Politics!

This forum is founded on discussions about T Campbell's work (alone and with artist partners).

Moderators: TCampbell, Gisele

Re: Politics!

Postby Valerie » Fri Nov 02, 2012 10:49 pm

Alice Macher wrote:Via Facebook:


Image


<3
Lia S wrote:Valerie is right.

As usual.


TCampbell wrote:Val has a harem, but it's chiefly structured online at the moment.
User avatar
Valerie
 
Posts: 3994
Joined: Wed Sep 10, 2008 11:18 pm

Re: Politics!

Postby Mr. Brightside » Sat Nov 03, 2012 1:31 am

mindstalk wrote:If there are 6 domain experts opining about their domain, and 5 say X while 1 says Y, and you have no reason to believe any are bought out or deliberately biased, who do you bet on, the 5 or the 1?


I shuffle chips as I look into it further.

Where probabilities are concerned, with the factors in play and the obvious biases all around, the wisdom of crowds being what it is, I think I'm more inclined to trust gamblers to come to a fair line than I am pollsters to assess the odds correctly. Obama's doing better than he was a few days ago by both measures (the polls you've seen): Intrade's at 2:1 right now, Pinnacle and Betfair at 3:1. I get the feeling the Intrade crowd is more serious about making money, but I may just be brainwashed by their avoidance of the B and G words (then again, that might actually have attracted a more sensible market). Hardly a tossup in any case, but hardly a lock, nor the 5:1 Silver is claiming.

EDIT: It just occurred to me that the difference is almost certainly because Intrade is the only one that takes US clients. So yeah, I'd say Intrade has it.

mindstalk wrote:So a national poll is sampling 1000 people across the country. A poll about Ohio sampled 1000 people in Ohio. Ditto for all the other state polls. If there are 12 toss-up states, then in a polling cycle their polls have 12,000 people, vs. 1000 for the whole nation. And their predictive power is focused on "who will win", since the electoral college goes by state, while the national poll is 'wasting' time asking people in Texas or California how they'll vote, which is irrelevant for EC purposes; we know how those states are going to go.


All true. I wouldn't have brought up Gallup except that you had - the point was just that the difference illustrates how different demographic adjustments are affecting the poll results.

mindstalk wrote:Add the fact that Gallup is an outlier even among national tracking polls.

It's possible that polling is totally borked this year, across half a dozen or more companies, and Gallup is uniquely right. But it's possible that it's Gallup that's borked. Which is more likely? They're an old and familiar name in polling, but I don't know of anything that says they're uniquely qualified.


It's probably actually neither. Gallup was an outlier, true, but not by as much as it looks like: they stopped polling when the hurricane hit, at which time most of the pollsters had the popular vote breaking Romney.
(There has never been a signature.)
Mr. Brightside
 
Posts: 2026
Joined: Thu Jan 11, 2007 2:44 am

Re: Politics!

Postby thebitterfig » Sat Nov 03, 2012 6:57 am

Adrishiana wrote:
Pink Freud wrote:Yeah, Seth Meyers pretty much summed up my opinion on polls last week: "...people who still have a land line phone and pick-up calls from "Unknown Caller" on caller ID."


Sometimes the caller ID actually says "political call" or, recently, "REPUBLICANS." I laughed (but did not answer the phone).


Someone ought to get the Caller ID "DO NOT ANSWER." I mean, I'd totally pick up the phone.
The notes of this paradoxalist do not end here, however. He could not refrain from going on with them, but it seems to us that we may stop here. - Fyodor Dostoevsky, Notes from Underground (trans. C. Garnett)
User avatar
thebitterfig
 
Posts: 622
Joined: Mon Jan 09, 2012 4:25 pm
Location: Maine, where it's probably snowing.

Re: Politics!

Postby mindstalk » Sat Nov 03, 2012 4:20 pm

Mr. Brightside wrote:
mindstalk wrote:If there are 6 domain experts opining about their domain, and 5 say X while 1 says Y, and you have no reason to believe any are bought out or deliberately biased, who do you bet on, the 5 or the 1?


I shuffle chips as I look into it further.


Say you can't. How do you bet?

Where probabilities are concerned, with the factors in play and the obvious biases all around, the wisdom of crowds being what it is, I think I'm more inclined to trust gamblers to come to a fair line than I am pollsters to assess the odds correctly.


Why? What data are the gamblers using to make their bets, other than polls?

mindstalk wrote:Add the fact that Gallup is an outlier even among national tracking polls.

It's possible that polling is totally borked this year, across half a dozen or more companies, and Gallup is uniquely right. But it's possible that it's Gallup that's borked. Which is more likely? They're an old and familiar name in polling, but I don't know of anything that says they're uniquely qualified.


