MATH DISINFO ALERT

Use the exact math on other polls, if you got different answers using the poll in the OP you may be missing a step.

They do, but all the percentages are just made up if they don't specifically state sample sizes per question.

In scientific research it would be something like:

When 1789 people were asked x:

10% (179) answered A
40% (716) answered B
etc

HOWEVER

Because the sample sizes have a clear bias when it comes to such political questions, equal numbers of biased people are necessary in order to give the best representation of that population possible. This is what the media isn't saying when they publish those BS percentages.

Normalizing is literally skewing the data given by the two sample sizes in the population and adjusting them proportionally to their representation with that population, which is exactly what the screencapped pic in the OP did. I'm trying to find another poll that published sample sizes to prove my point.

LOOKS LIKE THE SHILLS HAVE ARRIVED LOL

Source? Pic related makes it look like Hillary got more than him.

I hope you're right.

NO ITS ACTUALLY 9990998763678984hgei86688%%^253 OVERSAMPLING PRAISE GEG

A poll that doesn’t publicly post its methodology IS NOT A VALID POLL, and you can reject it offhand.

Reported for CTR.

I disagree with this idiots incorrect math and I'm ctr? Dem oversampling was +7 in that poll, not +20. But muh shills though right? Never mind that Dems will never get +7 in their wildest dreams, anything other than 100% blind faith in the rosiest of predictions is shilling

Then:

That poll had 33% repubs + repub leaning, 47% dem + dem leaning, and 11 + 9 memes

And with those voter percentages wound up with:

Hillary 44%
Trump 40%
memes 16%

(voter percentage) 47/33 = 1.4242…

1.4242 dem bias

40 percent for trump under bias / 33 repub voters under bias = 1.2121 trump got 21% more of the vote greater than the percentage of repub voters

so

1.2121 * 1.4242 = 1.7262 bias against trump

ergo Trumps actual score is:

33 * 1.7262 = 56.9670

Yeah, so it was a D+14 poll sample. That's what I'm trying to get you to understand.

Are there any besides LAT/USC that actually post their methodology? I know some may give you a peak at some of their demographic breakdowns, but I don't think any of the other ones ACTUALLY give you full demographic info, how they collected their sample, how they account for sampling bias, etc etc etc.

Hell, they may actually use a justifiable D+7 weighting, but then have all their "republicans" be women with college degrees who are the least favorable towards Trump in the republican party. This would all them to SEEM unbiased while actually being full of shit.

You can't directly compare Dem to GOP turnout like that. Democrats always have far higher primary turnout relative to election turnout, and vice versa for Republicans. Instead, look at how things change WITHIN the parties' primaries from election to election. Democrats had record low primary turnout. Republicans have record high primary turnout, with Trump now holding the record in total primary votes of any GOP candidate ever, despite the fact that he was running against a dozen+ other candidates, several of which stayed in the race until the very last minute.

wew

Hillary won he primary with fewer votes than what she lost the primary with in '08. Dems are demoralized. Primary voting for Dems down 20%. Republican primary turnout up 60%. This is going to be a blowout.

15.8 + 12 = 27.8

13.3 + 7.6 +4.1 + 3.4 = 28.4

more republicans.

And since then many demoshits switched to republican.

No, the screencapped answers "what if we scaled the results such that they sampled an equal # of D's, R's, and I's?" Which, I'm just pointing out, isn't really great either, because there aren't likely to be an equal # of D, R, and I voters.

The screencapped post made no attempt to quantify that actual distribution of D, R, and I voters and skew towards that.

Holy shit. Sauce? That's hilarious if it's true.

This guy has been making this error for months. He never gets it.

en.m.wikipedia.org/wiki/Results_of_the_2008_Democratic_Party_presidential_primaries

This election cycle she won with ~16m. In '08 she lost with 18m.

Not to mention that the Clinton machine probably went on and bussed niggers, spics and tumblrites around to inflate Shillarys vote count and the whole "Hillary beat exit polls by a fair margin wherever electronic voting was used" which is another pointer towards a rather unreliable primary.

If you want a much more accurate way of worst-case tier sampling, we need to look at 2012 data

ropercenter.cornell.edu/polls/us-elections/how-groups-voted/how-groups-voted-2012/

And interestingly, despite the worst case scenario, trump still wins in most polls.

