I love how they think they can predict what 12 people can do based on what they did last week. 

For starters, Silver says that if TCU had been 4 last week, then Ohio State would be favored to pass them, but since they were 3, it is more difficult. But this assumes a poll mentality. As we all know, they revote each week and simply winning isn't enough to keep you in your spot.

Frankly, trying to use simulated computer analysis of this stuff is a bit like voodoo. At least with polling for elections, you have scientific methodology on your side and can aggregate. For this, they only have a single poll of 12 people, a week old.