Debate Now MIT Analysis of Voting Machine "Fraud" Analysis Thread.

Hey man, you probably think I'm an idiot because I generally don't take things too seriously. But I have a graduate school research background and I know statistical analysis and regression pretty well. I know you understand this as well from your posts on climate change and what the data really shows.

Don't think you're an idiot and I'm not shocked that you're a STEM type of hooligan.. LOL...
Having said that my take on the MIT professor's data is this. I find it troubling, but I cannot say conclusive. Statistically virtually anything is possible given enough trials. Quantum mechanics comes to mind. The data set he presents is troubling because it does seem to correlate to potential voter fraud. But to prove that case in a Court of Law the correlation would need to be damn near 100%. He has not done that.

It's troubling because it CANT prove vote shifting. It's only variables used are SPVoting and "Other Trump votes".. HE LEAPS to conclusions about the number of votes "stolen" with NOTHING in either axis that relates to "differential partisan turnout" or "partisan race totals" or even the strength of Republicans in that precinct.. His use of Repub SPVoting as a proxy for Republican STRENGTH in that district is even a stretch. The two variables are VOTING CHOICES given to Republican voters in Mich. Not anything COMPETITIVE related to the race..

Of COURSE the slope of line is gonna go down.. Because he SET IT UP to go down.. ANY data spilled into those axes has a slope of (OTV - RSPV) / RSPV. BY DEFINITION it's gonna go down because as SPV GOES UP -- the Y point goes DOWN !!!! Because of the subtraction in the numerator and a LARGER reduction from the denominator..

Might work if the minus became a plus a bit better. And THAT would be a clear variable definition -- because with the + sign -- that's the Total Trump vote in that district..


OTV would only INCREASE in precincts that AREN'T predominately leftist,. And the 4 largest Mich counties - like any state -- ARE PREDOMINATELY left leaning by wide margin..
It's completely those 2 simplistic variables. His use of Repub SPVoting as a proxy for Republican STRENGTH in that district is even a stretch.. So -- it's easy math.. But the SET-up to the problem was bungled badly..

As a quick aside, me and a group of three other students in grad school did a lengthy study on adult children of alcoholics in helping professions. The hypothesis in the literature for years was these folks would gravitate to helping professions due to an externalized locus of control. That is, they are more comfortable externalizing their need for control in helping other versus dealing with their own shit. A fancy construct basically akin to the idea of codependency.

We used medical students and graduate social work students as the experimental group, and graduate business students as the control. Long story short, there was no statistically significant difference. The adult children of alcoholic business students perceived their role as "helping," that is, "I want to help people with their taxes." Hahaha. Funny stuff. There was a qualitative as well as quantitative measures. The data was rock solid.

I mention this because it involved an arrow issue which often happens in reasearch but rarely gets mentioned. Which variable is truly the dependent variable versus the independent. What is really affecting what? Which way is the arrow pointing. This is often overlooked or misunderstood.

This MIT professor's data sets up a construct that makes many assumptions that may not fully conform to the reality, and then there is the question of which variables are truly influencing which variables. That is, which way is the arrow pointing?

Your concerns about his data set are valid in my opinion, but I think there is enough there to warrant a closer examination of possible fraud.
 
Last edited:
One problem I have with these things is some of the untestable assumptions they are based on, such as the x axis of the scatter-plot claiming a percentage of Republicans in each precinct. Percent based on what? Percentage of total eligible voters? Percentage of registered voters? Then from that deriving that the balance then MUST be the other party! What about independents?

I WISH it was the %registeredRepubs in the precinct.. That's an unequivocal measure of partisan power.. But it isn't.. He used the RSPV (Repub Straight Party Vote) percentage.. Also correlated with partisanship -- but not DEFINITIVELY.. SPV is an option in several states where you check a box for party and the entire rest of ballot is voted that way..

I did not like that the y-axis of the chart DOES NOT GIVE any label to what it is plotting against % of republicans! +20% of WHAT? -30% of WHAT?

