Season Review: Centre-backs (Part 1)
Written by: Christian Wulff (@ahellofabeating)
Statistical data will never give you the full picture about a player or a team. It can only give a certain amount of information that then needs to be put into a wider context. This is true for all players but there is also a hierarchy within football statistics in terms of positions on the pitch and what we can reasonably infer from the available data.
As a rule, the more attack-orientated the player, the ‘easier’ they are to analyse. Attackers are ultimately judged upon the goals they score and the chances they create. There are a wide range of different metrics to cover this. Central and defensive midfielders can be a bit more tricky, but with advanced stats showing whether they had the second or third last pass before a chance was created, we can now more easily get information about their offensive contribution. The same very much goes for full-backs, who are now seen as much as attackers as defenders.
If we go all the way back on the pitch, it becomes easier again: the smaller number of set actions expected of a goalkeeper (stopping shots, dealing with crosses, distribution) have allowed us to better quantify their performances using stats, especially when it comes to shot-stopping.
That leaves us with what is still regarded as the biggest conundrum when in football statistics: the centre-back. While attacking data is very much focused on individual players (who took the shot, who set up the chance), the defensive data is overwhelmingly team based; it tells us many goals and chances a team has conceded but less about the individual players’ contribution to this.
There are individual stats available for centre-backs (which we will cover in this analysis) but they should be treated with some more caution than those for attacking players.
For example, when it comes to strikers we can mostly assume that more shots equals a positive contribution, but we can’t assume the same for how many tackles a centre-back have attempted. They might indicate bad positioning or being too rash, rather than being a positive defensive action.
These individual stats can also be heavily influence by a team’s quality and style of play; centre-backs in a team that has a high percentage of possession and press high up the field will likely challenge for a lot fewer headers than those centre-backs who play for a more counter-attacking team, who sit deep and are more likely to cede possession to their opponent.
So in analysing centre-back performances through stats, it is perhaps prudent to think about it just as much as an analysis of the type of centre-back a player is, and not solely as an exercise of ranking them per quality.
In this article we will look at centre-back performances in the 2017/18 SPFL Premiership season using three different types of statistics:
Team defensive performance
Player influence on defensive team performance.
Individual player actions.
Let’s start with defensive performances of the Premiership teams this season. The preferred method of doing this is not goals conceded but rather chances conceded. Whether a goal is scored or not is heavily influenced by the quality of the strike, a goalkeeper’s action and random variance. Hence, analysing chances instead of goals conceded is likely to give a better indication of the quality of the defensive performance.
The quality of a chance is measured through its Expected Goal (xG) value. In simple terms, xG show how many goals have historically been scored from different type of chances. Based on this average, each chance is given an xG value between 0 and 1. An xG value of 0.10 means similar chances have been converted 10% of the time previously, a value of 0.50 means it has been scored half of the time.
For defensive performances, this method of measuring the quality of chances conceded lets us determine how sustainable periods of conceding few goals are: if a team is conceding a lot more xG than actual goals, such streaks are very unlikely to continue for long.
This graph shows goals vs xG conceded per game for every team in the Premiership this season (excluding penalties)
Celtic and Aberdeen have both conceded the least amount goals and xG. While Hearts have the third least goals conceded, they are tied for 8th when it comes to xG against.
Only Dundee have ‘over-performed’ more than Hearts when it comes to xG against vs goals conceded (i.e. the quality of the chances they have conceded would on average result in a far higher amount of goals against).
Motherwell have the third lowest xG against and they are together with Ross County the highest ‘under-performer’ (i.e. the quality of their chances they have conceded should on average yield fewer goals against.)
With a baseline for the overall defensive record of each team, we can attempt to isolate the most relevant data in judging the centre-backs performances.
A few considerations come into play:
A centre-back’s main defensive aim is to prevent the opposition taking a shot. Blocking the shot is the next best thing and is likely to indicate a positive action from the defence. So we remove any attempts that were blocked from the analysis.
Penalties and direct-free kicks are also excluded. While centre-backs might not be ‘guilt-free’ when it comes to conceding such chances, there’s likely to be a significant amount of other factors in play.
Excluding set-pieces* is a tougher decision, but again there is plenty of other factors involved outwith a centre-back’s direct influence, such as very crowded penalty areas with a very different positioning and awareness demands than from open play.
The final consideration is the area of the pitch. We had two obvious options; the penalty box or the ‘danger zone’ (the area between each edge of the 6-yard-box, from the goal line up to the top of the penalty box).
For this analysis, I felt neither was ideal; the penalty box would include areas where centre-backs have limited influence, and danger zone would exclude areas where they reasonably should be expected to cover.
