/sci/ - Science & Math » Thread #12310605

387KiB, 1694x1500, Screen Shot 2020-11-06 at 5.51.45 pm.png

View Same Google iqdb SauceNAO

Anonymous Fri 06 Nov 09:04:37 2020 No.12310605 View Reply Original Report

Quoted By: >>12310607 >>12310629 >>12310770 >>12310907 >>12311051 >>12311182 >>12311332

>Plots of the first digits of counts in various precincts and wards for selected counties/cities.

>Befords Law states that in any large set of data the leading digit will be a 1 30% of the time with the rest of the numbers being a leading number dropping dramatically all the way down to 9. Deviations from this model are highly irregular. Large deviations are a mathematical impossibility. This has been used to prove fraud in financial fraud cases like Enron and election cases like the 2009 Iran elections

What it means?
By the laws of math, Joe Biden cheated, without question.

Anonymous

View Same Google iqdb SauceNAO Screen Shot 2020-11-06 at 5.52.0 (...).png, 384KiB, 1660x1536

Anonymous Fri 06 Nov 2020 09:05:49 No.12310607 Report

Quoted By: >>12310609 >>12310681

>>12310605

Anonymous

View Same Google iqdb SauceNAO Screen Shot 2020-11-06 at 5.52.0 (...).png, 190KiB, 1698x700

Anonymous Fri 06 Nov 2020 09:06:52 No.12310609 Report

Quoted By:

>>12310607

Anonymous

Anonymous Fri 06 Nov 2020 09:17:39 No.12310629 Report

Quoted By: >>12310647 >>12310764

>>12310605
>Befords Law states that in any large set of data the leading digit will be a 1 30% of the time with the rest of the numbers being a leading number dropping dramatically all the way down to 9.
No it doesn't. Why is every retard on this website suddenly pretending to be an expert on statistics?

Anonymous

View Same Google iqdb SauceNAO dumb.jpg, 104KiB, 1100x825

Anonymous Fri 06 Nov 2020 09:24:48 No.12310640 Report

Quoted By:

so what? the same shit happened in the primaries in 2016 and earlier this year and nobody gave a shit then, why do you think it makes a difference now? thats completely irrational
>b-b-but it s not fair
>muh math muffugguh
lmao most people can't even read, they can just pretend like they don't believe in your math and that will make it effectively fake.

Anonymous

Anonymous Fri 06 Nov 2020 09:28:49 No.12310647 Report

Quoted By:

>>12310629
at the very least Benford's non-conformity demands a deep investigation in to this data, given the importance of this election. If it's good enough to convict Enron, it's good enough for this.

Anonymous

Anonymous Fri 06 Nov 2020 09:35:07 No.12310655 Report

Quoted By: >>12310764

"...we find that conformity with and deviations from Benford's Law follow no pattern. It is not simply that the Law occasionally judges a fraudulent election fair or a fair election fraudulent. Its “success rate” either way is essentially equivalent to a toss of a coin, thereby rendering it problematical at best as a forensic tool and wholly misleading at worst."

https://doi.org/10.1093/pan/mpr014

If you're gonna make a thread for pretending to discuss stats, at least read the Wikipedia page first.

Anonymous

View Same Google iqdb SauceNAO bba4ce0c751038b86e3cc7285fbf6ec3.jpg, 30KiB, 745x743

Anonymous Fri 06 Nov 2020 09:39:47 No.12310659 Report

Quoted By: >>12310713

true if big and can believe they would cheat but how incriminating is this really?
Counts depend on population size, demographics and participation all of which conceivably combine to give tally distributions different to Beford expectation.
Cities usually lean donkey and would follow the leading-1 rule for population so if they vote around 1/2 for sniffy joe doesn't that give them all leading 5s?

Anonymous

View Same Google iqdb SauceNAO dilbert.gif, 44KiB, 495x151

Anonymous Fri 06 Nov 2020 09:52:28 No.12310675 Report

Quoted By: >>12310690 >>12310729

if you were going to cheat the whole us election wouldn't you hire a statistician to cover your tracks?

