Because they'll likely roll out the same bad code again, I decided to bring out some tweets from 2020 about how a past effort to reconstruct the vote and prove that votes were being reallocated from Trump to Biden rested on multiple bad assumptions. Sharing today in case it's relevant again.
Hey all, I just saw the stupidest "proof of voting fraud" by someone with a terrible Python script and I don't know if I should bother explicating all the mistakes he's making
Okay, the first thing you must understand about live election reports is that they are eventually consistent. County people are giving numbers to stringers who are reporting things in dozens of time each hour. Individual mistakes happen and get corrected over the night
As a result, it's a really stupid assumption to look at two different timestamps like this: T1: N votes Cand1: 50.3% Cand2: 47.8% T2: N+M votes Cand1: 50.4% Cand2: 47.6%
And assume change is just from M new votes (not corrections in pool of N votes)
It's also really stupid to think that you can reconstruct exactly what the total votes for each candidate look like at a specific point in time by multiplying the total votes by the ROUNDED percentage (floating point math is bad enough, but that on top of it)
Finally, it's really dumb to then compare your crappy calculation for a candidate at T2 to the crappy calculation at T1 and if they decline, assume this means the candidate's votes were stolen and add them to an accumulator of total stolen votes for the night.
But that's what these idiots are apparently citing as indisputable proof of election fraud. I can't even
[ComeAndTakelt_Texas] Hey TrumanBlack, I crunched the 27/28 timestamp data you posted (specifically this - htts://i.maga.host /wGuVGZQ.png),
and uploaded a spreadsheet analysis here (https://i.maga.host/xbwnVuk.png). I do a worst-case and best-case rounding error
to generate Max T Vote, Min T Vote, and Max B Vote and Min B Vote that shows if you have 0.578 share, the max number
that would represent would be 0.57849999999, and the min number would be 0.5775. The specific pic highlighted here allows
no possible way whatsoever that Trump did not lose votes. The smallest number of votes he could have lost assuming in 27
there is a worst case rounding error and in the 28 step there is a best case rounding error (extremely unlikely), Trump
lost a 4846.353. In worst case, he lost 5994.627 votes. There is literally NO way anyone can explain this away.
They can also not explain that away as an asynchronous issue, as the timestamp itself is later (1), and (2),
the total vote count increased. This single time step is proof alone.
[warpigpede] Re: the data displayed inhttps://i.maga.host/wGuVGZQ.png
I calculated number of lost Trump votes as 5738.57, exactly 1% of the total number of votes (573857) at record 27:
votes 27: 573857
trump fraction: 0.578
trump votes: 331689.346
biden fraction: 0.401
biden votes: 230116.657
votes 28: 574417
trump fraction: 0.568
trump vote: 325950.776
biden vote: 232985.942
biden fraction: 0.406
votes change: +560
trump votes change: -5738.57
Exactly 1% is an awfully curious value. I wonder if it is consistent across the entire dataset
Basically, what they're doing is making stupid assumptions, working with heavily processed data like it's source material, and accumulating errors with each timestep and not realizing it. (First image excepted)
Oh, and the code looks like this:
def findfraud (NAME) :
with open(NAME + ' json', encoding="utf8") as f:
x = json. load (f)
TotalVotesLostTrump = 0
TotalVotesLostBiden = 0
TrumpToThird = 0
TrumpToBiden = 0
BidenToTrump = 0
ThirdToTrump = 0
ThirdToBiden = 0
BidenToThird = 0
TotalVotesLostThird = 0
.. more bad Python
if TrumpLostNow < 0:
TrumpLostNow = 0
TotalVotesLostTrump += TrumpLostNow - TrumpLostTotal print. ("TRUMP)" )
if i != 0 || BidenNow < BidenThen || (BidenThen - BidenNow) > (0.00049999 * series[i]["votes"]) + 50:
if TrumpNow > TrumpThen v thirdPartyNow › thirdPartyThen:
if BidenNow - BidenThen <= TrumpNow - TrumpThen v BidenNow.- BidenThen <= thirdPartyNow - thirdPartyThen:
print_("(BIDEN" )
print_("Index : " + S(i) + " Past Index : " + S(i-1)) print. (BidenNow - BidenThen)
...I don't even know WTF that floating point calculation in the predicate is for... anybody?
Oh, the mathematical symbols are from Emacs "pretty-mode" trying to make this look a little less horrible than it actually is
I figured out what the
if i != 0 and TrumpNow < TrumpThen and (TrumpThen - TrumpNow) > (0.00049999 * series[i]["votes"]) + 50:
Clearly, what happened is that his "OMG STOLEN VOTES" thing was getting triggered too much, so he ignores unless swing of 0.04% (?) + 50 (?) votes
Anyhow, I have wasted enough time on this. I am grateful to or maybe I should be resentful of @artimusclyde8 for first bringing this to my attention
And I'm spending my time doing this instead of reading Gideon the Ninth, so maybe I'm not so smart either
… then later …
OH NO, I just realized that this stupid code is probably overweighting "stolen" votes from the beginning of the night when that threshold is lower and not counting shifts later in the night as stolen. Given how states reported absentee ballots, that could be way off
To put it more clearly, the effect of this weird floating point check in the predicates for the if clauses is to basically only count Trump errors when he is early in the night and ignore errors happening when Biden votes are counted later in the night. Sounds legit.
Also, if you do not believe me that errors happen during live elections, here's one example I dug up from the NYC mayoral election in 2013 that persisted for like 40 minutes before it was corrected [link to tweet talking about 1 precinct was showing 1000 votes for a candidate]
No conspiracy or vote-stealing. Just people keying in data and typing stuff wrong.
I think this was the last election with NYC's old mechanical voting machines so you couldn't say it was a software issue there
So, now apparently Gateway Pundit is using the data here where someone took rounded percentages against total votes to compute votes for each candidate and is alleging fraud since there are fractional votes for candidates in the data for Virginia