The Best Defence or Easiest Job?
The Best Defence or Easiest Job?
Best defence or easiest job?
It is a widely held belief that the best striker is the one who scores the most goals. This may be the case, but it is not a given. The same thinking has been applied to the best defence, namely the one which concedes the fewest goals. This article seeks to analyse the factors that help determine the relative strength of any given defence and thus allow for better, i.e., fairer, comparison between teams.
This article is broken into four sections:
- Traditional measures of defence
- Moving towards standardisation
- Attack is the best form of defence
- Best and worse defences in Europe
1.Traditional measures
Who gets the credit?
Below are some typical arguments found on the internet (including mainstream media and football writers/pundits):
The pronouncement from Squawka Football seems to credit the goalkeeper for keeping the clean sheets. Jan Oblak may be the best goalkeeper statistically, but this metric is barely worth the breath spent on reading it. The second from Castle Of The Kop seems to credit two specific defenders for the clean sheets. Notwithstanding the ridiculous sample size and bias of the latter, both ignore individual contributions made towards these statistics. Here lies a big problem when using clean sheets to quantify a goalkeeper/defender’s performance: who should get the credit? And more importantly, how much credit?
To illustrate this point, let’s ramp up some extreme scenarios. Oblak may never have had to save a single shot in those 100 clean sheet games or he may have saved 100s of shots he was not expected to save. Realistically, it will be somewhere between these two extremes, but where exactly? For the two defenders, Van Dijk and Gomez, they may have let their goalkeeper face 100s of shots in those five games, but it just happened that the goalkeeper saved every single shot against him or that the pair of them stopped every single attack. Again, realistically, it will be somewhere between these two extremes, but where exactly?
Furthermore, most football fans would acknowledge that a team’s defence is a combination of the whole team with the goalkeeper and defenders having most responsibility. When making comparisons such factors need to be considered. To do this, metric distortion needs to be removed. To keep it simple, the graphics below illustrate two teams, A and B, from the same league who both conceded 25 goals. Thus, suggesting that both defences or goalkeepers or both performed equally. This may be true, but hugely unlikely.
The numbers are the percentage the individual units (GK, defence, midfield and attack) contributed in restricting the opposition to scoring goals. Whilst it is ultimately the goal (no pun intended) for football metrics, currently, it is not known what the actual contribution is per unit (and individual) towards team metrics. Nevertheless, it must be acknowledged that variation within teams exists. For example, Team A’s defence contributed 42.5% in preventing the opposition from scoring, whereas Team B’s defence contributed only 30% in preventing the opposition from scoring. If this information was available, then when making comparisons – based on goals conceded – Team A’s defence deserves more credit for the low amount of goals conceded. However, metric distortion hides this.
Similarly, when comparing goalkeepers, Team B’s goalkeeper contributed 35% in preventing the opposition from scoring. Thus, when comparing to the goalkeeper for Team A, Team B’s goalkeeper deserves more credit for the low amount of goals conceded. Again, metric distortion hides this. A further problem for defence comparisons is that a defence consists of multiple players: sometimes three, four or five defenders. Again, did they all contribute equally? Possibly, but very unlikely. Player contributions vary within defences. More worryingly, is that these distortions (i.e., inaccurate measurements), can happen in either direction, exacerbating the problem.
This article’s aim is to somewhat improve the metrics used to compare defences. It will solely focus on the defending aspect and not account for the attacking contribution that defences do make.
2. Moving Towards Standardisation and Fairer Comparisons
Expected Goals Against
Team – as well as individual – performances are not necessarily reflected in the final match score, far from it in fact. This is true for all team sports, but especially in a low-scoring game such as football. Expected Goals Against (xGA) is a statistical measure of the quality of chances conceded. It removes the variability of the opponent’s finishing, i.e., whether they score or not from that chance, thus measuring the opponent’s chances created, regardless of the outcome of that chance. This metric, xGA, will be used to help compare defences by evaluating the quality of shots they allow their opponent to take. The xGA data in this article is taken from understat.