Anonymous

Anonymous Fri 06 Nov 2020 09:59:51 No.12310681 Report

Quoted By: >>12311030 >>12311886

Should be obvious to non-brainlets, but here's an explanation from the Wikipedia page:
>Thus, real-world distributions that span several orders of magnitude rather uniformly (e.g. populations of villages / towns / cities, stock-market prices), are likely to satisfy Benford's law to a very high accuracy. On the other hand, a distribution that is mostly or entirely within one order of magnitude (e.g. heights of human adults, or IQ scores) is unlikely to satisfy Benford's law very accurately, or at all.

According to https://github.com/cjph8914/2020_benfords the data for >>12310607 is from
https://county.milwaukee.gov/EN/County-Clerk/Off-Nav/Election-Results/Election-Results-Fall-2020
90% of the numbers are between 203 and 1370.

Anonymous

Anonymous Fri 06 Nov 2020 10:13:50 No.12310690 Report

Quoted By:

>>12310675
It's not an organization that does the election fraud. It's just an idea.

Anonymous

Anonymous Fri 06 Nov 2020 10:39:17 No.12310713 Report

Quoted By: >>12310887

>>12310659
It by itself does not provide a guarantee that rigging took place, it simply means there is a high probability that it happened.
>Counts depend on population size, demographics and participation all of which conceivably combine to give tally distributions different to Beford expectation.
You are looking at the leading digit, not average or median values. Unless polling stations are refusing people after a certain number or there are hundreds of cities with less than 10 people being used in data, you can assume it will still tend to Benford.
>so if they vote around 1/2 for sniffy joe doesn't that give them all leading 5s?
No because a person adding their vote to the tally does not influence the next person's ability to vote. If I use my vote (regardless of who I vote for), me handing in my vote won't stop the next person from being able to. You are looking at it in terms of the entire number, but this isn't about the whole number. It is about the frequency at which one particular digit at the end of a number is found.
If I start counting from one to infinity and then end at a random point, the number of times I end at a 1 will always be either equal or greater than any other number for the very simple reason that we count in ascending order. Similar to ending with a 2 but for it we can never get a higher number of times of getting a 2 than a 1, only equal or less. Other than that 2 has a greater or equal number of times than the following ones. That results in a probability distribution that ends up causing numbers to follow Benford.

Anonymous

Anonymous Fri 06 Nov 2020 10:56:41 No.12310729 Report

Quoted By: >>12310758 >>12310777

>>12310675
i guess it is really hard to micromanage things... if you were to cheat in an election, what is the procedure? fake ballots? counting errors? filling ballots to make up for the gap? if there is any manual procedure like this, it would be really hard do micromanage and not leave some statistical anomaly somewhere i suppose.

Anonymous

Anonymous Fri 06 Nov 2020 11:02:13 No.12310737 Report

Quoted By: >>12310758

i guess im retarded, but what does "leading digit" refer to with regards to the election

Anonymous

Anonymous Fri 06 Nov 2020 11:14:01 No.12310758 Report

Quoted By: >>12310769 >>12310771

>>12310729
Well also, the rigging wasn't a sort of clear-cut plan of dropping X fixed number of votes and then bailing. They were doing it while they were in the process of counting and so they were imparting their own bias into how many votes need to be added just in case.

>>12310737
X X X X [ X ]
The last digit in this sequence is the leading digit.
>area A has 34851 votes for candidate alpha and 34912 votes for candidate beta
>area B has 21346 votes for alpha and 59287 for beta
>area C has 59273 votes for alpha and 29384 for beta
>A has a leading digit of 1 for alpha and 2 for beta
>B has a leading digit of 6 for alpha and 7 for beta
>C has a leading digit of 3 for alpha and 4 for beta
And by doing this for a shitload more areas you can plot the distribution of these leading digits and see if there is anything unusual. Since when you count you do it in ascending order, you can expect a certain distribution. But only if there is no rounding or any of that sort of limiting factor that would modify the number.

