Generally, it is as Temptation says.
There are things to improve for the future. For instance, a differentiated xG that takes into account the personal efficiency (based on the history of strikes from same positions) of each player is something to look forward in the future.
As for the game states, yes that is true. But that is the job of tactical analyst to take into account in a piece, and has no bearing on the method used for calculating xG
Finally, I would say there are different xG calculation models. No all are equally good. For instnace, understat is one of the worst in the industry. Opta and Statsbomb are top
PS: Let me add this again for people (like Serghei and others) that keep repeating this on purpose:
Total xG of a game says nothing, when you don't look specifically at the composition of that aggregate.
You always have to look at xG shotmap and look at the big (or at least semi-big) chances to draw conclusions.
That's over 0.3 xG or around that number.
Thanks, lots of that makes sense. And yeah, you give good explanations for most of the flaws I pointed out. Good point too that cumulative xG is more misleading and it's more about the bigger chances. I assume a lot of nothing chances would often accumulate a bigger xG than two very very good chances. So like you say it's more about xG for the individual chances.
I have some other thoughts I might put down on paper at some point.
Just one thing - you didn't reply to what I said about that Spurs game and the City game at the weekend, both against United (as examples).
I will just take the game at the weekend as it is fresh in the memory - I assume you watched it, or at least have seen highlights and saw the copious one on ones City either converted or missed.
How can xG come to the conclusion that their expected goals was fewer than 2? Haaland scored two chances that were very good chances, for starters. He also missed pretty much a sitter off the post (admittedly someone, I think it was Luke Shaw, was in the way putting him off and he tried to avoid him hence the miss) and Reijnders was through on goal too. That allied to a few other decent chances, and Foden's goal might have been low xG but still was a fairly good chance - how does xG come to the conclusion it was fewer than 2 xG?
What I am asking is, when you see the xG figures (for the sites you reference as good), do you trust them implicilty? Or do you ask yourself if it reflects what you seen in the game with your eyes (if you watched the game or seen the highlights).
My problem isn't with xG as a use, my problem is that some of the figures the models come up with don't seem to match what I watched. Admittedly I could be being biased in the entire eye test, but I think anyone who watched that game at the weekend would say an xG of 3 for City would be fair - not less than 2 xG
