Lesson on statistics and probability.
This is how things looked in August 2012.
States in medium-to-strong red either had been polled, or were understood to be sure things for Barack Obama in 2012. States in medium-to-strong blue either had been polled, or were understood to be sure things for Barack Obama in 2012. States in gray showed signs of being swing states in 2012. They were legitimate swing states in 2008, the republican nominee who won the state won by a lesser margin than is typical of a Favorite Son (Arizona) in 2008, and if there was an absence of convincing polls.
At this point, Barack Obama had 257 electoral votes sown up and Romney had 158. OK, so Obama could win Arizona and nothing else among the states in gray, Indiana and no other states in gray, or Missouri and no other states in gray and still lose. But even with that I can simplify a few things by taking those off the table. Obama wasn't going to win Indiana without also winning Ohio, Arizona without winning both Colorado and Nevada, or Missouri without also winning one of North Carolina, Ohio, and Virginia. Another simplification: Colorado and Nevada were going to vote together. You can also ignore the odd district in Nebraska that voted Democratic contrary to the rest of Nebraska because it does not matter. So that basically reduces the states in play to Florida, North Carolina, Ohio, Virginia, and the combination of Colorado and Nevada.
Now it gets simple:
Putting Arizona, Indiana, Missouri, and the Nebraska district best described as Greater Omaha in light blue takes them out of consideration.
The combination of Colorado and Nevada or any one of the states (Florida, North Carolina, Ohio, and Virginia) wins the election for President Obama. Because (except for Colorado and Nevada together) are different enough from each other and (except for neighboring North Carolina and Virginia) are scattered across the country that no single appeal can easily win them all together without putting some of the states in red in doubt, one can describe four states and the combination of Nevada and Colorado as 'independent events' to the extent that coin tosses are, an assumption of equal chances of an Obama or Romney win in a given state or combination of states gives President Obama 31 in 32 chances of winning re-election and 1 chance in 32 of losing.
31 to 1 is a long-shot. The chance of Obama winning was 0.96875; a sure thing is 1 and no chance is 0. Chances of independent events all going one way are multiplicative.
I was telling people that Mitt Romney had a chance to win in August 2012, but it wasn't very high.
OK, so coin-tosses are easy. What happens when the chances drop off?
Let's suppose that the chance of Obama winning North Carolina drops to .25 Then what do we get?
The chance of a Romney win rises to .5*.5*.75*.5*.5, or 0.046875... or roughly one chance in 22. Poor, but that may be the best that one can hope for. Now what if North Carolina drops out of reasonable contention as a small lead for Romney seems to not go away while time runs out? We get a 0.0625 chance of a Romney win.
We then get a chance of President Obama winning re-election reduced to one chance in 16.
On the other side, what if a lead for President Obama suggests an 85% chance of the President winning the state as time runs out as North Carolina spirals away from President Obama?
.5*.15*.5*.5 = 0.01875 The odds go to roughly 53-1 for President Obama.
Evidence suggested that Barack Obama was more likely to win both Colorado and Nevada than any of the states in white. So what happens if those two states go completely out of reach for Mitt Romney even as the others swing decisively toward him, let us say .75 chances each as North Carolina slips completely away from President Obama?
1- 0*.75*1*.75*.75 = 1. Obama wins.
This is a statistical model. It does not work the same way as the assumption that "Barack Obama is so horrible that I can't imagine anyone voting for him". The polls would show that that assumption was thoroughly wrong.