Major League Soccer 2019: Statistically Best Central Defensive Midfielders Part Two – data analysis
In the last continuation of this series, we covered the best attacking central defensive midfielders in the MLS. However, there are different styles in the MLS, it makes sense to also cover the defensive-minded central defensive midfielders in the MLS.
In this data analysis, using data and statistics, we will conduct an analysis of the central defensive midfielders in the MLS – figuring out the best defensive central defensive midfielder in the league.
Setting the Guidelines with PCA
Before in this series, one of the troubles I encountered was just how difficult it was figuring out different styles with the myriad of the statistics we have available. For some players and some positions, using statistics to find the best player is relatively easy. Looking at wingers, attacking statistics, in general, can lead us to the find the best such as Leroy Sane of Manchester City and some upcoming players such as Nicolas Pépé of Arsenal. However, looking at positions such as central defensive midfielders – using statistics of any kind quickly becomes very difficult as these players have various components to their game.
Making scatterplots of various metrics not only becomes monotonous but also can lead to us missing out some players simply due to our monotonous activity.
In lieu of this, the Principal Component Analysis becomes very helpful. A PCA, as it is referred to, is simply a helpful guide that can tell us, mathematically, the different types of players that exist in a bigger pool of players. In doing so, the PCA also tells us what indicators most correspond with the types of players.
As such, players like Bayern Munich’s Thiago Alcantara and Chelsea’s Mateo Kovačić would be characterized as attacking central defensive midfielders that are press-resistant. The statistics corresponding to their mould would include metrics like dribbles, progressive passes, and through balls.
I applied this PCA to central defensive midfielders in the MLS who had played more than 20 games and here were the results.
Here we see four types of midfielders that are the most common in the MLS. Since our focus is on a defensive midfielder, we want to search for the type of midfielder that has a high correlation with defensive statistics. What this allows us to do is find other statistics that, at first don’t seem correlated, but upon a mathematical inspection, are positively correlated.
In this heatmap, the more blue the colour is, the greater correlation while the opposite is true for the shades of red.
We immediately find the midfielder we are looking for in the Midfielder Four category. This type of midfielder has a high correlation in aerial duels, possession-adjusted interceptions, possession-adjusted sliding tackles, and shots blocked per 90. This is as expected. However, looking closely, we see correlations to other stats such as forward passes, long passes, and progressive passes to name a few. Now we, mathematically, have narrowed down our statistic list to look at the most relevant metrics.
Statistical analysis into the league
We will start with the traditional defensive metrics such as aerial duels per 90 and make our way to the metrics that PCA told us that we “missed”.
Here is a beeswarm plot of aerial duels per 90 with the accuracy of it being plotted in the size of the circle. We can clearly see that most central defensive midfielders average about two aerial duels per 90 minutes which is reasonable as central defensive midfielders aren’t as renowned for their aerial ability.
In this light, we find Andy Rose leading the way in aerial duels per 90 near 5.5 aerial duels per 90. In addition to the high frequency, Rose also does it a high accuracy rate. Other names were also taken for further analysis.
Next up are defensive duels per 90. As expected, central defensive midfielders perform much better on average with seven defensive duels per 90. Defensive duels are a much natural ground for central defensive midfielders as their main job, in defence, requires them shielding, man-marking, and intercepting attacking plays.
Our previous outlier, Rose, finds himself bottom on the list. Instead, San Jose Earthquakes’ Judson tops the list with almost 12 defensive duels per 90. We also see that with the size of the dot that Judson does so with above-average accuracy.
Here we have quite possibly the most important defensive metric available to us: possession-adjusted interceptions per 90. Some teams average more possession, like LAFC, and others are content to leave it, like the Seattle Sounders. What is interesting is that possession and defensive actions are inversely related – teams with more possession generally tend to make less defensive actions as their possession serves as a form of defence. The opponent can only attack when they have the ball.
