Here's my final prediction.... Obama 311 - McCain 227
to recap my methodology:
I gave Obama any state where he has >50% in the weighted average of the fivethirtyeight.com polls and he has a +5 or more advantage over McCain. Where McCain has >50% he gets the state. Where both candidates are under 50%, I split up the undecideds 3:1 for McCain, assuming that the Bradley effect this year will be seen primarily not in voters who say they'll vote Dem voting Rep, but in those who say they're undecided going Rep.
I have eliminated tossups --- hey, I don't get paid the big money to waffle! As it turns out, according to 538.com there is no state where Obama has >50% and <5+ on McCain. So I don't have to call any tossups anyway.</p>
Similar to my last prediction - that gives Obama CO and VA, along with PA. McCain gets MO and now also gets FL by 0.3% and NC by 0.6% based on the undecided split. But Obama gets NV and OH. This gives Obama a total of 311 EV's -- and that's what I'm going to go with.
I would like to point out that MSNBC and CNN, who were still including tossups matched my 286 prediction up until this morning. Also, David Gregory of MSNBC used the same algorithm I used in his predictions after I made mine (I don't think he reads my diary sadly).
Now, a note about correlation. I've been a bit conservative in my predictions by giving McCain every state where he's over 50%, even if by a hair and by splitting the undecideds 3:1 in his favor. It has to do with my estimate of correlation.
If you flip 6 coins, one at a time, and you win a prize only if all six come up heads, you have a 1 in 64 chance of winning. Each flip is independent of the previous flip and so the six flips are uncorrelated. Thus, it is very hard to win the prize. But, if I taped all six coins together in a line, with all heads on one side, and flipped that line once, my chance of winning the prize has just zoomed to 50-50. That is because all the coins are perfectly correlated in that they are ensured of having the same result.
So it is, I believe with most of the battleground states in the industrial midwest/east. I believe there is a correlation (not a perfect one, but to some degree at least) of racial bias. That is, if McCain pulls out VA because of racial bias in the western part of the state, then I think that same effect will occur in Southeast OH and western PA. Maybe not enough to tip PA, but at least possibly so. Therefore I really expect that all three states are likely to go the same way, and because the polls are so close, it really could go either way. This is the one thing that makes me the most nervous about tomorrow. I am just hoping that Obama's lead in PA is large enough (+9) to overcome the racial effect.
With FL, CO and NV, I expect much less correlation because of the demographics. Therefore, I think even though all these states are still relatively close, it is statistically very unlikely that McCain could flip all three and the PA/OH/VA triplet.
So, in conclusion, if McCain wins the PA/OH/VA triplet which is still unlikely, then Obama needs FL or NC and NV. Either seems reasonably likely to me. Therefore, it really becomes exceedingly difficult for McCain to win this shooting match.......