they are putting forward an argument that do not hold up to statistical scrutiny.