Heh, this code is the most interesting of the lot since it *seems* wrong.

If the splice line had been

splice @array, rand(1+@array ), 0, $_ for map{ /^..(.*$)/ } <DATA>;
[download]

it would obviously have been fair. Now it seems to have a slight preference for left, so that makes you worry that e.g. maybe the last two elements don't get paired often enough. But that impression is wrong. The bias only causes the order inside one pair to be lopsided. The element inserted first becomes the last element in @array, and will remain last during the loop. The other n-1 elements will get a fair shuffle. So after pairing the result is fair, but the first element inserted will always be last in its pair.