Not quite. Each shot’s xG is a probability, so they run a Monte Carlo simulation (with several thousand attempts), to work out the probabilities of scorelines based on the probability of each individual shot. From there, let’s say that 80% would be wins, 10% draws and 10% defeats, that would work out to 2.5 expected points, from a maximum possible of 3.
Understat’s model for xG is somewhat limited, it counts a shot and a rebound from the same move separately, which doesn’t reflect overall chances created/granted as well as models that average the xG in that kind of scenario. It also doesn’t take into account goalkeeper positioning and then there’s non-shot xG, which is about the times where a dangerous situation doesn’t result in a shot being taken. Think of a low cross across the six yard box that the striker isn’t quite alive to and misses the ball. That wouldn’t register as xG but should register as good chance creation.
All that said, the eye test would back up the expected points table pretty well. City are a fair way ahead, with the next five teams (Arsenal, Newcastle, Brighton, Liverpool and ourselves) being pretty close to each other and then another big gap to Brentford, Tottenham, Villa and Chelsea.