I don’t mind that he asked his own questions–as moderator he’s entitled–but as moderator he’s also entitled to cut off a candidate that’s running off at the mouth, as Gwen Ifill did in the VP debate, and she only had to do it once.

Brokaw chose not to do that. Fine, but then there’s no sense in constantly whining about candidates running over time. I hope he was embarrassed that McCain had to actually try and do his job for him by suggesting a wave!

There were two poor performances last night, only one of which was from a candidate.

Sam’s analysis is a statistically correct interpretation of the input polling results.

My question is: what happens when the input weighting is wrong?

However, what happens when the weighting factors for inputs, such as party affiliation and age, are misleading, or the turnout bears no relation to expectations? Anticipating turnout is what skewed your results in 2004, and also skewed results in 2000. GOTV vs. Sit This One Out might affect outcomes more than expected.

Your poll model abstracts away the weightings. This, by itself, doesn’t matter if they’re wrong, because they can be dialed back in later. An interactive map that dials in GOTV or dials out sit-homes would be much like the one you have already, albeit rather than a vote going from one voter to another it would be one vote additional vs. one more taken away.

This could expose the potential for GOTV or the looming results for stay at home in places like Texas, which are hard red in your tables.