538

[whoman69]

The latest 538 simulations have Obama winning 90.9% of the time with an average of 313.0 EC. The popular vote is seen as 50.8-48.3 or a 2.5 point lead for Obama.

The closest state contests are:
Florida 50.3% Obama
North Carolina 74.4% Romney
Virginia 79.4% Obama
Colorado 79.7% Obama
Iowa 84.3% Obama
New Hampshire 84.6% Obama

The following battleground states are now considered safe Obama meaning more than a 90% probability:
Michigan
Pennsylvania
Wisconsin
Nevada
Ohio

Obama has 303 EC at least 79%
Obama has 332 EC at least 50%

[Donger]

Originally Posted by AustinChief

I agree with your take on this 100%... and as the know it all I will weigh in further...

What Nate Silver does is numerical entertainment. That's it. It has ZERO value when correlated to reality. It's so funny... people judge his results based on his FINAL projection ONLY... yet they try to prop up his work RIGHT NOW at this stage as if it is predictive and meaningful. OK, fine if it is predictive and meaningful, let's see metrics and judgements based on his assessments 30 days out, or 28 or whatever.

I have zero problem with Nate Silver, but it is pretty clear there are quite a few here who don't understand that at this point, my "guesses" have as much credibility as his "work." His is just far more entertaining and looks oh so scientific!

You want to hold me accountable for my predictions 30 days out.. that's fine. Then let's hold Nate accountable for his as well. What? We only care about his final prediction, up until then it is just guesswork? Yeah, same here.

I have calculated a 68.35% probability that I agree with you here.

[whoman69]

If the endgame was just to predict who was going to win the Presidency, then he would do a one article a year. If that were the goal of polls, they would only have one as well. That's not what he is doing at all. He is bringing sense to the myriad of conflicting polls out there. You have national and state polls with different numbers, sometimes even different leaders. While he does make a prediction that based upon the polling numbers at that time that Candidate X would win the election and also a separate if the election were tomorrow, the main goal is to track the ebbs and flows of the election.

Are you really so against 538 because he is giving an answer you don't like? Most services out there will agree that Obama has the easiest path to an electoral victory.

[WoodDraw]

Originally Posted by AustinChief

I agree with your take on this 100%... and as the know it all I will weigh in further...

What Nate Silver does is numerical entertainment. That's it. It has ZERO value when correlated to reality. It's so funny... people judge his results based on his FINAL projection ONLY... yet they try to prop up his work RIGHT NOW at this stage as if it is predictive and meaningful. OK, fine if it is predictive and meaningful, let's see metrics and judgements based on his assessments 30 days out, or 28 or whatever.

I have zero problem with Nate Silver, but it is pretty clear there are quite a few here who don't understand that at this point, my "guesses" have as much credibility as his "work." His is just far more entertaining and looks oh so scientific!

You want to hold me accountable for my predictions 30 days out.. that's fine. Then let's hold Nate accountable for his as well. What? We only care about his final prediction, up until then it is just guesswork? Yeah, same here.

This is complete bullshit, and I heavily doubt we'd be seeing so much bold and caps from you if he was showing Romney up by whatever percentage.

People in this thread fundamentally don't understand statistics or modeling - it's just bizarre. And they're ANGRY about it.

[KC native]

Originally Posted by WoodDraw

This is complete bullshit, and I heavily doubt we'd be seeing so much bold and caps from you if he was showing Romney up by whatever percentage.

People is this thread fundamentally don't understand statistics or modeling - it's just bizarre. And they're ANGRY about it.

Uh oh prepare for a barrage of captilaized words and bad math, but remember he did stay in a holiday inn last night .

[patteeu]

Originally Posted by WoodDraw

This is complete bullshit, and I heavily doubt we'd be seeing so much bold and caps from you if he was showing Romney up by whatever percentage.

People is this thread fundamentally don't understand statistics or modeling - it's just bizarre. And they're ANGRY about it.

Do you think this criticism applies to AustinChief's post, because I don't see the connection.

[KC native]

Originally Posted by patteeu

Do you think this criticism applies to AustinChief's post, because I don't see the connection.

Of course you don't.

[patteeu]

Originally Posted by KC native

Of course you don't.

AustinChief didn't say anything that brings into question his understanding of statistics or modeling. He alleged that Nate Silver's model hasn't been adequately validated for it's assessments 30 days out from election day. I don't know if that's true or not (although I suspect it is), but you can have a rock solid understanding of statistics and modeling and still make this type of criticism. In fact, you could have a better understanding of Nate Silver's model than Nate Silver does and still make that criticism.

[WoodDraw]

Originally Posted by patteeu

Do you think this criticism applies to AustinChief's post, because I don't see the connection.

It's like saying hurricane modeling is just "meteorological entertainment". You can criticize his methods, as many have done. There is a lot of research in political science about election models and the relative strengths and weaknesses of the different ones. But to equate what 538 does to someone on a message board making a prediction is political dumb****ery.