To further illustrate the importance of using xGA when comparing defences, imagine two different defences, A and B, that allow the opposition the same chances during a match, yet Team A loses 2-0 and the Team B draws 0-0. Going further, let’s say that it was against the same team and the same chances fell to the same opposition strikers in the same order and timings of match. Now, using Goals Against (GA) to evaluate that Team A’s defence is worse is wholly misleading, and more importantly factually incorrect. It did the same job/performance of Team B. The only difference is that the opposition’s striker/goal shooter performed differently. It is important to note, there will always be variation within an individual’s performance, this is to be expected.
Premier League Expected Goals Against
It is clear from the graph below that there is a considerable (>%5) difference between GA and xGA – supposedly measuring the same thing. Only two teams in the Premier League in 2019/20 had a difference of less than 5%, Norwich (4.5%) and Bournemouth (2.5%). Not only are the majority of teams’ xGA considerably different from GA, but they can differ in either direction. For example, Brighton conceded 6.4 (-11.9%) fewer goals than expected whereas Everton conceded 6.8 (12.1%) more goals than expected given the opportunities they allowed their opponents.
If Brighton and Everton’s defence were compared using GA, then Everton conceded two more goals (a 3.7% increase compared to Brighton). However, if a much fairer comparison was made using xGA, then it was Brighton who were expected to concede 11.2 more goals (a 22.8% increase from Everton). This illustrates the huge metric distortion caused by using GA, and yet so many more comparisons between defences below would contain even greater distortion. The point is: goals conceded should never be used to make comparisons between defences (and keepers).
During the 2019/20 season, a lot of media exposure highlighted Chelsea’s poor defence. This was mainly borne out by the fact they had conceded many more goals than their nearest rivals. Chelsea conceded 12.1 more goals than expected given the opportunities made by their opponents, meaning the strikers/opposition shot better than normal from the same positions whilst playing Chelsea. Therefore, the criticism towards Chelsea’s ‘poor’ defence was certainly not proportionate to their actual performance. Conversely, the team that benefitted the most from the opposition having poor days in front of goal was Sheffield United. They conceded 13 fewer goals than expected.
Overall in the Premier League, seven teams conceded more goals than expected. Five of these ended in the bottom half of the table. Furthermore, 50.1 more goals should have been scored across the league over the season given the opportunities presented to the attacking team, thus shooting was below average (assuming an average goalkeeping performance).
LaLiga Expected Goals Against
A number of teams in LaLiga conceded considerably fewer goals than expected: Real Madrid 8.2 fewer goals (-32.6%), Athletic Club 14.4 fewer goals (-37.9%), Levante 11.3 fewer goals (-21.2%) and Real Valladolid 10.9 fewer goals (-25.4%). All these teams benefited from the opposition squandering chances to score against them. The bottom seven teams, albeit only a little, all conceded more goals than expected given the opportunities their opponents had. Across LaLiga over the season, 30.7 more goals should have been scored given the chances presented to the attacking teams, thus the shooting was below average.
Serie A Expected Goals Against
The biggest winners were Lazio. They conceded 42 goals (2nd) but given the chances presented to their opponents, they should have conceded 10.3 more goals (xGA 52.3, 8th). Five of the bottom-placed six teams conceded more goals than expected. Across Serie A over the season, 17.5 less goals should have been conceded given the chances presented to the attacking teams, thus the shooting was above average.
More interestingly, Serie A had the highest xGA team average (56.8). This was 2.6 more than the Premier League, 3.7 more than the Bundesliga and a massive 8.2 more goals than LaLiga. In other words, regardless of the opponents taking their chances better than the Premier League and LaLiga, more goals per team were expected to be conceded than the other leagues. Whether you credit superior attacking or inferior defending enabling these chances, it certainly questions the narrative of Serie A being a defensive league.
Bundesliga Expected Goals Against
Freiburg benefited the most from the opposition fluffing their chances, conceding 16.4 (-34.8%) fewer goals than expected. Interestingly, the bottom 10 teams all conceded more goals than expected, meaning that the opposition took their goal-shooting chances better than expected against these lower-positioned teams. In contrast, only one top-eight team conceded more goals than expected. Across the Bundesliga over the season, 24.72 fewer goals should have been scored given the opportunities presented to the attacking team, thus the shooting was above average.