Anonymous

Anonymous Fri 06 Nov 2020 11:17:27 No.12310764 Report

Quoted By: >>12310797

>>12310629
Fine, it should read.
>Benford's law states that in any large dataset, the probability of encountering 1 as the leading digit will be 30%, with the probability of encountering subsequent digits dropping off logarithmically to 5% at 9.

So, 30% of precincts should have 1 as the leading digit as the number of votes cast for either candidate as long as you take a sufficiently large number of precincts.

This is the first I've heard of Benford's law and I don't think it's a very complicated concept. Still, I think it's vastly beyond the level of comprehension your average CNN viewer is capable of.

>>12310655
So produce your own charts showing how off benford's law is.

Anonymous

Anonymous Fri 06 Nov 2020 11:20:44 No.12310769 Report

Quoted By: >>12310778

>>12310758
isn't the leading digit the first digit?

Anonymous

Anonymous Fri 06 Nov 2020 11:21:08 No.12310770 Report

Quoted By: >>12310923

>>12310605
>pvalue = less than smallest float in python

what did they mean by this?

Anonymous

Anonymous Fri 06 Nov 2020 11:21:58 No.12310771 Report

Quoted By: >>12310778

>>12310758
So that should read
>>A has a leading digit of 3 for alpha and 3 for beta
>>B has a leading digit of 5 for alpha and 2 for beta
>>C has a leading digit of 3 for alpha and 4 for beta

Anonymous

Anonymous Fri 06 Nov 2020 11:24:13 No.12310777 Report

Quoted By:

>>12310729
>micromanage
i doubt it would be anything more than do-gooder vote counters putting a finger on the scale

Anonymous

Anonymous Fri 06 Nov 2020 11:25:10 No.12310778 Report

Quoted By: >>12310781

>>12310769
>>12310771
Fucked that one up, you are right. Although couldn't you get the same thing for the last digit as well? I would assume that it would be pretty much the same pattern, just flattened.

Anonymous

Anonymous Fri 06 Nov 2020 11:28:58 No.12310781 Report

Quoted By: >>12310785

>>12310778
no. the last digit would be completely random, so you'd get 10% for each digit

Anonymous

Anonymous Fri 06 Nov 2020 11:32:08 No.12310785 Report

Quoted By: >>12310788

>>12310781
Not entirely random since you are still counting in ascending order. The number of times you come across 1 is always either greater or equal to the number of times you come across any other number when counting.

Anonymous

Anonymous Fri 06 Nov 2020 11:35:57 No.12310788 Report

Quoted By:

>>12310785
fine. im, not a mathematician or statistician and this is the first ive heard of any of this stuff.

i should think any scientists or engineer should have no trouble figuring out the basic concept though.

Anonymous

Anonymous Fri 06 Nov 2020 11:42:45 No.12310797 Report

Quoted By: >>12310805 >>12310813 >>12311886

>>12310764
I just checked and like 99% of adults have a height that starts with 1 if measured in cm.
What are (((they))) hiding from us?

Anonymous

Anonymous Fri 06 Nov 2020 11:45:09 No.12310805 Report

Quoted By: >>12310820

>>12310797
Nice attempt at trolling.
You are, of course, a jewish democrat, right?

Anonymous

Anonymous Fri 06 Nov 2020 11:46:36 No.12310811 Report

Quoted By: >>12310887

Fucking idiots.
Assuming that all votes are equally distributed random variables? We have a situation where we have two clearlt distinct methods of voting, with different statistical characteristics, you fucking chuds.

Anonymous

Anonymous Fri 06 Nov 2020 11:47:06 No.12310813 Report

Quoted By:

>>12310797
Americans don't measure in cm. So what you will find is that 90% have a height that starts with a 5 instead. So really, you europoors are just short.

Anonymous

Anonymous Fri 06 Nov 2020 11:49:32 No.12310820 Report

Quoted By: >>12310826

>>12310805
Like 5 billion people is not a large enough dataset for Benford's law to apply? That sounds very limiting.

