Baseball for the Thinking Fan

Login | Register | Feedback

btf_logo
You are here > Home > Baseball Newsstand > Baseball Primer Newsblog > Discussion
Baseball Primer Newsblog
— The Best News Links from the Baseball Newsstand

Wednesday, October 31, 2012

OTP November 2012 - Moneypoll! The Pundits vs. The Election-Data Nerds

Come next Tuesday night, we’ll get a resolution (let’s hope) to a great ongoing battle of 2012: not just the Presidential election between Barack Obama and Mitt Romney, but the one between the pundits trying to analyze that race with their guts and a new breed of statistics gurus trying to forecast it with data.

In Election 2012 as seen by the pundits–political journalists on the trail, commentators in cable-news studios–the campaign is a jump ball. There’s a slight lead for Mitt Romney in national polls and slight leads for Barack Obama in swing-state polls, and no good way of predicting next Tuesday’s outcome beyond flipping a coin. ...

Bonus link: Esquire - The Enemies of Nate Silver

Joe Kehoskie Posted: October 31, 2012 at 11:42 PM | 11298 comment(s) Login to Bookmark
  Tags: mr president, off-topic, politics, sabermetrics, usa

Reader Comments and Retorts

Go to end of page

Statements posted here are those of our readers and do not represent the BaseballThinkFactory. Names are provided by the poster and are not verified. We ask that posters follow our submission policy. Please report any inappropriate comments.

Page 11 of 114 pages ‹ First  < 6 7 8 9 10 11 12 13 14 15 16 >  Last ›
   1001. snapper (history's 42nd greatest monster) Posted: November 03, 2012 at 02:04 PM (#4291737)
Missing the turnout ratio is a systemic error and not an error on the part of aggregators.

Well the aggregators could adjust the polls for their own turnout estimates.

But, it's not a systematic error, to my way of thinking. You're asking the pollsters to know the unknowable.
   1002. Joe Kehoskie Posted: November 03, 2012 at 02:04 PM (#4291738)
So has anyone in this thread gone on record as predicting a Romney win? (I couldn't check the thread at work yesterday and am too "busy" to go through and try and figure it out)

I meant to do this yesterday but I have way too many windows open and never got back to it.

This is my prediction for the 2012 presidential election:

Romney - 315
Obama - 223

I've been predicting a Romney win for months and feel increasingly confident that he'll win. I know the polling is slightly adverse for him right now, but I just can't see Obama overcoming all of the headwinds he faces, from the economy to unemployment to the enthusiasm gap.

(By the way, I know my map is the same as Michael Barone's. I thought about changing a state or two just to be different, but I really don't want to.)
   1003. The Id of SugarBear Blanks Posted: November 03, 2012 at 02:04 PM (#4291739)
Intrade's a good indicator -- real money put on the line by real people with an interest in getting it right -- and it has Obama having about a 66% chance of winning.
   1004. Jolly Old St. Nick Still Gags in October Posted: November 03, 2012 at 02:05 PM (#4291740)
Sure, I know Obama is going to win, I agree with Silver there. But if you believe that you can put the chance of Obama winning at 83.7%, and believe it means something, you deserve to get beaten up in a dark ally by Nassim N. Taleb.


So if Nate said something like OBama has roughly a 75 to 85% chance of winning the election based on the polls he would deserve a pat on the back instead of deserving a beating in a back alley?

Nate can explain his model a thousand times, and some people are going to think that that 83.7% means that Obama is going to win 83.7% of the vote, rather than winning 837 out of 1000 elections where the data are where they are at this point. Of course the micro-precision is to an extent a gimmick, but it's also what his computer model projects. Should he then change that number just to please people who don't understand what he's doing?

EDIT:
I guarantee you some people will wake up Wednesday, if Obama wins with 50.1% of the vote to 49.9%, and say "Hah! That fanboy Nate Silver said it would be 83 to 17!"

Exactly. As if the fault there lies with Nate, rather than with people too clueless (or too lazy, or too ideologically blind, or all of the above) even to grasp what he's patiently explained over and over again.
   1005. greenback likes millwall Posted: November 03, 2012 at 02:05 PM (#4291742)
So, I just looked at Silver's blog. HE'S USING THREE ####### SIGNIFICANT DIGITS!!! The man is a fraud and a charlatan. And using four digits for the elector seats is stupid on several levels.

It's over-precise to be sure, but I doubt it's stupid. Silver demonstrated back in his PECOTA days that he's got a good feel for the kind of #### that the sophisticated-but-not-too-sophisticated fan loves to see (90th percentiles! breakout rates! collapse rates!). It's eye candy really, but from my standpoint it's an educational opportunity for how to present lots of numbers in a way that doesn't make the target audience's eyes glaze over.

Sure, I know Obama is going to win, I agree with Silver there. But if you believe that you can put the chance of Obama winning at 83.7%, and believe it means something, you deserve to get beaten up in a dark ally by Nassim N. Taleb.

There's got to be some way to make fun of the typo here.
   1006. AROM Posted: November 03, 2012 at 02:05 PM (#4291743)
So then what is the point of his site? Should he lie? Should he say the race is 50/50 to make people happy? Should he say the race is 60/40 because people can't understand odds?


If I were to lower myself to predicting elections instead of baseball stats, the way I'd do it is show an electoral map and the projected EC totals for each candidate, and leave the percentages off entirely.

I haven't followed the Nate Silver saga, but what added value does he claim to bring over just looking at the polls themselves?


Time. You can do what I did for half a morning, look up the poll numbers in 50 states, put them in a spreadsheet, and run the math yourself. If you have few hours. Or you could just take a few seconds, look at his site, and see pretty much the same conclusion.

Now what value does he add over RCP? Nothing that I can see. As I compared earlier, the only difference is that Nate's model has Virgina going democrat and RCP has the state going republican by a margin of 0.1 percent. So there's a 50/50 shot that if another poll is done in Virginia, all else being equal, the two models will be saying the exact same thing. As it is, Virginia by itself doesn't come close to swaying the election.

One difference though is that RCP is using a simple average of polls, and Nate has all the black boxy adjustments. If you get pretty much the same result, I prefer the simpler method, sort of like some of the recent articles about using FIP or even kwERA (just strikeouts and walks) instead of some complex ERA estimator like SIERA.

