2) Ignore exit polls. The basic problem with how they work is that the polling organizations have to extrapolate from the weird sample they get trying to harangue a semi-random sample of people at a chosen set of polling places based on expectations of what the electorate looks like, but if they knew that, they'd know everything, and they wouldn't have to do a poll. Take every difficulty with doing a telephone-based survey, magnify it, and take away our ability to do polling aggregation, and that's what you've got with exit polls. Ignore them.
3) Ignore any effort to pick some narrow slice of the electorate and declare it the key demographic for this year. Why are 47-year-old women who majored in 16th Century Persian philosophy and sleep on the right side of the bed, but stretch out diagonally when their significant others travel on business the key demographic this year? They aren't. This kind of demographic slicing is bullshit, and not just the snarky version I'm doing here. You know exactly what I'm satirizing. Elections turn on big trends, not some media-chosen, arbitrary demographic.
So, what now? I dunno. It's all about turnout.