Sorting by a random comparison function is a crappy way to shuffle.

crappyshuffle.py

# Sort [1,2,3,4] with a cmp function that returns -1, 0, or 1 with equal probability.
# Do it 10^5 times.
h={1:0,2:0,3:0,4:0}
for i in range(99999):
  a = sorted([1,2,3,4],
             lambda a, b: random.sample([-1,0,1], 1)[0])
  h[a[0]] += 1
h
# => {1: 60443, 2: 17374, 3: 10990, 4: 11192}  
# The first element of `a` was `1` 60% of the time!

# Even if the cmp function only returns -1 or 1, it's still bad:
h={1:0,2:0,3:0,4:0}
for i in range(99999):
  a = sorted([1,2,3,4],
             lambda a, b: random.sample([-1,1], 1)[0])
  h[a[0]] += 1
h
# => {1: 36170, 2: 13965, 3: 18735, 4: 31129}
# Now we get a decent number of 4's, but still nowhere near uniform.