Anonymous

Anonymous Fri 06 Nov 2020 11:56:39 No.12310826 Report

Quoted By: >>12310900

>>12310820
Obviously when the data is tightly clustered around the first significant digit (i.e. small coefficient of variation), the law does not apply.

Anonymous

Anonymous Fri 06 Nov 2020 12:41:03 No.12310887 Report

Quoted By: >>12310891 >>12311280 >>12311877

>>12310713
>Unless polling stations are refusing people after a certain number
They sort of do this by design, you have one polling station per x thousand people.
>>12310811
why is Trump and other candidates' data all Benford consistent and Biden's isn't?

Anonymous

Anonymous Fri 06 Nov 2020 12:44:12 No.12310891 Report

Quoted By:

>>12310887
Because the percentage of milin votes for Trump is negligible, so you can assume all votes follow the same pdf and it kind of fits a single rv model.
Biden clearly has a two rv model, since inperosn and mailin votes have distinctly clear different pdf's.

Anonymous

Anonymous Fri 06 Nov 2020 12:50:33 No.12310900 Report

Quoted By: >>12310973 >>12311894

>>12310826
>Obviously, when the data doesn't fit my narrative, the law doesn't apply
Holy shit, how deep is the cognitive dissonance here?

Anonymous

Anonymous Fri 06 Nov 2020 12:53:51 No.12310907 Report

Quoted By:

>>12310605
It's bullshit.

Anonymous

View Same Google iqdb SauceNAO benford2.jpg, 342KiB, 1710x855

Anonymous Fri 06 Nov 2020 13:05:44 No.12310923 Report

Quoted By:

>>12310770

Anonymous

Anonymous Fri 06 Nov 2020 13:18:44 No.12310955 Report

Quoted By: >>12310965 >>12311003 >>12311010

Just separate the numbers for direct votes and for mailvotes and Benford's law will apply to both.

Goddammit, /sci/, why are you so retarded?

Anonymous

Anonymous Fri 06 Nov 2020 13:22:15 No.12310965 Report

Quoted By: >>12311009

>>12310955
that doesn't make any sense at all, you moron

Anonymous

Anonymous Fri 06 Nov 2020 13:25:50 No.12310973 Report

Quoted By:

>>12310900
He's right though. It only applies to data that spreads over an order of magnitude or more.

Anonymous

Anonymous Fri 06 Nov 2020 13:27:50 No.12310979 Report

Quoted By: >>12311024

I can post it later, but already did a small analysis on this.

If you extract the number of leadings digits if Total Vote. It doesnt follows Benford law. You will see an increase at the digit 6 then gradually go down.

Conaidering the amount of vote is 960 per ward, that Biden won 90% of them, and that in average he had 73% for each ward.

It is normal that he doesnt have tbat many 1, 2,3 because he fucking 73% of the votes on an average of 960 votes.

Anonymous

View Same Google iqdb SauceNAO biden fraud benfords twitter.jpg, 2MiB, 614x4909

Anonymous Fri 06 Nov 2020 13:28:14 No.12310980 Report

Quoted By: >>12310983 >>12311053 >>12311109 >>12311206 >>12311459

what are the odds of 130,000 ballots delivered at 4am all being for biden and 0 for trump?

Anonymous

Anonymous Fri 06 Nov 2020 13:29:22 No.12310983 Report

Quoted By:

>>12310980
https://streamable.com/xgsnl3

Anonymous

Anonymous Fri 06 Nov 2020 13:36:24 No.12311003 Report

Quoted By: >>12311009

>>12310955
Yeah when you combine two sets that both follow Benford's law so does the result.

Anonymous

Anonymous Fri 06 Nov 2020 13:40:34 No.12311009 Report

Quoted By: >>12311027

>>12310965
>hurrr durrr I don't understand it therefore you're a moron
Please rid yourself off this sorry place. Preferably permanently.
>>12311003
If both sets are of the same range of orders of magnitude. It's not the case here.

