Coffee in Hell

They sat down around the table. Objectively speaking, Gerald was not as young as he had been, Miranda would be classified as mature, but not in a derogatory way. She wore her physicality in a pretty and attractive way. Christine was almost a juvenile in this company.

Notwithstanding objective realities, when they seated themselves, the situation smelled of eldercare. Three crumbling seniors quavers to the table and fumbles with cup and coffee. They all sigh in sequence following a clockwise motion around the table.

M: “This is hell. We’re in hell 🔥🔥🔥”

G: “I want to immediately spell out my frustration with the fact that most Republicans think that the election in 2024 went fine unlike in 2020 were there was widespread fraud. I wonder how many forensic copies of EMS servers were made this time? How many logs were examined?”

M: “And I want to immediately point out how the supposedly logical conclusions which Jeffrey O’Donnell arrives at in his video series Fingerprints of Fraud are almost cartoonish. Not the first episodes, but when he ends up quoting the Sermon on The Mount and building an organisation chart based on Godfather.”

C: “Both of you, Gerald: Then you should also state that the same Pew report observes how a small fraction of Democrats expressed the same concern as the Republicans last time. Not at all the same amounts, I will grant you that, but the change is visible. And Miranda, be careful if you try to hint that Jeffrey’s follies towards the last segments invalidates his first arguments.”

M: “I’m not, and I agree. He presents at least a valid argument in the beginning of the movie. I believe this is what Tina Peters could be referring to when she impressed upon the judge that if he had seen the documents, he knew that she was right.”

G: “The remark that made judge Barrett scold her for trying to put his compliance on record.”

M: “He got mad, yes. I believe him when he says he has read most of the material handed to him. But I don’t believe him when it comes to Jeffrey’s statistical suspicions.”

The score so far

G: “Alright folks. Can we make a status report? How big a mess is it?”

C: “Mesa County: Dominion’s machines seem to be easily hackable. Easy to add WiFi connection to (judging by the purchase specification given in their manual - Dell machines). Doug Gold tried connecting to the central SQL server from an iPhone. Weak passwords, open ports, firewall not configured properly.”

G: “Several points: 1) We don’t know what Dominion would respond to that. 2) It’s interesting that Dominion and other vendors few years earlier testified in front of the House Administration Committee on 2020 Election Security that they did not have wifi or remote access enabled. Didn’t Doug Gold establish the presence of a Dell remote access system, IDRAC?”

C: “Unknown. Some install logs hinted at the possibility, but only close examination of the real hardware can answer that. But the wifi testimony sounds smelly to me, compared with Doug’s findings. Again, he concludes from a purchase specification in the manual. The actual hardware can be tailor made. Finding an official answer from Dominion is not easy.”

G: “Then there’s Jeffrey O’Donnell’s videos where he presents a short and catchy version of the full report. I’ll get back to that later. His absolute centrepiece of an argument is that he claims to have gotten hold on full datasets with Cast Vote Records, i.e. anonymised per-vote data for many counties. Apparently the information is obtainable. Using those, he can see how the election progressed minute for minute in those counties. He points out that given how the data should represent a random sample of the county, and therefore the count should converge around a constant percentage of Trump vs Biden voters. But they clearly do not.”

M: “Did you look into the unnatural statistics?”

Statistics

G: “Let’s talk about those for a second. There’s a huge report he compiled that contains many many counties. All he and his buddies could get a hold on. Jeffrey’s full report, 281 pages. Next time we meet, I may have been able to digest it.

However, I want to talk about his arguments for a second. The argument clearly revolves around the assumption that we can consider the incoming votes completely randomised. He goes through great pains to make sure they are.

In short: If he is correct and the votes are truly arriving in a random fashion, then I agree, something looks strange.

If not randomised, then it’s a completely different matter.”

C: “I need pictures.”

G: “Don’t crucify me if I make a mistake here. I am still reading the report.

First, here is how he statistically expects a vote count to look like if all is fair:”

Election percentage if votes are truly randomly distributed:

G: “We see that the balance between Trump and Biden very quickly establishes themselves on a fixed value. As more and more votes are counted, you get a stable curve that expresses the likelihood of the highlighted candidate in that county.

And here is the famous Mesa pattern which he first discovered when initially studying the hard disk images Christine had been reading about:”

The Mesa Pattern:

G: “Here’s my own simulation of a completely random pick amongst hypothetical votes in an imaginary district with a 67% chance of people voting Trump:”

Simulated in Python:

G: “The hazy outline represents the outcome of ten trials. Ten other trials would have produced different squiggles. You get the point.