The Y axis is OTV (Other votes for Trump) minus SPV. A rather odd construct. Because OTV + SPV is the TOTAL TRUMP VOTE and much cleaner definition.. By SUBTRACTING he's CAUSING the downslope on those lines. Because as SPV goes up -- The Y value is GUARANTEED to go down.. Regardless of all the other competitive and numerical metrics of the race.. It's garbage..
 
Last edited:
I did not like that the y-axis of the chart DOES NOT GIVE any label to what it is plotting against % of republicans! +20% of WHAT? -30% of WHAT?

It's the simple difference of what I said in the last post to you.. ALL he is plotting is DIFFERENCE of %OTV - %SPV.. It's a predictor of NOTHING -- other than the way people CHOSE to cast a ballot.. Not turn-out, not actual partisan registrations. Just a voting PREFERENCE.. It's TOO simplistic to FIND "machine fraud"..

AND - its GUARANTEED to make Y go down as X increases.. Simple math/graphing. EVEN WITHOUT any machine fraud...

Here's what the Doc probably thought.. In any race with a CHOICE of voting SPV -- as the partisanship of the district (X) increases -- the SPV INCREASES.. And that means that Repubs in this case would be casting FEWER FULL ballot votes and net balance would be ZERO change in the Trump vote across the X axis.. But that's not what how his choice of Y axis functions.

THAT thinking FAILS in a HOTLY Partisan race where MORE Repubs tend to vote MORE party line all the way down the ballot.. They check the SPV box and call it a day.. They vote FULL PARTY LINE..

So -- test the limits of OTV - SPV here.. If 100% of them go full gonzo partisan SPV -- OTV will ONLY be the number of Dems crossing party lines and Independents. ITS GUARANTEED to be a negative number as long as Repubs outnumber Dem crossers/Indies!!!!!

To be a specific example -- say the TOTAL trump vote in a precinct (OTV + SPV) is 45% --- 80% of Republicans choose SPV and 10% of the OTV is Dem crossers and Independents. Subtract the 10% for the crossers and indies = 35%.. The rest are Repubs. So 80% of 35% is 28%..

The other 20% of Repubs = 7%.. The OTV = 10% + 7% = 17%. (OTV - SPV) = 17% - 28% = -11%..

GUARANTEED TO BE A SIGNIFICANT NEGATIVE NUMBER WITH NO MACHINE ISSUES !!!!!

THE ONLY THING this MIT hothead is measuring is WHICH METHOD the voters chose to cast their vote !! Either STRAIGHT PARTY or filling out the entire ballot.. It's a FAILURE to find any data forensics for machine issues..


You can do the math if the SAME district, 45% vote for Trump total finds only 40% of voters choose RPV.. With the same assumption of a 10% Dem crosser/Indie vote.. You'll find that whole negative sloping line is nothing other than a PREFERENCE for a particular method of casting their vote..

45% - 10% = 35% Repub vote. 40% SPV of 35% = 14% 60% of 35% = 21%

The 21% goes to OTV so OTV is 10 +21 = 31% --- SPV is 14% --- SO Y = 31 - 14 = +17%..

So in the 1st example. The X value 80% ,,, Y value is = -11%
In the 2nd example X = 40% and Y = +17%


THERE's YOUR NEGATIVE SLOPE LINE WITH NO MACHINE ISSUES !! ALL THAT CHANGED was how many voters CHOSE to use SPV..

It changes from + to - when the SPV crosses about 50%.. Depending on the assumption about the Dem Crossers and Indies.. You WOULD ASSUME that LIKELY those crossers and indies would be fairly flat REGARDLESS of the strength of Republicans in the district..


Also crosses zero differently if you assume a differ TOTAL TRUMP VOTE % in that precinct.. But it STILL slopes negatively..
 
Last edited:
The Y axis is OTV (Other votes for Trump) minus RSVP. A rather odd construct. Because OTV + SPV is the TOTAL TRUMP VOTE and much cleaner definition.. By SUBTRACTING he's CAUSING the downslope on those lines. Because as SPV goes up -- The Y value is GUARANTEED to go down.. Regardless of all the other competitive and numerical metrics of the race.. It's garbage..

I am not sure I understand what you are trying to express.
It doesn't matter that the line is going down because they subtract (if they added it would go up).

