The Shrine

It took forty-five minutes and I was the only minority that wasn't working as an election official, but I voted. I think I once again understand how Republicans win elections in my state now.

I've heard a lot of stuff about Obama winning today and just heard that the guy who predicted all the states in the last election has recently predicted Obama'll win again. So here's hoping.

I'm not going to be a smug prick about voting so I didn't take an "I voted" sticker. I am going to be a smug prick about not having one though. So there.

That guy was in Sine's link in the other thread and on Colbert last night. Nick something. Isn't he the one behind the math that was the basis for sabremetrics used in the moneyball days of the A's?

Oh, right, that's him. Nate Silver.

Shrinweck wrote:Oh, right, that's him. Nate Silver.

Took a quick look at the blog after I posted that and it's up over 90% for Obama overnight after some of the early voting numbers came in. He leaves a very slim chance of R-Money winning (the chances of hitting an inside straight in poker). This goes against everything we've been hearing from everywhere else.

I'm not so sure I'm with him as his basis is the polls themselves (he averages all of them) and they're always biased, but it's an interesting technique he has, something very different than what you see on TV with the standard polls. It will be interesting to see if he successfully accounts for the inherent biasedness of the polls themselves with his statistical averaging of them all together and remains relatively accurate.

Well these polls generally come from land line phone calls and no one our age really even has one of those any more. One would think basically everyone with a land line would be a Republican (money for phone line on top of cell phone bills, age, less inclined to use technology), so the fact that the election polls very closely means that among people under 40 Obama could potentially be destroying Romney. This is kind of a gross generalization since I'd assume most of you with a family would have a land line for the sake of reliability or some such, but I'm just saying that polls would probably favor Romney in this regard.

Zeus wrote:

Shrinweck wrote:Oh, right, that's him. Nate Silver.

Took a quick look at the blog after I posted that and it's up over 90% for Obama overnight after some of the early voting numbers came in. He leaves a very slim chance of R-Money winning (the chances of hitting an inside straight in poker). This goes against everything we've been hearing from everywhere else.

That's because everybody else is:

1. Trying to pretend it's a horse race, which it isn't, just so the news media can cash in on it being some dramatic thing.
2. Not looking at the electoral college data, and just looking at national averages. Which is dead wrong.
3. Is wishfully thinking R-Money will win, despite the mathematical and statistical evidence.

Nate seems to be one of the few guys who actually gives a shit about not being biased in the statistical analysis. And that means his shit is accurate, to which he can point to his past history as a record of his accuracy. What few mistakes he's made, he's analyzed and explained them, and adjusted things to make his model work better. He's very good at predicting this stuff, and I see no reason to not believe him now.

Not all of us live in the US. The selection of the President by the electoral college will have a great impact on the rest of the world, because the places we live definitely dictate our foreign policy around the US's. This might leave some of us feelings helpless, but watching on to see what happens. As a poli sci man, I am glued to the coverage of the incoming numbers and the punditry.

Now we may not all be able to vote for the next President of the US (but Obama's going to win, so the world will be happy), but there is something that we can all do. The PSN is having a sale on the Persona games for the PSP (1, 2, and 3) and the PS2 version of Persona 3 FES! P3 FES is five goddamn dollars! I love that game.

Also:

I don't really see how you can possibly have a meaningful sample with an event that happens once every 4 years. Yes there's more than one state but only a handful of them are ever contested per 4 years.

At any rate predicting these things might as well boil down to whether you can predict who wins the few states that really matter. I think the winner of Ohio has taken the presidency for the last 6 times or something like that?

The BBC, CNN and The Daily Show are already projecting an Obama win. Supposedly Romney hasn't even prepared a concession speech. Can't wait to see this.

Nobody's going to admit they started out planning to lose. I'm sure Romney has a speech ready for that.

