PPP vs. Nate Silver and potentially cooked polls

[Lurker]

Nate Silver getting 50/50 states right isn't that impressive at all. You would have gotten exactly the same result - with Florida as the only possible exception - by doing a simple average of all poll, something not exactly requiring great mathematical skill.

For all the complicated prediction models of political scientists, there is no way to get a more accurate result than a simple aggregate of the polls.

A main reason Nate Silver's predictions appear so impressive are all the Republican pundits basically living in fantasy land. Rather than trusting the polls in November, they went with their "gut feeling" with the predictably laughable results (the "respected" Republican analysts of course being the worst of all).

What would be interesting to study is how many of Silver's long-term predictions prove true - does candidates who are given an 80% chance in the year before an election win 80% of the time?

[Lurker]

Nate Silver getting 50/50 states right isn't that impressive at all. You would have gotten exactly the same result - with Florida as the only possible exception - by doing a simple average of all poll, something not exactly requiring great mathematical skill.

For all the complicated prediction models of political scientists, there is no way to get a more accurate result than a simple aggregate of the polls.

Exactly.  Nate Silver is overrated.  The notion that President Obama ever had over an 80% chance of winning is ridiculous.  Granted, he was reelected, but his chances were never that high.

Obama chances were that high on election day - in fact much higher if you'd based your prediction on the state polls. This is were Nate Silver is very useful: by looking at earlier elections/polls he can tell that a candidate with a 2% lead on election day tends to win 85% of the time, for instance.

(That number is just a guess on my part, but hopefully you see the point.)

[Lurker]

Nate Silver getting 50/50 states right isn't that impressive at all. You would have gotten exactly the same result - with Florida as the only possible exception - by doing a simple average of all poll, something not exactly requiring great mathematical skill.

For all the complicated prediction models of political scientists, there is no way to get a more accurate result than a simple aggregate of the polls.

Exactly.  Nate Silver is overrated.  The notion that President Obama ever had over an 80% chance of winning is ridiculous.  Granted, he was reelected, but his chances were never that high.

I should point out that Nate Silver's rise to fame was based on a statistical model he did during the 2008 primaries that completely ignored polling and was based only on demographics...and was actually more accurate than ANY polling company.

This quote makes it appear that I'm agreeing with OldiesFreak, which is obviously not the case (other than on the fact that Silver's 50/50, so hailed by everyone, was not that impressive).

As to your assertion that Silver built a statistical model "that completely ignored polling and was based only on demographics" - please back this up with a source. I took a quick look at his 2008 Super Tuesday predictions, and that claim does not seem accurate.
http://www.dailykos.com/story/2008/02/05/450099/-Final-Super-Tuesday-Projection-2-5-08

[Lurker]

The states that Silver nailed better than any pollster through his demographic model were North Carolina and Indiana.

Your claim that Silver made a statistical model "that completely ignored polling and was based only on demographics" is still wrong though.