Languish Presidential Election!

[Viking]

What is wrong with Indiana anyway?

Gary, mostly. That place is a hellhole. It's like Mordor on Lake Michigan.

That just makes me wonder what Michael Jackson's "precioussss" was ... Macauley Culkin or Bubbles?

[Eddie Teach]

Should US voters consider the world's view, which gives Obama the most backing internationally?

No, we should re-elect him anyway. 

[Eddie Teach]

What is wrong with Indiana anyway?

Gary, mostly. That place is a hellhole. It's like Mordor on Lake Michigan.

That just makes me wonder what Michael Jackson's "precioussss" was ... Macauley Culkin or Bubbles?

If LOTR references make you think of Michael Jackson, well, I just don't know what to say...

[Scipio]

Gary Johnson 2012!

[Razgovory]

What is wrong with Indiana anyway?

Gary, mostly. That place is a hellhole. It's like Mordor on Lake Michigan.

That just makes me wonder what Michael Jackson's "precioussss" was ... Macauley Culkin or Bubbles?

If LOTR references make you think of Michael Jackson, well, I just don't know what to say...

:secret:Jackson is from Gary Indiana.

[Eddie Teach]

Ah, that makes it alright then I guess.

[The Brain]

I don't see how people can vote against a Nobel laureate. Boggles the mind.

[garbon]

The colors are all wrong though.

Aren't the locked in colors a relatively new thing?

[garbon]

Obama said he wants music to be more political. 

[OttoVonBismarck]

It's sort of like picking the March Madness bracket, if you just go on ranking all the way you'll do terribly. Upsets will happen.

Missed that the first time.  If you pick strictly on ranking, you will indeed do terribly.  However, that's the least terrible prediction available to you.  That's because you are expected to do very terribly no matter what bracket you go with.  However, if the tournament were repeated millions of times, going by rankings will likely do better than any other strategy. 

Deviating from rankings on the theory that upsets have to happen somewhere is a stupid strategy.  You should deviate from the rankings only if you have a superior insight to the people or system that generated the rankings.  (You should also deviate from rankings if you split money for ties in your pool, but that's a unique shortcoming of the betting structure rather than probability theory).

I didn't say you pick random upsets just because "upsets have to happen." Really if we're going to talk about it at length, what was intended to be a simple analogy just doesn't work. March Madness is fundamentally different from U.S. state results in Presidential elections. My point wasn't you should just randomly predict upsets in some states, but only that it is highly unlikely that for say, Nate Silver, every state he predicts as more likely to vote Obama than Romney actually votes for Obama. We have one election where Nate Silver made a name for himself picking 49 out 50 states correctly, and the one state he picked incorrectly was a very close race. The problem is in 2008 there were not nearly as many close races as there are going to be this year. I don't even need a statistical model and I can probably guarantee 35 state election results and be 100% correct come election day.

My only point was some people act as though all these predictions will come to be 100% accurate, the reality is with so many close races it's all but certain we'll see a few surprise results. Is it your position that the statisticians are perfect, and every state prediction they make will come to pass?  Silver is relatively new to this game, there are older statistical models like the Fair Model and others that get it very close to right, but they aren't perfect. I see no reason to assume just based on his success in the 2008 election that Silver has stumbled on some perfect model that will never make mistakes. Strong  "advocates" of the models preach it like gospel, most of the statisticians are a lot more realistic--for some reason you seem to be informed on the science of statistics but fall into the gospel camp when talking about it. I had an argument with someone about the Fair Model once where they said he was never more than a few percent wrong in any election, and I pointed out that "in a Presidential election a few percent is what you're fighting over and spending hundreds of millions of dollars on."

There's a good article Stephen Jay Gould wrote years ago, in which he basically said for some reason people look at simple statistical measures related to life expectancy with cancer, like median life expectancy after diagnosis, and assume that is the "real hard science." They assume the "noise", or the people who live longer and those who die earlier are just some abstraction that doesn't really mean anything. But the statistics is actually the abstraction, the "noise" is reality. No serious statistician believes they can use a model to predict something 100% accurately. Many will tell you they have a good model that gets "really close" and getting "really close" is very important and all that matters. (Actuarial tables for example, as you should know, get really close but aren't perfect--that is why they change over time.) In a Presidential election, I'm saying you can get very close to right but still blow the overall call, because in a close election a few thousand votes here and there can end up being the difference.

[Ed Anger]

Obama said he wants music to be more political. 

Ugh.

[Josephus]

Apparently Jenna Jameson is supporting Rommey because, "when you're this rich you vote for Republicans." Cant believe I wanked to that whore once

[CountDeMoney]

Apparently Jenna Jameson is supporting Rommey because, "when you're this rich you vote for Republicans."

Finally, the first honest endorsement yet.

[Phillip V]

Apparently Jenna Jameson is supporting Rommey because, "when you're this rich you vote for Republicans." Cant believe I wanked to that whore once

I wanked to her when she was poor and amateur. Not when she became rich.

[DGuller]

I didn't say you pick random upsets just because "upsets have to happen." Really if we're going to talk about it at length, what was intended to be a simple analogy just doesn't work. March Madness is fundamentally different from U.S. state results in Presidential elections. My point wasn't you should just randomly predict upsets in some states, but only that it is highly unlikely that for say, Nate Silver, every state he predicts as more likely to vote Obama than Romney actually votes for Obama. We have one election where Nate Silver made a name for himself picking 49 out 50 states correctly, and the one state he picked incorrectly was a very close race. The problem is in 2008 there were not nearly as many close races as there are going to be this year. I don't even need a statistical model and I can probably guarantee 35 state election results and be 100% correct come election day.

My only point was some people act as though all these predictions will come to be 100% accurate, the reality is with so many close races it's all but certain we'll see a few surprise results. Is it your position that the statisticians are perfect, and every state prediction they make will come to pass?  Silver is relatively new to this game, there are older statistical models like the Fair Model and others that get it very close to right, but they aren't perfect. I see no reason to assume just based on his success in the 2008 election that Silver has stumbled on some perfect model that will never make mistakes. Strong  "advocates" of the models preach it like gospel, most of the statisticians are a lot more realistic--for some reason you seem to be informed on the science of statistics but fall into the gospel camp when talking about it. I had an argument with someone about the Fair Model once where they said he was never more than a few percent wrong in any election, and I pointed out that "in a Presidential election a few percent is what you're fighting over and spending hundreds of millions of dollars on."

There's a good article Stephen Jay Gould wrote years ago, in which he basically said for some reason people look at simple statistical measures related to life expectancy with cancer, like median life expectancy after diagnosis, and assume that is the "real hard science." They assume the "noise", or the people who live longer and those who die earlier are just some abstraction that doesn't really mean anything. But the statistics is actually the abstraction, the "noise" is reality. No serious statistician believes they can use a model to predict something 100% accurately. Many will tell you they have a good model that gets "really close" and getting "really close" is very important and all that matters. (Actuarial tables for example, as you should know, get really close but aren't perfect--that is why they change over time.) In a Presidential election, I'm saying you can get very close to right but still blow the overall call, because in a close election a few thousand votes here and there can end up being the difference.

I understand all that you're saying.  I understand that models aren't perfect.  However, as a statistician, I also know that nothing is perfect, it's a science full of limited information.  Therefore, your job is to make the best estimate, or more realistically the least worst estimate, and go with it.  All outcomes are highly unlikely, but one of those highly unlikely outcomes has to come to fruition, and statistics is all about figuring out which outcome is the least unlikely.