Many people have messaged me asking about the electronic voter fraud allegations of @va_shiva (who is currently on twitter's naughty list - they are shadowbanned).

I am going to address his allegations now. If you don't know what I'm talking about already, pay this no mind.

Outside of the specific allegations of voter fraud, Dr Shiva makes complaints about voting record retention and data integrity. These are more or less uncontroversial.
They are outside of the charge that fraud was committed, and are instead about process and data integrity.
So if you wish to discard his claims that fraud has occurred during this election, you can still consider his thoughts on increasing transparency and integrity of voting procedures. There is nothing particularly odd about his asks.
Moving on to the meat of his fraud claims as laid out in his original presentation from Nov 10th, there are two feature-set claims and then a data analysis. The feature-set claims are assumed to be factual, but I would honestly have no direct way of validating it myself.
The first claim is that the database of some of these systems store votes as a data type called "double."

For those who aren't tech-savvy, when designing computer software you have to tell the system how to treat data.
If you woke up one day and found your bank account with a balance of 'CUPCAKE', you would be concerned, because your bank account should have a number value, not a word.
Similarly, if you hear the average family has 2.5 children it makes sense, but to hear your neighbors have 2.5 children would be alarming, unless it's a joke involving a pet.

So software must take these things into consideration - and set expectations about data.
You would not want a system to record a specific family as having 2.5 children. That would be nonsensical.

The allegation Dr Shiva makes is that the databases /can/ store a fraction of a vote.

For most regular, US based, elections, that is odd.
If true, it means the system /can/ store a fraction of a vote. Hypothetically, a candidate could wind up with fractions of a vote in the database. In the context of an average US election, this is odd.
The designer of the database system has a choice to restrict this data to whole numbers. Decimals would not be permitted. If this allegation is correct, then there was likely a reason fractional number storage was chosen.
That reason does not have to be nefarious. I deliberately said this would be odd for 'the average US election,' because I am not privy to every possible customer and every election format the database is being designed for.
So what matters next is the software that interacts with the database. The same data considerations would be made on the software level.

Dr Shiva alleges the software has a "weighted mode," where if turned on, a vote for candidate A would counted differently than a vote for B.
If true, this could be where the ability to save a fractional vote could be employed.

Imagine a vote for candidate A being saved as 1.1 and for candidate B as 1.0. After 10 "votes" for each, A would total to 11 and B 10. This is the plot of office space
But the existence of the feature could likewise be innocent. I cannot pretend to know how many customers the software is being designed for and what types of voting systems are being considered.

So what ultimately matters is if that feature was turned on.
So if a) the claims about the database are true and b) the software features exist, then the ability to audit the electronic systems to know what a machine's configuration was at the time of voting is of high importance. The system is only as secure as your ability to do that.
Therefore either we know how to get these answers, and getting them a regular part of election auditing and certification, or we don't know and skepticism and rumors are impossible to control, because nobody can claim to know. It's not a dumb question.
Now I'll get to the salacious accusations, which is that an algorithm is being employed to "steal" votes from Trump and give them to Biden.

If you haven't already watched the presentation or know the argument, you may want to do so before reading further.
Dr Shiva's analysis hinges on comparing 'straight ballot' results to 'split ballot' results. That is, by comparing the percentage of people voting all Republican versus all Democrat, and the people voting a mix of R and D.
By arranging counties by the ratio of ballots all R to all D, and setting that % ratio as a baseline of 0, Dr Shiva plots the arithmetic difference in split ballot vote% for Trump compared to baseline.
If 50% of the people voting all one-party vote Republican, and 50% of the mixed ballot vote for Trump, then at the 50% mark on the horizontal access, the "blue" data point is on the red baseline at 0. Dr Shiva alleges this is the normal expectation.
Dr Shiva claims one should expect "more or less" the people voting a mix of R and D to vote for Trump around the same rate as people who are voting completely Republican.
The viewer is then told this pattern is a "red flag"
And then provided what are purported to be actual results
It is then articulated by Dr Shiva that his interpretation of this data is that the electronic voting systems /must/ have an algorithm employed that "takes" votes from Trump "gives" them to Biden.
Dr Shiva makes the following strong claim.
So let's explore this claim, logically and rationally.
First, without knowing anything about how people actually voted, we can evaluate properties of his analytical setup to learn more about how to expect.
Let us take this step by step. Let us assume it is voting day. And now, on your horizontal axis, you have counties in no specific order, and on the vertical axis you have the percentage of straight R ballots versus straight D ballots. That graph would look like this.
Because there is no order, there values fluxuate like static. Some places nobody voted straight R, some completely R.

