Champions League

Yannik

Senior Member
If it's the same kind of chance, why ever not?

That Spurs example is useless. Total xG being more or less even doesn't imply the match should have ended in a draw. It just means Bayern scored from chances that usually do not end up as goals or tougher positions..

But the point is: It doesn't measure that reliably yet.
xG values were originally attributed simply to the position from where a shot was taken.

events0.jpg


Over the years a few more factors were put in. 1on1s, headers or weak foot strong foot were added and inconsistently applied.

But too many factors are still not present in current xG evaluation
- individual ability
- how much time does the shooter have
- how much space does the shooter have
- where is the goalie positioned (here back to the Lewy example: Goalkeeper lies flat on the ground, unable to do anything except of hoping to be shot at, then Berisha on the other hand shooting from the SAME position, but against an upright standing Neuer who closed him to less than a meter. And still Berisha had a slightly HIGHER xG for his chance than Lewandowski had)
- were are defenders positioned (inbetween scorer and shot? Close enough to pressure the shooter?)
- does the shooter fully control the ball or not
etc.


There was a very decent article written about this over a year ago evidenced through the game Leicester-Wolverhampton
https://thevideoscope.com/2019/08/13/expected-goals-xg-vardy-jota-fpl/
 
Last edited:

El Gato

Villarato!
But the point is: It doesn't measure that reliably yet.
xG values were originally attributed simply to the position from where a shot was taken.

Yeah Im aware, in general the further away you are, the harder it is for an average striker to get it on target & score.

But too many factors are still not present in current xG evaluation
- individual ability

It most likely never will involve individual ability because it varies over time with hundreds of players depending on form.
Nor should the value vary depending on who's on the end of it IMO. You're only assessing whether a striker 'should' score irrespective of talent. Not whether an elite one will be more likely to do so against a shit keeper.

- how much time does the shooter have
- how much space does the shooter have

Would be nice to have. Pretty sure there is some manual adjustment on the end of stat producers. But I dont know for sure.

(here back to the Lewy example: Goalkeeper lies flat on the ground, unable to do anything except of hoping to be shot at, then Berisha on the other hand shooting from the SAME position, but against an upright standing Neuer who closed him to less than a meter. And still Berisha had a slightly HIGHER xG for his chance than Lewandowski had)

So why is xG lower? We're kind of debating here as to what goes into and what doesn't, but do you know what actually made it lower? Because from the sounds of it you're just miffed by the value because Lewa scores these in his sleep.
 

Yannik

Senior Member
So why is xG lower? We're kind of debating here as to what goes into and what doesn't, but do you know what actually made it lower?

Yes, Lewandowskis shot was 6m out Berishas was 5m out, so it was valued higher as explained above.
But the fact that Lewandowski practically just tapped it home against a goalkeeper lying flat on the floor - whereas Berisha had Neuer stretching practically 10cm before him - is not a factor considered in the current algorithm.
 
Last edited:

Birdy

Senior Member
Over the years a few more factors were put in. 1on1s, headers or weak foot strong foot were added and inconsistently applied.

But too many factors are still not present in current xG evaluation
- individual ability
- how much time does the shooter have
- how much space does the shooter have
- where is the goalie positioned (here back to the Lewy example: Goalkeeper lies flat on the ground, unable to do anything except of hoping to be shot at, then Berisha on the other hand shooting from the SAME position, but against an upright standing Neuer who closed him to less than a meter. And still Berisha had a slightly HIGHER xG for his chance than Lewandowski had)
- were are defenders positioned (inbetween scorer and shot? Close enough to pressure the shooter?)
- does the shooter fully control the ball or not
etc.

Apart from individual ability, all the rest have started being incorporated gradually.
Many models even today account for space, position of the goalie, position of defenders.
I guess the time the shooter has, the speed of the pass before shooting, the full control of the ball will be the last to get incorporated (they are the most difficult to account for)

Your point about individual ability still stands, but I think it's not science fiction to say that with the algorithm learning techniques that already exist, in the near future the models will be able to: track the record of a certain player and measure his effectiveness from all different positions (and in regards to all the factors you mention) and 'learn' the value to be assigned to each player individually (instead of the generic one).
Form is a factor, but these models (when they arrive) will be 'learning' constantly and thus adjusting continuously the relevant value.

IMO, it would be interesting to see for a given game both the 'should' according to generic values, and the 'should' according to individuated values.

PS: The article you shared focuses on the preparatory stages before a play gets converted into a chance. This is not supposed to be captured by xGs but other sophisticated stats
 
Last edited:

El Gato

Villarato!
Odegaard instead of Casemiro. Going offensive I see.

Some dude mentioned he has 2 yellows and they don’t want to chance him being out for Gladbach. Still, mad run for Modrić at this age.

Benzema, unsurprisingly, is shit. Means he’ll probably fall over and score somehow.
 

Home of Barca Fans

Top