A question about XG and how it works

[Chipper]

Random question. FBRef has xG stats for women's football. Does anyone know if they use a different data set? One thing people often point out is that goalkeepers in women's football don't have the size or reach of male 'keepers so it can be a little easier to score.

I'm guessing they don't seeing as:

9 out of 12 teams in the WSL have scored more than their xG. 349 goals from 291.5 xG.
14 out 16 in Liga F have. 546 from 481.6.
10 out of 12 in Ligue 1 Feminine have. 402 from 341.
10 out of 12 in the Frauen Bundesliga have. 315 from 269.2.

It's only women's Serie A and A-League where half the teams have scored less than their xG. Even then there's more goals than xG. 261 from 233.2 and 405 from 395.4.

Looks like they need a new model for women's football. I only thought of this because I started considering if there should be different models for different standards of league. Women's football seemed like an obvious testing ground.

Should the Premier League use a different xG model to Eredivisie or MLS? I don't have an answer to that but I don't think they do as a rule.

 

[giorno]

So it is like predicting the weather.

It is a predictive model yeah. It was developed by gamblers for gambling purposes afterall

 

[PaulScholes99]

A problem we have when talking about xG is that instead of looking at the quality of individual chances, we have a tendency to look at a summation of xG - i.e 10 x 0.1 chances = a goal. Which is simplistic and just plain wrong. It is an analytical tool to help determine if a team is creating good chances and to determine if a player is capable of putting away high quality chances regularly.

Idk what you mean exactly, tbh. Of course it is simplistic if one says 10 x 0.1 chances = a goal, but on average it would be normal to score one goal out of these 10 chances. The same would be true for 2 x 0.5 chances. The only difference is, that the variance for the 10 x 0.1 chances is way higher than for the 2 x 0.5 chances.

It is bernoulli/binomially distributed (each chance is either goal (=success) or no goal (=no success)). For the 2 scenarios, with 2 x 0.5 chances, you would score out of these 2 chances in 25% of the cases 0 goals, in 50% 1 and in 25% 2. On the other side, with 10 x 0.1 chances you would score in 35% of the cases 0 goals, in 38% 1 goal, in 19% 2 goals. But while you simply can't score more then 2 in the first scenario with 2 x 0.5 chances, in the second with 10 x 0.1 chances you also score in 5.7% of the cases 3 goals, in 1.1% 4 goals, and so on (5 or more is then very very unlikely with 0.16% accumulated).

I agree that it is to simplistic to say 10 x 0.1 = 1 goal. But from your text i get the impression you think few high xg chances are better than more low xg chances, when the accumulated xg is equal, and i would not totally agree with that (maybe it was just my impression though).