Upside down explanation to Leave-one-out cross-validation in section 3.4.6

[shinhongwu]

Leave-one-out cross-validation would have very low bias since its analysis set is only one sample apart from the training set.

I though "Leave-one-out cross-validation" should have very "high" bias instead of "low" bias. Am I wrong or is it a typo?

[topepo]

The bias increases as the amount of data held out in the assessment set increases. That's why the bootstrap (with leaves out about 36% out on average) has such a bad bias problem.

Section 10 of this paper summarizes it well (where "cross-validation" in this quote is LOO):

It's more complicated that this on the variance side IMO but pretty clear about the bias.