In Search Of a Midfield Destroyer

By Conor Murphy –  @murphycon

 

The role played by N’Golo Kante in winning back to back league titles for two different clubs has brought attention to the value of a midfield destroyer. This type of player is proactively looking to win the ball back for his team. He’s not happy to sit and wait, screening in front of the defence and hoping to make an interception (I’m looking at you Michael Carrick). The midfield destroyer is going after the ball and he’s going to get it back. He puts in a lot of tackles and sometimes makes interceptions. Then, when he has the ball, he plays a simple pass to a team mate and his team retain possession. There’s no point in winning the ball back just to give it away again.

We can certainly expect some clubs to be in the market for a midfield destroyer this summer. It’s easy to pick out someone like Saul Niguez or Fabinho on the basis of watching them play a few Champions League games, but can we use stats to identify talent that might be a little under the radar?

If you’re not interested in the methods I used to identify some players, skip forward to the ‘Results’ section to find out why Valentin Rongier is a name you might want to know.

Method

The first challenge is to identify what type of output we are looking for. Tackles, obviously, are high on the agenda. Specifically we will want to look at successful tackles. Interceptions are another good way of winning the ball back, so if our midfielder can tackle and intercept, that would be great. Next, we need to make sure our new player doesn’t commit a lot of fouls. We don’t want someone who’s going to give away 5 fouls against Swansea and watch Gylfi Siggurdson rack up a hatful of assists. As this is just an initial list of prospects we are looking to put together, we will add just one more requirement. We need the player to keep the ball, so we’ll look for a good pass completion rate.

In summary, we’re looking for:

  • Successful tackles
  • Interceptions
  • Low number of fouls
  • High pass success rate

The next question is how many tackles do we want the player to make? Is 2 per 90 minutes good? 3, 4, 5? What about interceptions? What’s considered a low number of fouls for a player who will be attempting a lot of tackles? One way to answer this question is to benchmark. Let’s look at what Kante himself gives to his team and then see if we can find someone else who could offer a similar output.

Club Season pAdj Tackles p90 pAdj Int. p90 Combined Fouls p90 PS%
Chelsea 16/17 3.87 2.69 6.55 1.6 89%
Leicester 15/16 4.66 4.12 8.78 1.3 82%
Caen 14/15 4.47 2.68 7.15 1.6 79%

All stats are per 90 minutes. Tackles and interceptions are adjusted for possession for reasons explained on StatsBomb.

Before I go any further, I think it’s important to point out the limitations of using simple metrics to identify talent. It’s very possible that a player can have good stats, but have weaknesses in areas not captured by statistics, like positioning and decision making. I do think stats can identify players that might be worth looking at, but it’s important then to get eyes on the player to see if he is actually good at football, or just good at racking up tackles, for example.

The Search

Firstly, I’m going to rank players based on their successful tackles per 90 minutes. I’m going to filter for players who currently play as central midfielders or defensive midfielders. We could bring in a full back, for example, and train him to play central midfield, but it brings in another layer of risk. I’m also only going to look at players who have played more than 1000 minutes this season to ensure a decent sample size. I now have a list of 70 players from the top 5 leagues who average 3 successful tackles per game or greater.

The next step is to cut this list down a bit further using our other requirements of interceptions, fouls and pass completion rate. Let’s look for less than 1.8 fouls per 90 mins and a pass completion rate of 78% or greater. I’ll also take out players who are probably unavailable. Out goes Sebastian Rudy, who is promised to Bayern Munich, Ander Herrera and N’Golo Kante himself, among others. I’m also saying cheerio to Kevin Strootman and his knee injury history. Finally, I’m going to look for players who are 24 years old or younger. I’m not looking for a short term solution.

Using interceptions as a filter is slightly more difficult. Let’s say we have one player who completes 4 tackles per 90 and 1 interception, and another player who completes 3 tackles per 90 and 1.5 interceptions. If we set our requirement as 1.5 interceptions per 90, we’ll miss out on that first player who is posting exceptional successful tackle returns. Instead of filtering based on interceptions, I’m just going to add in a column for interceptions so they’re visible, and then add a new metric which I’ll call ‘win backs’, which is the sum of interceptions and tackles per 90 minutes.

Results:

Now I’m left with a list of seven potential midfield destroyer superstars. Here are the seven players, along with three seasons of N’Golo Kante as a reference, sorted by win backs per 90:

Player Club Season pAdj Tackles p90 pAdj Int. p90 Combined Fouls p90 PS%
Kante Leicester 15/16 4.66 4.12 8.78 1.3 82%
Kante Caen 14/15 4.47 2.68 7.15 1.6 79%
Rongier Nantes 16/17 5.23 1.74 6.97 1.7 82%
Kante Chelsea 16/17 3.87 2.69 6.55 1.6 89%
Llorente Alaves 16/17 3.60 2.49 6.09 1.4 87%
Paredes Roma 16/17 3.52 2.53 6.05 1.3 89%
Pulgar Bologna 16/17 3.17 2.88 6.05 1.5 83%
Højbjerg S’oton 16/17 3.29 2.54 5.83 1.3 88%
Doukoure Metz 16/17 3.32 2.42 5.75 1.4 85%
Brozovic Inter 16/17 3.25 1.63 4.88 1.4 83%

Perhaps unsurprisingly, N’Golo Kante is very good at consistently being N’Golo Kante. We can also see that Valentin Rongier looks like a player we should know more about, with the highest Tackles P90 on the list. And then there’s Marcos Llorente, who is on loan at Alaves from Real Madrid. For a club that loves buying Galacticos, Real have an amazing amount of young talent on the books at the moment. Kovacic, Morata, Isco, James, Danilo, Carvajal, Varane, Asensio and Lucas Vazquez are all 25 or under, and then they have players like Llorente who are out on loan.

