Thanks dude.
To be fair, it turns out I'm not 100% correct in my line of reasoning in this thread, because I misunderstood Y a bit. That makes my Desmos graph irrelevant, since it's showing something a bit different from Ayyadurai. The part about the downward slope being an artifact of his graphing is still true though: When you construct Y = whatever - X, the "whatever" part has to fight against a downward slope. That's where the perfect beautiful line comes from.
I understand where FormerGraveheart's is coming from though. He goes way too far I think, but I understand.
Over on Arfcom, most threads involving Trump before the election got brigaded by never-Trumpers and ShareBlue shills exploiting a fracture in the community over bump stocks. They know EXACTLY how to divide libertarian-leaning conservatives into useless infighting, and the mods let it happen, because Arfcom is a gun forum without an explicit pro-Trump mission statement. I see it happening so clearly, and I just want the mods to ban the usual suspects straight to Internet Hell, but they get away with it and succeed in demoralizing people into inaction...which is exactly their goal.
Arfcom has enough posters to function pretty well despite the smug shills, but they take their toll. Another less populated forum that I will not name has been totally destroyed by them. No matter how bad things get, the mods never fix the problem. You have to nip it in the bud before it takes over, and that's where FormerGraveheart is coming from.
Ever since November 3rd most of the demoralization I've seen has come from the "It's over. Give up" people instead, and I got so sick and tired of trying to motivate people that I came here specifically because of rule 1: "This is The Donald. Our community is a high-energy Trump rally. There are no exceptions."
I like that attitude, but you still have to separate the good evidence from the bad if you want to make a slam dunk argument in the courts. That creates a fracture point, because anyone doing due diligence against the grain can look suspiciously demotivating, especially if they're new. It sucks, but it is what it is.
This particular thread is REALLY complicated due to Ayyadurai's ambiguity about his axes, and most people only see one facet to each ambiguity. This makes it easy to mistrust someone who doesn't see the same thing you see. I do this too under other circumstances: Whenever anyone acts like the Democrats aren't cheating, I have to conclude, "They're either lying to themselves, or they're lying to me."
Tensions are really high for good reason. I honestly think it's not just our country at stake, but the fate of humanity. The fact that the New World Order put the Great Reset on the cover of Time Magazine last week was a huge eye opener that we are running out of time. Under the circumstances, I get why people are trigger-happy.
I'll just have to earn trust over time. It is what it is.
It's cool. There are a lot of ambiguities in how Dr. Ayyadurai is defining his axes, and it's responsible for most of the confusion in this thread. If we were all able to agree on those, I think this thread would've gone way better.
I got Y slightly wrong myself...not so wrong that the overall conclusion about the downward slope changes, but wrong enough that my argument about "running out of independent voters" on the right of the graph aren't as relevant. I wasted too much effort on that, because it was a dead end and not even important for explaining the cause of the slope, which is Y = anything - X.
I think my original assumption for Y was probably wrong, actually. The more I watch the video, the more I think that Y = (trump_votes_among_mixed_ballots / mixed_ballot_votes) - straight_r_votes / total_votes.
In that case, Y is "somewhat" more independent than I thought, because the percentage of Trump votes among mixed ballots -- the first half of Y -- is independent from X. Either way, subtracting X still creates the downward-right line.
As far as alternative graphs go, maybe the best way to get at what Ayyadurai was "trying" to do might be:
X = straight_r_votes / total_votes
Y = (trump_votes_among_mixed_ballots / mixed_ballot_votes) - (trump_votes_among_straight_r_ballots / straight_r_votes)
Here, Y would be totally independent of X, and we'd probably expect to see points clustered around a relatively horizontal line...and we can compare how that varies from county to county (keeping in mind that Wayne County is an outlier from all the known cheating, and from having very few points on the right side of X).
EDIT: I don't think we can do that with the data you have though, because it looks like the data you have doesn't account for "straight R with exceptions." The data you have seems to imply "every straight R voted for Trump," so the second part of Y might as well be " - 1.0," which is useless. That would go along with what another poster said, but it seems to contradict what Ayyadurai was getting at, and it definitely contradicts things like this:enter text
Based on the data you actually have, what you actually looked at seems like a reasonable approach. We can't really make inferences from a single county though; we'd have to see how they compare across counties, keeping in mind that Wayne County is the most wretched hive of scum and villainy. If we see that the slopes are consistent across most counties, but they're suddenly way different in another [which isn't undersampled], that would be a red flag.
I'm way too tired to do it myself though, and I should do some actual work...
