Go to 538. There are dozens of polls with outlandish projections -- both national and state polls. The key is to decipher which ones always lean one way, which ones have a track record, and which ones use proper methodology. That's basically what Silver's site does. He identifies the left-leaning and right-leaning outfits and puts them in a broader context, and he examines the underlying methodology of polls to explain their results. This poll in question seems to leap to Romney based on economic assumptions it has built in to its methodology, not based on polling numbers. That's fine, but it's a prediction, not really a projection. It's saying that the economy will dictate these results. That doesn't seem to hold up in the face of all the other polling data about how people are likely to vote, which 538 catalogs and incorporates into its projections. In the prediction game, Silver seems to cast the widest net and use the best formula to incorporate information, and thus usually ends up with the closest projections. At least he has in the last two cycles.