Ligue 1 Expected Goals Against
* PSG and Strasbourg had both played one game less, so to make them equal with the other teams, their GC and xGA averages were used for the 28th game.
To note, Ligue 1 ended early due to COVID-19 so the remaining fixtures the teams had left were not necessarily equal, i.e., some had more of the top teams to play whereas some had more of the bottom teams to play. Looking at the graph, it is quite noticeable that the bottom four teams all conceded more goals than expected: Saint-Etienne 7.3 (16.2%), Nimes 3.5 (8.0%), Amiens 11.6 (23.1%) and Toulouse 9.7 (16.8%). Across Ligue 1 over the season, 30.77 fewer goals should have been scored given the opportunities presented to the attacking team, thus the shooting was above average. Though, the four teams at the bottom significantly contributed to this.
Comparisons between leagues
As noted earlier, variation of an individual’s finishing is natural, thus expected. However, is the variation telling us something?
Overall Trends
In the top 5 leagues in Europe, the opposition tended to perform worse against teams higher up in the league. Against the top-of-the-table team, the opposition scored between 1.6 and 6.4 fewer goals than expected from the opportunities presented to them in the game. Whereas against the bottom-of-the-table team, the opposition scored between 1.3 and 8.3 more goals than they would have expected to from the opportunities presented to them in the game. One exception was the Premier League where all teams were expected to concede more, though still following the general trend. To give an indication of how much the league position affected the opposition’s goal-shooting, for every five league positions lower, the strikers scored 0.6 (Premier League), 1.2 (Serie A), 1.5 (LaLiga), 2.3 (Ligue 1), 4.1 (Bundesliga) fewer goals than expected.
To be clear, across all five leagues, the same striker with the same opportunity/chance to score in a game is less likely to score that same chance when playing against a higher-ranked team in the league. And remember, there is actually very little difference between the vast majority of strikers both between and within leagues. In other words, when an opportunity (a shot) presents itself to a striker, there is very little difference between Ronaldo scoring and Troy Deeney – it’s just that Ronaldo will have more shots (fact!).
Whilst the trend supports the above Twitter post, this trend is just for one season, and again, you would expect variation across seasons. So, does this trend continue across different seasons?
Six-year averages
Looking at differences between GA and xGA for the last six seasons for all five leagues, the overall trend is still there: a team concedes fewer goals than expected when near the top of the table and concedes more goals than expected when in the bottom half of the league given the opportunities the opposition had. In other words, when the opposition is presented with the same goal-scoring opportunities, they perform better than average against lower teams and perform more poorly against higher-positioned teams. To go further, regardless of the player’s ability, the majority of players shot better than their average against lower teams and worse than their average against top-of-the-table teams. So, yes, teams on the whole do fluff their chances against higher-positioned teams in the league.
In the 2019/20 season, the Bundesliga had the greatest difference between 1st position and last. However, over the last six seasons, LaLiga showed the greatest difference between GA and xGA for the top- and bottom-placed teams. The winning team in LaLiga, on average, conceded nearly 6 goals fewer than expected, yet, the team finishing last, conceded just over 6 more goals than expected. These two defences, given the opportunities that the opposition had scoring against them, conceded a difference of 12.3 goals through no fault of their own. Just the fact that the opposition took their opportunities more often against the bottom-positioned defence. In Ligue 1, the difference between top and bottom was 10.2 goals, the Premier League was 7.9 goals, the Bundesliga was 7.8 goals and Serie A was 6.8 goals.
Why is this happening?
