Who will win in Colorado?

[SPC]

Udall. Colorado polls have a history of under sampling Democrats, and now that the state uses all-mail ballots the traditionally Democratic constituencies with midterm turnout issues should vote in larger numbers. Basically, if the Democratic candidate leads, is tied with the Republican, or trails the Republican by 2-3 points, odds are they'll win.

http://en.wikipedia.org/wiki/United_States_Senate_election_in_Colorado,_2010#Polling_3

You're proving his point. No poll in the page you linked after October 2 showed Bennet leading Buck. Bennet won by 1.7.

If you remove the Rasmussen polls, the polling average showed a Buck lead of 1.5, which is within the margin of error of the actual result. But yes, applying a similar analysis to 2012 of taking the average of non-GOP hack firms shows Obama outperformed by about 2 points, and only about 1 point in 2008. On the other hand, Udall underperformed by 3 points in 2008; thus I am not sure whether recent Democratic "overperformance" in Colorado is truly attributable to systematic bias by legitimate firms as it is to random chance. Even then, the adjusted polling average is still fairly good, not missing any election in the last six years by more than 3 points<sup>1

1</sup> Yes, 2010 Governor was an exception to this rule, with Hickenlooper overperforming by 5 points relative to Tancredo, but part of that was attributable to Maes getting more votes than anticipated in order to maintain Republican ballot status. Absent the three-way dynamic the polling average error would have fallen within 3 points.

[SPC]

Probably Udall, but it should be noted that in 2008, polls significantly overestimated Udall against Bob Shaffer, at the same time as they underestimated Obama against McCain.

http://en.wikipedia.org/wiki/United_States_Senate_election_in_Colorado,_2008#Polling

No, they didn't.

Unfortunately, that article was largley absent of polls taken in the last two weeks before the election. Here's a more updated outlook:
http://www.realclearpolitics.com/epolls/2008/senate/co/colorado_senate-556.html

Indeed, polling did show Udall performing about three points better than he actually did.

[SPC]

Again, I'm not too worried about Colorado. If Gardner loses, he can challenge Bennet in 16

Gardner's chances at unseating an incumbent would be better with a midterm electorate.

[SPC]

I deleted my previous posts when I remembered how idiotic it was to defend the predictive value of polling leads within the margin of error. I apologize for suggesting that statistical ties are somehow indicative of a Udall advantage.