It's not just the simplicity though. If a specific poll has shown a bias in one direction in the past, don't you think the people who run that poll will see it and try to correct it? Probably more likely now than in the past, as poll aggregating has become more prevalent (which of course Nate is responsible for a lot of that.) Whatever adjustments worked 4 years ago may not work today.
   1007. Swedish Chef Posted: November 03, 2012 at 02:06 PM (#4291744)
So if Nate said something like OBama has roughly a 75 to 85% chance of winning the election based on the polls he would deserve a pat on the back instead of deserving a beating in a back alley?

Even better, knowing about significant digits could help him not fail lab courses in college. Seriously, this is basic stuff in statistics, he should know better.

Taleb would probably still think that he was a fool though.
   1008. Joe Kehoskie Posted: November 03, 2012 at 02:07 PM (#4291745)
Bingo. As I said before requiring that 20 to 30% of your respondents saying they aren't voting for you to have an accurate poll is beyond stupid and is only being argued for because some people don't like what the polls are saying. What matters is what the LV will do and it doesn't matter how many non voting pollees you go through to get your n sample.

At some point, I'm hoping McCoy will explain how an LV screen projecting 96 percent turnout isn't really an LV screen projecting 96 percent turnout.

If you talk to 1,000 RV and then pass through 960 of them as LV, then you're projecting a 96 percent turnout. This is basic math.
   1009. Monty Predicts a Padres-Mariners WS in 2016 Posted: November 03, 2012 at 02:07 PM (#4291746)
We don't have the data. The issue is, the number of incumbent Presidents who have run for re-election in poor economies is not large enough to make a statistical case. Elections like Carter-Reagan, and Bush-Clinton show the phenomenon.

You're trying to force "sabermetrics" into a situation where you don't have enough data, and should rely on "scouting" instead.


I think of 538 as an experiment. We're trying to find out if we have enough data to accurately predict election results based solely on the poll data. What's the harm in trying?
   1010. spike Posted: November 03, 2012 at 02:09 PM (#4291747)
All it really requires is money

Excellent point - in addition to the lack of available generators, difficulty in turning a gas station to generator power, and difficulty in getting a generator to a site, you'd also have ot pay some pretty significant up-front cost that a gas station owner likely wouldn't just have lying around. And then there's the fact that your window would only be a few days - as soon as the city powers up your competitors your generator is useless.
   1011. McCoy Posted: November 03, 2012 at 02:09 PM (#4291748)
But, it's not a systematic error, to my way of thinking. You're asking the pollsters to know the unknowable.

But you are then also asking the aggregators to know the unknowable as well. The aggregators are taking the data as supplied by the polls to come up with a number that incorporates all of the polls. If the polls make the wrong assumptions then the aggregators make those same wrong assumptions. Thus if the aggregators get it wrong, and they are all pointing to an Obama win, that means the majority of the polls got it wrong. It isn't like the aggregators are saying Obama is going to win while the majority of polls are saying Romney is going to win.

It might very well be a coin toss but it is a coin toss that almost all of the polls say Obama is going to win. That is something that is going to get picked up on by the aggregators and it is also why the aggregators are even more important during close elections like this one. People think this election is a toss up and in all probability it is not.
   1012. Morty Causa Posted: November 03, 2012 at 02:09 PM (#4291749)
Self-defense is a natural right, democracy isn't.


I see what you did to distort the issue. Going along with that, though, democracy is as much a natural right (as nebulous as that is) as self-defense if we see democracy as having a right to a say in one's destiny. Actually, they become inextricably intertwined. Thus it was so, even at the tribal level.
   1013. greenback likes millwall Posted: November 03, 2012 at 02:10 PM (#4291750)
Intrade's a good indicator -- real money put on the line by real people with an interest in getting it right -- and it has Obama having about a 66% chance of winning.

Intrade almost certainly is influenced by the polling aggregators. Maybe Silver could build a model for it.
   1014. Jolly Old St. Nick Still Gags in October Posted: November 03, 2012 at 02:12 PM (#4291751)
One difference though is that RCP is using a simple average of polls, and Nate has all the black boxy adjustments.

The biggest difference is that on the page that 95% of its visitors likely look at, RCP's average reflects only the national polls. They may wind up in roughly the same place, but they're not really the same thing at all.
   1015. McCoy Posted: November 03, 2012 at 02:12 PM (#4291753)
If I were to lower myself to predicting elections instead of baseball stats, the way I'd do it is show an electoral map and the projected EC totals for each candidate, and leave the percentages off entirely.

So how would you handle a state where the two candidates are seperated by a point or less? This is what Nate is doing. He's taking hard to read states and information and crunching the numbers to come up with a probability of the candidate winning the state and the whole enchilada.
   1016. snapper (history's 42nd greatest monster) Posted: November 03, 2012 at 02:13 PM (#4291754)
I think of 538 as an experiment. We're trying to find out if we have enough data to accurately predict election results based solely on the poll data. What's the harm in trying?

No harm in trying. It just makes you look silly to evince false precision while you're trying to figure it out.
   1017. AROM Posted: November 03, 2012 at 02:15 PM (#4291756)
Joe 1002,

Here are the differences between your map and Silver's, followed by the RCP poll margins:

CO D+1.0
IA D+2.0
NH D+1.8
OH D+2.9
PA D+4.6
WI D+4.6

If that happens it pretty much means the polls are too f***ed to be useful in any semi-close state.
   1018. snapper (history's 42nd greatest monster) Posted: November 03, 2012 at 02:15 PM (#4291757)
I see what you did to distort the issue. Going along with that, though, democracy is as much a natural right (as nebulous as that is) as self-defense if we see democracy as having a right to a say in one's destiny. Actually, they become inextricably intertwined. Thus it was so, even at the tribal level.

Not really. We prevent a 16-y.o. from voting, but he still has the right to kill an assailant in self-defense.

The vast, vast majority of societies in history have not been "one-man one-vote" democracies, yet they still, almost universally, recognized the right of self-defense.
   1019. McCoy Posted: November 03, 2012 at 02:17 PM (#4291758)
No harm in trying. It just makes you look silly to evince false precision while you're trying to figure it out.

The only people who think of it as false precision are the people who know the number isn't really precise while the people who think it is precise are too stupid (about math and probability) to grasp a less precise number.
   1020. snapper (history's 42nd greatest monster) Posted: November 03, 2012 at 02:17 PM (#4291759)
Intrade almost certainly is influenced by the polling aggregators. Maybe Silver could build a model for it.

