Only Rasmussen’s “likely voter” polls are accurate.
Even Ras should be taken with a grain of salt. There might be a D+1000 advantage in the populace, but Hugh Hewitt pointed out in his show that the response rates are way, way down.
Like, 1/4 of 2008 down with only 9% answering. My gut says that the people who aren’t answering aren’t a representative sample of the overall, thus creating skewed ideologies within the 9% who are answering.