Wed May 12 08:12:30 +0000 2021
[tweet] [link]
Alright, it's late, but screw it - he's my thread on some of the latest filings in Antrim. Specifically, the Lenberg reports:
https://www.depernolaw.com/uploads/2/7/0/2/27029178/[3]_ex_2_lenberg_2.pdf https://www.depernolaw.com/uploads/2/7/0/2/27029178/ex_1_-_lenberg_report.pdf
As usual, the Halderman report is helpful to reference:
[tweet] [link]
To understand the allegations in the Lenberg report, let's first go over what happened in Antrim according to J. Alex Halderman. We'll use Helena Township as our example (same township used by Lenberg)
[tweet] [link]
Antrim required a ballot correction in October. Ballot updates require both scanners and the Election Management System to be updated.
Per the excerpt, the new ballot definition was loaded into the EMS, but only 4 of the 18 scanners had the correct election definitions loaded
[tweet] [link]
These excerpts go into detail about the structure of the election definition files, but the important thing to remember is that individual ballot choices (ie. Trump, Biden) are assigned a unique, sequential number (internalMachineID).
The ballot choices are also tied to contests
[tweet] [link]
So you end up with a ballot definition that maps the IDs of ballot choices to specific contests, and that maps the position of markings on the paper to specific ballot choices.
So when Antrim updated their EMS, but only 4 of 18 scanners, things got wonky.
[tweet] [link]
The scanners recorded results using the old definition IDs (in blue). The EMS read those same results using the new definition IDs (yellow).
You'll see that after the school board contest was modified, all down ballot races incremented by 1 (blue -> yellow)
[tweet] [link]
This means a scanned vote for Trump is logged with ID 970...but the EMS thinks Biden is assigned 970. So the vote for Trump is logged for Biden.
[tweet] [link]
Votes for Biden are logged as ID 969 by the scanner...but the discarded by the EMS. Why?
The ballot choice is associated with a specific contest. A ballot choice with ID 969 for the presidential election doesn't exist in the EMS election definitions.
[tweet] [link]
Remember that ballot choices are mapped to specific contests.
If your contest has ballot choice IDs 960-965, a ballot choice ID 959 or 966 would be discarded.
[tweet] [link]
Here are the results for Helena Township before, and after, the issue was corrected. Both were taken from the results published on Antrim's website.
You can see that in all cases, the original counts were cleanly shifted up by one choice.
[tweet] [link]
The conservative narrative around what happened in Antrim tends to talk about a "vote switch"...which, yes, votes were switched, but it wasn't a clean switch just between Trump and Biden...all races down ballot of the updated contest were affected.
[tweet] [link]
So what's Lenberg's report say?
- SQL Studio was installed on the EMS
- Lenberg was able to use SQL Studio to modify the election configuration
- ElectionSource (county election vendor) have access to modify election files and configuration (aka, their job)
[tweet] [link]
RE: SQL Studio, it should not have been installed. Halderman pointed that out. Doug Logan pointed that out.
There's no evidence that was used to tamper or modify with the election configuration or results, but they're alleging its very existence was proof enough of fraud.
[tweet] [link]
They also demonstrate that by directly modifying the ballot definition files used by the scanner (and not updating the EMS), they can cause ballot scans to intentionally swap votes.
Which, yes. You could do the same by editing the election definition directly in the EMS.
[tweet] [link]
So, they're not presenting evidence that the definitions were modified with the intent of swapping votes.
Instead, they're basically saying that their demonstration is proof enough. I'm not a lawyer, so I don't know if that holds up.
[tweet] [link]
They layout their demo:
- First, they'll directly edit the election configuration
- Then, they'll burn the new configuration to the CF card
- They'll then load the CF card into a scanner.
[tweet] [link]
Then they'll try two test case:
- Swap Trump and Jorgenson
- Swap Biden and Trump, and Ferguson & Bergman (congressional), while leaving other races untouched.
[tweet] [link]
The end result should pretty much be this.
Notice how the arrows cross? Antrim was a complete shift of results because someone forgot to update all of the scanners.
Here, they're intentionally swapping votes and claiming it's exactly what happened in Antrim (spoiler: it's not)
[tweet] [link]
They make all sorts of claims about how this proves the vote shift was definitely not human error...
but...that would mean someone intentionally shifted a ballot choice up by one for every race that just so happened to be down ballot from the race that needed a correction..
[tweet] [link]
But why would someone intentionally shift all races, if they can selectively fix one?
Why shift Jorgensen's votes to Trump, when you know that'll result in unusually low votes for a very red county? Won't that stick out like a sore thumb? (in fact, it did)
[tweet] [link]
BTW, I want to point out that in both of their test cases, they swapped choices within the same contest. Keep that in mind while we look at the claims in Lenberg's 2nd report, released on 5/10 (ex_1_-_lenberg_report.pdf)
[tweet] [link]
Oh, will also be looking at the accompanying brief: https://www.depernolaw.com/uploads/2/7/0/2/27029178/supplemental_brief_051021_1_.pdf
First thing that stood out to me when I reviewed the new docs: DePerno's just learned about the word "subversion" and he's very excited about it.