But we should still assume the worst and look at 2012 data since it is something really important to look at.

What the hell kind of garbage math and theory is this? There's not an even split of republicans and democrats. And you wouldn't manipulate the numbers in that fashion to find the raw data on respondents anyway. Here:

The raw responses out of their total unweighted sample would be:
194+20+80=294 = 37% for Hillary
22+213+114=351 = 44% for Trump.

But you can't just use those numbers straight, because we don't have 1 republican for every 1 democrat.

In Pew Research's party identification surveys in 2014 and 2015 the split was 39:23:32 and 40:23:30, Independents:Republicans:Democrats, but this is the whole population, and not necessarily likely voters. IDB uses 35:28:35 as their projected voter mix. Exit-polls in 2008 and 2012 both showed a mix of about 29:32:39.

Pick whatever mix you want in this range. But don't just do 1:1:1, that's nigger tier whites murder as much as blacks shit.

Using:

http ://www.realclearpolitics.com/docs/2016/2016_Reuters_Tracking_-_Core_Political_10.18_.16_.pdf
(included in PDF form)

looking at the first picture something very strange already starts to happen. I'll give you a second…

Dems 723 (A) + Reps 643 (B) + Inds 210 (C) = 1576(X), which is different from both the number of registered (1501 or Y) and likely (1190 or Z)

STEP 1

To mitigate these differences I will normalize the sample sizes as a proportion of these population sizes:

Z / X = .755 (@)

Y / X = .952 (!)

STEP 2

These new variables will be used to morph the samples A, B, and C to normalize those samples to the corresponding polls:

LIKELY VOTERS

A * @ = 546
B * @ = 485
C * @ = 159
TOTAL = Z (1190)

STEP 3 THE GOODS

Refering to the second pic, I'm going to focus on the first poll on the top for likely voters. They're numbers are:

D 43%
R 39%
All others: 18%

Using the math from the OP, as Dems are over sampled again its as follows:

546/485 = 1.126 = 12.6% dem oversampling rep
546/159 = 3.434(WOW) = 343.4% dem oversampling of ind

39% * 1.126 = 43.9% normalized reps voting trump

Sadly it must be assumed that 18% won't change since ind voters likely to vote are still uncertain

SO

True Dems = 100% - 43.9% (Weighted normalized Reps for Trump) - 18% Other (Indies on the fence) = 38.1% of the normalized population of Z (1190) are Dems for HRC

You're conflating data with sample size. F, apply yourself.

Sorry I was I meant to tell to apply themselves. Your math looks good.

I'm a soldier not a dweebamatician

Also this is the fire monkey even if I'm not using the correct actuarial formulas I bet ya my numbers are very close to reality

Kek I'm on a different ip but both were me

first post was an approximation I came up with the other day, still it wasn't far off

Shoot using the same methodology from the OP I can apply the ind choice for meme candidates using the oversampling variable for them:

8% * 3.434 = 27.5% of normalized ind voting for meme candidates

So true dems should actually be: 100% - 43.9% Trump - 27.5% for ind voting for memes = 28.6% for HRC with normalized sample sizes lol

This bitch has NO CHANCE

I always thought that the polls were already normalized for this, and when Holla Forums complained about the oversample, they were just pointing out at the poor methodology, not that results were incorrect.

If you literally just take the number of people who say they will vote Trump/Hillary but like 80% of them are democrats or women, that's fucking retarded. I can't believe media passes this off as legitimate research.

There's probably not exactly equal numbers, but it's close enough for the time being. OP assumes the true split is 50-50 (not counting indep) and he will be off by however much that assumption is off.

Is there data for the total number of registered voters somewhere? Those would be a better estimate of what % identifies with what party. Seems like there's numbers for individual states like sec.state.ma.us/ele/eleenr/enridx.htm

We can just add up all the states and get pretty close, don't even need to track down minor overseas territories and so on since they have a very small effect.

Actually what we can do is this:

1. Define important categories like race, age, education level, sex
2. Take the poll data and break up into groups for every combination of this (eg. black single women with college degree etc.)
3. Weight the Trump/Hillary preference of each group by its proportion in the country

This should give much better correction for demographics. For smaller sample sizes, use only the few most significant factors - I'm guessing race and sex alone are probably pretty good based on that map that keeps getting posted.

