Ahem, sorry. I can see where my thinking isn't complete.
The math is right, but the setup is wrong. If you could walk up to 50 static drawings where your own inputs didn't matter then yes, this would be right and yes, I am wrong. But when you drop 50 into one slot people will be less likely to put their codes in that slot as opposed to the next slot or the previous slot. That's why the numbers for each time slot end up close to the same ammount (depending on time of day). So you have to do the odds of 50/800 vs. 1-(799/800)^50 The true answer is somewhere in between though as I can see that neither method accounts for the reality of randomness.
So the conclusion is simple... Getting into an off-peak time slot is more important than how you spread out your codes. I'm glad I read that post cause I thought I had this thing all figured out. Thanks "exint" for copy/posting that.