[tweet] [link]
He also breathlessly claims that Halderman's report is DEBUNKED because - per Lenberg - Biden's votes should have gone to the Natural Law Party in the straight ticket/partisan contest, which would have registered as a vote for their presidential candidate.
[tweet] [link]
We'll pause there and look at the Lenberg report. What I like about the Lenberg, is that between his two reports, you get to watch him try and figure out how the system works for the real time. Who needs actual experience when you have google! (and the Halderman report)
[tweet] [link]
tl;dr: They got the scanner to throw errors when tallying the results, but the EMS does not throw those errors, so SUBVERSION!
Also, Biden's votes should have got to the NLP candidate in the straight ticket contest. Instead, the Biden votes were tossed out entirely. SUBVERSION!
[tweet] [link]
They claim "Halderman indicated that the shift of votes that occurred would have changed the index of the candidate selection to cross the boundary from the Presidential contest to the Straight Party Ticket
contest"
I can't find that claim in the Halderman report.
[tweet] [link]
In fact, Halderman points out how contests affect ballot choices several times, and not a single instance describes a ballot choice working for contest to which it doesn't belong.
[tweet] [link]
While discussing why certain townships had issues loading memory cards on election night, Halderman once again mentions that unless ChoiceManifestations are associated with a ContestManifestation, the affected selections are ignored (and that no error is displayed)
[tweet] [link]
And again, Halderman confirms that selections for Biden were ignored - and not that they should have been shifted to the NLP choice for the straight ticket contest.
[tweet] [link]
I want to clarify that the straight party contest was also affected - but that votes did not shift across contests as Lenberg is claiming.
[tweet] [link]
There was also a weird edge case: Republican straight-party voters that voted for Biden had their vote interpreted correctly...because the Biden vote was discard, and the Republican straight party selection was shifted to the Democrat straight party.
[tweet] [link]
Similarly - Libertarians straight-party voters that split and selected Biden had their votes go to Trump: The Biden choice was ignored, and the Libertarian straight party selection was registered as a Republican straight party vote
[tweet] [link]
So back to Lenberg: I have no idea where Halderman supposedly mentions votes can be shifted across contests.
Still, Lenberg tried a little experiment
[tweet] [link]
For this test, he edited the file that contains the ballot choice definitions, and swapped the Biden and NLP straight ticket internalMachineIDs.
The expected result was that a Biden ballot selection would register as an NLP straight ticket selection, and vice versa.
[tweet] [link]
Instead, he received an error on the scanner when he tried to close the election and tabulate the results. He correctly points out that ballot choice internalMachineIDs have to belong to their associated contest
[tweet] [link]
He seems surprised by the error, but I think I have an idea of what may have happened.
[tweet] [link]
We can see an error in the scanner logs that says "cannot find choice instance 3016".
Lenberg had set Biden's choice instance to ID 3016, and the NLP straight party vote to 3015.
[tweet] [link]
The blue column shows the ballot definitions that were on the scanners on 11/3. The orange is the definition that was on the EMS. Notice that the shift is one number up, from scanner -> EMS.
Also notice that all numbers in both columns are within the ranges for each contest.
[tweet] [link]
Now let's look at what Lenberg did. He didn't simply shift everything up - he intentionally swapped two ballot IDs in different contests.
The straight party contest has a range of 3009:3015. But now there's a ballot choice with an ID of 3016 - outside of the contest range.
[tweet] [link]
So I believe the scanner threw an error because the straight party contest now includes a ballot choice ID that is not actually part of the straight party contest definition.
[tweet] [link]
Lenberg then loads the card into the EMS, which doesn't throw an error.
Instead, it loads the results, treating votes for Biden as undervotes (so basically not counting them).
And that's basically it
[tweet] [link]
So they're mad because they contend Biden's votes should have gone to the NLP and/or that the EMS should have thrown up errors.
Which, I agree with. Halderman also mentioned that Dominion should display additional errors/warnings to operators.
[tweet] [link]
But I'm not sure how you get from "it didn't break how I wanted it to when I intentionally tampered with it" or "it didn't throw an error" to "IT'S SUBVERSION AND PROOF OF FRAUD!"
[tweet] [link]
I'm really getting the sense that DePerno & co are grasping at as many straws as possible. Like I said, you can tell by the two different Lenberg reports that they're just screwing with the database files directly to see what breaks.
[tweet] [link]
The case seems to be hinging on SQL Studio being installed - it shouldn't have been.
But to think someone tampered with the config so votes were shifted up for every choice - which meant that Trump would very noticeably get the handful of Libertarian votes - is hilarious.
[tweet] [link]
If China were trying to commit election fraud in Antrim, I'm pretty sure they would have found a more subtle way to do it than to shift the Libertarian votes to Trump.
Maybe by using one of those "algorithms" Lindel's always talking about, so it wouldn't be so fucking obvious.