Expected goal difference
xGD°An estimate from a model we built. Useful, but not a fact — every modeled stat's page says where it's weak.
Definition
A team's expected goals created minus expected goals conceded, summed over the scope. Tracks the balance of chances rather than the balance of finishing.
What it refuses to measure
xGD does not know the score state — teams protecting leads concede chances deliberately, and it reads that as decline.
What would prove this wrong
If final-third-entries difference predicts end-of-season points as well as xGD does, the model adds nothing over territory.