Hah, that's a good rebuttal. Well it would be an excellent estimator if it was unbiased (or at least had very little bias). But if f is an estimator for the true average of a function F where E[F] = 100 (for example), it's very clearly not unbiased.

It is a bit tongue in cheek.