Predictions

Stephen Green posted 7 different potential scenarios.  Is it a bad thing that I didn't pick a single one of them?

Here's my call: start Obama with 201, Romney with 191, as per RCP.  That leaves the following swing states, with electoral votes in parentheses:

Colorado (9)
Florida (29)
Iowa (6)
Michigan (16)
Nevada (6)
New Hampshire (4)
North Carolina (15)
Ohio (18)
Pennsylvania (20)
Virginia (13)
Wisconsin (10)

There's an outside chance that Maine's second congressional district flips from blue to red as well, but we'll keep it simple for now.  Let's deal with each of the above.

Put Michigan in Obama's pile, and North Carolina in Romney's.  217-206, Obama.

Give Romney Florida, and Obama gets Nevada via Harry Reid's dirtydirtydirty turnout machine.    235-223, Romney.

Here's where it gets to be a combination of art and science.

The biggest debate out there has been about whether you can trust the state polls or not.  It's been discussed at length, and people can make a good case either way.  No links in here because I don't have time to go find them -- I may edit this later to include some.

I find it hard to believe that state pollsters would intentionally pick stupid ways to weight the electorate in an effort to get Obama re-elected, and that they are more likely to be doing so out of pure laziness by referencing the 2008 composition of the electorate.  Averaging these polls helps reduce any bias or failures of any single poll... of course, Nate Silver and others may be reinforcing the bias with their own weighting of the credibility of each poll.  But an average should still be useful, right?

At the same time... I find some if not most of the poll weighting in this cycle to be ridiculous.  A poll that puts the ratio of Dems to GOP at 2008 levels is suspect, and polls that put up a greater spread than 2008 should not be considered in making predictions.  RCP and others include them in their averages, but they are undecutting the value of the averages in their polls by doing so.  People can, with some justification, claim that Republicans are ignoring the polls at their peril and dismissing them because they don't like the numbers.  I think it's equally silly to consider polls with sampling issues like the ones that have been spotlighted as credible. 

Jay Cost has posited, with some good analysis, that the state polls seem to be settling into a bimodal distribution which makes the averages somewhat less useful; effectively, this means that the average is unlikely to be correct, in what Sean Trende classified as a "two guys enter, one guy leaves" scenario.  That still leaves Romney behind in state like Ohio in the polls, but a lot closer than the average or a lot further. Basically, Trende is saying Romney is actually down  about 4, or down about a point.

I'll posit a Raje's Rant Rule on polls that I think is borne out by historical data, save for 2008 and 2000.  Generally, the polls undercounted the GOP candidate's share of the vote, such that I feel comfortable taking any state in the RCP average that is within a point and putting it in Romney's column, and think anything that's within 1-2 points is a total toss-up. 2000 was offset by a huge Dem turnout effort and the October surprise of Bush's DUI arrest revelation, while 2008 was a wave election.  I don't think any of those apply here, even if the Dem turnout is solid, because I think the GOP turnout is going to be strong.  And I say this even as I hear that the lines in my local precinct are an hour long right now, in one of the more Democratic counties in northern Virginia.

But looking at that rule means the following:

Hand Virginia, Colorado, and New Hampshire to the GOP.  In the meantime, send Pennsylvania to Obama, by a bare margin produced by Philly's turnout operation.  Pennsylvania will finally go red again in 4 years, just not now.  This makes the electoral college count 261-243, Romney.  As an aside, if Virginia or Pennsylvania go the other way in an early call, the game is almost certainly over.

Three states remain: Iowa, Wisconsin, and Ohio.  Only the last two matter -- in my scenario, either candidate can win Iowa, but they each need one of the others (and in Obama's case, both).

Now it becomes faith-based on my part. 

I don't know if Romney has the horses to overcome the Dem turnout machine in Ohio, and I'm not sure whether Scott Walker's impressive turnout machine in Wisconsin can pull off a second straight win in a state with a solid blue Democratic registration advantage.  But two caveats on apply here -- Bush made Wisconsin close twice, and Romney's campaign operation is a lot closer to W's than it is to McCain's.  As for Ohio, the GOP candidate always runs a little bit ahead of the national average (even McCain did).  Obama's camp is making hay of an impressive turnout apparatus in Ohio.  But the GOP faithful are now revved up to vote for Mitt -- they weren't before, but they are now.

So my pick here -- give Mitt the sweep of Iowa, Wisconsin, and Ohio.  295-243.  That's probably overly optimistic, but I like being optimistic.  I can be a pessimist afterwards if the numbers don't work out.

Now, I'm going to go vote.  Two words: Go Romney.