The problem is that the line is skewed in reference to both the X and the Y axes.
Variation in the population of the data can always be expected, but the concern is more that the variation takes form in a pattern.

The thing I found most interesting was the consistencies in the variations and the patterns ... For Instance:

Data population lower on the X axis (preferably around 20%) all looked about the same and showed little skew.
Those precincts would favor more Democrats than Republican.
At the same time, for the most part, they are higher on the Y axis than the SPV, and even higher than that in regards to Republican Leaning Precincts.
Also the more a Precinct favored Republicans, the greater drop away from the SPV.
The increasing disparity along the X axis (skewing the data) and in relation to the saturation of the precincts with more Republicans ...



... Would not only suggest corrupted data, but targeted corrupt data.


.
 
The Y axis is OTV (Other votes for Trump) minus RSVP. A rather odd construct. Because OTV + SPV is the TOTAL TRUMP VOTE and much cleaner definition.. By SUBTRACTING he's CAUSING the downslope on those lines. Because as SPV goes up -- The Y value is GUARANTEED to go down.

Thank you, I thought maybe it was just me. Subtracting the OTV made the Y axis meaningless because it actually HURTS the candidate getting these votes, especially if the precinct is more red! Neither of which makes sense to me. I could not wrap my head around what he was plotting against the x axis that would give the point spread distribution any meaning and apparently that is why.
 
The Y axis is OTV (Other votes for Trump) minus RSVP. A rather odd construct. Because OTV + SPV is the TOTAL TRUMP VOTE and much cleaner definition.. By SUBTRACTING he's CAUSING the downslope on those lines. Because as SPV goes up -- The Y value is GUARANTEED to go down.

Thank you, I thought maybe it was just me. Subtracting the OTV made the Y axis meaningless because it actually HURTS the candidate getting these votes, especially if the precinct is more red! Neither of which makes sense to me. I could not wrap my head around what he was plotting against the x axis that would give the point spread distribution any meaning and apparently that is why.

I don't know the degree to which the "up or down" skew would matter.

To me, the exponential increase in corruption (skew along the X axis), in reference to the saturation of Republicans in the precinct ...
would suggest that whatever corrupted the data, was targeted to that saturation level.

The more a precinct was saturated with Republicans, the more the data was skewed, and in a linear representation.
That's not an accident ... That's a target.

.
 
Just thinking on-line on what Dr. Shiva MIGHT have intended.. If the - sign is replaced by + sign -- the Y function becomes OTV + SPV.. The SLOPE of that graph (if any) becomes

(OTV + SPV)/SPV or (OTV/SPV) + 1

Gonna redo the 2 examples above..

45% Trump vote in a precinct with 10% Indies and Dem crossers. 2 cases,.

SPV = 80%
Republican vote is 35%., 80% of 35% = 28% 20% of 35% = 7%
The 7% is added to the 10% Indie/Crosser = OTV = 17%
SPV = 28%

Y= 17% + 28% = 35%
X=80%

Second data point.

SPV = 40%
Republican vote is 35%., 40% of 35% = 14% 60% of 35% = 21%
The 21% is added to the 10% Indie/Crosser = OTV = 31%
SPV = 14%

Y= 31% + 14% = 45%
X=40%
 
The Y axis is OTV (Other votes for Trump) minus RSVP. A rather odd construct. Because OTV + SPV is the TOTAL TRUMP VOTE and much cleaner definition.. By SUBTRACTING he's CAUSING the downslope on those lines. Because as SPV goes up -- The Y value is GUARANTEED to go down.
Thank you, I thought maybe it was just me. Subtracting the OTV made the Y axis meaningless because it actually HURTS the candidate getting these votes, especially if the precinct is more red! Neither of which makes sense to me. I could not wrap my head around what he was plotting against the x axis that would give the point spread distribution any meaning and apparently that is why.
I don't know the degree to which the "up or down" skew would matter.