I mean sure they'd say like the vote projects 50% Obama 49% Romney in Ohio and that's within the margin of error of +-3% but 1% of the population in Ohio is 100K people. It's actually very unlikely for that many people to suddenly change their mind on the day of election but of course you wouldn't admit that if you're the side with less people. If you ever see some of the closer GameFAQs poll you can see it pretty clearly, like when FF7 goes against Zelda, you'd see Zelda have like 50.1% versus 49.9% but that gap never gets closer because anyone you can tell to go vote for these things already did so that 0.2% is something you can never make up unless there's a mysterious large block of FF7 fans that has never heard of GameFAQs before that is somehow reachable now.

It's a quote directly from him saying that he only prepared the one speech. Whether one was prepared in the background and not shown to him isn't what I was talking about. :P

Yeah I know he said that, but it's not like he's going to say, 'err I got nothing to say because I didn't prepare the other speech'. You can probably ad lib it anyway and it's not like people really care what the loser has to say to begin with. It's sort of like all the random guys guaranteeing their team won't lose knowing that if the team loses, usually no one would care what they have to say anyway.

People care very much about the concession speech. At the bar I worked at in 2008 more people (including me) were interested in what McCain had to say than Obama. Bringing it back to 2008, McCain's concession speech was very important in that it emphasized coming together and not being dicks about it. And that's why the concession speech is important. It's a marker that the country has voted and it's a symbol that it's time to come together despite how you voted. It sets the table for the future. And it finally lets me turn off coverage and get back to my god damned life. COME ON Romney, for pete's sake (supposedly due out in the next few minutes).

This concession speech is even more important because if Romney decides to bring lawyers into the close states and dispute people's votes it will set the country on fire with anger.
Edit: That doesn't sound likely, though, since he reportedly just called Obama to concede.

And the forces of evil are once again thwarted! =D

Don wrote:I don't really see how you can possibly have a meaningful sample with an event that happens once every 4 years. Yes there's more than one state but only a handful of them are ever contested per 4 years.

At any rate predicting these things might as well boil down to whether you can predict who wins the few states that really matter. I think the winner of Ohio has taken the presidency for the last 6 times or something like that?

Don, quit pretending you know more than a guy who's being doing this shit for years. Polling data comes out at an almost frightening pace, so barring some last minute boondoggle that crashes a campaign a day before election, you can predict this stuff pretty well.

http://elections.huffingtonpost.com/2012/results/

Wow, will you look at that? The map looks EXACTLY like Nate's prediction map. Some shade differences here and there, but he predicted ALL 50 MOTHERFUCKING STATES CORRECTLY!

This is a good day, NO ARGUING until midnight =P

It's a bright bright morning for your country =)

Or to illustrate it another way:

Don - don't kid yourself. R-Money and the Republicans believed their own hype and were "confident" enough to not have even prepared a full concession speech. What he said didn't sound like it was prepared in any way as it was pretty generic. I'm sure he woulda thought of it ahead of time but didn't prepare an actual speech.

Also, it's very, very important to make a good concession speech. It shows that you're all still Americans and need to work together regardless of your personal political beliefs. Has to come from the top down before it can become a reality (although I'm sure the Tea Party will have something very loud to say about it still)

Sine - gotta give Silver credit, his method worked extremely well the last two elections. Maybe this actuarial nerd is on to somethin'. I'll be lookin' at his predictions more closely during the 2014 elections to see if his method can stand up.

One of the things I liked about McCain's speech more is that he told people who were booing to stop. Romney didn't bother hushing booers, but he also didn't pick the most hateful, spiteful vice president candidate in the history of the United States.

Something that interested me that the BBC commented on was that out of the super rich presidents we've had, they've all had something humanizing about them. Teddy Roosevelt's wife died when he was young (26?), FDR was wheelchair-bound, and JFK was a war hero. Romney didn't have anything that brought him down to our level until he lost the election.

This guy obviously weighed some polls as more reliable than others, and unless God told him which polls are better it's obviously still based on some form of opinion. For that matter the fact he's right doesn't validate or invalidate his model. If Romney had won the election we can't just call up our friends in the nearest 9 other parallel universes to see if Obama won there to see if his model is accurate.

