[jackmercer]

The really funny thing about the whole Nate Silver thing is that what he is doing is really not complicated or difficult....it IS however novel. No one really weighted pollsters in their averages the way he did.

All you have to do is throw all the polls in an algebraic equation and weight each pollster based on history, polling method (auto vs live call), percent of cell phones included, fundamentals like economic conditions, what he calls “house effect” which in my opinion is the strongest, and whatever else he deems appropriate.

If you lay out this equation and multiply each poll (X, Y, Z, etc) with a .5, .75, .9 etc based on their history, house effect, etc, you get a really accurate average.

Many bloggers have already started to replicate this method and the funny thing is, you can almost reverse engineer some if his model since he posts the weights in bars right there on his blog. Of course he could be BSing those published weights to maintain the proprietary nature of this model but maybe not.

All I did this year is average the polls and threw out Rasmussen and Gallup (for reasons posted many time before), gave automated polls a < 1 weight, internet polls a < 1 weight and threw out any polls with no history. With that pretty crude method, I was able to call Obama +1.8% and 290 (and possibly 303) for Obama in the electoral college.

http://www.freerepublic.com/focus/chat/2955102/posts?page=117#117

I didn’t see Florida at all, the one I missed. I’m not sure how Silver saw that one but next time around, I hope to have fun building a better model of my own.

[jwalsh07]

There are 4 swing states, Florida, North Carolina, Virginia, Ohio. Thats it. The notion that Romney was going to win Wisconsin, Iowa etc was ridiculous. Republicans have not been a majority party since Reagan and may never be again. the other side has a bigger pool of votes and when the pubbies cave to amnesty there will be no contested states. 10 million new democrat voters have that effect in National elections. Republicans will still be able to win the house for a while because of the number of votes that the dems pull from urban precincts but that will go away as well. Call me a glass 1/32 full kind of guy but the takers, committed marxists and vaginacentric voters outnumber us.

[jwalsh07]

ps: NH is gone so the R’s will have to figure out a way to win Colorado or find a lefty Ross Perot to run. Barring that or a big ass war I don’t expect to see another republican POTUS in my lifetime. I am old though so the glass is still 1/32 full.

[major-pelham]

All you need is excel with a monte carlo plug-in
http://www3.wabash.edu/econometrics/EconometricsBook/Basic%20Tools/ExcelAddIns/MCSim.htm

[nathanbedford]

Jack, I have been remiss in not publicly complimenting you on the accuracy and professionalism of your posts throughout the campaign. I also have a confession, I dismissed the implications of your findings not because I did not believe they were accurate but because I did not think they were predictive. In other words, I got stuck on what people like Morris and Barone were saying concerning the weighting of the polls etc. and that comported with my prejudices which were the effect that people thought the way I thought and would not suffer their country to be destroyed before their eyes, much less destroy it by their own hands.

One of the reasons why I posted this article about Nate Silver is to demonstrate how treacherous it is for one to ignore people like you and believe our own worldview, which is precisely what I did.

There is a lesson to be learned. Before we decide in which direction to go to win the next election and save the country (assuming there is something left to be saved and something of liberty too), convinced of our virtue, demanding that we turn right, insisting that the election was stolen, blaming the candidate, even blaming the weather, we should turn to people like you and get some hard data and we should interpret that data as dispassionately as human nature will permit.

I do not assume that is easy, Michael Barone got it wrong, but there is no other way.

[gusopol3]

So, how does the turn out model work on election night when <120 m votes have been counted , if it’s based on “2008 model” when 130 m votes were cast? I understand how it can predict 100% Obama vote in certain Cleveland or Philadelphia precincts, but how does it predict St. Lucie county 140% turnout?

[Political Junkie Too]

I did analyze Rasmussen's polls all election long. Even with its flaws, the best Romney did was a 45% probability of winning.

-PJ