It doesn't because like Flacc said, it was merely reflecting voting HABITS or choices against an increasingly negative number made so by the fact that he was always SUBTRACTING the republican straight party voting which might merely correlate with an increase in how red the precincts were! Hardly any surprise. That and the plot of chiefly blue Detroit which showed a totally different distribution were only really reflecting natural causes (party affiliation) against his algorithm which was a function of HOW you voted, NOT whether there was any artificial vote flipping / vote stealing bias in the machines themselves, made more convincing by the fact that Detroit, his one control, simply had too few republican voters in it (~5%) to kick in the effects of his algorithm, rather than showing a freedom of machine cheating.

Not only did he choose a wrong formula to use against a wrong set of plot variables for the two axes, he would need to do the same for 2016 IMO, a year allegedly absent the machine bias if he wants any hope of detecting or proving software bias.

Then he needs to explain why it only affected Trump and not other key republicans. I mean, if you could do that to Trump, then why let other GOP members win either?

I set out to propose a different kind of graph and variables with zero at the center rather than the corner to plot in four directions instead of two, but I still don't think it could prove machine bias.

The way to prove machine bias is examine the machine code used and to conduct audits during the voting.

IMO, you shouldn't have to PROVE fraud to the court in order to FIND fraud! If fraud is suspected they ought to simply do a review of the ballot counting in select areas + an audit of the machine codes used to see whether any weighting was used, but then, you'd hope they would be doing this BEFORE the election in the first place.
 
Your concerns about his data set are valid in my opinion, but I think there is enough there to warrant a closer examination of possible fraud.

He found nothing about machine errors.. That's the point. Only a preference for choosing WHICH method to cast a ballot..


Agreed. But the frequency distribution is off. It does not "prove" machine error, but one can infer the data set does not follow a logical statistical pattern based on a predicted frequency distribution. That is the professor's point which I think you are missing.

In a frequency distribution based on human height the odds of 11 foot human beings showing up in the distribution are virtually nil. If they show up 3% of the time something is likely wrong with the data set. Understand?

The data the professor provides does that not fit a logical distribution pattern based on identified voter preferences. Now it could be a garbage in garbage out situation. That is my concern and yours if I am understanding you correctly. But without question the data as presented does not logically fit an anticipated frequency distribution. Something is off somewhere. It warrants further investigation.
 
Last edited:
The Y axis is OTV (Other votes for Trump) minus RSVP. A rather odd construct. Because OTV + SPV is the TOTAL TRUMP VOTE and much cleaner definition.. By SUBTRACTING he's CAUSING the downslope on those lines. Because as SPV goes up -- The Y value is GUARANTEED to go down.
Thank you, I thought maybe it was just me. Subtracting the OTV made the Y axis meaningless because it actually HURTS the candidate getting these votes, especially if the precinct is more red! Neither of which makes sense to me. I could not wrap my head around what he was plotting against the x axis that would give the point spread distribution any meaning and apparently that is why.
I don't know the degree to which the "up or down" skew would matter.

It doesn't because like Flacc said, it was merely reflecting voting HABITS or choices against an increasingly negative number made so by the fact that he was always SUBTRACTING the republican straight party voting which might merely correlate with an increase in how red the precincts were! Hardly any surprise. That and the plot of chiefly blue Detroit which showed a totally different distribution were only really reflecting natural causes (party affiliation) against his algorithm which was a function of HOW you voted, NOT whether there was any artificial vote flipping / vote stealing bias in the machines themselves, made more convincing by the fact that Detroit, his one control, simply had too few republican voters in it (~5%) to kick in the effects of his algorithm, rather than showing a freedom of machine cheating.

Not only did he choose a wrong formula to use against a wrong set of plot variables for the two axes, he would need to do the same for 2016 IMO, a year allegedly absent the machine bias if he wants any hope of detecting or proving software bias.

Then he needs to explain why it only affected Trump and not other key republicans. I mean, if you could do that to Trump, then why let other GOP members win either?

I set out to propose a different kind of graph and variables with zero at the center rather than the corner to plot in four directions instead of two, but I still don't think it could prove machine bias.

The way to prove machine bias is examine the machine code used and to conduct audits during the voting.

IMO, you shouldn't have to PROVE fraud to the court in order to FIND fraud! If fraud is suspected they ought to simply do a review of the ballot counting in select areas + an audit of the machine codes used to see whether any weighting was used, but then, you'd hope they would be doing this BEFORE the election in the first place.

