That’s not quite how it works though. Because we turn a low percentage chance into something that makes the keeper do something spectacular, it doesn’t alter the statistical probability of the chance.
So their goal is a tap in from a yard out, very high xg that our keeper doesn't get near saving (because he's busy fumbling around in thin air having just dropped the fucking ball) the wonder save from Dier's free kick has a teeny tiny X/G because of the inherently much h smaller probability of scoring a free kick from 25 yards etc etc.
In a nutshell, Great players (Kane, Son eg) doing great things is kind of helping us buck probability (Just look at Son's goal v Arse) whilst some shit finishing at the other end is helping us as well.
It's not impossible to buck those trends to the extent we have for the first 12 games for a sustained period, it happens occasionally, but generally you would expect us to adjust more toward the median, or get better and make those xg numbers better and adjust results accordingly.
Obviously as the fella says above, with most statistical metrics, the bigger the sample the more reliable/valid the outcome, but as a general guide (taken with context of the who's and how's etc) they are still a pretty good indices (probably the best right now) of figuring out without bias the quality/quantity of chances in individual games. And a great example is your (and most of our) perspectives of that game Sunday based around their keeper making 3/4 great saves and ours making maybe 1 ? The truth is, if Kane was heading a couple of their chances, or Son was shooting instead of Zaha/Eze they probably score or force fantastic saves.