I look forward to this applying to billionaires on both a personal level, and in relation to every single corporation and NGO they're connected with.
I can't wait to see the choir: enter text
The Y axis you describe is probably correct, because it matches 3/4 of the evidence I've seen. It matches what Ayyadurai says verbally, and in the slide at 19:32, and wherever the heck I saw the quote "% Trump non straight - % GOP straight ticket."
There's a slight ambiguity here, where his graph at 21:00 says "(Trump - Republican Straight Party) Vote %". I think this threw me off earlier, because it seems to imply the Trump percent there is Trump's percent of the total vote. That's what I based my Desmos graph on, but I think that was wrong. We'll move forward assuming your Y is correct.
We may be in disagreement about what the X axis is: You're using it as straight_party_r_voters / straight_party_voters, whereas I always thought it was straight_party_r_voters / total_voters. I think this is ambiguous, and the only way to resolve it for sure is to run the numbers both ways to see which matches what Ayyadurai did.
However, for the sake of this thread, we'll assume your understanding of the X axis is correct, so we can move forward under shared assumptions.
In that case, your math checks out, and we're meeting at the same point:
0.25 - 0.45 = -0.2
Proceed? Note that the important part that forces the "downward-right slope" as X increases is the part where Y = blah - X.
As far as my graph goes, I should clear something up: My 'r' was not "% of Straight Ticket Republican voters." My 'r' was "% of Trump votes, among Straight Ticket Republican voters," who are assumed to be able to override their straight ticket with a Biden vote, as described here: enter text
My 'd' was "% of Trump voters, among non-straight-ticket-R voters."
My 'x' was "% of Straight Ticket Republican voters." That's Ayyadurai's X too, although it's unclear if he means, the percent of them among all voters (as I believe), or the percent of them among all straight party voters only (as some others have argued).
By these definitions, the total Trump percentage (or ratio, as I calculated things) t = r*x + (1 - x)*d. I then calculated Y = t - x. Ayyadurai implies this is true in his Y axis label "(Trump - Republican Straight Party) Vote %", but he verbally says that it's only the percentage of Trump votes among ballots with individually filled Presidential picks, so who knows? It's a bit ambiguous, but it wouldn't really affect the downward slope either way.
EDIT: The more I look at things, the more I think my graph does NOT match what Ayyadurai did. Based on 3/4 of the available evidence, he's only including the percentage of Trump voters among individual candidate ballots, not among all ballots. I was wrong about that point. This doesn't affect the downward slope issue though.
In any case, your interpretation of Ayyadurai's graph seems to be:
X = straight_ticket_r_percent_among_straight_ticket_voters
Y = trump_percent_among_individual_voters - X
Is that correct? If so, we're still going to see the downward-right slope, because we're subtracting X as part of Y. Any definition of Y which includes negative X will show a downward slope as X increases, unless the part before the minus sign happens to increase faster than X, as X increases. Basically, the part to the left of the minus sign is always fighting a negative value which gets bigger and bigger as we move right on the graph.
For everyone's benefit, now that you've run the numbers yourself, could you weigh in on what you think the exact meanings of Dr. Shiva Ayyadurai's axes are?
I agree that we may still be able to find something in this data though. I don't have ideas at the moment for detecting algorithmic flips. Another thing we might want to look for though is, "How many people mailed in a straight-party R ballot, only for a poll worker to fill in the Biden bubble?"
To find that, I think we'd want to compare the percentage of supposed "never-Trumper Republicans" among all Republicans, across different precincts and counties. We'll probably be able to get a pretty predictable distribution as a rule, and any areas that have massive spikes not fitting the rest of the data might predict flipped mail-in votes. Comparing by county might be most fruitful, if there's enough data.
If nothing else, I think we should all be able to agree on one point: Dr. Shiva Ayyadurai could benefit from clarifying his axes and making sure to refer to them consistently throughout, to remove the ambiguity...
We're arguing over what Y means again here.
In my graph, Y is not "trump_vote_ratio_among_mixed_ballots - trump_vote_ratio_among_straight_r_ballots." In my graph, Y is "trump_vote_ratio_total - straight_r_vote_ratio_total_which_may_or_may_not_include_trump", and X = straight_r_vote_ratio_total_which_may_or_may_not_include_trump.
There are a number of ways in which we might argue that Ayyadurai's graphs are showing something different. The arguments tonight have included:
What X means: Straight R ballots as a percentage of the total vote, or as a percentage of the straight ballot vote. I argue the former.
