Nate Silver discusses the charge that the Rasmussen Poll, which is generally favored by Conservatives, is too biased to trust. I think it is…but I agree with Silver that the real bias may be in how selectively it's quoted. I don't think any one poll has a monopoly on the truth but if you only quote one and not a range, you oughta consistently quote just that one and not switch pollsters because you like someone else's outcome better one moment.
Silver, who knows more about this stuff than anyone I've come across, says that Rasmussen did a pretty good job of calling the 2008 election. My impression is that he's measuring their final polling against the final results…and that Rasmussen showed a much more favorable picture for McCain/Palin until just a few days before we all filled up those ballot boxes. Then, suddenly, they "noticed" a huge rush of Democrats who abruptly became Likely Voters and their polls suddenly fell in the line with all the others. Was this not the case? I seem to remember Mr. Zogby — whose polling is so far off that I don't believe it even when it yields results I like — on some news show arguing once that, based on the outcome of some election, his poll was "on target." And in rebuttal, someone else noticed that it was off for weeks before and only became "on target" about halfway through Election Day.