But of course this is all assuming the data was not tampered with from the start. I wouldn't be surprised if some of these interns that actually make the phone calls occasionally "forget" to enter a pro-Trump answer into their spreadsheet.

Is there a consolidated post/pic that explains this? I need it to help explain to some people that want to see the light, but I'm not numbers savvy enough to lay the math out correctly or even word it will enough to even fake it. I've been trying to tell this dude that all the polls that 528 or whatever site uses are all skewed

I've tried to explain it to someone without the numbers, but they just looked at me skeptically and asked for a source. Thank you in advance.

The fastest way I know is that for any poll to be statistically correct they need to have an equal amount of people who share each individual bias.

If there are 10 people in a group who are thinking of lunch, if 6 of them's favorite food is pizza then that group is biased towards pizza and the remaining members will never enjoy their preferred dish.

But that's the problem. This only works if the poll bothers to give a breakdown of EVERY SINGLE DEMOGRAPHIC METRIC possible. They could show a breakdown by party affiliation and race and gender and have all these metrics fairly represented, but still have it skewed by oversampling people with college degrees, and not reporting the demographic breakdown of education level when they publish the results.

If a poll doesn't show report certain demographic information, there's no way to know if it's skewed or not along that metric.

Even when they do like here The numbers don't add up and don't even bother trying to explain it to the layman.

Doesn't have to be every metric. Just use whatever they give. If they give more metrics and have enough samples to apply them all, then your correction will be more accurate, but even metric is effective as OP demonstrates.

I suspect there is a diminishing returns with each subsequent correction. Ie. 1st metric will remove 60% of bias, 2nd removes 30%, 3rd removes 5%, and then when you get to 10th factors like "favorite anime genre" they only correct by a tiny amount, so you don't lose anything by ignoring those.

Essentially, just one factor is already enough, 2-5 is great, anything past that I wouldn't worry about.

But these factors are all dependent anyway. In this example, if you correct only for race and gender, you end up correcting for some of the degree bias too. And race/gender are probably more important for this than degree anyhow.

You can't get a perfect correction, but you can remove the obvious biases for what it's worth. Doesn't mean the adjusted result is truth, but it is a method that will easily reveal low-effort rigging. I'm pretty sure most polls are very simple, superficial manipulations and very few are elaborate and difficult to detect because of how much effort they take.

Oh, that part I understand- I do understand how under/over sampling destroys the fairness of a poll. I am just trying to find the exact X and Y numbers of THIS specific poll so I can say "These are the numbers that were sampled, these were their party affiliation. This is the math they did to get to these fucked up 41% number. If you use this number to account for the disparity in the over polling (this is the math you use to get that number) you get 50% and 39% (or whatever the fuck number it was). I've got a lot of the numbers from the post just to plug and play, but I'm missing a lot of the math

Yes it does. Let's say I specifically do all my sampling from urban areas and college campuses. I could theoretically have an unskewed sample if you look at age, race, gender, and party affiliation. But it could still show Clinton having a falsely huge advantage, because the *type* of "republicans" I sampled were all NRO-reading free market lolbergtarian #NeverTrump cuckolds, rather than an actual representative sample of republicans. And there'd be NO WAY of knowing this if they didn't specifically report in their data that they collected their samples overwhelmingly from college towns and urban areas.

Sure, you can try to unskew by whatever variables they give you. And this will make the polls "more" unbiased. But there is still the possibility of them biasing it in other ways that you can't detect from the data they give you. And if they bias it in even ONE undetectable way like this that you can't account for, it invalidates the poll REGARDLESS of whether you unskew it for all the other metrics. If just one metric remains skewed, into the trash it goes.

This whole idea is an exercise in pointlessness. The polls are biased in both enumerated and unenumerated ways. You can't account for all the devious ways they fuck with the data.

Right. And the polls assume more Democrats, so the polls are only off by however much their assumption is off.

We're simply talking about changing the assumption to make the results look better for us. We have not even tried to estimate what the actual split will be.

That's dumb. By your logic if 99% of the population were D's and 1% R's, a poll should still poll 50% D's and 50% R's.