Honestly I just see this guy saying things what one party wants to hear and this time he happens to be right. If you say his model is absolutely correct there's still a 10% chance Romney would have won because that's what his model says, so what happens if Romney did win? Do you say this guy is totally wrong even though we started with the assumption that he is absolutely correct? By the way, things do change significantly all the time. You'd predict a landslide Obama victory prior to the first national debate because nobody thought Obama could've screwed up that bad. Hurricane Sandy likely had a large positive effect on Obama since it's not like Katrina where so much damage was done that you couldn't even move the supplies there because the infrastructure is gone. The recovery effort was definitely something you can just throw more money at it and Obama took advantage of that, while Romney can't do anything meaningful. This stopped Romney's momentum and I doubt he had a model that calculates a hurricane was going to stop Romney's momentum that nobody thought he would have in the first place.

Rachel Maddow analyzes the election.

Little bit of a crush on Rachel, not gonna lie.

Don wrote:This guy obviously weighed some polls as more reliable than others, and unless God told him which polls are better it's obviously still based on some form of opinion. For that matter the fact he's right doesn't validate or invalidate his model. If Romney had won the election we can't just call up our friends in the nearest 9 other parallel universes to see if Obama won there to see if his model is accurate.

Honestly I just see this guy saying things what one party wants to hear and this time he happens to be right. If you say his model is absolutely correct there's still a 10% chance Romney would have won because that's what his model says, so what happens if Romney did win? Do you say this guy is totally wrong even though we started with the assumption that he is absolutely correct? By the way, things do change significantly all the time. You'd predict a landslide Obama victory prior to the first national debate because nobody thought Obama could've screwed up that bad. Hurricane Sandy likely had a large positive effect on Obama since it's not like Katrina where so much damage was done that you couldn't even move the supplies there because the infrastructure is gone. The recovery effort was definitely something you can just throw more money at it and Obama took advantage of that, while Romney can't do anything meaningful. This stopped Romney's momentum and I doubt he had a model that calculates a hurricane was going to stop Romney's momentum that nobody thought he would have in the first place.

The model was...wait for it:

HE AGGREGATED STATE POLLS!

So, yes, he had a "model" for Hurricane Sandy. It's called poll numbers AFTER Hurricane Sandy. They release poll numbers on a daily basis in various states.

The 10% chance just means that, like I said earlier, "barring some last minute boondoggle that crashes a campaign a day before election", Obama is going to win.

God, just admit that you were mistaken and be done with it.

SineSwiper wrote: The model was...wait for it:

HE AGGREGATED STATE POLLS!

So, yes, he had a "model" for Hurricane Sandy. It's called poll numbers AFTER Hurricane Sandy. They release poll numbers on a daily basis in various states.

The 10% chance just means that, like I said earlier, "barring some last minute boondoggle that crashes a campaign a day before election", Obama is going to win.

God, just admit that you were mistaken and be done with it.

I'd hope you're aware that he assigns different weights to different polls based on some measurement of their historical accuracy because you sure could not have come to the conclusion he did if you just averaged all the polls. And guess what this part is still guessing, no more than a guy running a basketball simulation believes you should assign a weight of 0.8 instead of 0.9 to this variable. None of these are anything beyond educated guesses because it's not like there's a physical law of basketball or election voting you can discover. I mean if you want to talk about hypothetical formula here's the one for PER that's widely used for baskeball:

uPER = (1 / MP) *
[ 3P
+ (2/3) * AST
+ (2 - factor * (team_AST / team_FG)) * FG
+ (FT *0.5 * (1 + (1 - (team_AST / team_FG)) + (2/3) * (team_AST / team_FG)))
- VOP * TOV
- VOP * DRB% * (FGA - FG)
- VOP * 0.44 * (0.44 + (0.56 * DRB%)) * (FTA - FT)
+ VOP * (1 - DRB%) * (TRB - ORB)
+ VOP * DRB% * ORB
+ VOP * STL
+ VOP * DRB% * BLK
- PF * ((lg_FT / lg_PF) - 0.44 * (lg_FTA / lg_PF) * VOP) ]

Most of the terms in the formula above should be clear, but let me define the less obvious ones:
factor = (2 / 3) - (0.5 * (lg_AST / lg_FG)) / (2 * (lg_FG / lg_FT))
VOP = lg_PTS / (lg_FGA - lg_ORB + lg_TOV + 0.44 * lg_FTA)
DRB% = (lg_TRB - lg_ORB) / lg_TRB

I'm sure if you look into Nate's model you'll find some kind of underlying equation that's similar to the one listed above since it'd be impossible to have a model without some kind of equation that calculates the outcome.

