 Quadratic risk

The risk of an estimator under squared Euclidean error is its expected squared distance from the true parameter. Pointwise risk domination can be strict even when the supremum risk agrees, because the supremum can be approached at parameters tending to infinity.