This picture finally looks better:

[Betty Karlson]

http://www.electoral-vote.com/

For days it had been forecasting a probable net loss of ten seats for the Democrats (which would land us at 49), now it's climbing back up at 51. If you click on the option to see the picture without the Rassmussen polls, the picture become even (marginally) better.

I'm not saying we will lose eight seats! I'm saying the avegare trendline is saying we are rebouncing! Now let's rebounce to the point where it drives the GOP mad.

[Stuart G]

1. K and R

[deutsey]

2. The map definitely looks better with more blue in it n/t

[Baalath]

3. Looks better but...

I worry it is overly optimistic for NE Ohio. I worry about both the Sutton/Ganley and Kucinich/Corrigan.

I saw my first Sutton lawn sign today and I drive around in her district every day. Usually Dennis signs are everywhere in his district, they are scarce. Dem signs are lacking for all races here and I am VERY worried about all of the races.

I think state congress will stay dem.

[uponit7771]

4. rAssMusen has been an outlier for years, why are they still counting them?!

[karynnj]

5. It is beginning to look better

and when you look carefully at WV's trends, I bet that will soon be leaning blue. The underlying information on the PPP poll really could show it moving that way. The reason that the D was behind was people want a R there, so they were willing to go with someone they liked less. But, his unfavorables are growing - and they may dislike him too much.

[Betty Karlson]

6. there appears to be more to it than just that!

It's about phones, if you can believe it:

http://politicalwire.com/archives/2010/10/13/is_cell_phone_bias_skewing_polls.html

[karynnj]

7. It could well be - a sample needs to fully represent the population

Within every demographic group they control for, an assumption is made that the people not included in the sampling frame (the list they can sample from) are the same as those on it that they actually can reach. In addition to landlines, even the fact that with caller id, there is a higher non response rate makes the assumption that the sample reflects the full intended population shakier.

Then the likely voter models are as much art as science. One thing I read questioned whether the model which already includes the fact that the type of people likely to be Republican already do vote as a higher rate might have modelled an "enthusiasm" factor that is to high, given they are already assumed to be very likely to vote.

[Betty Karlson]

8. kicked, because it looks even better today

the GOP is now at 47 seats, with Dems at 50 and 3 seats too close to call. (Without Rassmussen, it's Dem 52 versus GOP 48.) Now take in that the pollsters may be undersampling the cell phone brigade (as in: they only call land lines), and take in that the cell phone brigade is leaning Dem: I'd say we would lose only five senate seats of the elections were held today.

And the GOP lead keeps evaporating by the day, even in the land line biased polls, so let's keep this going until their gains are as naught.