Champions League

Richard.H

Senior Member
As shit as we are, we would have beaten Lyon last night.

giphy.gif
 

El Gato

Villarato!
Cornet shoots the 1-0 into an empty goal but since it's outside the box, xG accredits it with 0.1 or so.
Dembele is in a 1on1 with the keeper, but again it's a outside the box and it gets a 0.3/0.4.

Realistically this alone was a combined 1.5, unless you play on a level where your strikers can't hit a cows arse with a banjo.

It wasn't an empty goal, he did have to go around the keeper, even if it is essentially a FIFA target practice.

Isn't the distance factor in the description of the xG metric anyway?
 

Birdy

Senior Member
Because xG is flawed as fuck, and not processed enough to be unironically used to make a point. It accounts the position of a shot and gives it a value based on past shots going in from there. It doesn't take into consideration things like header/weaker/stronger foot finish, or the position of the keeper or defenders.

events0.jpg


Cornet shoots the 1-0 into an empty goal but since it's outside the box, xG accredits it with 0.1 or so.
Dembele is in a 1on1 with the keeper, but again it's a outside the box and it gets a 0.3/0.4.

Realistically this alone was a combined 1.5, unless you play on a level where your strikers can't hit a cows arse with a banjo.

Not quite so...
It's not at all solely based on distance.
Most models include positions of defenders, keepers, speed of pass, angle. And they are getting better and better.

Cornet goal is not as easy as people think, and Dembele's goal outside the box is accredited just the right value.
It's a probabilistic model, it doesn't have to be 1 or 0.9 to say it's a clear chance.
Any chance with value 0.3 and above is a clear-cut chance
 

Morten

Senior Member
xG doesnt factor in the quality of the players, does it?
Doesnt matter what xG says if the player in question is awful at finishing.
 

Tackle

Senior Member
xG are a bs indicator for now.

It is not very good in isolation or over a small sample size. Certainly should not be used to read into single matches. Over the course of a season however it can provide insight into the creative output and finishing ability of teams. For individual players it can be a decent measure of a striker's movement and composure in front of goal. Jesus is the perfect example. Excellent movement, below par finishing, xG bears this out. The imbalances @Yannik was talking about tend to even out in the long run. For every shot overrated by the stat, there will be one underrated by it.

xG doesnt factor in the quality of the players, does it?

Doesnt matter what xG says if the player in question is awful at finishing.

That is what makes it particularly useful. Comparing the xG of a player to their actual goal tally is generally a good indicator of how good their finishing ability is. Problem with the stat like many others is that people tend to misinterpret it and twist it to fit their own agendas.
 

Yannik

Senior Member
Not quite so...
It's not at all solely based on distance.
Most models include positions of defenders, keepers, speed of pass, angle. And they are getting better and better.

Cornet goal is not as easy as people think, and Dembele's goal outside the box is accredited just the right value.
It's a probabilistic model, it doesn't have to be 1 or 0.9 to say it's a clear chance.
Any chance with value 0.3 and above is a clear-cut chance

But the Opta one everyone refers to doesn't include that. True, headers or shots are indeed already being differentated, but they don't use some high precision multimillion dollar software that automatically tracks and processes pretty much everything on the pitch. And also don't have weak foot data for most players either. And noone else has any type of technology like that either. Opta doesn't give values based on how much time or space a player has on his finish. There's simply that dude Mark from Costco they pay a Pizza and 30$ the evening, to press a big red "is it a big chance" button, that lifts the xG in a pre calculated way and that is all the magic so far in 2020. All of the "complex" factors of these stats are simple selected by a dude who pushes all sorts of buttons, then all that gets punched into an algorithm that throws out the most inconsistent of values. And the issue being they have 50 guys doing that with inconsistent evaluations of what a big chance really is and how big it really even is.

xG.png


This is xG map for Leicester vs Wolverhampton a year ago. The score was 0-0, so admittingly the xG seems pretty fitting right?
That's until you watch the match.


I will add timestamps:

3:48 Vardy chance xG evaluation: 0.00 Because it was not a registered shot. No shot on target means no xG registered. The analytics might say that a goal wasn?t expected for Vardy, but the eye test sure does.
5:00 Jota chance xG evaluation: 0.00 Same reason for Vardy. Yeah he stumbled there, but still.. it isn't a shot, so xG says it wasn't a chance.
5:41 Jimenez chance xG evaluation: 0.17 Well admitted, this shot was blocked but Jimenez could literally square it to Jota for an easy tap in.
8:25 Jota chance xG evaluation: 0.3 I mean with all due respect to Jota and him apparently not being particularly good with his finishing from what I can tell in this game, but another issue here is that these stats do assume all players are born equally. Messi or Ronaldo get the same 30% score from there as Jota did. And the same goes for the goalkeepers they face, they are all equal in Optas eyes.

There was a total of 0.47 so a 47% probability for all these chances I just named to be scored. They'd have expected almost 1(!) goal if I had named double as much chances of this calibre here. I mean, common..

It all in all just doesn't really come out realistic. If anything even, they would tell you how good the actual finishes were, not the chances as a whole.
The whole xG thing is an interesting project, but it's still in its baby steps.
 

Home of Barca Fans

Top