Just like elections, and turnout are likely influenced by polls. We've got a big observer effect going on.
   1021. Monty Predicts a Padres-Mariners WS in 2016 Posted: November 03, 2012 at 02:17 PM (#4291760)
No harm in trying. It just makes you look silly to evince false precision while you're trying to figure it out.


See, I think he only looks silly if it turns out he's wrong. Once that happens, feel free to heap abuse on him. But in the last couple elections, he's been pretty close.

And you don't know it's false precision. You're just asserting that it is.
   1022. McCoy Posted: November 03, 2012 at 02:18 PM (#4291761)
What is it about 3 and 4 way ties in Diplomacy that make them a natural right for players?
   1023. snapper (history's 42nd greatest monster) Posted: November 03, 2012 at 02:18 PM (#4291762)
while the people who think it is precise are too stupid (about math and probability) to grasp a less precise number.

They're too stupid to understand "likely Obama victory"?
   1024. Howie Menckel Posted: November 03, 2012 at 02:19 PM (#4291764)

Christie just said that he is allowing early voting this weekend due to the crisis.

Also, he said that the real problem in northern NJ is that too many of those gas stations aren't equipped to handle a generator. He said the ones that are equipped, are getting generator help, not sure if that's state or federal.

So the supply indeed is here, and the issue isn't having enough generators as much as stations having the current capability to have one brought in, if I understood it all correctly.

   1025. Joe Kehoskie Posted: November 03, 2012 at 02:22 PM (#4291765)
Here are the differences between your map and Silver's, followed by the RCP poll margins:

CO D+1.0
IA D+2.0
NH D+1.8
OH D+2.9
PA D+4.6
WI D+4.6

If that happens it pretty much means the polls are too f***ed to be useful in any semi-close state.

The topline poll numbers might be wrong, but that doesn't mean the polls were useless. People laugh at "poll truthers," but all they do is take a set of polls and play with the party ID split and/or the turnout rate.

Beyond that, in four of the above states, the polling is within the margin of error. If Nate feels confident enough to make projections based on such thin polling margins, then he shouldn't complain about bad polling data afterward, especially if an Obama +1.0 poll turns into a Romney +1.0 victory.
   1026. Tilden Katz Posted: November 03, 2012 at 02:23 PM (#4291766)
People laugh at "poll truthers," but all they do is take a set of polls and play with the party ID split and/or the turnout rate.


I.e. change the results so the polls show what you want them to show.
   1027. Joe Kehoskie Posted: November 03, 2012 at 02:27 PM (#4291768)
I.e. change the results so the polls show what you want them to show.

How is that any different than pollsters claiming a 96 percent LV screen is representative of the entire electorate?
   1028. McCoy Posted: November 03, 2012 at 02:28 PM (#4291769)
They're too stupid to understand "likely Obama victory"?

Yes. Hell, I'm too stupid to grasp what that would actually means.
   1029. Monty Predicts a Padres-Mariners WS in 2016 Posted: November 03, 2012 at 02:29 PM (#4291771)
How is that any different than pollsters claiming a 96 percent LV screen is representative of the entire electorate?


They're reporting the results of the polls. You're reporting what you think the results of the polls should have been. Maybe the polls really are wrong, but they're clearly different approaches.
   1030. greenback likes millwall Posted: November 03, 2012 at 02:30 PM (#4291773)
How is that any different than pollsters claiming a 96 percent LV screen is representative of the entire electorate?

Are they coming up with that screen before or after their model spit out the results?
   1031. Joe Kehoskie Posted: November 03, 2012 at 02:31 PM (#4291776)
They're reporting the results of the polls. You're reporting what you think the results of the polls should have been. Maybe the polls really are wrong, but they're clearly different approaches.

So "poll truthers" are dumb to theorize that the 2012 electorate will be a couple points less Dem than 2008, but it's OK for pollsters to claim there will be 96 percent turnout despite turnout in 2008 being ~72 percent? Who's taking the bigger statistical leap here?
   1032. Tilden Katz Posted: November 03, 2012 at 02:32 PM (#4291779)
Is there any evidence that the LV screen is different from what it was in 2008, when the polls were highly accurate?
   1033. Tilden Katz Posted: November 03, 2012 at 02:33 PM (#4291780)
So "poll truthers" are dumb


Yes
   1034. Chokeland Bill Posted: November 03, 2012 at 02:36 PM (#4291783)
Are they coming up with that screen before or after their model spit out the results?


Most of them don't screen by party. They screen based on race, age, and gender.
   1035. Monty Predicts a Padres-Mariners WS in 2016 Posted: November 03, 2012 at 02:37 PM (#4291785)
So "poll truthers" are dumb to theorize that the 2012 electorate will be a couple points less Dem than 2008, but it's OK for pollsters to claim there will be 96 percent turnout despite turnout in 2008 being ~72 percent? Who's taking the bigger statistical leap here?


I didn't call anyone dumb. You asked how what the pollsters were doing was different from what you're doing. The answer is: they're asking people who they're going to vote for and if they're going to vote, and then basing their reports on that. And you're regressing the answers much more heavily, based on historical results. Do you not see how there's a fundamental difference in your approaches?
   1036. Joe Kehoskie Posted: November 03, 2012 at 02:37 PM (#4291786)
Is there any evidence that the LV screen is different from what it was in 2008, when the polls were highly accurate?

There's not a standard LV screen. For PPP, their entire LV screen is the statement, "If you don't plan to vote, please hang up now." For Gallup, it's much more intricate.
   1037. Jim Wisinski Posted: November 03, 2012 at 02:40 PM (#4291787)
but it's OK for pollsters to claim there will be 96 percent turnout despite turnout in 2008 being ~72 percent?


I have a hard time believing that you're actually so dense and so hyper-partisan that you can't/won't understand why 96% of poll respondents a few days before an election are labeled as likely voters. If it's not party ID it's LV screens or biased polling or whatever, anything to argue that Romney is actually leading the race. Never mind that the reason that Obama might be outperforming fundamentals is that Romney is a terrible candidate running on a terrible platform of more money for the rich, less help for the poor, more wars in the Middle East, interference in women's healthcare, and regression in social progress. If the fundamentals were neutral for Obama he would be crushing Romney because the right wouldn't be able to use scary economic doom-saying to hide how their platform is crap for the majority of Americans.
   1038. Joe Kehoskie Posted: November 03, 2012 at 02:42 PM (#4291789)
I didn't call anyone dumb.