It's probably actually neither. Gallup was an outlier, true, but not by as much as it looks like: they stopped polling when the hurricane hit, at which time most of the pollsters had the popular vote breaking Romney.


...I'm not aware of any period when the polls were breaking to Romney, rather than getting really close.
User avatar
mindstalk
 
Posts: 601
Joined: Mon Jan 18, 2010 10:02 am

Re: Politics!

Postby Mr. Brightside » Sat Nov 03, 2012 5:43 pm

mindstalk wrote:
Where probabilities are concerned, with the factors in play and the obvious biases all around, the wisdom of crowds being what it is, I think I'm more inclined to trust gamblers to come to a fair line than I am pollsters to assess the odds correctly.


Why? What data are the gamblers using to make their bets, other than polls?


It's not about information. The job of pollsters and poll aggregators isn't to assess probabilities so much as to give a broad sense of them - when they say "this one's a 5:1 favorite, this one a 2:1 favorite, this one a 14:1 favorite, and this one a 9:2 favorite" (the latter two being Silver's and Intrade's respective lines for Warren), they have no incentive to be accurate. When the election's over, probably someone, somewhere is going to add up all the upsets and see how Silver's probabilities stack up, but no one's going to care. Then on the other hand, you've got people who live and die by it.

mindstalk wrote:...I'm not aware of any period when the polls were breaking to Romney, rather than getting really close.


Honestly, I tried to hunt it down, but all I can find now is Rasmussen at Romney +4; most of the tracking polls have been overwritten. Silver's aggregate gets within 1%; the RCP aggregate turns to 1 over. I'm having a hard time getting into Silver's method, though, where the popular vote is concerned (the state-by-state's much more transparent), getting more a sense of "this is my analysis, which is mine."
(There has never been a signature.)
Mr. Brightside
 
Posts: 2026
Joined: Thu Jan 11, 2007 2:44 am

Re: Politics!

Postby mindstalk » Sat Nov 03, 2012 6:16 pm

Mr. Brightside wrote:
mindstalk wrote:
Where probabilities are concerned, with the factors in play and the obvious biases all around, the wisdom of crowds being what it is, I think I'm more inclined to trust gamblers to come to a fair line than I am pollsters to assess the odds correctly.


Why? What data are the gamblers using to make their bets, other than polls?


It's not about information. The job of pollsters and poll aggregators isn't to assess probabilities so much as to give a broad sense of them - when they say "this one's a 5:1 favorite, this one a 2:1 favorite, this one a 14:1 favorite, and this one a 9:2 favorite" (the latter two being Silver's and Intrade's respective lines for Warren), they have no incentive to be accurate. When the election's over, probably someone, somewhere is going to add up all the upsets and see how Silver's probabilities stack up, but no one's going to care. Then on the other hand, you've got people who live and die by it.


I doubt anyone's living or dying by betting markets yet. And aggregator reputations are based on their accuracy; it's pontificating pundits who get to bloviate without check. Poll watchers make state by state predictions too, so a single election actually gives lot of data by which to evaluate their accuracy. Who adds up their upsets? *They* do.

Honestly, I tried to hunt it down, but all I can find now is Rasmussen at Romney +4; most of the tracking polls have been overwritten. Silver's aggregate gets within 1%; the RCP aggregate turns to 1 over. I'm having a hard time getting into Silver's method, though, where the popular vote is concerned (the state-by-state's much more transparent), getting more a sense of "this is my analysis, which is mine."


Yeah, Silver's pretty complicated, maybe too much so. But then, simple averaging like electoral-vote or realclearpolitics give similar results. Silver talks about Gallup here http://fivethirtyeight.blogs.nytimes.co ... the-world/

http://cheddarfortress.blogspot.com/201 ... -isnt.html says Gallup claims a huge PV lead for Romney in the South... which he's going to win anyway. So even if it's true it'd boost the likelihood of a split result, but not of Romney winning the EC.
User avatar
mindstalk
 
Posts: 601
Joined: Mon Jan 18, 2010 10:02 am

Re: Politics!

Postby thebitterfig » Sun Nov 04, 2012 11:10 am

Mr. Brightside wrote:It's not about information. The job of pollsters and poll aggregators isn't to assess probabilities so much as to give a broad sense of them - when they say "this one's a 5:1 favorite, this one a 2:1 favorite, this one a 14:1 favorite, and this one a 9:2 favorite" (the latter two being Silver's and Intrade's respective lines for Warren), they have no incentive to be accurate. When the election's over, probably someone, somewhere is going to add up all the upsets and see how Silver's probabilities stack up, but no one's going to care. Then on the other hand, you've got people who live and die by it.