Oversampling only skews a poll if it's oversampling relative to the actual population. An ideal poll would randomly sample the entire population with a high enough number of samples. In practice they probably don't want to do that many samples so they approximate by saying "the population is X% D's, Y% R's, and Z%, so we will take N samples, X% from known D's, Y% from known R's, and Z% from known I's". If they sample 50% D's but the real population is only 40% D's, then they've constructed a biased and misleading poll.

I wonder if OP will learn his math lesson this time

No, you make sure your sample size is 99% D and 1% R and state your sample and population sizes in your maths to corroborate that if you want to be taken seriously on the world stage.

Agreed with your premise, but:
How many polls do you think are rigged in this perfect, undetectable manner? I doubt it's all of them, there's many incompetent pollsters just like with any other sphere.

But also we get to demonstrate exactly how the polls are rigged. I bet even most normalfags know they aren't trustworthy, just based on past ability to predict results such as the primary. But not everyone carries through with the mental exercise to figure out exactly how they are biased, so they don't have concrete proof and only a suspicion.

Op's pic finding 50.8% Trump doesn't have to be correct in the sense that currently 50.8% of likely voters support Trump. The important thing is that it shows exactly how to see the bias for yourself and the result after correction is much more realistic. So it's very normalfag-friendly.

Actually that'd be a bad methodology, because different demographics have different turnout percentages. This isn't reflected by surveying everyone.

The higher and more randomized a sample, the better it is, but you'd still need to weight it to fit estimated turnouts. And this issue of estimated turnout is where things get really hazy. A rigorous methodology would, at the very least, assume the same demographic turnouts as in 2012. But Trump is most likely going to have much higher turnout from republicans/whites (same thing) and Hillary is most likely going to have much lower turnout from blacks compared to obama, for obvious reasons. But these things still wouldn't be reflected in the polls. And there's really no way to account for them.

...

Nobody in this thread has even tried to prove that the polls are rigged/biased. To do so you'd have to make a compelling argument that they are oversampling D's compared to the actual number of D's in the population, not just complain that they're sampling more D's than R's.

(I did see some ZeroHedge articles making the argument for oversampling, but I didn't review them in enough detail to say whether they were compelling or not).

What you're pointing out is that the population = likely voters, not all American citizens. No disagreement there.

Doesn't matter. At the end of the day, you have to trust the pollster at some level. If the pollster doesn't report ANY demographic data, you have to trust them to accurately weight/sample for race, affiliation, gender, etc. And even if they list these things, you still have to trust that they were accurate in their sampling of income level, education, rural vs urban, etc. And there are a wide variety of other ways they could make shit up beyond even these things. At some point, you have to say "okay, I'm going to unskew what I can, but I HAVE to trust you on everything you don't report"

Not every polls needs to be rigged. But there's enough doubt that I still wouldn't trust a single damn poll even after you unskewed it by all the variables you could.

Incidentally: archive.fo/7GRtV (zerohedge)

So we know they try to go for race, age and location.

Nice find! Its one thing to doubt, another to show, and the worst to know that there's evidence of foul play.

I'm the last "shill" as you call me.
I'm sorry my math did not include proper explanations but I didn't have time (i also included 18 nonrespondents)

see pic related (slightly updated) and verify yourself

not burger but MAGA!

The thing I dont understand is why they oversample in the first place. Can't they just randomly ring 1000 randomly selected people in the state being polled?
I mean how can Trump win in the polls if they intentionally poll more confirmed democrats? It's ridiculous.

we both came to similar results. yay for math and logic.

Praise KEK.

THAT'S LITERALLY THE POINT

Its to try to persuade the public that Trump ISN'T winning in most battleground states.

Yeah dude but my observation and calculations took me like 10 minutes. You have a freaking spreadsheet my dude, that sort of autism is remarkable!

I hope we meme MAGA as the 21st century Heil Hitler. I literally just pronounce it Mahgah in public already.

winning internet arguments IS after all important enough to warrant turbo-autism inbetween cubicle workstuffs.

proTip: Spreadhseets make math easier.