I then decided to examine what things would look like if the votes were not randomised. For instance, this is a political arial photo of Mesa county for the 2020 election, give and take a little as Mesa County is not rectangular. As you can see, mostly Trump precincts, but not all.”

Mesa County 2020:

G: “So then I thought: What if his assumption does not hold up? What could happen? Look here:”

Simulation: Changing ground truth probabilities over time/counting:

G: “The blue line is the average probability for a Trump vote for a small segment of ballots. As you can see, some piles of votes have around 45% probability for Trump, others have as high as 90%. The resulting curve does interesting things.

Next I wanted to see what the curve could look like if they accidentally has started out with a pile of 50/50 % votes, just for the first say 50 out of 1000 votes, and after that, things randomised naturally around the expected 33/67 % split:”

Simulation: With initial 50/50 split:

C: “Pretty much the Mesa pattern.”

The group looked at his screen. Christine leaned back and looked thoughtful.

C: “But Jeffrey is convinced the lots are properly random? How?”

G: “I need to read his report thoroughly. The records contain precinct information, and they are absolutely not processed in order! What you see here - I assume! - is the precincts from which a specific ballot originates over time. Jeff is clear:”

Counties that do not show randomness are not further tested for the unnatural patterns seen elsewhere

G: “For completion, here is his thinking:”

To fairly judge a county’s Cast Vote Record mail-in results, I first test the results to see if they meet the randomness assumption. The assumption is that mail-in ballot results contained in the Cast Vote Records are randomized by the processing of the ballots themselves. As mail-in ballots are randomly requested, randomly sent out, randomly filled out, randomly returned or delivered by the voter, and not presorted by the county upon receipt, they become naturally shuffled and mixed.

G: “So he only examines rows in the CVR file marked ‘MAIL’, and he assumes they were shuffled into a randomised pulp in the postal system. He confirms that using the precinct column of the CVR file.”

Which precinct (Y) does which ballot (X) come from:

G: “Imagine Santa Claus picking up ballots all over Mesa County in a completely random fashion. That would certainly mix the red/blue colours from the map above into a purple-red pulp.”

M: “Hmm. Interesting, to say the least. Any comments from clerks or election officials?”

G: “I’m still looking.”

C: “I guess that X-axis in all those plots indicates the row in the table of CVRs, correct? In other words, not time specifically, but what we assume is a time-like property, such as if the ballots were recorded immediately in the database when it was received.”

G: “I know where you are going…”

C: “Exactly: All precincts could in theory undergo a non-random time sensitive evolution as well: First a lot of Biden votes send in their mail. Then comes the Trump voters.”

G: “Very true. I have looked for verification or rejection of that hypothesis, but so far nothing. "

M: “Keep us informed if you can gain any further insight.”

Rubenstein again

M: “Christine, does Rubenstein’s report mention any of this?”

C: “No! Doesn’t that fall under the purview of a county clerk or a statistician? Anyway, the Rubenstein report is useless. All the evidence he puts forth is a bunch of interviews and watching the camera feeds. Okay if he just wanted to establish if any of the workers had participated in criminal conduct.

His conclusions is a paragon of overreach:”

We have found extensive evidence that the conclusions in Report 3 are false. Finally, and most significantly, this investigation uncovered no evidence that would indicated outside interference with the election.

C: “I am at a loss as to where in the report he dug that conclusion out. As mentioned, he didn’t even bother to respond to reports 1 and 2 which examines wifi-capabilities.”

It should be noted that Report 3 indicates that elections staff has a strong recollection of the above events, and specify that DVS support was contacted at approximately 4:00pm on October 21, 2020. In fact DVS support was not contacted at all on October 21, 2020 or any day thereafter during this election. It is unknown what specifically resolved the apparent error Ms. Brown encountered, but she took one basic measure and two extreme measures on October 21, 2020. The remainder of the election has been unchallenged by Report 3, and no other anomalies appear to have occurred. A random audit sampling of ballot images that went through both adjudication sessions disclose no reason to question the cast vote record.

C: “I assume that they did an audit on a small sample is prescribed by the EAC. Incidentally, when he says ballot images, he suggests a paperless audit. This is exactly what the election scholars warn about. Andrew Appel describes the various methods and their susceptibility to hacker intrusion. Rubenstein’s line of reasoning is embarrassing: He rules out hacker intrusion by conducting an audit vulnerable to hacker intrusion!”

M: “Where are all the real experts when you need them!”

They all sighed.

M: “Well, at least we are doing a better job than most journalist out there will ever do.”

G: “That is no comfort. No comfort.”

/ПРИЗРАК