Agree with this.
Would love someone to tell me how the measurable metric for 'expected' is worked out.
Sounds bizarrely nonsensical.
Several different methods that all get to the same general outcome. I like this explanation: http://11tegen11.net/2014/08/07/expected-goals-2-0-some-light-in-the-black-box/
Basically, the xG model looks at match situation (open play, set piece, PK, etc), league level, scoreline, shot location/angle to goal, header vs typical shot, distance, and type of buildup (cross, through ball, counter attack, etc), and weights each of these based on how much they contribute to a goal, and how often.
<Nerd alert> For the good xG models, this is fed into a multivariate analysis based on a large set of previous goal attempts that generates a sort of equation that you can feed all the relevant info into for a particular attempt/game/season/whatever to determine the expectation vs actual outcome </Nerd alert>
Essentially, it weighs all the things that we might expect to contribute to a goal, based on how much they tend to impact actual goals being scored. For Sandro, we've seen from shot charts that last season he scored a lot of goals from tight angles--these shots by nature will have low expectations for conversion. Now, it could be luck or it could be genuine skill, so how he performs this season will go far to answering that question.