What "straight R" means: Ayyadurai is clear that this means ballots where the straight R bubble was filled. We've had disagreements over whether voters can mark exceptions though, or whether all of these voters necessarily voted for Trump. I argue the former, which may be necessary for Ayyadurai's argument to even work anyway.
What Y means: Ayyadurai is a bit unclear about this, but one thing is clear, that most people aren't getting: Whatever the terms are that he uses to define the left side of Y before the subtraction sign, the part after the subtraction sign is the same as his X axis (whatever that axis actually is). Negative X is baked into his Y axis, so no matter what else is part of Y, it has to fight against a downward-right slope as X increases.
Yeah, that makes sense to me, at least as "one" potential source of cheating. There were clearly larger sources in places like Detroit, but...
If we were to add X back to Y to get a less confusing chart, we'd see that Trump gets the most votes in the same precincts that have the most straight party Republicans. The further left you go from there, Trump's votes gently drop pretty linearly, as the straight ticket R voters also drop linearly....until the "bend" on the far left where they suddenly drop, which happens to correspond to a [coincidentally?] straight portion in the "Y = blah - X" graph. Something is happening there in the precincts that don't like straight Republicans.
I can think of three possible "somethings" here, but none of them fully satisfy my curiosity about the shape of the bend:
a. They're cheating
b. There's a threshold at which the peer pressure in that community suddenly makes a qualitative shift toward increasingly greater D conformity
c. There's a threshold at which some kind of "flight" instinct kicks in among R-leaning independents, such that independent Trump voters suddenly drop off as you get beyond a certain threshold of anti-Republican-ness in a community
That's actually the opposite of what other people are arguing.
Dr. Ayyadurai defines the "straight Republican Party vote" by the MI ballots which actually filled out that straight R bubble (which exists on their ballots). This includes Trump by default, unless someone makes an exception for the Presidential race. (Even if they make exceptions for all the races, Ayyadurai counts it as straight Republican, because that bubble was filled. That's the whole basis for this data: MI includes that bubble.)
Another poster said you're not allowed to make exceptions, in which case the straight party vote always includes Trump. This doesn't appear correct though, based both on the data and on information derived from voter resources, like: enter text
In any case, the axes are really, really, REALLY dependent, because Ayyadurai also defines Y as including -X.
I'd agree, if: Y = trump_vote_percent_among_mixed_ballot_voters - trump_vote_percent_among_straight_republican_voters
In that case, you'd be absolutely correct, because we'd be subtracting fundamentally unrelated values, and the value of Y would be independent of the value of X.
However, that's not what Y is. Dr. Ayyadurai defines the Y axis in writing in 2-3 separate places, in ways that might appear inconsistent with each other depending on how you read them, but none of them are what you're assuming above. All of the ways Ayyadurai defines Y shows that he's actually subtracting the X axis from it, which means that Y naturally decreases as X increases.
At 19:33 he writes, -- X-axis: Republican Straight Party (RSP) Vote Percentages (%)
-- Y-axis: Difference of %Trump Individual Candidate Votes minus %RSP votes
At 21:17, he writes Y as: "(Trump - Republican Party) Vote %" That's a little ambiguous, because it makes it look like he means "total Trump vote percentage, minus the Republican Straight Party vote percentage," but he says elsewhere that the first part refers to the Trump vote percentage among mixed ballots only, where voters picked Trump directly. The difference is confusing, but it doesn't matter for our argument.
Somewhere else (either in an image I saw, or I'm repeating telephone game stuff, because I can't find it in the video), he supposedly also used "% Trump non straight - % GOP straight ticket" for the Y axis. I repeated this elsewhere, but I can't find it anymore...argh.
In all three cases, it's clear that the subtractive part of Y is not Trump's percentage among any group, but the X axis. Because Y includes negative X, Y decreases with slope = -1 as X increases. The final slope isn't quite as negative as -1 though, likely because the percentage of independents who vote Trump is likely higher in precincts with more straight party R voters.
There's another debate about what the X axis means, but it's a bit irrelevant to our debate:
a. straight_r_voter_percentage_of_total_voters
b. straight_r_voter_percentage_of_straight_party_voters
I think it means 'a', but two other posters think it mean 'b', and I haven't found anything yet that tells them apart for certain.
I just rewatched a few minutes, and I'm more certain that X = straight_r_voters / total_voters based on another Ayyadurai's quote here: enter text
EDIT: Uhhh...actually no, crap. Ayyadurai's quote doesn't actually disambiguate there...both interpretations still fit what he says, in different ways. Argh. If you can find a good way to distinguish, let me know.