I am not thinking about the voter and what they may have done ... I'm thinking about the data.

Nut-shell version:
The equation/condition applied to all points along the Y axis is equal to all points.
The equation/condition applied to all points along the X axis is equal to all points.
The equations/conditions are applied to create a measurable variation.

When the variation skews, something has happened.
When the variation skews in an exponential linear fashion something specific has happened.

I mean if y'all want to think something is wrong with equation/condition they applied to the points, it's still equal to all points.
You could add and subtract all kinds of numbers to each point, and as long as you did it equally, you would still have the same representation.

The problem is that the representation is skewed (corrupted).
It wouldn't matter if we were counting votes or weighing chicken eggs ... The skew (corruption) is linear.





Oops, Sorry ... I don't know what I was thinking.
In fact, I going to forget everything I just posted.
I gonna leave this discussion alone, and y'all have a lovely evening.

.
 
Nut-shell version:
The equation/condition applied to all points along the Y axis is equal to all points.
The equation/condition applied to all points along the X axis is equal to all points.
The equations/conditions are applied to create a measurable variation.

When the variation skews, something has happened.
When the variation skews in an exponential linear fashion something specific has happened.


Thank you. You expressed this much better than I did. The data is not distributed in a logical or reasonable manner based on the expressed initial conditions. Now Flacaltenn I think has a problem with the initial conditions established in the data analysis. I understand that and It may be a valid criticism. My guess is it was the only data the professor had so he used it. That does not mean he was biased per se, but he likely worked with what he had.

But the bottom line, if you place any validity on the initial conditions set, then something is off. It is concerning and needs further investigation.
 
Last edited:
Thank you. You expressed this much better than I did. The data is not distributed in a logical or reasonable manner based on the expressed initial conditions. Now Flacaltenn I think has a problem with the initial conditions established in the data analysis. I understand that and It may be a valid criticism. My guess is it was the only data the professor had so he used it. That does not mean he was biased per se, but he likely worked with what he had.

But the bottom line, if you place any validity on the initial conditions set, then something is off. It is concerning and needs further investigation.

LoL ... One last thing that I am just now forgetting.

If I had to guess, the corruption that caused the skew would be machine driven.
If it was tried by human interference, there would be more variation in the skew.

But in all honesty, that's just an opinion.


.
 
Last edited:
The Y axis is OTV (Other votes for Trump) minus RSVP. A rather odd construct. Because OTV + SPV is the TOTAL TRUMP VOTE and much cleaner definition.. By SUBTRACTING he's CAUSING the downslope on those lines. Because as SPV goes up -- The Y value is GUARANTEED to go down.
Thank you, I thought maybe it was just me. Subtracting the OTV made the Y axis meaningless because it actually HURTS the candidate getting these votes, especially if the precinct is more red! Neither of which makes sense to me. I could not wrap my head around what he was plotting against the x axis that would give the point spread distribution any meaning and apparently that is why.
I don't know the degree to which the "up or down" skew would matter.

It doesn't because like Flacc said, it was merely reflecting voting HABITS or choices against an increasingly negative number made so by the fact that he was always SUBTRACTING the republican straight party voting which might merely correlate with an increase in how red the precincts were! Hardly any surprise. That and the plot of chiefly blue Detroit which showed a totally different distribution were only really reflecting natural causes (party affiliation) against his algorithm which was a function of HOW you voted, NOT whether there was any artificial vote flipping / vote stealing bias in the machines themselves, made more convincing by the fact that Detroit, his one control, simply had too few republican voters in it (~5%) to kick in the effects of his algorithm, rather than showing a freedom of machine cheating.

Not only did he choose a wrong formula to use against a wrong set of plot variables for the two axes, he would need to do the same for 2016 IMO, a year allegedly absent the machine bias if he wants any hope of detecting or proving software bias.

Then he needs to explain why it only affected Trump and not other key republicans. I mean, if you could do that to Trump, then why let other GOP members win either?

I set out to propose a different kind of graph and variables with zero at the center rather than the corner to plot in four directions instead of two, but I still don't think it could prove machine bias.