Remember, it is not the ability of the attacker because that has been controlled for. But surely, the best teams have the best attackers who finish better? No. In his book The Expected Goals Philosophy, James Tippett highlights that strikers do not consistently outperform their expected goals (xG). For example, in the last five seasons from 2015/16 season, Cristiano Ronaldo has only outperformed his xG once, and that was in 2019/2020. The fact he outperformed his xG was due to his 12 successful penalties (out of 13) which turned his xG from -1.07 to +1.57. Thus, his general shooting was still below the average. There are of course exceptions, but that is it: exceptions. Lionel Messi has outperformed his xG consistently – the last 6 seasons worth of data. In the last four seasons, Ciro Immobile has outperformed his xG in three out of four seasons. The fact is: finishing ability between the vast majority of strikers is small.
So, one theory is pressure. First, it is more likely that when playing against the higher-positioned teams that games will be tighter, meaning that the opposition’s shot could heavily impact on the result of the match. To put another way, a proportion of the higher-positioned teams’ opportunities will happen when the game / result it pretty much determined, i.e., already two or three goals ahead, thus less pressure on the goal shooter. Second, lower-positioned teams create fewer chances, which the strikers know about, so, again, putting further pressure on the opportunity when presented. However, these are just theories until analysed with the data (which I do not currently have).
3. Attack is the best form of defence
Unsurprisingly, in general, the more possession opponents have the more goals they are expected to score. If all teams had 50% possession, you can expect the opposition to score 1.24 goals in Ligue 1, 1.28 goals in LaLiga, 1.32 goals in both the Premier League and Serie A and 1.56 goals in the Bundesliga. Again, whether you attribute a greater number of expected goals against a team due to a more superior opposition or an inferior defence (or a mix of the two perhaps), it is more dangerous, in terms of expected goals against, to allow your opponents to have the ball in the Bundesliga.
What happens when a team has to defend 10% more? For example, what’s the difference between a defence working 45% of the time to 55% of them time, i.e., their opponents increase their possession from 45% to 55%? By giving the opponents an extra 10% of possession in a match (45 to 55), you can expect to concede 20.4% more goals in LaLiga, 22.2% more goals in the Premier League and Serie A, 24.8% more goals in the Bundesliga and 27.4% more goals in Ligue 1.
Not only are more goals conceded in the Bundesliga when not in possession, but also a greater increase (not percentage increase) when a team allows the opposition to go from 45% possession to 55%. In other words, a decrease in a team’s possession is more likely to increase the amount of goals conceded. Finally, there is quite a lot of variation between the teams within each league, except for the Premier League and Serie A. Therefore, each team in the Premier League and Serie A can expect to concede the same amount of goals when giving their opponent more possession, whereas in the other leagues a 10% loss in possession would not be the same for each team. Nevertheless, it would still result in more expected goals against.
More possession, less defending
Acknowledging that a player/team, i.e., defence, cannot completely fall asleep defensively whilst attacking and that defences do too contribute to attacking play, defending is nevertheless a far easier job when the opponents do not have the ball! Expected goals controls the opposition/striker’s shooting, but there are many more things to keep the same when making fairer comparisons. By using a team’s average possession from whoscored, the minutes the opposition had the ball was calculated.
The above graph shows the 20 teams that had the most possession. As a result, it shows the total number of minutes the opposition had the ball. In 2019/20, Barcelona had 63.2% possession, meaning that their opposition only had the ball for, on average, 33.1 minutes per game. In effect, Barcelona’s defence had the least amount of defending to do over the course of the season. Manchester City’s defence had the second least amount of defending to do with their opposition only having the ball for 33.7 minutes per game. Third was PSG’s defence (34.4 minutes) and fourth was Bayern Munich’s (34.7 minutes). In Serie A, Napoli had the least defending to do (38.5 minutes), and 9th overall.
Less possession, more defending
The above graph shows the 20 teams that had the least possession. As a result, it shows the total number of minutes the opposition had the ball. In 2019/20, Augsburg in the Bundesliga had 41.3% possession, meaning that their opposition had the ball for, on average, 52.8 minutes per game. Consequently, Augsburg’s defence had the most amount of defending to do over the course of the season. Newcastle’s defence had the second most defending to do with their opposition having the ball for 52.4 minutes per game. Third was Brescia’s defence (51.4 minutes) and fourth was Alaves’ defence (51.1 minutes).
