Today's Football

[jamesedwards]

OK I confess to not being a student of such things as xG, I use my own algos but I can't see how the xG for Liverpool can possibly be 0.49 AFTER they scored a goal! Surely after the goal is scored their xG must be 1 + 'something' or does xG live in complete denial of actual goals scored?

xG is based on chances created rather than goals scored. So a speculative effort goal from outside the area might generate 0.1 xG while a missed penalty might create 0.8 xG.

OK, so the only way I can resolve this in my mind is the xG is saying this is from 'now' until the final whistle in which case we are both right The expected goals at the final whistle is 1 + current xG

No. xG is the expected goals based on the quality + quantity of chances created.

eg last night's game, the total xG was PSG 1.82 v 0.27 Liverpool, despite the final score being 0-1.

https://theanalyst.com/2023/08/what-is- ... d-goals-xg

[wearthefoxhat]

Comparably, I gave Liverpool 0.20 xG adjusted, before the goal..

Before the goal.png

------

After the goal, a little more reasonable.

Full time.png

OK I confess to not being a student of such things as xG, I use my own algos but I can't see how the xG for Liverpool can possibly be 0.49 AFTER they scored a goal! Surely after the goal is scored their xG must be 1 + 'something' or does xG live in complete denial of actual goals scored?

Those sheets I posted are the way I produce my own xG. I don't, (can't possibly), emulate the way it is scientifically calculated, so set upon a way of calculating my own, based on the standard metrics that can be found on sofa-score.

One those metrics I use is shots on target. As Liverpool only achieved 1 (and scored from it) then it didn't impact my baseline xG that much, however, the scientific xG should also (it doesn't) take into account an actual goal scored. I circumvent that by averaging my xG with the any goal(s) scored in the game.

[firlandsfarm]

xG is based on chances created rather than goals scored. So a speculative effort goal from outside the area might generate 0.1 xG while a missed penalty might create 0.8 xG.

OK, so the only way I can resolve this in my mind is the xG is saying this is from 'now' until the final whistle in which case we are both right The expected goals at the final whistle is 1 + current xG

No. xG is the expected goals based on the quality + quantity of chances created.

eg last night's game, the total xG was PSG 1.82 v 0.27 Liverpool, despite the final score being 0-1.

https://theanalyst.com/2023/08/what-is- ... d-goals-xg

I appreciate that James but once an event happens it has a probability of 1. So after a team has scored their Expected Goals must be "their current goals scored" + "expectation of future goals". To me it is utterly nonsensical that after Liverpool have scored a goal there is an 'expectation' they will finish the match with only 0.27 of a goal!

[sionascaig]

I appreciate that James but once an event happens it has a probability of 1. So after a team has scored their Expected Goals must be "their current goals scored" + "expectation of future goals". To me it is utterly nonsensical that after Liverpool have scored a goal there is an 'expectation' they will finish the match with only 0.27 of a goal!

It is indeed confusing but if there is a, say 0.2 chance of a goal and its missed then that event has also passed with O goals so using the same logic the expectation should be reset to zero.

Maybe think of it as "chances created in the past" from which a goal may or may not have been scored.

For me the issue is more: is the past any guide to the future )

edit - although from a goal perspective it maybe make sense to reset xG after a goal is scored to monitor the new state of game... I suppose you could do this by deducting xG when a goal was scored from the current value?

edit edit - a good example of how careful you have to be with this when there is a penalty, which has around 0.75 chance. That 0.75 will stay in the xG figure whether or not a goal is scored.

[henbet22]

Its called expected goals not expected goals plus actual goals.

PSV vs ARS xG 1.38 vs 1.91 so clearly needs some work.

[wearthefoxhat]

Its called expected goals not expected goals plus actual goals.

PSV vs ARS xG 1.38 vs 1.91 so clearly needs some work.

FWIW, my baseline xG, before the actual goals are factored in, indicated that PSV had the opportunity on stats, but the woeful defending/goalkeeping told a different story and gets adjusted to reflect that. Instead of a 2-3 score-line, its adjusts down to 1-5 as a fairer score than the final 1-7.

IMO, if this was representative of a league game, the form tables with the goals F/A and goal diffs, could be markedly different along with any predicted scores going forward. If, and it's a big if, this may constitute an edge over time.