I know. I used your comment as a jumping-off point for a more general reply. "Poll truthers" have been a hot topic here, as evidenced by other replies.

You asked how what the pollsters were doing was different from what you're doing. The answer is: they're asking people who they're going to vote for and if they're going to vote, and then basing their reports on that. And you're regressing the answers much more heavily, based on historical results. Do you not see how there's a fundamental difference in your approaches?

No. If a pollster asks 1,000 RV if they plan to vote and 96 percent of them say yes, and the pollster knows that 2008's very high turnout was only ~72 percent, then the pollster is taking a huge leap of faith. Again, there's a big difference between a poll being accurate for those specific people and pretending that the sample is representative of the entire electorate.

No matter how you slice it, an LV screen that passes 96 percent of RV respondents is projecting a 96 percent turnout rate.
   1039. Random Transaction Generator Posted: November 03, 2012 at 02:42 PM (#4291790)
We have hundreds of thousands of data points.

Pardon? We've had <25 Presidential elections since nationwide polling was even attempted.


So, I just looked at Silver's blog. HE'S USING THREE ####### SIGNIFICANT DIGITS!!! The man is a fraud and a charlatan. And using four digits for the elector seats is stupid on several levels.


You do realize that Silver runs hundreds of thousands of (I assume Monte Carlo method) election simulations, right?
That's why his numbers have significant digits.
He's not saying that someone is going to win a fraction of an elector seat. He's saying that the "average" result is xx.x% or xxx.x EV.

I really thought that was obvious for anyone that paid even a small amount of attention to what Silver says/does.

   1040. McCoy Posted: November 03, 2012 at 02:44 PM (#4291791)
but it's OK for pollsters to claim there will be 96 percent turnout despite turnout in 2008 being ~72 percent?

I'm guessing Joe said this. 96% LV screen does not mean there will be a 96% turnout or even that the poll thinks there will be a 96% turnout. It is either beyond stupid to think that or Joe trolling again which is why I ignore his posts.
   1041. McCoy Posted: November 03, 2012 at 02:44 PM (#4291792)
Pardon? We've had <25 Presidential elections since nationwide polling was even attempted.

Every single person who takes part in a poll is a data point.
   1042. Joe Kehoskie Posted: November 03, 2012 at 02:45 PM (#4291794)
I have a hard time believing that you're actually so dense and so hyper-partisan that you can't/won't understand why 96% of poll respondents a few days before an election are labeled as likely voters. If it's not party ID it's LV screens or biased polling or whatever, anything to argue that Romney is actually leading the race. Never mind that the reason that Obama might be outperforming fundamentals is that Romney is a terrible candidate running on a terrible platform of more money for the rich, less help for the poor, more wars in the Middle East, interference in women's healthcare, and regression in social progress. If the fundamentals were neutral for Obama he would be crushing Romney because the right wouldn't be able to use scary economic doom-saying to hide how their platform is crap for the majority of Americans.

Is there a math argument in there or just a general rant?

For about the tenth time, the only way a poll with an LV screen that passes 96 percent of RV through as LV is representative of the entire electorate is if the pollster is projecting 96 percent turnout.
   1043. Jim Wisinski Posted: November 03, 2012 at 02:46 PM (#4291796)
To be fair, a lot of blame for the decent success of the Romney campaign needs to be directed at the Democratic Party and Obama campaign because they have generally done a pretty lousy job of effectively explaining how bad the Republican platform is for a lot of people and they've allowed Romney to jump back to the middle without doing much to highlight his flip-flopping and the obvious unreliability of anything he says.
   1044. Joe Kehoskie Posted: November 03, 2012 at 02:48 PM (#4291797)
I'm guessing Joe said this. 96% LV screen does not mean there will be a 96% turnout or even that the poll thinks there will be a 96% turnout. It is either beyond stupid to think that or Joe trolling again which is why I ignore his posts.

Please explain the math of a 96 percent LV screen not translating into 96 percent turnout.
   1045. greenback likes millwall Posted: November 03, 2012 at 02:48 PM (#4291798)
No matter how you slice it, an LV screen that passes 96 percent of respondents is projecting a 96 percent turnout rate.

You can't be this dense.

You do realize that Silver runs hundreds of thousands of (I assume Monte Carlo method) election simulations, right?

He can run a billion simulations of a coin flip, and he won't get exactly 50%, but it's not particularly useful to publish a number to five decimal places.
   1046. Matt Clement of Alexandria Posted: November 03, 2012 at 02:49 PM (#4291799)
The gap between the LV/RV rate in polls and the voted/didn't vote rate in the general population probably has a lot to do with response rates. People who can't be bothered to vote are significantly less likely to respond to poll.
   1047. McCoy Posted: November 03, 2012 at 02:50 PM (#4291800)
What more do they need to say in order for people to understand that Romney has no real plan? At this point in the game the only people who think Romney or Obama have some great plan or a good plan, or even a realistic plan are biased partisans. People are basically voting for the lesser of two evils at this point.
   1048. McCoy Posted: November 03, 2012 at 02:51 PM (#4291802)
You can't be this dense

Yes, yes he can.
   1049. Monty Predicts a Padres-Mariners WS in 2016 Posted: November 03, 2012 at 02:52 PM (#4291803)
My prediction for the election results:

Obama wins the popular vote by 2-4 points and the Electoral College with 304-307 votes.

I determined the EC number through the scientific means of "looking at FiveThirtyEight, which currently says 305.3, and adding an arbitrary two-vote margin." I am aware that the votes come in bigger chunks than that, but this was as much work as I was willing to do.

Oh! Also, I think the voter turnout will be higher than 2008 (due to more vote-by-mail possibilities), but probably not 96%.
   1050. Joe Kehoskie Posted: November 03, 2012 at 02:52 PM (#4291804)
You can't be this dense.

Please explain how a poll with a 96 percent LV rate is representative of the entire electorate when turnout is only ~72 percent. Thanks in advance.
   1051. Jim Wisinski Posted: November 03, 2012 at 02:53 PM (#4291807)

Is there a math argument in there or just a general rant?