The difference between Barcelona’s amount of defending and Augsburg’s is, on average, 19.7 minutes per game. This equates to 748.6 minutes over the course of a 38-game season, which is equivalent to 8.3 games. That’s defending for over eight whole games more! Whilst these teams are in different leagues, the difference between Barcelona and Alaves is 18 minutes per game, 684 minutes over the course of a season, equivalent to 7.6 games. Now, across all five leagues it has been shown, unsurprisingly, that the more possession your opponent has, the more goals they can expect to score. Thus, the longer a team does not have the ball, the more goals they can expect to be scored against them. If teams/defences have different amounts of defending to do, is it fair to compare them without factoring this in?
The next five graphs, one for each league, controls the opposition’s shooting ability (xGA) and the time that the opposition have the ball. Consequently, this is looking at the efficiency of a team’s defence, i.e., how often and how good are the goal-shooting opportunities given the amount of time they defend for. For example, Team A defending for 40 minutes with an xGA of 1.5 is the same rate (efficiency) as Team B defending for 54 minutes with an xGA of 1.8. Both Team A and B’s defences performed equally well despite Team B expecting to concede more goals, but only because they were defending for a greater amount of time.
Best Defence in the Premier League 2019/20
The best defence in the Premier League was Wolves. If all teams had equal possession (45 minutes), opposition teams against Wolves would create the lowest expected goals per match (0.96). They were the only team that could expect to concede just under one goal per match. Manchester United had the second-most efficient defence, expecting to concede 1.1 goals. From third to eighth there is little difference. Norwich had the least efficient defence, which was nearly twice as ineffective as Wolves.
The top two finishing teams, Liverpool and Manchester City, who conceded the fewest goals (33 and 35 respectively), now have their defences ranked 6th and 7th best. As previously mentioned, Chelsea received plenty of criticism for their defending in the 2019/20 season, yet, their defence is actually very similar (1.31) to both Liverpool (1.29) and Manchester City’s (1.30). Thus indicating, that all three defences offer the opponents the same accumulative opportunities whilst factoring in the opposition’s possession. In addition, the defences for Liverpool, Manchester City and Chelsea were all around 35% less effective than Wolves.
Best Defence in the LaLiga 2019/20
The best defence in LaLiga was Atletico Madrid’s, expecting to concede only 0.81 goals had the opposition had the ball for 45 minutes. For anyone who follows Spanish\European football, this is not a surprising result at all. Atletico Madrid are renowned for their superb organisation whilst defending. Here, is where the term ‘defence’ takes a bigger picture: the whole team. Is it that Atletico’s midfielders, and strikers, help out their defenders more than other teams? If that was the case, then should Atletico’s defence be regarded as the best? Easier job perhaps? Notwithstanding how you define ‘defence’, Atletico Madrid’s defence was quite considerably more efficient than the rest. It was 25.3% more effective then second placed Real Madrid.
Real Madrid can expect to concede 1.02 goals and both Granada and Getafe 1.04 goals each game had the opposition had the ball for 45 minutes. To remember, Real Madrid conceded only 25 goals (8.2 goals less than expected), but Getafe 37 goals (6 goals less than expected) and Granada 45 goals (1.6 more goals than expected). The reason for such similar defensive metrics now is that all three defences conceded a very similar rate of accumulative goal-scoring chances for their opponents. So, when the teams did not have the ball, the opposition were able to create similar chances at the same rate against the three defences.
Finally, Leganes must be pointed out. Their defence ranked 5th in LaLiga, despite being relegated by finishing 18th in the table. In contrast, Barcelona’s defence ranked 15th despite finishing second in the league. This is due to the fact that Leganes only expected to concede 7.8 more goals than Barcelona over the course of the season, yet they defended 10.9 minutes more per game, totalling 413.8 minutes, which is the equivalent of 4.6 games.