You think they conduct polls in order to show an accurate representation of where the candidates are. This is incorrect. The two purposes of these polls are
1) To demoralize trump supporters. Nobody likes being on the losing team. The polls allow the media to fill their news cycles with stuff like "Trump is totally finished and we're only reporting on him because he's still technically running, but President Clinton is already confirmed, goyim!"
2) To justify electoral fraud. It'd be a lot harder to mask voter fraud of the polls all showed Trump +3 and then Clinton wins +6. But if they craft the narrative of Clinton being way ahead in the polls, then the vast majority of people won't question the rigged voting response. They can just say "these allegations of fraud are ridiculous! The vote fell right in line with the polls! Conservatives are just crazy conspiracy theorists!"

Yurop here.
Agree with that. But, dont you have in all USA one (1) fucking polling agency, whitch would just call 5000 random people all over the place and show us results? 65%Trump 35%Cunt

Presumably all the major pollsters are subverted. Sounds a bit out there until you consider that if you were a wealthy (((globalist))) buying up every single pollster is exactly what you'd do, and make sure your media organs ignore the ones you couldn't buy.

Would have been cool to see Holla Forums organize its own independent poll operation. You just have to figure out where to get random phone numbers, but it can't be that hard. Too bad it's pretty late in the game now.

Yup. Any notion of plausible deniability goes out the window from how we've seen literally every branch of the media attack Trump, and all the wikileaks and O'Keefe videos that confirm that the corruption is even worse than we thought. There are no lengths these people wouldn't go to. There is no longer any doubt. No one can say "well they could technically do this, it's highly unlikely and it would require far too much corruption and people in on it" any longer. If something is at all possible, they WILL do it.

This method assumes the population is made up of equal percentages of democrats republicans and independents. And this false according to past elections and current gallup polls on party affiliation. Now if you can give mathematical proof that you are right then I will be forced to believe you.

Oh and your math is wrong.
Proof: Turn it around look at republican bias:

235/291=0.807

Clinton: 41.2%*0.807=33.2%=/=37.4% (OP number)

Meme candidates 235/271=0.867; 11%*0.867=9.53%=/=11.8% (OP number)

Trump: 1-0.118-0.332=55%=/=50.8%

As you can see the numbers dont check out. And there are so many other things that are wrong with your method, however this short mathematical proof should convince anons that your method is BS.

I "unbiased" a poll a while ago and made a shitty video about it. Vid related. If you listen carefuly you will find out that you cant unbias the polls unless you know how many republicans are voting for Trump Clinton, how many dems are voting for Trump/Clinton and how many indies are voting for Trump/Clinton.

You forget that more reps than dems showed up for the primaries.

You forget that since then lots of dems switched to rep.

You forget that historically dem turnout is higher for the primary than the actual election and the opposite for the reps.

Clinton is going to get so btfo.. I can't wait for the nov 9th headlines "rash of suicides in democrat cities"

No you can't get an accurate number unless you conduct a no biased poll. But statistically you can get a good approximation and the salt will shine through on the 9th when everyone sees that the 56% number was the most accurate.

Remember how shrill and butthurt the media was from 2000-2008? Every day was anudda fuckin' shoah. Every little thing or every pronunciation mistake from GWB was the week's news cycle. They were fucking hysterical, both literally and figuratively.

And that was with GWB, the compassionate cuckservative. Just imagine how the media will act with Trump at the helm. And while GWB took the media onslaught like a little bitch, Trump is going to be firing back at the lugenpresse every step of the way. It's going to be yuuge.

Its not about equal percentages but equal representation. Even with your math it shows that when its normalized Against Reps it would be

D 33.2%
M 9.5%

Meaning that R would have to be 57.3%, showing an R/D spread of 14.1, very similar to the OP picture's math of 13.3%.

I don't think neither you or that user are wrong if the spread is within one standard deviation, especially when the goal was proportion.

Nigga, it's clear you don't know anything about math. Stop making it so obvious.

Shoo shill, leave.

2+2=5
I'm not wrong, everyone who disagrees is CtR.

Thank you for contributing nothing to the thread and lowering the standard for posting on this board, you cockmongering flaming faggot.

Can someone please debunk this for me?

electionbettingodds.com/

No one cares.

No sample and population sizes mean that percentages can literally be made up, which is why you can't trust lugenpresse polls at all.