Looks like we might have some bigger things going on now but I went down a thread today on Twitter from Mr. Taleb and it seems like it might not be the best way to detect fraud.
Just curious if any of the autists in the group may disagree. Nassim Taleb seems to give it to people straight.
Even more details I found going through his Twitter feed.
No math forensic fraud detection is pointing out guarantees of fraud, its just looking for signs of manipulation which can then be levereged into investigations of fraud. Sure, its POSSIBLE that you have a voter turnout around 90% in WI, but its not likely. Therefore you investigate to determine which side of that coin it lands on.
His "this number needs to be more likely than that number" argument loses teeth when you consider there isnt a clear "likely" number (like 5 ft in his height example) because who you vote for isnt a numerable trait, it's usually one of 3 things and lawfully only one vote per person. the larger the sample size, the more the data should conform to predictions and we are using numbers in the hundreds of thousands.
a math logic pede in there, trying to show them fuckery...
and they STILL are like:
LOL
I been looking into it some more. It sure does look like it was used in other elections.
https://www.kdnuggets.com/2020/09/diy-election-fraud-analysis-benfords-law.html
Looks like we might have some bigger things going on now but I went down a thread today on Twitter from Mr. Taleb and it seems like it might not be the best way to detect fraud.
Just curious if any of the autists in the group may disagree. Nassim Taleb seems to give it to people straight.
Even more details I found going through his Twitter feed.
http://www-personal.umich.edu/~wmebane/inapB.pdf
I am sure there is 27 other avenues to check on fraud. But just wanted to share to get/give another angle.
No math forensic fraud detection is pointing out guarantees of fraud, its just looking for signs of manipulation which can then be levereged into investigations of fraud. Sure, its POSSIBLE that you have a voter turnout around 90% in WI, but its not likely. Therefore you investigate to determine which side of that coin it lands on.
His "this number needs to be more likely than that number" argument loses teeth when you consider there isnt a clear "likely" number (like 5 ft in his height example) because who you vote for isnt a numerable trait, it's usually one of 3 things and lawfully only one vote per person. the larger the sample size, the more the data should conform to predictions and we are using numbers in the hundreds of thousands.