Anonymous

View Same Google iqdb SauceNAO ????.png, 864KiB, 905x789

Anonymous Fri 06 Nov 2020 13:40:44 No.12311010 Report

Quoted By:

>>12310955

Anonymous

View Same Google iqdb SauceNAO 4chan.jpg, 77KiB, 1575x609

Anonymous Fri 06 Nov 2020 13:46:23 No.12311024 Report

Quoted By: >>12311168

>>12310979
On the right is the frequency of each wards have < then x votes or between the previous one and the current number.

100 -> "0 to 100"
200 -> "100 to 200"
300 -> "300 to 400"
etc

Right is the leading digits per candidate and the total number of votes per ward. In yellow, you got the Benford curve.

You can clearly see that even the total number of votes doesn't respect the Benford curve. Simply because Ward population is a manmade create that here is this case averages 960 votes per Ward. So if you want to win in a Ward, you'll need about 400 to 500 votes at least (Assuming 2 candidates)

Then you got your Ward with populations Above 1000, as we can see, not that many goes above 2000 so either you win with a lot of votes and get over 1000 votes (Leading 1) or you get under 1000 votes, but still have the majority so anything between a leading 5 to 9.

Anonymous

Anonymous Fri 06 Nov 2020 13:47:28 No.12311027 Report

Quoted By:

>>12311009
Even if they're different magnitudes. If you added village populations to cities you wouldn't even change the city leading digits.

Anonymous

Anonymous Fri 06 Nov 2020 13:48:35 No.12311030 Report

Quoted By:

>>12310681
Trouble is that it's getting harder to tell the real retards from the pretend ones, for the same reason there can be no liberty without law, or standards without enforcement of rules that are comparably explicit.

Anonymous

Anonymous Fri 06 Nov 2020 13:57:39 No.12311051 Report

Quoted By:

>>12310605
Are these numbers actually scale invariant? Does benford's law even apply here?

Anonymous

Anonymous Fri 06 Nov 2020 13:59:25 No.12311053 Report

Quoted By:

>>12310980
Greater than 0.

Anonymous

Anonymous Fri 06 Nov 2020 14:35:39 No.12311109 Report

Quoted By:

>>12310980
Exactly zero. Only probabilities over ranges yield nonzero results.

Anonymous

Anonymous Fri 06 Nov 2020 14:58:58 No.12311168 Report

Quoted By: >>12311188 >>12311195

>>12311024
But why don't we see the same distortion for Trump's case? He won less smaller wards?

Anonymous

View Same Google iqdb SauceNAO chisquare2020.jpg, 396KiB, 2500x1000

Anonymous Fri 06 Nov 2020 15:05:30 No.12311182 Report

Quoted By:

>>12310605
I did a county by county analysis of each state. Here are my results.

Some interesting finds:
More statistically significant results in states that went for Trump
A couple of scenarios in states where the candidate who lost had significant results (Ohio, Oregon, New Mexico, Wyoming, Indiana, Washington, Vermont)
Biggest candidate discrepancy in Iowa (4.2 Biden, 27.8 Trump). Though I think their numbers were funny in 2016 too.

Anonymous

Anonymous Fri 06 Nov 2020 15:08:50 No.12311188 Report

Quoted By: >>12311195

>>12311168
He only won 10% of the wards, and only got 27% of the votes. He basically got a bunch of 100, 200, 300 votes... (He lost, bigly).

If anyone has the same kind of data, but for counties where Trump won by a big margin, it would be interesting to compare.

Anonymous

Anonymous Fri 06 Nov 2020 15:12:51 No.12311195 Report

Quoted By: >>12311211 >>12311262

>>12311168
>>12311188
To be a more precise, this doesn,t count the leading digits where Biden or Trump won.

It counts the leading digits for all wards by candidates, even if they lost or won the ward.

Anonymous

Anonymous Fri 06 Nov 2020 15:19:28 No.12311206 Report

Quoted By: >>12311261

>>12310980
>still making that claim
It was proven wrong within minutes and was a typo on one single news station. God, you people are retarded.

