Reading’s Defence and Valuing Defensive Midfielders

Trying to value defensive midfielders

One position that often gets overlooked is the defensive midfielder – particularly for dominant sides that possess lots of exciting attacking players. With that being said, this has become a bit of a cliché and it’s often people will still claim a player is underrated or that a player rarely gets mentioned despite them getting lots of praise, but there’s no need to get into all that. What I’m trying to get at is on dominant sides it can be hard to quantify their worth with basic statistics, as they’re unlikely to have huge numbers for defensive actions, especially compared to sides who sit back and face wave of attacks and have more opportunities to win the ball back.

With this in mind, I wanted to try and find a way to try and see how these players’ jobs could be measured using basic numbers (as I have no idea how to go about possession adjusted numbers or things like possession value models). There’s still a lot of problems and assumptions which I’ll mention, but again I found a lot of the stuff below quite fun to look at.

Putting out fires

A phrase that I see used to describe what defensive midfielders to is to put out fires. It’s a nice way to describe what a defensive midfielder could do on a more front-footed side. They’re unlikely to do something headline-grabbing, but constantly being able to stop counters or the opposition progressing the ball can have huge value – particularly if they’re stopping them at the source, rather than after running back into their own half.

To try and find players who fit the bill I defined a set of rules, similar to how I looked at transitions above. I used the same dataframe that contained successful turnovers and shots. The assumption is that if an event isn’t a shot and the previous event also isn’t a shot and the two events are from different teams, it means the first team lost the ball and then won it back with the second action.

I then set rules to say the team winning the ball back must have lost it in the final third and then win it back within 10 seconds of losing it. It also must be won back in the opposition half, so it’s stopping the counter at the source rather than dealing with it later on. A further rule I added was to say it has to be further back the 18-yard box, just because some scrappy sequences out wide could slip through the cracks.

It’s quite specific, so the numbers players put up aren’t huge, and there are also a lot of other problems with it. It’d be better to look at how a team is attacking and then see who breaks it up rather than hacking it together this way, but I wasn’t sure how to do that. Then there’ll be plenty of times where a team isn’t countering but just clear it and a player mops it up. It’s not what I wanted to find, but without looking at the type of attack from the opposition, I wasn’t sure how to rule these out.

So after taking a while watching clips and trying to decide on the rules, it turns out the player with the highest number of these actions isn’t even a defensive midfielder, it’s Katie McCabe from Arsenal. She makes 1.84 of these actions per 90 – showing that this is probably too rigid a definition to be of much use.

The good news is after her we get some defensive midfielders, mostly from front-footed sides. The next five are Kim Little also from Arsenal, Man City’s Caroline Weir, Birmingham’s Chloe Arthur, Reading’s Remi Allen and Man City’s Keira Walsh.

A few things popped up from this though. Chelsea make around the same number of these actions per game as Arsenal and City but they don’t have an individual amongst the top few names. Sophie Ingle has the most for them and isn’t far behind, but it is an area where I expected to see her near the top. Second, the top three midfielders are all Scottish, yet Scotland had some problems defensively in midfield during the World Cup.

Stopping progression

To try and get a better look at who stops the opposition from progressing, I thought I’d use the same dataframe as before that contains successful turnovers and progressive passes.

First, to try and find players that stop progressive passes from being completed, I said if a progressive pass is incomplete then the turnover following has to be what stopped it from being completed – which was double-checked with the time and duration of the pass and the time of the defensive action.

Looking at this, it was disappointing. It was mostly centre-backs as an attempted progressive pass can just be a long-ball upfield. However, removing these centre-backs, we get some similar names from above and some new ones as well.

In first place is Caroline Weir, followed by Maéva Clemaron from Everton, Tottenham youngster Chloe Peplow, Liverpool’s Jade Bailey and Chloe Arthur again.

While stopping progressive passes is obviously good, there is less of a front-footed approach to this. A player can sit back and make these interventions, which doesn’t really help in trying to find ways to value defensive midfielder on dominant sides.

My next idea was to look at players who win the ball back after a progressive pass, rather than just stopping it being completed. Similar to the putting out fires above, trying to find which players regain possession after the opposition progresses the ball.

It’s the same as above, but rather than the time being right after the pass, it’s now within 10 seconds of a completed pass. The top five for winning the ball back in these instances are Arsenal’s Lia Wälti, Chloe Arthur, Caroline Weir, Kim Little and Chloe Peplow.

Combining the two, finding players who both stop progressive passes and win the ball back following a progressive pass gives the following top five. I’ve decided to call them STOPP’s, mostly because it spells stop, but it did stand for successful turnovers following a progressive pass when I was messing around with the data.

PlayerTeamSTOPP’s p90
Caroline WeirMan City2.99
Chloe ArthurBirmingham City2.69
Chloe PeplowTottenham2.69
Lia WältiArsenal2.63
Maéva Clemaron Everton2.23

From the above, how impressive Caroline Weir has been out of possession is highlighted once again. She’s been the most active player in the division, which makes it no surprise she’s popped up so much here, but also doesn’t make it any less impressive. Her evolution from last season where she had City’s 3rd highest xG contribution, behind only top scorer Nikita Parris and Claire Emslie who left, is an intriguing one. She’s still been City’s 3rd most creative player, but that’s largely thanks to her strong set-piece taking ability, assisting 0.16 xA per 90 just from her corners.

When looking at the transition figures for the Reading section, Birmingham were also poor, which makes it interesting to see Chloe Arthur 2nd here. The Blues had the second-highest xG per game after Reading. Reading conceded 0.36 per game, Birmingham 0.33 per game and the next worst was Bristol City with just 0.20 per game. Arthur has popped up throughout all these figures, yet Birmingham still allow the opposition to complete the most progressive passes within 10 seconds of winning the ball back, with the 4th best completion rate in the league.

21-year-old Chloe Peplow is a nice name to see pop up. Playing at the base of Tottenham’s midfield, she seems adept defensively and she also has good numbers for completing passes under pressure – albeit with most of them going backwards or sideways. Even with that caveat, having a player capable of shielding the defence and keeping the ball under pressure seems a great combination, if she could add more progression to her game – she makes just under 3 progressive passes per 90 – it seems like she’d tick all the boxes for the position.

I haven’t seen much of Maéva Clemaron, so Lia Wälti is the last player I’ll comment on. Wälti gets a lot of praise for being an important counterweight to balance out Arsenal’s high number of attacking midfielders and these numbers back up why. She stops opposing sides progressing well and also makes a good number of recoveries in counter-pressing situations. On top of this, looking at xG of shots in transition situations and allocating the xG to the player who wins the ball rather than the shooter, the Swiss midfielder has the best numbers in the division. 0.26 xG per 90 is created from shots in the 10 seconds following her winning the ball back.

Another fun stat I found for Wälti involved the pass completion of opposition progressive passes within 10 seconds of winning the ball back when Wälti played compared to when she didn’t. Of course, there are lots of variables in here that have nothing to do with Wälti, like the opposition and the type of passes being played and whether it’s even Wälti making the intervention, but there is a pretty significant leap. In games where Wälti started Arsenal’s opponents completed these passes 47% of the time, without her they completed them 60% of the time. That 60% would put Arsenal as the 5th worst side in the division for these kinds of passes, the 47% would put them as the 2nd best.