People keep missing my sarcasm... I am well aware of the problems of basing virtually anything on two observations. I'd be a poor economist if that weren't the case.
Extrapolating the change from PPP's last poll, Roberts should take the lead on November 10th.
This is generally a very bad idea, from both a common sense and statistical standpoint.
Really?
Of course it is in this context. The slope up when where one is tends to increase as you close the gap. That is of course related to fundamentals. Booker in NJ has lost a ton of points in lead, a rather steep decline. Will he lose based on extrapolations? Of course not. Now Virginia, where the same thing is happening, with different fundamentals, is a tad more interesting. 
I would imagine that any state "wants" to regress to its conditional mean more or less and the creation of "momentum" can create a positive feedback loop to accelerate that process. There is more than ample reason to suggest that Roberts could accelerate his rise moreso than Orman could do in a reversed situation. I was merely pointing out that the here-to-date closing would not be sufficient for a Roberts victory.