[patteeu]

Originally Posted by WoodDraw

It's like saying hurricane modeling is just "meteorological entertainment". You can criticize his methods, as many have done. There is a lot of research in political science about election models and the relative strengths and weaknesses of the different ones. But to equate what 538 does to someone on a message board making a prediction is political dumb****ery.

OK, that's a fair criticism. "Numerical entertainment" was an unfair description.

[WoodDraw]

There are three main issues with election models. And they all kind of relate to each other.

First, there just isn't much data. If you're talking about about presidential elections and limit your data to post WWII, as I think pretty much everyone does, you just don't have much to work with.

Second, for that reason, a lot of models tend to "predict" the past, and not the future. It's easy to cherry pick a bunch of different data points that explain every election. But if you have to update your "model" after every election, you're not really accomplishing anything. If you had hundreds of years of election data to use, you could have more success here. But as it is, complexity isn't always your friend.

And lastly, bias. Republicans have been hostile to 538 lately because it shows Obama winning, while Democrats have been fawning over it. People don't like models that don't affirm their beliefs, especially with politics.

That works outside of political ideology too. If you're a political scientist that believes the economy is the best predictor of election results, you're probably not going to create a model that shows otherwise.


So there's a lot of room to criticize these things. But I think it's helpful to at least know what you're talking about before doing so.

[cosmo20002]

Originally Posted by whoman69

Are you really so against 538 because he is giving an answer you don't like? Most services out there will agree that Obama has the easiest path to an electoral victory.

Originally Posted by WoodDraw

This is complete bullshit, and I heavily doubt we'd be seeing so much bold and caps from you if he was showing Romney up by whatever percentage.

Pretty much nailed it there. But remember that's Austin "Cliff Clavin" Chief you're talking to.

[patteeu]

Originally Posted by cosmo20002

Pretty much nailed it there. But remember that's Austin "Cliff Clavin" Chief you're talking to.

I agree that he nailed it (since he expressed a corollary to the point I made earlier in the thread).

Originally Posted by WoodDraw

Democrats have been fawning over it. People don't like models that don't affirm their beliefs, especially with politics.

And adore the ones that do.

Originally Posted by WoodDraw

So there's a lot of room to criticize these things. But I think it's helpful to at least know what you're talking about before doing so.

Likewise for the people treating it like electoral gospel.

[WoodDraw]

There's another point I'll make and then I'll shut up and attempt to not get fired from my job.

Nate Silver's model is a combination of two different things - polls and fundamentals. Fundamentals are things like economic and foreign policy measures, while polls are obviously...polls. Early in the political season, he biases towards fundamentals, because the polls don't mean much at that stage. People haven't gotten into election mode yet, and many may be unfamiliar with the candidates. So the unemployment rate, or amount of US soldiers killed, or other similar stats tend to give a clearer idea towards the view of individuals and the chances of the incumbent party/candidate.

Over time, weight is shifted more to polls. The idea being that voters have become more informed (whether rightly or wrongly) and have made up their mind on issues. In other words, history may tell us that in a poor economy, the incumbent should lose. But if polls show the incumbent out-polling the challenger, even on the economy, then it becomes harder to weigh theory over current reality.

There's some complexity to it, of course. Not all polls are created equal. Some favor democrats, some favor republicans, and some are just shit. So the model tries to adjust that noise. It is also slow to recognize trends - which caused a lot of Republicans to bitch. So if I go down to Texas and run my own poll, and find Obama winning by +15, I should be ignored. If three other well known polls after me find the same thing, things change.

There are other theories in political science that tend to ignore polls and focus on fundamentals. The idea being that you can plug in data on the health of the economy, the ideology of the candidates, and current wars, and make prediction solely off of this. I think these are interesting, but largely useless for a New York Times blog on the state of the election. They also don't tend to be all that accurate.

So the criticism of 538 isn't completely wrong. If I run a poll the day before an election and then predict the winner, wtf cares? So 538 does get some criticism for that. But I think those people misunderstand what it's trying to accomplish.

[cosmo20002]

Originally Posted by WoodDraw

There's another point I'll make and then I'll shut up and attempt to not get fired from my job.

Appreciate that. We don't want you screwing up the unemployment numbers. You'll be the one to push it back over 8%.

[BigRedChief]

Originally Posted by AustinChief

I agree with your take on this 100%... and as the know it all I will weigh in further...

What Nate Silver does is numerical entertainment. That's it. It has ZERO value when correlated to reality. It's so funny... people judge his results based on his FINAL projection ONLY... yet they try to prop up his work RIGHT NOW at this stage as if it is predictive and meaningful.

wellllll today Nate has his nowcast and Nov.6th pretty close to the same now.

Obama 66%
Romney 34%

Pretty impressive movement by Romney. From a 3% nowcast and 22% Nov. 6th forecast a couple of weeks ago.