And you seem to be making the mistake that just because this formula has a lot of numbers and variables in it it's got to be scientific. For example assists in per is given a weight of 2/3. Why is an assist worth 2/3? It's not because the basketball God told you that this is the law of basketball. People came up with this value because it seems to capture the value of an assist but absolutely nobody can prove this value is correct. Whatever factors he used to account for how *good* a given poll is (because there's no way he gives them all equal weight, or he'd come to the same conclusion as anyone else) is still just a guess. Someone can easily go to the formula and change the scaling factor to 1 instead of 2/3 and there's nothing less scientific about that. You run the values again through this new formula and presumably one of those two value is better than the other in their predicative value. Further, all these weights are based on some kind of historical extrapolation and while history is useful it cannot predict everything. It is easy to see the NBA pass some rules that favors assists more than they do now and what would previously be correct scaling factor is no longer so. I believe one of the key difference this guy had was that he modeled who is likely to vote a lot different from the traditional models and it looked like he's right this time, because it turns out that voting behavior of Americans have changed significantly (minorities now count for far greater portion of the votes compared to other historical periods). This is behavior, not law. There isn't some law that tells you minorities now weight 0.7 instead of 0.6 compared to 20 years ago.

At best you can say the parameters he picked to model things like voter behavior are certainly very good, though he isn't the only statistical guy who got all 50 states right. Not to mention getting all 50 state right itself doesn't even validate or invalidate by itself. If there's a completely correct, all knowing model that says Obama has 51% chance to win Florida, you obviously should trust that over Nate's 'too close to call'. But if you picked Obama with a model that is all-knowing there's still a 49% chance you'd be wrong predicted by the model itself, and had Romney went on to win Florida it doesn't mean your all-knowing model was wrong, because that outcome had 49% chance of happening. By the way 90% chance of something happening is nowhere 'barring some miraculous whatever'. In XCOM if they tell your snapshot has a 90% chance to hit and then you miss you don't start questioning if the game lied to you because there's still a 10% chance to miss. Maybe if you missed every single time when it says you've 90% chance to hit you should question what's going on, but election is an event that's inherently unrepeatable. Had Romney won the election it wouldn't have invalidated the model because his chance of winning is small but nowhere at a point where you've to question the model itself. If Romney went on to win the election 10 times in a row then there's something wrong with the model but this event isn't even repeatable so it can't even be observed either.

Don wrote:I'd hope you're aware that he assigns different weights to different polls based on some measurement of their historical accuracy because you sure could not have come to the conclusion he did if you just averaged all the polls. And guess what this part is still guessing, no more than a guy running a basketball simulation believes you should assign a weight of 0.8 instead of 0.9 to this variable. None of these are anything beyond educated guesses because it's not like there's a physical law of basketball or election voting you can discover.

Yes, I am aware, and no, it's not an "educated guess". It's more than that. Getting 99 out of 100 right in the last 4 years is not an "educated guess". Comparing this to basketball is apples and oranges. Basketball has many different variables that affect how the game will turn out. Voting is a simple matter of putting down a binary choice.

And the polls TELL YOU exactly what people will put down as their binary choice. And you aren't polling the candidates, but the score makers themselves. The closer the day before the election, the more accurate the results.

To compare this to basketball, it would be like checking the score 10 seconds before the game was over. Unless there's a 3 point or less lead, you already know who's going to win.

Don wrote:I'm sure if you look into Nate's model you'll find some kind of underlying equation that's similar to the one listed above since it'd be impossible to have a model without some kind of equation that calculates the outcome.

State = Poll Results * Weight

And the weight is going to be damn near 1 in most cases, else it's worthless. Even Rasmussen doesn't have a wild swing like 0.5. It would be closer to 0.98 or something like that.