The solution was a hybrid which I’ve titled the Centre-Back Area. It terms of the width of the pitch it starts and end at the point halfway between the six-yard box and the edge of the box, and goes three yards beyond the penalty area lengthwise.
So for the purpose of reviewing the league’s centre-backs we have narrowed the data down to:
Non-blocked attempts from open play within the centre-back area.
The graph below show the xG conceded from these non-blocked, open play attempts from inside the area, together with the average xG value of each attempt conceded.
Eyes are immediately drawn to Motherwell; not only are they tied with Aberdeen for the second lowest xG against, the average quality of the chances they concede is considerably lower than any other team. It’s also worth noting that while Celtic have conceded the least amount of xG, the average attempt they allowed in open play from this area is the highest in the league.
These numbers are a good reminder of how overall team style and performance are likely to impact on centre-backs’ stats and vice-versa. Conceding the least amount of xG will be partly due to the qualities of Celtic’s centre-backs, but a likely big factor is the fact that they are simply the best team in the league, one that controls possession and dominates other teams offensively.
Similarly, while they are the team in the league that concedes the biggest open play chances in this area, that should not automatically reflect badly on their centre-backs. How much of this is simply a consequence of the way Celtic is set up? Playing a high line and an emphasis on building play up from the back is fertile ground for counterattacks, with centre-backs often exposed.
The opposite can be applied to Motherwell; their more direct style and lower press help restrict space behind the centre-backs, making them are lot less likely to concede big chances on the counter attack. However, the stark difference between Motherwell and all other teams when it comes to the average quality of these chances is definitely very interesting.
The other outliner is Dundee; they’ve not only conceded the most xG, but the average quality of each chance they concede is second in the league.
Time to drill down further and look at how the data changes when adjusting for when the individual centre-backs have been on the pitch. For this review, I’ve set two criteria for a centre-back to be included; at least 1000 minutes played in a centre-back position, with at least 70% of total playing time being at centre-back*, giving us a total of 40 centre-backs.
We use the same criteria of non-blocked, open play chances conceded in the centre-back area to see how the teams performed when each of the players were on the field, playing in that position.
Another consideration to make at this point is the quality of opposition faced by each player. Both within and between teams, a centre-back’s stats can be influenced by the quality of opposition faced. To adjust for this, I’ve calculated the average quality of the teams each centre-back has faced, based on the Club Elo Ratings at the end of this season.
Since they were the three most significant outliners in the team ratings, I’ve colour coded the centre-backs of Celtic, Motherwell and Dundee throughout the analysis (Jack Hendry’s combined numbers for Celtic and Dundee get a lovely cyan colour, but his individual contributions to both teams is also colour-coded on the graph). Hearts' centre-backs are also specifically marked (in pink); with a defensive record that was generally lauded throughout the season but whose underlying stats triggered warning signs. They’re worth following closely through the analysis.
The graph below shows the team xG conceded from non-blocked attempts in open play, within the centre-back area, per every 90 minutes each player was on the pitch this season:
Kristoffer Ajer comes out on top at the ranking, having the lowest xG conceded in the league this season. He has also faced, on average, a tougher opponent than all the other Celtic centre-backs (note Nir Bitton on the far left – he’s faced the weakest opponents of all the centre-backs in the analysis).
Josh Meekings had the highest xG conceded - by quite a distance - followed by Jack Hendry during his time at the same club. Both Meekings and Hendry faced a tougher opponent on average then their teammate Darren O’Dea, whose numbers are also pretty bad. Steven Caulker is the last blue point on the graph, and there was definitely an improvement in Dundee’s ability to restrict chances in open play in this area when Caulker was on the pitch. The Motherwell centre-backs’ contribution to the team’s overall defensive record can clearly be seen within this metric, all of them in the top third when it comes to restricting chance. There’s quite a big range in quality of opponent faced, with Tom Aldred’s having the weakest numbers of the group with the other three very even, especially adjusted for the opposition faced.
It’s also worth noting the Aberdeen centre-backs: there is a big difference between the chance quality conceded when Kari Arnason was on the pitch, compared to Anthony O’Connor. Some of that is likely to be down to quality of opponent, but it’s a significant variance. The mainstay in the team, Scott McKenna, is logically in the middle of the two.
Another notable player is David Bates – his xG conceded is significantly lower than two of his Rangers teammates, Bruno Alves and Danny Wilson, and also better than Russell Martin, while the opponents Bates faced were the toughest of any of his (now former) teammates.
The Hearts centre-backs were average at best; Christoph Berra ranks 23rd, John Souttar 28th and Aaron Hughes 36th out of the 40 centre-backs. Hughes especially has faced tough opposition in his games, but these are not great numbers for any of the three.
