• Hi Guest Just in case you were not aware I wanted to highlight that you can now get a free 7 day trial of Horseracebase here.
    We have a lot of members who are existing users of Horseracebase so help is always available if needed, as well as dedicated section of the fourm here.
    Best Wishes
    AR

XG technology

I 'm new to this so can you help.
I 'm trying to produce a powerful odds calculator for football matches. The objective is to predict better than the bookies, or at least a little better.

I know maths is not everything in football, because things off the pitch always happen and when we bet we must always be on our guard.
For example what if our friend Pep decides to field an amateur squad in a match ?
Does n't that throw maths and statistics out of the window for that particular match ? Yes it does and things less serious than that can also have an effect (such as a family row involving the goalkeeper maybe).
But nevertheless we must have a strong basis and build our attack from there.

So I tried using Elo Ratings and also using the Poisson method.
Elo ratings performed better and I reckon it is superior to Poisson.
You can find details of these methods in wikipedia, other sports websites and in mathematical papers from Academia.

But something better is needed.
At first I was confident that by taking account recent form I could improve things.
This proved wrong. Recent form -the various WWDLW, WWDWW sequences you see around- produce exactly zero additional information.
It turns out that other people who tried came to the same conclusion.
It appears "recent performance" is already factored in and so adds nothing.

But xG does something more likely than not, is the consensus of opinion.
xG is essentially the expected goals in a match.
So for example if Liverpool beat (today's) Manchester United by the single goal of the match, it earns them little in terms of Elo ratings. In terms of Poisson theory averages the same.
But xG looks at the details of the match. How many opportunities did Liverpool create ? If they did n't then it was indeed a drab performance from them but if they actually did then it is something positive. So therefore we have extra information -hidden to the elos and to Poisson- and it is extra information we are looking for.

How is xG precisely defined however and how is it calculated for an entire match ?
More importantly where can I find histories ?
For elo ratings and results we have histories, going back to 1955. I laughed looking at their 1955 ratings for the teams, as I know a few things about those years. Of course that does n't matter, because the ratings progressively correct themselves and we are in the year 2025 now.
But for the xGs where can I find stuff ?
 
Last edited:
XG appears to be an imaginary scoreline taking into account chances.
So it is deemed to be more representative of the teams performances than the real scoreline.
Data exist from 2014-15, some of them compiled retrospectively, but we don't seem to have day to day xg scores - like we do have the real scores from the various stats websites.

For the last two EPL matches (16 May 2025) I read this:

Chelsea - Man. United xg score 1.30-0.34
Aston Villa - Tottenham xg score 1.68-0.62

In reality Chelsea won by 1-0 and Aston Villa won by 2-0.
So I find it strange Aston Villa's xG is less than 2. Don't they count the real goals ?

The xG metric is considered a better statistic for prediction purposes.
However it is not necessary.
From the existing data, real scorelines, ratings, I have now managed -after some effort- to derive formulas beating the bookmaker benchmarks and computationally feasible.
Fairly soon I will publish more about it in my blog and let you know,
 
Back
Top