Nate is an excellent statistician. But his problem is the polls he is using as an input into his model. If a poll is wrong, that makes the entire model suspect. This is the difference between his baseball work, and politics. In baseball, we know the players data (hits, RBI, HR, etc) to put into the model. In baseball, it would be like polling all the other coaches/players. How many RBI’s do you think Jay Bruce is going to hit this year? That is the fault with his polling model.
And DUH.... if the turnout swings D+3 to R+1, even I can predict it will greatly increase one parties chances.
And if Nate is proved wrong, he will blame the polls. ‘Hey I was just working with what I had and based on that I had a 125% chance of Obama winning’. He gets all the credit for being right and none of the discredit for being wrong.