For about the tenth time, the only way a poll with an LV screen that passes 96 percent of RV through as LV is representative of the entire electorate is if the pollster is projecting 96 percent turnout.


That was a general rant. The math argument is what McCoy keeps saying, that a few days before the election not many RVs who aren't going to be LVs are responding to polls. The fact that it's a poll call right before the election IS AN LV SCREEN IN ITSELF! Nobody is claiming that the poll respondents who make it through the screen as LVs are 96% of registered voters, everyone who bothers to think about it knows that the percentage is irrelevant because of the damned obvious underlying factor of the election being days away.

You know this though, you have to know. You're just ignoring it because you can use horribly faulty logic to twist it into another claim of poll bias and ignore the fact that your candidate is losing.

   1052. Monty Predicts a Padres-Mariners WS in 2016 Posted: November 03, 2012 at 02:53 PM (#4291808)
He can run a billion simulations of a coin flip, and he won't get exactly 50%, but it's not particularly useful to publish a number to five decimal places.


I like this post!
   1053. McCoy Posted: November 03, 2012 at 02:54 PM (#4291809)
Screw 538, 506 is the site that matters the most.
   1054. Gold Star - just Gold Star Posted: November 03, 2012 at 02:55 PM (#4291810)
Rupert Murdoch is saying Christie has a lot of explaining to do if Romney loses.
   1055. Random Transaction Generator Posted: November 03, 2012 at 02:57 PM (#4291812)
He can run a billion simulations of a coin flip, and he won't get exactly 50%, but it's not particularly useful to publish a number to five decimal places.


If he said that heads came up 499,998,992 times out of a billion flips, is it wrong to say that heads came up 49.9998992% of the time?

In Silver's simulations, there are LOTS of different results (not a coin flip) about the amount of EV.
If he runs a million simulations, is it better to list all million results, or group them together (175,392 of them had Obama winning 302 EV, 169,232 of them had Romney winning 271 EV, etc), or can he list the average/median/whatever of the entire set?

And if 80.3% of those simulations have Obama with 270+ EV votes, is it wrong to say that (based on these simulations) Obama has n 80.3% chance of winning the election?

If the whole "simulation" thing bothers you, then why even visit/care about Silver's results?
   1056. cercopithecus aethiops Posted: November 03, 2012 at 02:57 PM (#4291813)
For Gallup, it's much more intricate.


And didn't Gallup wind up projecting an 82% white electorate based on that more intricate screen? That seems pretty loopy to me.
   1057. 'Spos Posted: November 03, 2012 at 02:58 PM (#4291814)
He can run a billion simulations of a coin flip, and he won't get exactly 50%, but it's not particularly useful to publish a number to five decimal places.


If the number is far enough from 50% I'd think a close look at the coin might be in order.
   1058. Matt Clement of Alexandria Posted: November 03, 2012 at 02:59 PM (#4291815)
Rupert Murdoch is saying Christie has a lot of explaining to do if Romney loses.
Neh. I mean, who knows what moves undecided voters, I certainly don't. But Christie is very much a northeastern pol, and I doubt his name-recognition or personal appeal are particularly significant in the swing states of the midwest, southwest, and upper south.
   1059. 'Spos Posted: November 03, 2012 at 03:00 PM (#4291816)
If he runs a million simulations, is it better to list all million results, or group them together (175,392 of them had Obama winning 302 EV, 169,232 of them had Romney winning 271 EV, etc), or can he list the average/median/whatever of the entire set?


He does something like that in a sidebar.

Anyone else read his book yet?
   1060. Tripon Posted: November 03, 2012 at 03:00 PM (#4291817)
Why the heck would a governor of a state that leans heavily democratic have a 'lot of explaining' to do? Did Christie's comments move NJ from 70-30 Obamato 72-28 Obama or something?
   1061. Joe Kehoskie Posted: November 03, 2012 at 03:01 PM (#4291818)
That was a general rant. The math argument is what McCoy keeps saying, that a few days before the election not many RVs who aren't going to be LVs are responding to polls. The fact that it's a poll call right before the election IS AN LV SCREEN IN ITSELF! Nobody is claiming that the poll respondents who make it through the screen as LVs are 96% of registered voters, everyone who bothers to think about it knows that the percentage is irrelevant because of the damned obvious underlying factor of the election being days away.

This is another rant without any math. I agree that an LV is much more likely to participate in a poll than an RV, but the idea that only LV are answering their phones right now is silly. With ~28 percent of RV not voting, the pollsters should be getting more hang-ups or non-responses than they seem to be getting.
   1062. Matt Clement of Alexandria Posted: November 03, 2012 at 03:01 PM (#4291819)
ARoM -

A note on RCP. They have a mostly arbitrary, possibly somewhat partisan poll screening method that gives them a rightward lean in their averages compared to Pollster, TPM, and 538. They don't include about 25% of publicly released polls, and their selection methods appear entirely subjective. I think that by making use of every existing poll not released by a campaign, the other poll averages are better options. If there were an RCP that handled all polls equally, I'd be interested to see how it compared to the more complex aggregation models at 538 and Pollster.
   1063. Morty Causa Posted: November 03, 2012 at 03:04 PM (#4291820)


Not really. We prevent a 16-y.o. from voting, but he still has the right to kill an assailant in self-defense.


Plant your flag. Which is it: voting or democracy. They are not coextensive. Just because you cannot vote doesn't mean you can't have rights. Moreover, someone who can vote can be voting on your behalf. You know, like parents voting in the interest of their children. Not only is your standard a floating one, it doesn't comport with an avowed (I believe) interest in the inviolability of the family, and of individuals as their rights relate to the family. Modern democracy isn't the only kind that ever existed, and tribal comminities are heavily democratic.

And simply because self-defense with guns is not exactly on par with other rights doesn't diminish or negate those other rights. They don't have to be the same to have value, value essential to the vibrancy of a community.
   1064. Random Transaction Generator Posted: November 03, 2012 at 03:05 PM (#4291821)
If he runs a million simulations, is it better to list all million results, or group them together (175,392 of them had Obama winning 302 EV, 169,232 of them had Romney winning 271 EV, etc), or can he list the average/median/whatever of the entire set?


He does something like that in a sidebar.


He does, but it's a bar graph without any detailed numbers, so it only gives a feel of where the results landed.

