Bad poll for Romney.
A tie in Gravis is basically a 2-3 point lead for Obama due to their house effect.
Perhaps.
I am just going by Silver's assessment on Gravis.
Oh, I do think on average they lean R. But where Rasmussen leans R because of methodological choices he's made, and is thus pretty reliably a little bit to the right on almost every poll he does, Gravis just seems to be bad at their job. Their average poll is a little to the right of the polling average, but sometimes they're drastically to the right, like when they had Smith beating Casey by 19 points in PA-Sen, and sometimes they're just a little D-friendly compared to the average, as with that Iowa poll with Obama up 4 (a couple of points better than the polling average).
If the only information I had about a race was that Gravis showed it tied, I'd assume that Obama was ahead by a couple of points -- unless he was ahead by 20, or behind by 4.