Best Defence in the Serie A 2019/20
The best defence in Serie A was Inter Milan’s, expecting to concede 1.10 goals each game had the opposition had the ball for 45 minutes. The second-best defence was Atalanta’s, though it was still 12.2% less effective as Inter’s. The least effective was Lecce, expecting to concede a massive 2.24 per game if the opposition had had the ball for 45 minutes. This was over twice as ineffective as Inter Milan’s defence. Interestingly, the second-least effective defence in Serie A was Sassuolo’s. Despite this, they finished 8th in the league. Why? It’s a combination of two things: Sassuolo had a high xGA whilst only allowing, on average, their opponents to have the ball for 43.9% of the match.
Best Defence in the Bundesliga 2019/20
The best defences in the Bundesliga were Wolfsburg and Leipzig’s, expecting to concede 1.20 goals per match when controlling for possession. Leipzig came 3rd in the league, conceded 37 goals (2nd), xGA 37.58 (2nd), with their opposition having possession 45.9% of the time. Wolfsburg came 7th in the league, conceded 46 goals (6th), xGA 41.91 (3rd), with their opposition having possession 51.4% of the time. Despite an xGA difference of 4.33 goals, due to Wolfsburg’s defence defending nearly five minutes more per match than Leipzig’s, it shows that when both teams are without the ball their opposition accumulate the same rate of goal-scoring opportunities against them.
To highlight the distortion in using goals conceded or even just xGA as a metric to compare defences, Dortmund came 2nd in the league conceding 41 goals (4th), xGA 42.04 (4th) and opposition possession of 41.5%, whereas Augsburg came 15th in the league conceding 63 goals (13th), xGA 59.12 (14th) and opposition possession of 58.7%. Again, despite an xGA difference of 17.08, Augsburg’s defence were defending for about 15.5 minutes more per match. When equalising the time – a fairer comparison – the two defences were nearly identical at allowing their opposition to create goal-scoring opportunities. In fact, Augsburg would then expect to concede 1.48 goals to Dortmund’s 1.49 goals.
In contrast to the other leagues, the best defences in the Bundesliga are expected to concede over a goal a game. This is further exaggerated when looking at 3rd position Bayern Munich who expect to concede 1.39 goals when the opposition have the ball for 45 minutes. This is in comparison with Ligue 1 (Lyon, xGA 0.98), LaLiga (Granada, xGA 1.04) and Premier League (Sheffield United, xGA 1.23). This is mainly explained by the earlier findings from the Bundesliga: first, more goals are expected when the opposition have possession, and two, when the opposition increase their possession, the expected goals increase at a quicker rate. In other words, possession is more valuable in the Bundesliga, with goals being produced more often when in possession as well as being even more valuable when the opposition increase their time with the ball.
Best Defence in Ligue 1 2019/20*
The best defence in Ligue 1, before the league ended due to COVID-19, was Angers, expecting to concede only 0.93 goals per match had the opposition had the ball for 45 minutes. Two other teams could have expected to concede under one goal per match: Reims (0.94) and Lyon (0.98). Again, this does certainly need to be taken into context as some teams would have played more higher-positioned teams than others before the cancellation.
Before moving onto the best and worst defences across the leagues, is controlling possession a fair comparison? Why not the number of attacks in the final third? Or the number of possessions a team has? Herein lies the problem with measuring defence/defenders. For example, the metric attacks in the final third, could be argued that if the defence are doing a good job then there is less chance of an attack in the final third, thus not being accounted for. This is similar to the tackling statistic. Tackling is often seen as a last resort, with the better players reading the game better and avoiding tackling, or certainly making less of them. Less is more…
4. Best and worst defences in Europe?
The graph below shows the most efficient 25 defences across the top five leagues in Europe for the 2019/20 season. These are the same metrics used in the previous section in that they look at the rate of goal-scoring chances created by the opposition.
Not only is Atletico Madrid’s defence the best, most efficient, defence in Europe, but by some margin too. Four of the top 6 defences are from Ligue 1 and 10 of the top 25 are also from Ligue 1. Due to the fact that Ligue 1 did not complete its season, and the fact these metrics are created within the leagues and that Ligue 1 is the weakest league (arguably by some way), the following graphs are made without Ligue 1.
