It's two-fold, really. The first is that its sample skews so relentlessly Republican it can mislead at times; the second that its results are often used very conveniently by the rightwing blogosphere to push back on difficult narratives. Like, er, that Obama has had a hell of a few weeks of gains, and is now leading solidly. So go to the polling table. Which one of these does not belong?
I'm not saying they rig anything. Or that they did not have a good record in the last election cycle. I just don't trust them enough – and the news cycle spin they always seem to be focused on – to include them in my poll of polls.
The Dish 