So there his slide says: -- X-axis: Republican Straight Party (RSP) Vote Percentages (%) and he says out loud, "So plotting on the graph, again to be clear, on the X axis we're going to have Republican straight party vote percentages, and on the Y axis..."
So I'm interpreting this as "Republican straight party vote percentages out of all votes," and you're interpreting it as "Republican straight party vote percentages out of straight party votes."
To disambiguate, look at what he says around 21:17, which only makes sense in the case of what I'm saying: "Because this means Trump actually got 65%, minus the 60" when he refers to a 5% differential, and the difference is the Republican straight party vote % (== the X axis, == the amount subtracted from the Y axis).
EDIT: Actually no, I may have misunderstood that. 21:17 doesn't actually disambiguate, because both interpretations still fit what he said. Crap! Well, hopefully we can find a good way to tell them apart.
In any case, have a good night!
dixond, according to Dr. Ayyadurai's video at 19:14-19:33: "On the X axis, we're going to look at the Republican straight party votes. Okay? Per precinct, how many Republican straight party votes were there? And on the Y axis, we're gonna look at the direct Trump individual candidate votes, which means how many people voted directly for Trump."
That is, X = straight_r_voters / total_voters. You seem to be saying X = straight_r_voters / straight_party_voters.
Are you sure? Can you cite where this is made clear? Liver_Kick is saying the same thing, but I don't know where he's getting that. If that's actually the case, it changes some things.
EDIT: I looked a little deeper here in my conversation with Liver_Kick: enter text
Agreed. Even though the negative slope is a natural consequence of Ayyadurai subtracting the X axis, I think his final conclusion about a vote flipping algorithm might still be legit, details aside. It would just change the slope.
I think someone should be doing a statistical analysis of the various counties' slopes throughout all of Michigan. I'm not a statistician though, so it shouldn't be me.
My first guess was cheating: The extreme blue precincts, where straight party R votes don't happen, would be more likely to start flipping Trump votes to Biden.
RStroud has a more benign explanation: "The change in slope could also be a confounding variable. I would suspect demographics differences in heavily democratic precincts may make them more sticky, and less likely to swing republican."
What are your thoughts?
I went back, and I couldn't find what you're referring to. Could you go through the video and get me the quote that you think overrides the one I posted from 19:xx?
Agreed, he theoretically can, although the margin by which he can do so decreases as the chart moves right. He's still fighting the negative X built into the Y axis.
I think people may be able to see the data more clearly if they add X back to Y.
Ironic. You should see all my posts on Arfcom where I'm arguing that people need to stand up for Trump and stop being weak demoralizing quitters.
I'm not your enemy. We're in this together, and we both agree the Democrats are massively cheating. We're just looking at Ayyadurai's data differently.
I'm not a doomer.
Democrats blatantly cheated and tried to steal the election from Trump, and Trump will win. The evidence is overwhelming. Also, Trump MUST win, because we must stop the Great Reset if we don't want a thousand years of darkness. I'm more invested in Trump winning this than you can possibly imagine. My work performance is suffering from being so obsessed with this, because the very fate of humanity depends on it.
I'm just saying that people are misinterpreting these particular charts.
Yeah, I'd like to see a larger sampling of counties.
There are two problems with Dr. Shiva using Wayne County as his "control" setup:
- Wayne County includes Detroit, where we already know for a fact the Democrats were massively, massively cheating from all the other evidence...statistically impossible vote counts by Benford's Law, statistically impossible turnout, statistically impossible Biden mail-in advantage compared to everywhere else, back-dated USPS ballots, eye-witness testimony (sworn affidavits?) of trays of Biden-only votes, sworn affidavits of poll watchers being disallowed...yeah, Wayne County is not a good control group for what things should look like without cheating.
- Sample size, as you said. There were too few precincts with very many straight Republican voters to make much sense of the data there. He drew a line through it, but it didn't actually fit the data points besides being "kind of halfway between the closest data points that actually did exist," which were ultra-sparse.
If that were right though, the data would have to show it: As you go to the right, the Y axis would have to converge toward 0, because no deviation would be allowed if 100% straight R == 100% Trump.
The charts only go up to ~80%, and the dots only go up to ~75%, so maybe you could say they "would" be curving back toward zero, but...
According to this article, the machine should have accepted the ballot for the guy in front of you: enter text
Yeah, if you could run the second visualization (yours) on many counties, do linear regressions, and plot out the statistical distribution of all the slopes and offsets, you might find something interesting.