Still, Atletico Madrid’s is the most efficient but now by an even further margin of 18.5%. Wolves from the Premier League were the only other team that expected to concede less than one goal per match had their opposition had the ball for 45 minutes. Interestingly, five of the top six defences come from LaLiga, suggesting that LaLiga contains the opposition by having the most effective defences. There are of course many possible reasons for this, and until further analysed would remain speculative.
One possible reason is that teams in LaLiga build up attacks with possession more than the other leagues, thus taking longer to create a chance. Another would be that the attacking strength of a team is closer to the opposition’s defensive strength. A limitation of this way of analysis is that they are calculated within leagues, thus not accounting for league differences. For example, more goals were scored in the Bundesliga whilst the team had possession, thus meaning that teams, on average, will concede more per minute without possession. Hence why only two teams, Wolfsburg and Leipzig, ranked in the top 25 defences in Europe’s top 4 leagues.
From the top four leagues, none of the champions had the best defence. In the Numbers Game, the authors pointed out that the best defence will also be crowned champions 46% of the time, with a low of 40% in the Premier League compared to a high of 55% in Serie A. Unfortunately, they [the authors] fall into their own criticism they gave to Charles Reep in which they stated: ‘He simply did not have the open mind or the techniques required to make sense of the wormball of information that every football match, every tournament, every season provides us.’ They simply regurgitate the same old mythical rhetoric that the best defence is the one which concedes the fewest goals. To be clear, this article’s method is no way near the level of sophisticated statistics that football needs, which other sports have. It is though, without doubt, a marked improvement on today’s dismal goals against analyses.
As previously mentioned, football struggles to capture what defenders actually do in terms of metrics. Over the last couple of seasons, Liverpool have been receiving plaudits left, right and centre for their defending. The graph above ranks Liverpool’s defence in 21st position in Europe during the 2019/20 season, and 6th best defence in the Premier League. In the 2018/19 season, Liverpool should have conceded 32.5% more goals than they actually did, meaning the plaudits given for conceding only 22 goals in the Premier League was not warranted. That season, Virgil Van Dijk just lost out to Lionel Messi on football’s greatest individual award: the Ballon d’Or. In 2020, up until the Ballon d’Or was cancelled, Virgil Van Dijk was still a Top 6 favourite amongst most bookmakers. What metrics supported this individual’s rapid rise? Was it the case of a team’s success overshadowing the true performance of an individual?
Lecce in Serie A had the least efficient defence in 2019/20. Even if they had the ball for 45 minutes, they could have expected to concede 2.24 goals per match. This was the highest rate by quite some margin. A reason for this is that Lecce expected to concede 91.97 goals, which was 13.59 more goals than the 2nd worse xGA of Brescia, also in Serie A. It is unsurprising that they were relegated, though were only four points off safety. Interestingly, Sassuolo who had the second-least efficient defence in Europe, 2nd worse in Serie A, came 7th in the league. One reason for this is due to the fact they scored the 5th most goals in Serie A, thus demonstrating the inextricably attacking-defensive relationship on final league position.
Four out of the seven least efficient defences in Europe were from Serie A, further questioning that the Italian league is the most defensive league. Of the 12 least efficient defences in Europe, six were from Bundesliga, though not surprising from the fact that earlier findings showed teams scoring at a better rate whilst in possession compared to the other leagues.
The Stat Squabbler says:
- Never use goals conceded or clean sheets alone to compare defences and/or goalkeepers. Must use expected goals.
- Opponents shoot worse when playing against a higher-placed team. Thus, flattering higher-placed defences/teams and not punishing them when given opportunities to score.
- Defences do not do the same amount of defending. This should be factored when making comparisons.
- Better defence measurements needed. This is not specific to football. Many defensive actions that are measured do not necessarily reflect the best defending.
Do you agree with the Stat Squabbler: Do the best team’s defenders have the easiest job? Same can be said for attackers, they get the most chances. How would you measure the best defence?
Comment below.
