As I've warned before, a clever pollster can produce any result he's paid to produce, and will, up to the last minute when having an accurate prediction can secure continued employment.
In that spirit, here's the Billlls poll shamelessly based of educated guesswork and somebody elses actual data. Here is a look at a CNC/ORC (who ARE these people?) poll featuring the usual questionable breakout. I'm getting around all this by assuming that at heart, about 1/3 of Americans are Dems, 1/3 are Republicans, and 1/3 are independents. I will also go out on a limb and assume that people will vote their normal party. Variations happen when some percentage of the electorate from one party or the other are disenchanted with the party pick and simply stay home.
This leaves the independents to actually decide the election. These folks are split between those who know nothing and vote their long time favorite party, and those who actually know something and don't want to be affiliated with either side, for good reasons. The poll in question assumes a lot of Dems, but ignoring that and calling the base vote even, they note that the independents are breaking 14% in favor of Romney. This adds 18.8 pts to the 33 point base for R and 14.2 pts to the 33 point bade for the D, hence the above figure.
If I wanted to get carried away, I would state that the D's will stay home at the same rate they stayed away from the DNC which would take 20 points away from the D's and from the overall turnout giving R 33+18.8=51.8 and D 13+14.2=27.2 out of a total of 80. This makes the final vote 51.8/80=64.7 to 27.2/80=34.0. Some rounding error is included.
If either of those numbers proves accurate, I'm charging big bucks next election to do the same thing.