PSV v Arsen.png

[Euler]

Couldn't make it up!

[wearthefoxhat]

Couldn't make it up!

Jimmy Tarbuck, John Lennon, Maggie Thatcher, Jamie Carragher, Kenny Dalgleish, your boys are taking one hell of a beating....(at HT anyway)

[firlandsfarm]

It is indeed confusing but if there is a, say 0.2 chance of a goal and its missed then that event has also passed with O goals so using the same logic the expectation should be reset to zero.

Maybe think of it as "chances created in the past" from which a goal may or may not have been scored.

For me the issue is more: is the past any guide to the future )

edit - although from a goal perspective it maybe make sense to reset xG after a goal is scored to monitor the new state of game... I suppose you could do this by deducting xG when a goal was scored from the current value?

edit edit - a good example of how careful you have to be with this when there is a penalty, which has around 0.75 chance. That 0.75 will stay in the xG figure whether or not a goal is scored.

Thanks for your explanation Sionascraig. I think my problem is that it is called 'expected goals' because if a team scores a goal I 'expect' them to have that goal at the final whistle and maybe score more. It seems to me to only be of value pre-off. It's value in-play is mathematically meaningless. And thank you for the example of the missed penalty ... what a nonsense!

To be of value in-play with the possible final score in mind a better calculation might be 'goals scored' + ('xG at the off' x 'Time Decay')

[firlandsfarm]

Its called expected goals not expected goals plus actual goals.

PSV vs ARS xG 1.38 vs 1.91 so clearly needs some work.

Sorry Henbet but if a team has scored a goal in say the 60th minute how can you in the 61st minute 'expect' them to score 0.27 of a goal? I can see the value of xG pre-off but there is no value in ignoring the actual in-play.

[henbet22]

Just re-name it shot quality metric (ignoring whether a goal is scored or not) and you will be fine.

[henbet22]

Its called expected goals not expected goals plus actual goals.

PSV vs ARS xG 1.38 vs 1.91 so clearly needs some work.

FWIW, my baseline xG, before the actual goals are factored in, indicated that PSV had the opportunity on stats, but the woeful defending/goalkeeping told a different story and gets adjusted to reflect that. Instead of a 2-3 score-line, its adjusts down to 1-5 as a fairer score than the final 1-7.

IMO, if this was representative of a league game, the form tables with the goals F/A and goal diffs, could be markedly different along with any predicted scores going forward. If, and it's a big if, this may constitute an edge over time.


PSV v Arsen.png

imo the eyecatching result is more of a reflection of PSV stuttering home form of late and Arsenal playing with the handbrake off as the Prem has long gone. Also the strength differentials in the two leagues. Nice work though fox always thought provoking stuff. Which is what its all about.

[firlandsfarm]

Just re-name it shot quality metric (ignoring whether a goal is scored or not) and you will be fine.

Yes but it's still a useless indicator in-play no matter what you call it! The value is meaningless nonsense once the actual goals exceeds the 'expected' goals.

[jamesedwards]

Just re-name it shot quality metric (ignoring whether a goal is scored or not) and you will be fine.

Yes but it's still a useless indicator in-play no matter what you call it! The value is meaningless nonsense once the actual goals exceeds the 'expected' goals.

PSG 0 Liverpool 1, suggests that Liverpool were the better team and perhaps they would be likely to win again if the game was to be replayed under the same conditions.

PSG 1.82 xG Liverpool 0.27 xG, presents a more accurate reflection of the game and would, arguably, be a much more useful statistic than goals to use when judging form.

[ShaunWhite]

Yes but it's still a useless indicator in-play no matter what you call it! The value is meaningless nonsense once the actual goals exceeds the 'expected' goals.

Always thought that about xG. Shoulda woulda coulda doesn't mean much.

But sanity checking my view with gpt I got ...
An elite set-piece taker vs. someone who balloons it every time will both get the same xG from a good position, but the actual outcome is miles apart.

For trading, xG can be misleading because markets move on actual goals, not theoretical ones. A team "deserving" to score means nothing if they’re wasteful, lack confidence, or keep running into an in-form keeper.