I'd say Silver's got a pretty big incentive to be accurate: readership. If he's dead wrong, if his presumptions are bogus, people won't read his blog. He's doing this from a statistical model direction, which means that the structure of estimate methodology needs to be properly working. If it's shown to be nonsense, then the grab he's using won't grab readers.

Of course, Nate's probably got a different definition of accurate than the rest of us. Even an 85%/15% probability split reduces to a phrase I'm sure anyone who's played RPGs or tabletop miniature games understands: Don't Roll Ones (16.7% chance on a d6). I'm far to familiar with Don't Roll Ones to see a Romney win as a complete repudiation of the model. If that's the outcome on Nov. 6 or later (since it might take a while...), then Silver's task is to go into the weeds of his model, see what aspects were right and what were wrong. Tweak a few presumptions, see how this impacts the model, and if the estimates now match the election. If he's right about 19 out of 20 internal factors, and that 1 factor causes the electoral outcome to flip, that's not really proof the model is bullshit.

An example of someone who doesn't give a damn about accuracy and the consequences of it's lack is Larry Cramer. He's got a prediction of a 440/98 Obama win in the electoral college. That pretty much means the old south needs to be Red/Blue swiss cheese, and Obama sweeps the southwest except for Utah. That's just complete nonsense.
The notes of this paradoxalist do not end here, however. He could not refrain from going on with them, but it seems to us that we may stop here. - Fyodor Dostoevsky, Notes from Underground (trans. C. Garnett)
User avatar
thebitterfig
 
Posts: 622
Joined: Mon Jan 09, 2012 4:25 pm
Location: Maine, where it's probably snowing.

Re: Politics!

Postby Trefle » Sun Nov 04, 2012 11:34 am

...2 days left.

Just coming here to say, good luck, Americans. I can only pray the righteous one win.
Trefle
 
Posts: 961
Joined: Fri Aug 22, 2008 2:16 am

Re: Politics!

Postby Mr. Brightside » Sun Nov 04, 2012 3:45 pm

thebitterfig wrote:I'd say Silver's got a pretty big incentive to be accurate: readership. If he's dead wrong, if his presumptions are bogus, people won't read his blog. He's doing this from a statistical model direction, which means that the structure of estimate methodology needs to be properly working. If it's shown to be nonsense, then the grab he's using won't grab readers.

Of course, Nate's probably got a different definition of accurate than the rest of us. Even an 85%/15% probability split reduces to a phrase I'm sure anyone who's played RPGs or tabletop miniature games understands: Don't Roll Ones (16.7% chance on a d6). I'm far to familiar with Don't Roll Ones to see a Romney win as a complete repudiation of the model. If that's the outcome on Nov. 6 or later (since it might take a while...), then Silver's task is to go into the weeds of his model, see what aspects were right and what were wrong. Tweak a few presumptions, see how this impacts the model, and if the estimates now match the election. If he's right about 19 out of 20 internal factors, and that 1 factor causes the electoral outcome to flip, that's not really proof the model is bullshit.


That's just it. People aren't really going to be looking at his probabilities, only his margins. For that reason, I'm guessing his model leans blue more than Intrade leans red.

(This thread brought to you by my abortive attempt to contribute to the Halloween thread involving a vampire playing poker. And Cyndi.)

thebitterfig wrote:An example of someone who doesn't give a damn about accuracy and the consequences of it's lack is Larry Cramer. He's got a prediction of a 440/98 Obama win in the electoral college. That pretty much means the old south needs to be Red/Blue swiss cheese, and Obama sweeps the southwest except for Utah. That's just complete nonsense.


Okay, that... yeah.
(There has never been a signature.)
Mr. Brightside
 
Posts: 2026
Joined: Thu Jan 11, 2007 2:44 am

Re: Politics!

Postby Adrishiana » Sun Nov 04, 2012 4:54 pm

Seriously, Jon Husted, quit trying to gift wrap Ohio for Romney.
No more zoos!
User avatar
Adrishiana
 
Posts: 1457
Joined: Thu Jan 11, 2007 8:37 am

Re: Politics!

Postby thebitterfig » Sun Nov 04, 2012 10:26 pm

Voter suppression is fucking sickening.

Anyhow, out of 12 national polls released today, 9 were Obama-leading, 3 tied. Saturday's battleground polls had 16 Obama leads, 3 ties, 2 Romney leads.

