and kept his opinions out of it. It’s the same problem I have with BP nowadays. They’re too concerned with giving the numbers a voice and personality rather than the sticking to facts.

so forgive the fact that I’m going out of line.

If you don’t think BP has strayed from the numbers than I’m not sure what to tell you. John Perotto has zero statistical insight in his articles — they’re pure reporting. Christina Kharl and Joe Sheehan use sketchy analysis to supplement the narrative of the day they’d like to construct. Rarely do we hear from Silver on BP or Davenport on reasoning for translations.

The same thing is happening at 538 where he offers suggestions for an advertising campaign or why he thinks a certain demographic will split from traditional voting lines. Having numbers available doesn’t inherently mean that they are being used. I don’t care what his personal opinion is on a set of data — tell me why the data points to a certain outcome via math and empirical evidence not why person X thinks event Y would help or hurt candidate Z.

Opinion is not a bad thing. This isn’t a personal high-horse where I pretend like I never offer an opinion. A lot of the prospect analysis at FR is opinion based on what I’ve seen first hand and my interpretation of the numbers. I don’t market myself, however, as a mathematical model for election/baseball prediction. If you like Nate (or Sheehan’s or Goldstein’s) opinion great. I don’t find him insightful (or accurate) when he discusses things other than factual math.

And I want to try not using this forum to share them, but it is useful to realize that while 538 offers good analysis, they aren’t able to offer much more than a snapshot of the race based off others polling work.

First, this thing is more fluid than people tend to think. It’s a volatile situation economically and that makes it volatile politically.

Secondly, it is a very difficult year for polling. People assume the rules regarding party ID are broken, but the question is by how much. Republicans usually outperform October polling in the elections. Part of the reason is that, historically, young voter turnout almost never lives up to the hype. This year is certainly different, but that is being priced into the spreads in party ID these polling outfits are using.

Party ID has been pretty consistently around a 3 point spread, in favor of the democrats, historically. It was right around there, even in 06’, when the republicans took a bloodbath. Again it certainly has changed in the Dem’s favor, but by how much? Many of these spreads are up to 6,7,8 points. Even more in some cases. What if they are off by a percent or 2? Rasmussen, for example, is currently using a 6.7% percent spread, I believe, and has Obama up 4. What if they are overshooting a bit in their samples and it is really a 5% spread? Presumably that’s about 3.5% gain for McCain, states like Virginia, Missouri, Colorado, etc become flippable, and the thing is a toss up.

Adding to the confusion, the Republican primaries were decided pretty early while the Dem primaries dragged on forever. This created an incentive for republicans to vote in democratic primaries even if they had no intention of voting democratic now. This could impact the ease with which polling outfits can measure party id in their state pollings as well.

That’s a long way of saying that the certitude that sites like 538 have right now is very misleading. I do not think this thing is over. The historical rules of sampling do not apply, but it is hard to say with certainty what the new rules are. That type of analysis doesn’t generate traffic, but it may be closer to the truth than what these sites would suggest.