I actually cover that in the post (well, more or less - I talk about hashing to remove bias, after talking about the 'min element' algorithm. According to the papers cited, though, taking the count of leading zeroes is more space efficient, allowing you to have more buckets in the same amount of space.