Anonymous

View Same Google iqdb SauceNAO 4chan2.jpg, 42KiB, 1149x384

Anonymous Fri 06 Nov 2020 15:21:09 No.12311211 Report

Quoted By: >>12311262 >>12311283

>>12311195
So I did the same thing, but I only extracted the leading digits where each candidate won a ward.

Neither are close to the Benford law. See post above to understand why it doesn't apply.

Anonymous

Anonymous Fri 06 Nov 2020 15:50:50 No.12311261 Report

Quoted By:

>>12311206
FWIW, fivethirtyeight (nate silver's website) said that sometimes they batch counts by candidate due to sorting, and any counts by Trump for that same area were included in another batch later on that day.

I think there was more than one case though; one was the "oh we accidentally put an extra 0 at the end" and the other was just batching.

Anonymous

Anonymous Fri 06 Nov 2020 15:51:05 No.12311262 Report

Quoted By:

>>12311195
>>12311211
Ah, I get it now. Thanks for the taking the time.

Anonymous

Anonymous Fri 06 Nov 2020 15:56:37 No.12311280 Report

Quoted By:

>>12310887
You have a station per 1000 people but you do not have stations instantly shutting down after reaching their 1000th voter.

Anonymous

Anonymous Fri 06 Nov 2020 15:57:12 No.12311283 Report

Quoted By: >>12311302

>>12311211
So in a nutshell, Benford's Law won't apply if a place is biased 60-40 for one candidate over the other, and an even bigger bias like a city for Biden would be even less "cooperative" with Benford's Law?

Anonymous

Anonymous Fri 06 Nov 2020 16:02:59 No.12311302 Report

Quoted By: >>12311939

>>12311283
Consideting that the average ward have just under 1000 votes total, then yes it is very unlikely that the benford law would apply.

If 60% of the votes go for candidates A, and 40% go for Candidate B. Then Candidate A would get a bunch of 6 has leading digits, and Candidate B a bunch of 4.

If we think of big wards over > 1000 votes, then yes the winner would probably end up have a leading digits of 7, 8, 9 (or 1 in the case we have very close to 2000 votes, or if the candidates wins by a big majority in that ward)

Anonymous

Anonymous Fri 06 Nov 2020 16:14:38 No.12311332 Report

Quoted By:

>>12310605
Wth are you trying to show here? The share of voters split in boxes of thousands, the first digit in that number?

Anonymous

Anonymous Fri 06 Nov 2020 16:47:23 No.12311459 Report

Quoted By: >>12311842

>>12310980
What's the probability that 5 seconds ago I stood up, did a cartwheel, sat on the floor, and went back to my bed?
Ex post facto probabilities are meaningless

Anonymous

Anonymous Fri 06 Nov 2020 18:34:47 No.12311842 Report

Quoted By:

>>12311459
I get the strange feeling you have no background in statistical analysis.

Anonymous

Anonymous Fri 06 Nov 2020 18:41:27 No.12311877 Report

Quoted By:

>>12310887
>why is Trump and other candidates' data all Benford consistent and Biden's isn't?
neither was found to be consistent retard

Anonymous

Anonymous Fri 06 Nov 2020 18:44:11 No.12311886 Report

Quoted By:

>>12310797
Because the heights are around the same order of magnitude and Benford's doesn't work in that case, you retard, see >>12310681

Anonymous

Anonymous Fri 06 Nov 2020 18:45:36 No.12311894 Report

Quoted By:

>>12310900
retard

Anonymous

Anonymous Fri 06 Nov 2020 18:53:42 No.12311939 Report

Quoted By:

>>12311302
Good post

Capcode	All Only User Posts Only Moderator Posts Only Admin Posts Only Developer Posts
Show Posts	All Only With Images Only Without Images
Deleted Posts	All Only Deleted Posts Only Non-Deleted Posts
Ghost Posts	All Only Ghost Posts Only Non-Ghost Posts
Post Type	All Only Sticky Threads Only Opening Posts Only Reply Posts
Results	All Grouped By Threads
Order	Latest Posts First Oldest Posts First

Your latest searches