About three weeks ago Fbref dropped a massive update to their site releasing a whole slew of new advanced data courtesy of Statsbomb. One blogger called it “the biggest moment for public analysis since at least 2015.” I can’t argue with him. It’s really hard to explain how big a deal this is to nerds like me.
The newly available public data gave me an excellent opportunity to fine tune and update my radars for analyzing players. Over the past few weeks I’ve been collecting tons of data from Fbref, revising calculations, and upgrading the radars.
The template for all of them is still from Statsbomb. They’re most recent radar primer can be found here. For the most part, I kept the stats as similar as I could, but in some cases I made some changes.
That’s what I’m writing this for, to be a reference for explaining the changes I made and what the new stats I put in are.
For any stat that I used solely Statsbomb data, I was able to use the same parameters on the radars as they did. But for several others I didn’t have exactly Statsbomb’s data. In those cases I took all the data that Fbref has on the top five European league’s and made the calculations for the 5th and 95th percentiles myself. On the (rare) occasions that I still had to use understat, I did the same thing.
I’m still tinkering around with some of them, trying to make them as best as they can be, but I’m pretty sure we’re very close to the final product.
Here I’ll explain all the stats that I use that are different than Statsbomb’s. Anything I don’t mention can be found in Statsbomb’s original primer. Let’s get to it.
Shot Touch % – This is the number of shots a player gets as a proportion to their total amount of touches. Statsbomb has this too but if you’ll notice, on Statsbomb’s radar, the numbers actually get lower as you get closer to the outside.
My original thought was more shots is good. I want my center forward getting shots. I reached out to someone at Statsbomb inquiring why this was. The answer I received was “directionality was a decision made by the CEO due to his preference. The thinking was a lower number would indicate a less one dimensional player. I personally would maybe lean the other way. YMMV.”
For the record, when I did the calculations the players that came out with the highest shot-touch % were Harry Kane and Lionel Messi. I don’t think they’re one dimensional so I made the call. Higher number = better. I want my strikers getting shots.
Deep Progressions: Statsbomb defines those as any time a player progresses the ball into the final third via a pass, carry, or anything. I don’t have access to all that data, I only have access to how many progressive passes they make (progressing the ball 10 yards further than it’s been in the last six passes) so my Deep Progressions are just “Progressive Passes per 90.” Need to change the title.
Turnovers/Touch: Turnovers as a proportion to the amount of touches you take per game.
I skipped over this for the strikers but it’s the same concept here. Statsbomb and most other places do turnovers per 90, but I don’t think that’s a great stat so I changed it up.
For example, player A turns the ball over 3.48 times per 90. Player B turns it over 2.96 times per 90. Player B has better ball retention right? But it turns out, Player A touches the ball a lot more, 92.61 times per 90 compared to Player B’s 63.39. Therefore wouldn’t it make sense that Player A turns the ball over more?
Therefore I decided to look at how often do you turn the ball over per how many touches you take. I use this stat throughout instead of turnovers.
OP SCA: Open play shot-creating actions.
I scrapped open play expected assists because I thought SCA is a better stat overall. Fbref defines SCA as “the two offensive actions leading directly to a shot, such as passes, dribbles, or being fouled.”
That makes a lot of sense. Creativity isn’t just made with a pass but sometimes with a dribble. More importantly it accounts for the previous two actions, not just the final action before the shot. A lot of times a play wouldn’t happen if it wasn’t for the pass before the pass. Just look at this Mason Greenwood goal from the pre-season.
This goal doesn’t happen without Paul Pogba’s initial pass, but with xA he wouldn’t get any credit for it. I kept it to open play because for attackers, Statsbomb keeps their xA to open play, probably because not everyone takes free kicks.
Defender radars are much less useful because their positions are so system based. Some teams have their fullbacks contribute heavily towards the attack, while others don’t contribute at all. This is entirely from the Statsbomb template. Their last update added pressure stats and dropped crossing stats – which are valuable but are also system based.
The radar obviously skews more towards defensive stats, which is why Wan-Bissaka comes out looking much better than Alexander-Arnold.
Clearances to teammate: Pretty self-explanatory. How many times did you clear the ball to a teammate and retain possession as opposed to just hoofing it out and letting the opponent mount another attack.
Pressured Passes: Same concept. How many times did you pass the ball while under pressure instead of just hoofing out a long clearance.
I fully anticipate these radars to continue evolving as we go forward and I’ll do my best to keep this page updated whenever that happens.