The OG individual defense post is definitely from Thom Lawrence, called PATCH. He was one of the great posters ever in football analytics and a really nice and smart guy also, sadly been blocked for years on twitter so I don’t know what exactly he’s up to nowadays. But please read that post, it’s great and lots of what will follow is attempting to recreate it in some way.
I tried to recreate it a while ago with something catchily called DUTRAW that somehow did not become a mainstay on Match of the Day. It fell by the wayside a bit over time so we go again.
The downside of PATCH is the numbers are not available to me, or to anyone else. So we have stuff like xG conceded while on pitch and basic defensive numbers (pressures, tackles, blocks) which we’ve all agreed for a long time don’t give a full picture. So I wanted to add to my canvas of metrics and stats and deepen our hues of understanding with an extremely convoluted metaphor that will confuse more than enlighten. I want to be able to speak more deeply about individual defenders, so I want to develop statistical reports that can at least tell me now the inverse of what I often like to measure with the ball.
Concept
What I wanted to know was
how well do players stop the opponents from progression volume
how well do players stop opponents from efficiently progressing (yards gained per ball loss)
and how well do players stop shot production through their zone
notably what is missing is how well do players break up play (that is easy to measure and I already do so) and how much xG do possessions through their zone create (this I think will not tell us too much as high value chances (high value chances) tend to dominate this metric in a way to make it even less repeatable than the metrics I am creating now of unknown repeatability.
So the parts of this metric
Define the zone
the question is whether to determine the zone simply as where you defend or where you are the entire game. Right now I started with where you are the entire game, because it gives a bigger sample. Some players have 3 or 4 defensive actions the entire game, and go huge chunks of time without one, which makes it tough to map out. I think conceptually this makes some sense in the era of counter pressing, if you are passing the ball standing 45 meters from the opposition goal and lose the ball, when the opponents rush 20 meters upfield past you, you shouldn’t be off the hook because your defensive zone is actually 30 meters back should you?
So I am defining the zone as any Block (field is broken into 25 equal chunks) that a player has at least 10% of his touches in. For example here is Allan Saint-Maximin’s defensive zone against Chelsea…Blue is what he covered the whole game and orange he covered sporadically…during those times he was in that zone he gets the credit or demerit for what Chelsea did against Newcastle.
If we tried to find his zone from his defensive touches only it would be difficult as he had just 9 defensive touches counting ball recoveries in that. While you sometimes get some weirdness like ASM switching over or touching ball in sort of left defensive midfield, I think that makes more sense than just giving players assigned zones based on position. Let them show you their position.
Define progression
Here it gets simpler. I just look at the average teams progression through each zone and compare what the opposition gets. So if the average team progresses through Zone 12 at say .8 Fields Gained per 90 and Brentford just went through at 0.7, thats +0.1 fields gained for everyone responsible for Zone 12.
Fields Gained is my favored progression metric: it is simply % of field toward goal progressed. So if you start 50 meters from goal in a straight line from middle of goal and end up at 35 meters from goal, you’ve gained 30% of the field toward goal so get credited with 0.3 fields gained. There are no negative fields gained as I think backwards passes do not give away the yardage gained generally as the team usually keeps some sort of forward control as backwards passes are generally completed at 95+% rate.
Define Efficiency
Look at how often the opposition loses the ball from that zone compared to how much progression they gain (fields gained divided by ball loss). Compare that to the league average through that block…not too difficult again
Define Shot Activity
Look at how often passes and carries through the zone lead to shots taken in the next 20 seconds and compare how often the touches allowed in that players zone lead to shots in the next 20 seconds. So if a team has 20 passes or take ons through a zone and 3 of them lead to shots in the next 20 seconds but the league average rate was just 10%, they allowed 1 more shot creating action than average. Sum this up for all zones and all times and you get a total number.
Weaknesses of this, room to improve
Lots of what you do defensively is not factored in here: center backs in particular might be valued for their ability to prevent high value chances by turning what would have been a .25 xG shot into a .07 xG shot. Blocked shots, tackles, interceptions, pressures, etc are not directly counted in here but are simple enough to track separately.
Zone definition can be haggled over and likely improved.
Speed of opponent attack can be factored in.
Is it repeatable at all? Right now I’m just looking at current season data.
Current Season Data
Let’s look at a few of the current season breakdowns…I’ll just paste a table of percentiles here
starting at AC on we get the defensive metrics and percentiles. Click around and find out
A few standouts are
Conor Gallagher, Rodri, Joao Cancelo, Ben Chilwell, Mason Mount, Ruben Neves, Bernardo Silva, Federico Fernandez, Caglar Soyuncu.
Those that standout in a bad way: Junior Firpo, Curtis Jones, Dele Alli, Sergio Reguilon, Adam Webster, Jadon Sancho, Cristiano Ronaldo, Luke Shaw, Jonny Evans, Kelechi Iheanacho, Martin Odegaard, Vladimir Coufal, Alex Iwobi, Joao Moutinho.
Not sure exactly what all this means but we will keep working through it. Ideas, comments, improvements, suggestions, criticisms, questions, etc comment below.