You can hear that roundtable I did with David Weigel and David Freddoso for AFF here.

As I said, we were all asked for an "upset pick" for this year's elections; I picked Michael Steele in Maryland's Senate race.

Weigel picked Tom McClintock in the Lieutenant Governor's race in California, which he speculates may give him a leg up in the race to be Ah-nuld's successor.

Freddoso had some  interesting comments, seeing Michigan as a state that may be bucking the national trend as fertile ground for Republicans this cycle. He also has a fascinating look at some of the House races that are getting less attention, spotlighting the new rock-ribbed conservatives who may be the new faces in the House GOP.

Weigel wrote in shortly after my Burns comments, noting:

I give you this snippet from the Hotline, from 10/19/2000.

A Great Falls Tribune *poll;* conducted 10/13-15/00 by A&A Research; surveyed 402 likely voters; margin of error +/- 5% (Great Falls Tribune, 10/19). Tested: Sen. Conrad *Burns* (R), rancher Brian *Schweitzer* (D), and Reform Party candidate Gary
Lee.

General Election Match-Up
*Burns* 48%
*Schweitzer* 40
Lee 1
Undec. 11

Burns went on to win the race by 4 points after outspending Schweitzer 2-1 and as Bush was carrying the state with 58%. He's probably going to outspend Tester by the same margin, but Burns has never been behind like this, so close to the election, in a bad cycle. And his negatives have never been higher. Get the fork, stick it in.

That's worth recalling (although I note the 5 point margin of error in the poll makes the 8 point lead less useful than it first appears); having said that, the registered voters polls look terrible for Burns; the likely polls show him still down, but not insurmountably. I guess the question is, when the Registered Voters numbers show the GOP down significantly more than the Likely Voters, is it a sign that a lot of first-time voters are energized to vote for the Democrats? Or does this mimic previous years that have shown the Democrats doing better among un-likely voters, voters who never seem to show up on Election Day?

It's still an open question, that we won't quite have answered until the votes are counted. Erick at RedState noted a great example of how polls can't effectively measure the likely turnout from 2002:

On November 4, 2002, one day before the election, the Atlanta Journal/WSB-TV poll showed Roy Barnes at 51% and Sonny Perdue at 40%. The internal polling of the Barnes campaign showed similar results. So did the internals of the Perdue campaign, except the Perdue campaign was no longer looking at its internal polling. Instead, the campaign was looking at its GOTV ground game data and knew Perdue would win.

On November 5, 2002, Sonny Perdue beat Roy Barnes 51.4% to 46.3% with the Libertarian taking 2.3% of the vote.

If a hitting-all-cylinders GOTV ground game - and Georgia 2002 was one of the all-time greats - can make an 11 point lead the day before the Election meaningless, it means that most of the leads in polls that we're seeing this year are meaningless.

By the way - how much do you think the GOP is studying its GOTV in Georgia's 8th and 12th Congressional Districts?