QuincyAbeyie But when this just keeps happening, surely there must be some reason for it other than bad luck.
Funnily enough, I read an analysis today that helped me to understand this distortion for this particular match. It seems to lead to a more general "blind spot" in xG-based interpretations of matches.
Potter selected a tinker-man's back three of Cucurella / Koulibaly / James, having trained on Friday with a different line-up. So Villa went ahead through Watkins' early goal after a big defensive miscommunication.
With De Zerbi coaching Villa into a more defensive posture soon after they went into the lead, Chelsea had more latitude to push up and create chances for most of the match. That went on because their hapless forwards never actually put any of the chances away and Martinez played a blinder. Meanwhile Villa ended up pulling further ahead.
The broader conclusion (I think?) is that xG accumulated from a losing position will often be less analytically important—for example whenever the side that's leading is trying to shut up shop. The dynamic will tend to go further if a side is losing by two or more goals.
Similar logic applies to measures like field tilt. All these metrics are affected by the current scoreline in a range of different ways contingent on tactics. I think I've read these ideas before, but there must be some stats folk out here who've got a more systematic way of looking at it.