Now we will order them by the % of R votes.
As you can see, we get a nice 45 degree line from left to right, as you would expect because we are putting the lowest straight R areas to the left and the highest to the right. This is a "1:1 mapping" and it could not look any other way.
Now consider the mixed ballots and % vote for Trump per county in this view. In any county, as many as 100% of mixed ballots could vote for Trump and as few as 0% could vote for Trump. Those are our logical boundaries for the data.

The top of the graph and the bottom.
It would be impossible for 101% or -1% to vote for Trump in any area, which means logically we are between 0 and 100%.

But our graph doesn't look like Dr Shiva's yet, and that's because we haven't performed his transformation.
In order to make our graph match Dr Shiva's setup we have to turn the 45 degree diagonal line into a horizontal baseline using subtraction.
So we will take our maximum value of 100% and our minimum value of 0% for the mixed vote % going to Trump and subtract the StraightR% vote /exactly/ like Dr Shiva does for his data. And this is the result.
The blue lines are our 100% and 0% boundaries inside of his visual framework. This means it is logically impossible for any data points to go above or below the blue lines.
Let us pretend for a second that Donald Trump was insanely popular. Astronomically popular. And that on average 80% of split ticket votes went to Trump no matter what county. Because of this setup that graph would look like
This means that at a minimum, his portrayal of the algorithm that he says "must" be in use is flawed in this presentation, as even making Trump universally loved yields the same type of linear slope he claims must be algorithmic.
In a followup video Dr Shiva changes his argument of what a "normal" situation should look like this. He now argues the results should be parabolic - which is different than his previous claim which suggested that the results should roughly parallel to baseline.
Returning to his original graphs you can see him draw straight horizontal lines prior to the downward diagonal trajectory.
This interpretation is one of choice. One could easily fit a parabolic line to this data and therefore eliminate the claim they violate the normal case.
In conclusion, I am not only unpersuaded by his analysis, I am dissuaded by it in its current form.

I can agree wholeheartedly with his pleas for more transparency and some of his recommendations for process enhancement. But in its current form his analysis is very weak.
Specifically, he has failed to establish what the expected "signal" should be - even though he spends a good deal of time making analogies to signals - and therefore has failed to establish the results do not fit the expected "signal."
If Dr Shiva would like to persuade people like me, who are open minded about the possibility and willing to seriously consider the arguments, he will need to retool his analysis. It is very weak in its current presentation.
*Their values

• • •

Missing some Tweet in this thread? You can try to force a refresh

Keep Current with President-Elect Dr RollerGator PhD

President-Elect Dr RollerGator PhD Profile picture

Stay in touch and get notified when new unrolls are available from this author!

Read all threads

This Thread may be Removed Anytime!


Twitter may remove this content at anytime! Save it as PDF for later use!

Try unrolling a thread yourself!

how to unroll video
  1. Follow @ThreadReaderApp to mention us!

  2. From a Twitter thread mention us with a keyword "unroll"
@threadreaderapp unroll

Practice here first or read more on our help page!

More from @drrollergator

21 Nov



Read 6 tweets
20 Nov
Read 7 tweets
3 Nov
Good morning, everyone. Today is the day the United States votes for the next citizen Commander in Chief.
2020 has been an amalgamation of all the forces, pressures, and stresses that resulted in Donald J. Trump winning election in 2016, on overdrive, bombarding us in an unrelenting fashion, daily.
We are worse off than we were four years ago in many important ways. This is not all Trump’s fault. It is not even necessarily mostly his fault.
Read 25 tweets
2 Nov
Read 8 tweets

Did Thread Reader help you today?

Support us! We are indie developers!

This site is made by just two indie developers on a laptop doing marketing, support and development! Read more about the story.

Become a Premium Member ($3/month or $30/year) and get exclusive features!

Become Premium

Too expensive? Make a small donation by buying us coffee ($5) or help with server cost ($10)

Donate via Paypal Become our Patreon

Thank you for your support!

Follow Us on Twitter!