Because we select the pivot at random, we are equally likely to get each possible split.
We can express the expected running time as:
k is the number of ``lows.''