Commentary's Noah Pollak notes that CNN trumpeted a poll where debate-watchers declared Obama the winner, 51-38, and reported that "fact" as news.
Only thing is, it turned out more Democrats than Republicans were watching the debate -- the split was 41% Democrat, 27% Republican, 30% independent (and, as Pollak notes, who knows how many of the latter were Obama "leaners"?). Given (as I've pointed out) that there were no clear "knockout punches" in the debate, each side naturally gives the advantage to its guy (and, if each candidate won all his party supporters, 10% of the independents went wtih Obama, 11% with McCain). In a poll loaded with Democrats, that translates to an Obama "win."
Interestingly, a CBS poll of 483 uncommitteds -- defined as people who haven't decided for whom to vote, or who might change their minds -- also gives the edge to Obama. But even setting aside the built-in disadvantage of weekend polling for Republicans, it would be interesting to know how many of the "uncommitteds" identify themselves as Democrats, Republicans or neither. After all, there's a big difference between Barack winning back hitherto-disaffected Democrats, and winning independents (or winning over Republicans).
But then again -- as the CNN example shows -- you can stack a poll in order to obtain the results you want to report, can't you?