Originally Posted by cdcox
I put a lot of weight on the methodology. If 538 proves wrong, I don't think it will be by much, because the methodology is strong. If he did turn out to be way off it would be due to some fundamental change in the electorate. I wouldn't switch to unskewedpolls no matter what. A bad method that gets the right answer for the wrong reason isn't likely to be reliable in the future.
I did't find the National Review article that convincing. For example, they criticize the low weight given to the Rasmussen poll, but they didn't mention why it is given a low weight. It is because they don't include any cell-phone-only users in their polls That was fine in 2000, but this is 2012. That poll has a weak methodology and therefore gets a low weight. The National Review has a strong right tilt, so its not surprising that they don't like a method that shows their guy down. If they are going to criticize Nate Silver, they are going to have to at least step up to his level of nerdiness.
The guy who wrote the piece is @numbersmuncher on twitter.... Can you be more nerdy?