David L. Burkhead

Speaking as a computer scientist... Note also that the numbers you referenced aren't all just 64-35, they're all 64.6 + or - 0.2, and 35.4 + or - 0.2. That's even more unlikely. This is what happens when you don't pay attention in math class. Likely some corrupt programmer put in a random factor that he thought would alter the numbers, without making it look like it altered the numbers. Like (brain-dead example) "If this ballot was for the opponent, and the random number generator is greater than 0.6, then change it.". But he didn't bother to do the math, or he would have realized that if you use whatever that random element is, over enough cases, then the result converges on 64.6 and 35.4. At least if I was going to vet the program that did this, then that's what I'd most expect to find. This is a typical undergrad mistake. Just because your algorithm has a randomizing element, that doesn't mean that the result isn't going to converge, over a large enough number of trials. Also note that this kind of manipulation almost always leaves a footprint in the data. Statisticians who specialize in analyzing voting results know what properties honest results should have. Unless the corrupt programmer really knew what he was doing, they'll spot the strange results, and the nature of the strangeness will give them a strong clue as to what the programmer did.