Searching through Ted Knutson’s Twitter timeline I couldn’t find a radar for Valentin Rongier, but I did find a radar for Marcos Llorente and one for N’Golo Kante for comparison.

13

14

From a Premier League point of view, I’m intrigued to see Pierre-Emile Hojbjerg as the only remaining Premier League player on this list, but of course not surprised that Southampton has made another great signing. We know Premier League clubs place a premium on players who have Premier League experience, so don’t be surprised to see Hojbjerg move to one of the top 6 (Liverpool maybe!?).

So, who missed out?

It’s interesting to look at the players who were cut out of the list during the filtering stage and see why.

Quite a few players were cut out when I set the bar for fouls per 90 mins to less than 1.8, including Fabinho, Casemiro, Saul Niguez, Diego Demme and Naby Keita. These are undoubtedly good players, but I’m specifically looking for someone who can put in a lot of tackles and not give away fouls, so unfortunately they can’t make the list. People with access to more detailed data may be able to break the fouls down by location, and place a heavier weight on fouls conceded in or close to the box versus in the opposition half, for example. A player who commits a lot of fouls in the opposition half could be considered an asset if your aim is to break up counter attacks.

Moving on, one of the players signed by Leicester to replace Kante, Wilfred Ndidi, missed out because of his pass completion rate of 76%. He’s only 20, so maybe that will improve in time. Rodri at Villareal missed out because he has only played around 600 minutes, but he’s one to keep an eye on for next year. My age limit of 24 has no strong logic behind it, and if we set it to 25 then Oriol Romeu makes the cut (Southampton strike again).

Also players from outside the Top 5 leagues were not included in the analysis, so maybe Portugal, Holland or the Championship are currently home to a gem of a midfield destroyer that we’ve missed.

Caveat Emptor

As I pointed out above, just because a player has good defensive output, it doesn’t mean they are a good player. It may only take an experienced scout 10 minutes to look at someone like Valentin Rongier and find key faults in his game. However, at least by using stats, we can draw attention to players who may not have been considered previously.

In 2015, when Leicester signed Kante, Charlton Athletic also made a signing from Ligue 1. At the time, he was making more tackles and more interceptions per 90 than Kante. In the mean-time, Kante has won two league titles, and El-Hadji Ba was released by Charlton in February, and is now plying his trade with Stabaek in Norway.

Player Club Season pAdj Tackles p90 pAdj Int. p90 Combined
El-Hadji Ba Bastia 14/15 6.30 0.90 7.20
Kante Caen 14/15 4.47 2.68 7.15

Conclusions

  • Not many players have the same skill set as N’Golo Kante. Goal-scorers are hard to find, but maybe top quality midfield destroyers are also a rare and valuable commodity?
  • Midfield destroyers are probably quite good value as their skills are less attention-grabbing than a striker.
  • This time last year Valentin Rongier would not have made the list, having only completed 741 minutes in 11 appearances in 2015/16 season (less than 1,000 minute cut off), but his stats were looking good then already, with 3.9 tackles per 90. Right now, he could be the next N’Golo Kante, or he could be the next El-Hadji Ba.

All data is from whoscored.com.

5 thoughts on “In Search Of a Midfield Destroyer”

  1. Hi. How do you make possession adjustment? Do you use sigmoid or simple method (2*x/(oppossition average possession%)?

    And what possession% stats? Season average? Or you look at every separate match, where needful player was involved?

    Like

    1. Reply from Conor:

      I didn’t use the Sigmoid method (I have read it and I must learn how to). My formula for adjusting for possession is as follows:

      Tackles * (1+((Pos%-50%)*2))

      Where Pos% is the team average Pos%

      I used season average but yes, looking at the specific games the player was involved in would be more accurate.

      Like

  2. Also when you use “fouls” filter, it’s very important to consider that there are aerial fouls and ground fouls. And it’s possible that you’ve excluded somebody, who makes lots of aerial fouls.

    Like

    1. Reply from Conor:

      The data I have doesn’t break up the type of fouls, only the total number. I am curious as to why you would separate for ground and aerial fouls and why you would assign different value to each?

      I would like to break down fouls by location, just like shots and xG. A foul in the centre circle is much less damaging than a foul in the penalty box and should be counted as so.

      Like

      1. I think we should separate aerial and ground fouls, because if player has 2 fouls per 90 (1.5 aerial and 0.5 ground), it doesn’t mean that he is not good enough as tackler. Maybe he was involved too much in aerial duels. It’s important component, but it’s not about tackles.

        Like

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s