Also, SC more likely to flip than MO? GA more likely to flip than IN? Really?
A lot of the weird juxtapositions are explained by the fact that the model adds in a bunch of uncertainty when there is no little or polling conducted, and is much more certain when polling has been conducted.
Anyway, Silver is a terrible pundit but a decent statistician. When he shows his work (like now) it's probably fine. When he pulls numbers out of his ass, like he spent most of 2015 doing, you shouldn't put too much stock in them.