I'd love it if he released his full set of results in a spreadsheet/text file. I don't need to see his black box calculations, but the list of all million random runs (including state-by-state results that lead to the EV totals) would be interesting (for me) to pour over.
   1065. Random Transaction Generator Posted: November 03, 2012 at 03:06 PM (#4291823)
Not really. We prevent a 16-y.o. from voting, but he still has the right to kill an assailant in self-defense.


16-year olds can own guns in the US?

Yikes.
   1066. snapper (history's 42nd greatest monster) Posted: November 03, 2012 at 03:07 PM (#4291825)
You do realize that Silver runs hundreds of thousands of (I assume Monte Carlo method) election simulations, right?
That's why his numbers have significant digits.
He's not saying that someone is going to win a fraction of an elector seat. He's saying that the "average" result is xx.x% or xxx.x EV.

I really thought that was obvious for anyone that paid even a small amount of attention to what Silver says/does.


You can run hundreds of millions of Monte Carlo simulations, but that doesn't make your results and more accurate than the underlying data they are drawn from.

MBS and ABS CDOs that were rated AAA based on 100,000 Monte Carlo simulations (sometimes without a single loss in the 100,000 runs), ended up worthless b/c the underlying assumptions (no worse than 0% housing appreciation) were off by about -30%.

If you have a sample size of a few dozen elections, you can't get to even one decimal point of precision.
   1067. 'Spos Posted: November 03, 2012 at 03:07 PM (#4291826)
I'd love it if he released his full set of results in a spreadsheet/text file.


I wonder what he'd say if someone asked?
   1068. Joe Kehoskie Posted: November 03, 2012 at 03:07 PM (#4291827)
They don't include about 25% of publicly released polls, and their selection methods appear entirely subjective.

The only major polls not being used by RCP are polls that didn't exist in 2008 or 2010 (or even, in RAND's case, in June 2012).
   1069. spike Posted: November 03, 2012 at 03:08 PM (#4291828)
Why the heck would a governor of a state that leans heavily democratic have a 'lot of explaining' to do?

Because if Romney loses the long knives will be out for some RINOs to purge. It just won't occur to the party faithful that they lost because a majority disagreed with them.
   1070. snapper (history's 42nd greatest monster) Posted: November 03, 2012 at 03:09 PM (#4291829)
Every single person who takes part in a poll is a data point.

But you never know how they vote, so that data is useless. All we have as real data is actual vote totals vs. projected vote totals for the X number of elections that are still relevant.
   1071. snapper (history's 42nd greatest monster) Posted: November 03, 2012 at 03:10 PM (#4291830)
16-year olds can own guns in the US?

Yikes.


Yes. In most states, they can not purchase them, but can own them if gifted to them (usually by their parents).
   1072. 'Spos Posted: November 03, 2012 at 03:10 PM (#4291831)
16-year olds can own guns in the US?


I bet more 16-year olds kill people with cars than guns.
   1073. Joe Kehoskie Posted: November 03, 2012 at 03:10 PM (#4291832)
MBS and ABS CDOs that were rated AAA based on 100,000 Monte Carlo simulations (sometimes without a single loss in the 100,000 runs), ended up worthless b/c the underlying assumptions (no worse than 0% housing appreciation) were off by about -30%.

It's really funny that, on a site like this, the concept of challenging assumptions is mocked as outlandish and even "stupid."
   1074. McCoy Posted: November 03, 2012 at 03:12 PM (#4291833)
But you never know how they vote, so that data is useless. All we have as real data is actual vote totals vs. projected vote totals for the X number of elections that are still relevant.

And we can compare projected to real which is what we also do in sabermetrics. Thus the data isn't useless. Are you trying to argue that there is no correlation between polling data and election numbers?
   1075. Morty Causa Posted: November 03, 2012 at 03:12 PM (#4291834)
"16-year olds can own guns in the US?"

What's corresponds with the sophisticate urbanite to "provincial"?

If you are from a big city, that might seem scary, but where I'm from--the rural south--many boys get their first shotgun when they are about 8.

Man, doesn't anyone ever get tired flagellating those polls?
   1076. Random Transaction Generator Posted: November 03, 2012 at 03:15 PM (#4291836)
You can run hundreds of millions of Monte Carlo simulations, but that doesn't make your results and more accurate than the underlying data they are drawn from.


So you're saying that Nate Silver's predictions could be wrong because of the data (polls) that he gets might be wrong?

Isn't that the start of this whole argument in the beginning, that Silver's excuse of "bad data" was a cop-out?
Yet here, you're saying that it is an excuse.
   1077. snapper (history's 42nd greatest monster) Posted: November 03, 2012 at 03:15 PM (#4291837)
Plant your flag. Which is it: voting or democracy. They are not coextensive. Just because you cannot vote doesn't mean you can't have rights. Moreover, someone who can vote can be voting on your behalf. You know, like parents voting in the interest of their children. Not only is your standard a floating one, it doesn't comport with an avowed (I believe) interest in the inviolability of the family, and of individuals as their rights relate to the family. Modern democracy isn't the only kind that ever existed, and tribal comminities are heavily democratic.

And simply because self-defense with guns is not exactly on par with other rights doesn't diminish or negate those other rights. They don't have to be the same to have value, value essential to the vibrancy of a community.


I'm sorry, I don't get your question/complaint.

All I said is that certain rights are "natural rights" that everyone, everywhere has, and any law that restricts them is unjust (e.g. self-defense, freedom of speech, freedom of worship, property ownership). Voting/democracy is not one of those rights. A monarchy can be a completely legitimate form of gov't.

I simply used the example of the 16 y.o. to show that even a highly free democracy restricts voting rights, while we do not restrict the rights to speech, religion, self-defense, etc.

   1078. Monty Predicts a Padres-Mariners WS in 2016 Posted: November 03, 2012 at 03:16 PM (#4291840)
It's really funny that, on a site like this, the concept of challenging assumptions is mocked as outlandish and even "stupid."


Nah, this site just has a different set of sacred cows. Try challenging OBP or something and see how far you get.
   1079. Random Transaction Generator Posted: November 03, 2012 at 03:18 PM (#4291841)
What's corresponds with the sophisticate urbanite to "provincial"?


