> Split any number of people into any arbitrary groups and some of those groups will be higher than average and other lower than average.
Wait, what? Surely they should all be (sans a bit of noise) the same? (Split 1000 random numbers into two groups. The sums should be almost equal).