Parsing out the math, Wisconsin, Nevada, Iowa, and Pennsylvania are all pretty solid Obama. North Carolina and Indiana are solid Romney. New Hampshire is tied but essentially irrelevant. Colorado is a wildcard - will the pot voters show up? If they do will they vote Obama (he's been pretty shitty on medical marijuana)? It comes down to Virginia, Florida, and Ohio. Romney needs to sweep all three. Obama leads Ohio by about 2.5% in most polls, Virginia by 1%, Romney leads Florida by 1%. I'm going to guess Romney wins Colorado (polling sorta leans Obama, but Romney's had good early voting results so far), Florida; Obama wins NH, Ohio, and Virginia; 294-244 electoral college map.
The notes of this paradoxalist do not end here, however. He could not refrain from going on with them, but it seems to us that we may stop here. - Fyodor Dostoevsky, Notes from Underground (trans. C. Garnett)
User avatar
thebitterfig
 
Posts: 622
Joined: Mon Jan 09, 2012 4:25 pm
Location: Maine, where it's probably snowing.

Re: Politics!

Postby mindstalk » Mon Nov 05, 2012 12:36 am

Huh, just learned that Nate Silver is gay.

Not too relevant in the grand scheme of things, but human interesty, especially for this site where we're talking politics in the shadow of gay-friendly comics. Also, he's been attacked (verbally) recently for being effeminate.
User avatar
mindstalk
 
Posts: 601
Joined: Mon Jan 18, 2010 10:02 am

Re: Politics!

Postby thebitterfig » Mon Nov 05, 2012 7:29 am

Oh, if Tuesday rolls around and you're still not tired of voting, you can vote Vikings punter Chris Kluwe to the probowl.

http://www.nfl.com/probowl/ballot?modul ... ng_probowl

He's only an average punter, but fuck, what do punters matter anyhow. He's one of the few football players sticking up for equality.

http://www.huffingtonpost.com/2012/11/0 ... ref=topbar
The notes of this paradoxalist do not end here, however. He could not refrain from going on with them, but it seems to us that we may stop here. - Fyodor Dostoevsky, Notes from Underground (trans. C. Garnett)
User avatar
thebitterfig
 
Posts: 622
Joined: Mon Jan 09, 2012 4:25 pm
Location: Maine, where it's probably snowing.

Re: Politics!

Postby Mr. Brightside » Mon Nov 05, 2012 4:59 pm

As a favor to the dying scientist of my youth, I'm going to give myself an opportunity to get egg on my face. Here are Silver's odds against an upset, where he gives a chance less than 98% one way or the other, to two sig figs:

Nevada: 9:1
Colorado: 2.3:1
Iowa: 4.3:1
Wisconsin: 17:1
Ohio: 6.6:1
New Hampshire: 4.1:1
Maine 2: 9.8:1
Warren: 15:1
Murphy: 12:1
Casey: 29:1
Kaine: 5.7:1
Manchin: 8.7:1
(Sherrod) Brown: 28:1
Donnelly: 2.1:1
Baldwin: 3.4:1
McCaskell: 7.5:1
Heinrich: 13:1


North Carolina: 3.4:1
Florida: 1.2:1
Nebraska 2: 7.1:1
Heller: 3.3:1
Flake: 4.1:1
Rehberg: 2.2:1
Berg: 8.5:1


King (N/A): 12:1

I'm going to guess that a hypothetical gambler putting down at these odds some amount of money on every blue upset, plus King (since his Republican opponent is leading the Democratic by a wide margin), then using the equivalent total amount of money to lay the red proportionally to the odds offered, will win a fair amount. That is to say, if Kenny Rogers stakes one boondollar each on every bet in blue above, and the black, for a total of eighteen, then lays 2.05 against Obama taking NC, 0.72 FL, 4.29 Nebraska 2, 1.99 against Shelly Berkley, 2.48 against Richard Carmona, 1.33 against Jon Tester, and 5.13 against Heidi Heitkamp, I'm fairly sure he'll come out ahead if he doesn't break even. (...I take the leftover cent as vig.)
(There has never been a signature.)
Mr. Brightside
 
Posts: 2026
Joined: Thu Jan 11, 2007 2:44 am

Re: Politics!

Postby retrophrenologist » Mon Nov 05, 2012 6:18 pm

thebitterfig wrote:
Adrishiana wrote:
Pink Freud wrote:Yeah, Seth Meyers pretty much summed up my opinion on polls last week: "...people who still have a land line phone and pick-up calls from "Unknown Caller" on caller ID."


Sometimes the caller ID actually says "political call" or, recently, "REPUBLICANS." I laughed (but did not answer the phone).


Someone ought to get the Caller ID "DO NOT ANSWER." I mean, I'd totally pick up the phone.

"IT'S COMING FROM INSIDE THE HOUSE"
"I am the spirit that denies. All that you call sin and destruction-- or evil, if you will-- that is my proper element."
retrophrenologist
 
Posts: 1485
Joined: Fri May 22, 2009 12:37 am
Location: Melbourne, Australia

PreviousNext

Return to General Discussion

Who is online

Users browsing this forum: Google [Bot] and 6 guests

cron