I'm a Canadian, so the whole idea of teenagers with guns scares the crap out of me.
Out of all demographic age groups, teenaged boys seem the least mentally stable of them all.
With hormones, peer pressure, sexual awakening, and educational stress, giving those guys a gun seems like a bad idea.

I remember being a teenager. I wasn't the most rational person...
   1080. snapper (history's 42nd greatest monster) Posted: November 03, 2012 at 03:19 PM (#4291842)

So you're saying that Nate Silver's predictions could be wrong because of the data (polls) that he gets might be wrong?

Isn't that the start of this whole argument in the beginning, that Silver's excuse of "bad data" was a cop-out?
Yet here, you're saying that it is an excuse.


I'm not criticizing Silver for potentially being wrong; the nature of the beast is that any prediction is only an educated guess in a close election.

All I'm criticizing is the overly false precision he is claiming (implicitly or explicitly) when he's working from a data sample of 4 or 5 Presidential elections. There's just no way to get better than 50:50, likely, highly-likely, with that sort of base data.
   1081. snapper (history's 42nd greatest monster) Posted: November 03, 2012 at 03:19 PM (#4291843)
I bet more 16-year olds kill people with cars than guns.

Probably by orders of magnitude.
   1082. cercopithecus aethiops Posted: November 03, 2012 at 03:20 PM (#4291844)
Latest Ipsos/Reuters Ohio poll had 77% of RV pass the LV screen. Seems more in line with what Joe is looking for. Obama led 48-40 among RV and 47-45 among LV. Seems broadly in line with CW about RV vs LV in general.

Interestingly, the same pollster polling Virginia during the same time frame had 89% of RV pass the LV screen. Basically, I have no idea what any of this is supposed to mean.
   1083. Swedish Chef Posted: November 03, 2012 at 03:20 PM (#4291845)
You do realize that Silver runs hundreds of thousands of (I assume Monte Carlo method) election simulations, right?
That's why his numbers have significant digits.
He's not saying that someone is going to win a fraction of an elector seat. He's saying that the "average" result is xx.x% or xxx.x EV.


My whole ####### point is that those are not significant digits, they're noise. And he's publishing them, no error bars in sight. That's creating an impression of precision that simply isn't there.
   1084. McCoy Posted: November 03, 2012 at 03:20 PM (#4291846)
Fox News latest poll had a LV of 92%. The one they did in the middle of August had a LV of 83%. As I said before at this point in the game the only people picking up the phone and doing the polls are people who are going to vote. If you are not going to vote you don't take the poll. Plain and simple. So easy a troll can understand it.

In both cases it was Romney +1.
   1085. Jolly Old St. Nick Still Gags in October Posted: November 03, 2012 at 03:21 PM (#4291847)
16-year olds can own guns in the US?

Yikes.


I realize it's not the same thing, but BITD half the 8 year olds on my block in DC owned "Daisy / Red Ryder" BB guns that were advertised on the back cover of comic books. They're not deadly or anything, but they can sure put some kid's eye out. They still make them, and I have no idea whether or not gun laws apply to them, but they sure didn't back then.
   1086. Joe Kehoskie Posted: November 03, 2012 at 03:21 PM (#4291849)
And we can compare projected to real which is what we also do in sabermetrics. Thus the data isn't useless. Are you trying to argue that there is no correlation between polling data and election numbers?

Unless we're talking about turnout, in which case using polls with a 96 percent LV rate (compared to ~72 percent turnout) is perfectly acceptable.
   1087. Random Transaction Generator Posted: November 03, 2012 at 03:24 PM (#4291854)
I bet more 16-year olds kill people with cars than guns.

Probably by orders of magnitude.


I also don't think kids should be driving at 16, either.
   1088. Joe Kehoskie Posted: November 03, 2012 at 03:24 PM (#4291855)
Latest Ipsos/Reuters Ohio poll had 77% of RV pass the LV screen. Seems more in line with what Joe is looking for. Obama led 48-40 among RV and 47-45 among LV. Seems broadly in line with CW about RV vs LV in general.

Thanks; perfect example of what we've been discussing: An 8-point Obama RV lead becomes a 3-point LV lead, and that's with turnout still projected to be ~5 points higher than 2008.
   1089. Morty Causa Posted: November 03, 2012 at 03:24 PM (#4291856)
All I said is that certain rights are "natural rights" that everyone, everywhere has, and any law that restricts them is unjust (e.g. self-defense, freedom of speech, freedom of worship, property ownership). Voting/democracy is not one of those rights. A monarchy can be a completely legitimate form of gov't.


And all I'm saying is that you are wrong. And I've explained why you're wrong. They are either both natural rights or neither is. And, moreover, they are related. Insisting on a myopic, horse-blinders view of "democracy" as only having a political sense allows you to evade seeing this connection. Voting (or democracy) has larger philosophical import. What you're really talking about is having a say in your existential condition. If that isn't a natural right, there are none (well, they are none-but). That's the first step in a bedrock social contract. And that exists even in monarchies or dictatorships.
   1090. Random Transaction Generator Posted: November 03, 2012 at 03:25 PM (#4291857)
They're not deadly or anything, but they can sure put some kid's eye out.


So I've heard.
   1091. snapper (history's 42nd greatest monster) Posted: November 03, 2012 at 03:25 PM (#4291858)
And we can compare projected to real which is what we also do in sabermetrics. Thus the data isn't useless. Are you trying to argue that there is no correlation between polling data and election numbers?

Correct. We can compare them, and assess a correlation. My only point is that with a working sample of 4 or 5 elections, maybe 20 if you bring in Gubernatorial and Senatorial races, the correlation just can't be that strong.

Any estimate of likelihood is going to have a huge error bar around it, maybe +/- 10% (i.e. a 65% likelihood means somewhere between 55 and 75%). So, to express that in decimals is simply ridiculous and misleading.
   1092. DKDC Posted: November 03, 2012 at 03:26 PM (#4291859)
I'm too lazy to learn the ins and outs of how the pollsters conduct their surveys and how they adjust and report their results. But it seems like the RCP simple analysis of those polls and Nate's overly complicated (and possibly biased) analysis of those polls are telling us the same thing.

The criticisms I've seen of those analyses seem to hinge on very personal and anecdotal views of turnout. We'll see on Tuesday, but my bet is that the pollsters know what they are doing.