This is where possession-adjusted statistics help us. By adjusting for more or less possession, these metrics allow us to measure the actual defensive capabilities of players. It is also more comparable among players. As such, possession-adjusted interceptions are the important metric here as it is not only actually providing us with the most accurate defensive data but also the most important one. Interceptions, arguably, are the most important metric as they tell us the ability for a player to be defensively aware and spot a dangerous pass or move and intercept.
In this metric, we see two new outliers: Alexander Ring and Sean Davis. Both nearly record eight PAdj interceptions per 90. Other names such as Rose were also taken for further analysis.
Here we have possession-adjusted sliding tackles per 90. These metrics are also possession-adjusted and this time tell us about central-defensive midfielders who are rather excellent at making more old-fashioned defensive actions. This is a style that is rather going out of style and we see this reflected in this plot where we see a high number of players recording less than 0.8 sliding tackles per 90.
In this mess, we find the experienced Jonathan dos Santos leading the way with 2.7 sliding tackles per 90. When considering the team he plays for, LA Galaxy who are a sloppy team in defence and transitions, it makes sense why dos Santos would require such high figures.
The last of the defensive metrics that I have included here is shots blocked per 90. It is not as accurate of a defensive metric as it is influenced by team style. However, the PCA showed that there is a decent positive correlation and as such, it merits putting this metric to the test.
We find old faces such as Rose and Judson leading the way in this metric with 0.70 and 0.57 shots blocked per 90. However, as these numbers themselves show, this metric is not as powerful as some of the other metrics. Blocking seven shots in a total of 10 games, a much more realistic explanation of 0.70 shots blocked per 90, doesn’t tell us much about their prowess. However, it is still a decent metric to analyze but its findings must be taken with context.
Now in the PCA, the analysis lit up other metrics as forward passes, long passes, and through passes per 90. However, looking up the correlation factors, only one statistic really showed up strongly: long passes.
We do see an outlier in this trend with the name Eduard Atuesta popping near the top of this plot. However, this speaks to his exceptional talent – recording high numbers in a statistic that was shown to be for identifying defensive midfielders. However, we see that Atuesta is the outlier with other attacking central defensive midfielders like Lee Nguyen and Jackson Yueil are near the back and/or average.
In this metric analysis, we see Bryan Acosta leading the way with 8.25 long passes per 90 with a high accuracy per match. Other names such as Ring were taken for further analysis.
These two metrics that were not included in PCA also have enough merit to analyze. In previous instalments of this series, I have talked about the graph between fouls per 90 and successful defensive actions per 90. While all midfielders make fouls – we don’t want a midfielder who makes many as losing the ball between the attack and defence is the most criminal act a midfielder could do.
So we are looking for average to below-average fouls per 90 and successful defensive actions per 90. In this context, we see Davis lead the way with 13.5 successive defensive actions per 90 with almost 1.2 fouls per 90. We also see with the colour that David records a high number of PAdj interceptions. Other numbers such as Ring were taken into account for the final list of central defensive midfielders to closely analyze.
Picking the best defensive central defensive midfielder
After much analysis, I chose the four following central defensive midfielders to fixate upon: Andy Rose from Vancouver Whitecaps FC, Alexander Ring from NYCFC, Judson from San Jose Earthquakes, and Luis Alberto Caicedo from New England Revolution.
We’ll start from looking at their most basic defensive statistics first as those are the more important statistics.
Here we have the seven defensive statistics in line with the four players in the order we previously analyzed.
Rose performs the most aerial duels – however doing it with bad accuracy as shown by the colour. Ring, Judson, and Caicedo all perform excellently in the defensive duels per 90 metric and their accuracy rate can be particularly forgiven as defensive duels are more often lost and those who have better success rates, typically, commit less defensive duels, to begin with.
All midfielders perform excellently in their PAdj interceptions with only Ring and Caicedo succeeding the PAdj Sliding Tackles metric. However, as we stated before, the PAdj Sliding Tackles isn’t as descriptive in terms of how defensive a midfielder is.
Rose and Judson perform excellently in blocking shots with Ring, Judson, and Caicedo doing exceedingly well in recording successful defensive actions per 90 – a good indicator of defensive prowess. However, quite paradoxically, the same players also commit the most fouls per 90. While this is slightly troublesome – it must be regarded that these players are also engaging in high defensive actions.
