From an earlier thread:
"The whole point of "expected goals" or "xGoals" is to get rid of the luck factor. If you shoot from one spot 5 shots, and 5 go in, your actual shooting percentage is 100%. Should we derive from that set of data that you will always shoot 100% from that spot, or should we calculate what everyone else is shooting from that shot, to calculate your xGoals number?
If we don't think you will shoot 100% from that area, we can look at what everyone else has done in that area, and then determine what is the league average percentage for that shot to go in. Then we have a situation where the actual production is 100%, but the value we should give more weight to is "only 2 shots go in from that area in general for every other player". So the xGoals number for that performance is 2 instead of 5.
And that makes it a better stat, especially when you fine-tune it to the player in question (and the goalie, and the pass that he just received, and how many defenders are screening, and probably what the player ate the last day with today's technology).
And then later, if we figure out that you got to that same spot, but did not score in 10 shots straight, we can determine that you should be worth at least 4 "expected goals", even though you scored none.
And that is what makes analytics useful. It tries to take the luck (or random variation) out of it."
So it's basically: What is the probability of any given shot going in if we do a league-wide average result of said shot.
There were serious doubts about Laines shooting ability because his shots are not supposed to go in.