I’ve also noted Harry Souttar’s on the graph; he’s 18th in the league having played for the league’s bottom team. More on the younger Souttar brother later. Unfortunately for him, we’ll also get back to Alexandros Gogic…
While stopping the opposition from getting to shots and headers in the centre-back area is a key metric in this analysis, it’s also a critical task to prevent your opponents from creating chances in the same area
Our data from Stratagem includes the location of each pass that set up a chance in the Premiership this season (called a key pass). The quality of chances set up from such passes are also measured in xG, most commonly referred to as Expected Assists, or xA
Fellow contributors to Modern Fitba, Jason from The Rangers Report and Matt Rhein from The BackPass Rule have both done great work analysing the influence of where a chance is set up. Both found that the more central the area the chance was created from, the higher the scoring chance usually is, underlining the importance for centre-backs to stop such key passes in this area.
The graph below show the xA value of the open play chances created with a key pass from within the centre-back area (the actual attempt that followed might have been taken from outside the area and could also have been blocked), together with quality of the opponent faced.
The graph is quite similar to the previous one; the amount of chances a team and player concede from this area seem to align well with the amount of chances are actually created from within the same area.
There are a few notable changes; Anthony O’Connor is now level with Scott McKenna and David Bates is now the Ranger centre-back with the worst numbers, roles reversed with Bruno Alves and Danny Wilson compared to the previous metric.
Josh Meekings can console himself with the fact he’s not bottom of the league in allowing chance creation, as that ignominy goes to Danny Devine, who has also faced a slightly weaker opponent than the league average. It’s not pretty, to say the least. Neither are Darren O’Dea’s numbers, who is second worst in the league, followed by Steven Caulker. While Caulker’s presence seems to have improved Dundee’s ability to restrict chances, Jack Hendry’s departure is felt in the ability to stop chances being set up from the same area.
Hendry’s (almost) six full games at centre-back for Celtic is really too small a sample to be drawing significant conclusions from (they are only included here as a counter-reference to his Dundee numbers). This is seen in him being the player with the best numbers on chance creation allowed, but 29th on chances conceded in these games – smaller samples often lead to such swing in numbers. However, Hendry’s numbers for Dundee (on a much larger sample) does reflect the same pattern; his team is better at suppressing the pass that set up a chance rather than the chances themselves, when he’s been playing centre-back.
There is also another notable point on the Celtic centre-backs. They were evenly clustered on chances conceded and are mostly the same here, Nir Botton and Dedryck Boyata edging slightly ahead of Kristoffer Ajer. However, when Jozo Simunovic have been on the pitch, Celtic numbers are actually worse than the league average for allowing open play chance creation in this area. The quality of opposition Simunovic has faced is very similar to Ajer and Boyata, and they both have far better numbers than the Croat.
The pattern between the Motherwell centre-backs is very similar (again, Tom Aldred has the worst numbers), and they actually perform even better as a group compared to chance quality conceded.
On both these metrics, the Motherwell players have outperformed the centre-backs from teams such as Rangers, Aberdeen and Hearts. Their numbers might well have been helped by their team’s more risk-free attacking approach, but there is no denying the centre-backs influence on team’s performance.
The Hearts’ centre-backs continue to underwhelm, now dropping even further down the rankings with Berra highest ranked on 29th out of the 40 centre-backs. Harry Souttar is 11 places above brother John , who is way down in 32nd.
So what tentative conclusions can we draw looking at these team metrics and how the influence of individual players altered them?
As expected, the Celtic players are towards the top in each category, with Kristoffer Ajer edging the others slightly based on chances conceded and having faced on average the toughest opposition.
Motherwell’s overall defensive record is reflected in the fact that their centre-backs performed well, while similarly Dundee’s centre-backs struggled in the ratings, Darren O’Dea probably the worst in a bad bunch.
Heart’s no-more-than-average defensive record is reflected in their centre-back’s performances in this area; they gave up more chances and let the opposition create more than over half of all the other centre-backs in the league. In a future piece for Modern Fitba I’ll look at the shot-stopping of Premiership goalkeepers this season, and that will show that Jon McLaughlin had a lot to do with the difference between the chances Hearts allowed and the actual goals conceded.
For Rangers, there is no clear pattern. They conceded the least chances in open play from the centre-back area when David Bates was on the pitch, but the largest amount of chances was created from there at the same time. The opposite was true with Bruno Alves and Danny Wilson, while Russell Martin was in the middle in both metrics.
In part two of this season review, we’ll be focusing on the centre-backs’ individual actions on the pitch, with one very unique player appearing…