Defensive actions, unlike attacking actions, are not as much in control by the person. In attack, you can choose to go where you want, for the most part, and have control over your passes. In defence, much of your actions are in part yours, in part your team’s defensive structure, and in part the opposite team’s evolving positions.
Having analyzed the defensive characteristics, we’ll analyze the characteristics that the PCA told us were also associated with defensive central defensive midfielders. In this part of the analysis, I have chosen to include other factors to give us more width in our analysis as we have now gotten specific and don’t have to narrow down our pool.
Here we see three metrics that the PCA tells us are related to a defensive central defensive midfielder: long passes, forward passes, and progressive passes.
We have already discussed why the inclusion of long passes makes sense here. Forward passes also deserve their place here as defensive as a central defensive midfielder can be, the manager will still task them at getting the ball out of the defence which is where forward passes come into play. Since the central defensive midfielders have to advance the team most, it also makes sense to include progressive passes per 90 – defined by WyScout as forward passes that are 30m long when the pass starts in the team’s own half or at least 10m in length in the opponent’s half.
We see Ring not only play the most long passes but also do it with the best accuracy among the four. Ring and Judson play the most forward passes with Judson recording the best accuracy among the players. Lastly, both Ring and Judson, yet again, make the most progressive passes with both recording the highest accuracies.
At last, we’ll have a look at the attacking metrics related to these midfielders. It is important that our midfielder be able to attack in some sort of sense. In a surprising trend, Ring excels in every metric recording the highest passes and high accuracies. Following him on the trend is Judson who records above-average performances in every metric. Rose and Caicedo are the poor performers who do not make a lot of attacking passes and don’t record great accuracy rates.
Looking at these charts, the final defender that comes out is Judson of the San Jose Earthquakes. He records high PAdj Interceptions, defensive duels, and successful defensive actions per 90. While he does make many fouls – an addition of 0.5 from the average of 1.25 – he more than makes up for it with his excellent command of passing.
Here we have a standardized metric view. A standardized metric view is a term that I’ve made that describes a graphic that shows the standardized version of metrics. What do I mean by this?
Well in statistics, there is a common problem – how to compare a player against a player or against a league? There are various ways to do that but what gives you a single number for the one metric?
The process of standardization answers the problem. It is the process by which numbers are converted into a form where the comparison is allowed. The mathematics is not complicated but not the focus of this article and as such, I won’t delve into the specifics of it. Simply put, standardization of metrics allows us to objectively see how much better/worse a metric is compared to the average.
Simply put, this graphic gives us a nice, objective look at what things Judson is good at and bad at – compared to the average. Analyzing the green spokes, we find that those are mainly linked to the defensive aspects – which is what we want of course. A lot of the red spokes are mainly linked to the attacking aspects – which is something that would be nice to have but it is not necessary.
A key to notice is that in all the defensive aspects’ Judson’s spokes extend far meaning that he is above-average in that metric. That performance in various of the defensive metrics backs up our previous claim of Judson being the best central defensive midfielder.
As we have seen from this data analysis, Judson stands out as the best defensive central defensive midfielder in the MLS! Of course, this bears stating, this is not the final product. Merely, it is an in-depth look at statistics and data to narrow our future scouting and recruitment.
Of course, the eye-test would need to be followed up to truly determine whether the player could be suitable for one’s team’s playing style. And in addition, players change over time and period and this was only for the 2019 season for the MLS. Additionally, there is also a variability factor in the analysis as when I looked at trends in the MLS, the 2019 season was an outlier in terms of its goals scored. In other words, most defences faced lesser shots save for when they encountered LAFC and as a result, the defensive metrics we have at hand may not be the most representative of the league.
In any other case, our backline for the statistically best midfielders is now done. In the next instalments of the series, we’ll look at the remaining four positions in our 4-2-3-1: the attacking midfielders, wide midfielders (two of them), and the striker.