The way to prove machine bias is examine the machine code used and to conduct audits during the voting.

IMO, you shouldn't have to PROVE fraud to the court in order to FIND fraud! If fraud is suspected they ought to simply do a review of the ballot counting in select areas + an audit of the machine codes used to see whether any weighting was used, but then, you'd hope they would be doing this BEFORE the election in the first place.

You make a good case that the more saturated a precinct is with Republicans the more likely they are to vote straight ticket.
The only problem with that would be when the skew doesn't start at 0, but starts at 20-40%.

If you can explain how that threshold in the data is introduced, then I can clearly accept your interpretation.
When I look at the data, my concern is the the scatter plot is in control at one point, and out of control at another.

I am not saying the the model could not answer a question.
I am not saying that the model asks the wrong question, and gets the wrong answer.
I am suggesting that the model asked one question, and arrived at two conflicting answers.

With that interpretation, I am also suggesting, that if we were looking at control data associated with injectable pharmaceuticals,
I wouldn't be comfortable with them sticking a needle in your mother's arm.

.
 
Last edited:
You make a good case that the more saturated a precinct is with Republicans the more likely they are to vote straight ticket.
The only problem with that would be when the skew doesn't start at 0, but starts at 20-40%.
If you can explain how that threshold in the data is introduced, then I can clearly accept your interpretation.


At this point I simply have better things to worry about and don't think the problem really worth my time. Without the cooperation of the government and election officials and a much wider study done over more than one election cycle, I don't think the original goal of the study is obtainable.

Since there will always be a RANGE of degrees in which voters apply the OTVs, I am content to assume that it takes until about 20% GOP voter by precinct in order for the Shiva algorithm to affect the random range of the OTV enough to begin to overcome that to the point of introducing the noticeably downward chute in the plot without going through the bothersome task of trying out the theory with myriad sample calculations in order to test the efficacy.
 
I'm calling a halt here. Need some "me" time on this.. I'm confused by how he DESCRIBED certain metrics that were selected to graph.. Particularly, the thing he calls "Individual Trump Votes".. Which, because he segregated SPV voting and Full Ballot Voting -- I took to mean the percentage of Trump votes cast by Full Party Ballot..

His examples PRIOR to loading the data into the graph seem to indicate it's the FULL trump vote at that precinct.. More like (OTV + SPV) for ONLY the Republican side -- than what I thought was (OTV - SPV).. When I watched with the sound off and didn't USE his terminology -- I discovered this..

SO -- reviewing this again.. MIGHT start over -- dunno.. There would STILL be issues with a couple of his assumptions, but MAYBE there IS an anomaly here.. Went from hoping to 85% negative and now I'm back at about 40% negative.. LOL...

See y'all later.
 
The devil is only in the data if you either put it there, or try to ignore it.

I think I see something in the data,
I think I know how it could be done.

I am sure my theory could be debunked if the machines in question had open source code.
I am sure that if the code is not reviewed, we will never know if I am correct.

.
 
Last edited:
The devil is only in the data if you either put it there, or try to ignore it.

I think I see something in the data,
I think I know how it could be done.

I am sure my theory could be debunked if the machines in question had open source code.
I am sure that if the code is not reviewed, we will never know if I am correct.

.

My thought is the voting machines could randomly assign a fractional value to any R Presidential vote as long as the ultimate value is a whole number.

Very easy to do and we already know software exists is the Dominion machines to assign fractional votes.

So, on every even vote count from lets say 2 to 24 for the R Presidential candidate the voting machine assigns a .5 value for the vote. Only program this for the Presidential vote, not down ballot. You don't want everything to look hinky....just kill Trump.

Admittedly, it is only a theory, but my theory is supported by the professor's data. There is a variance in the data that does not look like it can be supported by random chance. However, to make this leap you would have to support the professors initial conditions. That is debatable, but then in any research the initial conditions set by the researcher is almost always debatable. That's why it's called research. :).

I am not a computer expert, but I would love to see someone who is look at the source code of some of these voting machines. I think it could be interesting.
 

Forum List

Back
Top