Personally, I think this is one of the least important elections of my lifetime. The power of the executive is generally overstated, and both of these guys seem like pragmatists who know how to pander to their base but govern largely from the center.
   1093. cercopithecus aethiops Posted: November 03, 2012 at 03:27 PM (#4291860)
A monarchy can be a completely legitimate form of gov't.


Listen, strange women lyin' in ponds distributin' swords is no basis for a system of government. Supreme executive power derives from a mandate from the masses, not from some farcical aquatic ceremony.
   1094. Morty Causa Posted: November 03, 2012 at 03:27 PM (#4291861)
I simply used the example of the 16 y.o. to show that even a highly free democracy restricts voting rights, while we do not restrict the rights to speech, religion, self-defense, etc.


That's cloud talk. All those rights are restricted.
   1095. Morty Causa Posted: November 03, 2012 at 03:28 PM (#4291862)
Listen, strange women lyin' in ponds distributin' swords is no basis for a system of government. Supreme executive power derives from a mandate from the masses, not from some farcical aquatic ceremony.


Exactly. Kings and tyrants have been known to be deposed.
   1096. DA Baracus is a "bloodthirsty fan of Atlanta." Posted: November 03, 2012 at 03:28 PM (#4291863)
I'm too lazy to learn the ins and outs of how the pollsters conduct their surveys and how they adjust and report their results.


Don't worry, not knowing what you are talking about hasn't stopped people before.
   1097. Joe Kehoskie Posted: November 03, 2012 at 03:28 PM (#4291864)
As I said before at this point in the game the only people picking up the phone and doing the polls are people who are going to vote. If you are not going to vote you don't take the poll. Plain and simple. So easy a troll can understand it.

There's a major difference between "participating in a poll" and "answering the phone." Are you really claiming that almost 100 percent of the people who aren't going to vote have been not answering the phone for the past three months?
   1098. McCoy Posted: November 03, 2012 at 03:29 PM (#4291866)
Latest Ipsos/Reuters Ohio poll had 77% of RV pass the LV screen. Seems more in line with what Joe is looking for. Obama led 48-40 among RV and 47-45 among LV. Seems broadly in line with CW about RV vs LV in general.

Interestingly, the same pollster polling Virginia during the same time frame had 89% of RV pass the LV screen. Basically, I have no idea what any of this is supposed to mean.


Ispos wasn't also conducting these polls 3 months ago so we can't compare them to past numbers. These are also online polls which will result in lower LV numbers.
   1099. snapper (history's 42nd greatest monster) Posted: November 03, 2012 at 03:30 PM (#4291867)
And all I'm saying is that you are wrong. And I've explained why you're wrong. They are either both natural rights or neither is. And, moreover, they are related. Insisting on a myopic, horse-blinders view of "democracy" as only having a political sense allows you to evade seeing this connection. Voting (or democracy) has larger philosophical import. What you're really talking about is having a say in your existential condition. If that isn't a natural right, there are none (well, they are none-but). That's the first step in a bedrock social contract. And that exists even in monarchies or dictatorships.

I have no idea what "having a say" means.

All I'm talking about is the explicit form of gov't. Living in a representative democracy is not a natural right.

When people talk about human rights violations in China, they're almost universally talking about forced abortions, and religious persecution, and displacement of peasants from their land, not the lack of representative democracy.
   1100. cercopithecus aethiops Posted: November 03, 2012 at 03:30 PM (#4291868)
An 8-point Obama RV lead becomes a 3-point LV lead


Of course, the other point is that it's still an Obama lead, and in line with his lead in most other polls of LV. Including the ones that you say are passing too many of the respondents.
Page 11 of 114 pages ‹ First  < 6 7 8 9 10 11 12 13 14 15 16 >  Last ›

You must be Registered and Logged In to post comments.

 

 

<< Back to main

BBTF Partner

Support BBTF

donate

Thanks to
Guts
for his generous support.

Bookmarks

You must be logged in to view your Bookmarks.

Hot Topics

NewsblogOMNICHATTER FOR APRIL 18, 2014
(48 - 7:31pm, Apr 18)
Last: Joyful Calculus Instructor

NewsblogEscape from Cuba: Yasiel Puig’s Untold Journey to the Dodgers
(21 - 7:27pm, Apr 18)
Last: Joe Bivens, Minor Genius

NewsblogFletcher: Foes have slowed Trout in one category so far - steals
(7 - 7:22pm, Apr 18)
Last: Los Angeles El Hombre of Anaheim

NewsblogRB: Carlos Beltran: more of a center fielder than Mickey Mantle, Ty Cobb or Duke Snider. So what?
(31 - 7:21pm, Apr 18)
Last: toratoratora

NewsblogOTP April 2014: BurstNET Sued for Not Making Equipment Lease Payments
(1665 - 7:21pm, Apr 18)
Last: spike

NewsblogDoug Glanville: I Was Racially Profiled in My Own Driveway
(318 - 7:17pm, Apr 18)
Last: Joe Bivens, Minor Genius

NewsblogPrimer Dugout (and link of the day) 4-18-2014
(29 - 7:10pm, Apr 18)
Last: Eric J can SABER all he wants to

NewsblogGleeman: Mets minor league team is hosting “Seinfeld night”
(153 - 7:00pm, Apr 18)
Last: TerpNats

NewsblogMitchell: Now Playing First Base for the Yankees
(51 - 6:54pm, Apr 18)
Last: DKDC

NewsblogMinuteman News Center: Giandurco: This means WAR
(75 - 6:50pm, Apr 18)
Last: DanG

NewsblogNightengale: Pujols nears 500 home runs...and no one seems to care
(82 - 6:18pm, Apr 18)
Last: Rob_Wood

NewsblogOT: NBA Monthly Thread - April 2014
(320 - 5:58pm, Apr 18)
Last: tshipman

NewsblogDesign Room: Top 10 Logos in MLB History.
(75 - 5:29pm, Apr 18)
Last: Zach

NewsblogOT: The Soccer Thread March, 2014
(898 - 4:58pm, Apr 18)
Last: Grunthos

NewsblogDaniel Bryan's 'YES!' chant has spread to the Pirates' dugout
(93 - 4:19pm, Apr 18)
Last: Joe Bivens, Minor Genius

Demarini, Easton and TPX Baseball Bats

 

 

 

 

Page rendered in 1.2694 seconds
52 querie(s) executed