the Air Vent

Because the world needs another opinion

More ‘no-evidence’ to Ponder

Posted by Jeff Id on November 27, 2020

I’ve been continuing to work with this New York Times voting data and have found a number of bits of data which are a bit interesting.

From my previous post, this data is presented as live information to the news outlets. It contains rounding errors which are a bit frustrating but we can still use this data.

In order to avoid critique of this rounding issue, below is Wisconsin data using only data greater than 4x the resolution limit of the rounded data. Most states counted the majority of mail in ballots after the primary count was completed. I used an arbitrary time of 3am local time on the 4th as a divider for the votes.

Green bars below are the number of vote reporting events before 3am on the 4th of larger than 4x the rounding limit in size that fell into the calculated ratio. For example there were 25 entries of votes that landed between .5 and .55 ratio of the Trump/Total Votes. So X axis numbers less than 0.5 are movements toward Biden.

The red bars are the number of votes having larger than 4x the rounding limit after 3am. All three were heavily weighted toward Biden but the one on the farthest left edge was a massive single reporting event which was of a magnitude of 118,000 votes which is 6 times more than enough to flip the vote in Biden’s favor. It is also of a size that is 53 times the rounding error in the Wisconsin vote so you can trust this entry for its rounded accuracy.

Focusing on this single extreme event, the odds that this event would occur at any size level of entry greater than 4x the rounding error were 0.41% based on a normal curve probability distribution.

The story gets even more interesting though. Below is the a histogram of the count size of significant vote reporting events (some of these are small because in early voting there were few votes yet tallied).

Can you guess which event we are talking about?

I didn’t think you would have much trouble. The single largest number of votes reported in Wisconsin is also by percentage, the single most extreme pro-Biden voting event reported in Wisconsin. It is also one of the last events, posted well after the magnitude of Biden’s vote deficit was well known.

Odd’s anyone?

Wisconsin doesn’t allow counting to stop once it is started by state law. This is interesting because many of the mail in ballots would be counted earlier in the day for smaller areas.

Anyway, the magnitude of this event is so large that there are only two possible counties it could have come from in Wisconsin, Milwaukee has 451,000 votes and Dane had about 338,000

Despite Milwaukee having 1.33 times more votes, Dane gained almost 50% more Biden votes than Milwaukee so I am suspicious that this entry came from Dane county. From the linked article:

In Dane and Milwaukee counties, more than 68,000 voters said they were indefinitely confined in 2020, compared to about 17,000 in 2016.

Just to re-explain the details of the mail in vote concerns of normal thinking folks. In Wisconsin, ballots can be submitted by folks to polling places across the entire state. It is up to election workers to validated addresses and signatures. If they don’t match, they are supposed to be rejected. Ballot rejections are a critical part of the voting process and if the rejections don’t happen at all, the vote count can be cheated. If the rejections happen differently from county to county, this can also skew the record. Critically, once the ballots are separated from the envelopes, the forensic history is lost because the names and addresses on the envelopes are no longer associated with the unidentifiable ballots. If fraud were occurring, you could surreptitiously separate the ballots from the envelopes, or perhaps not even use envelopes and create completely untraceable votes. Recounts are useless.

What we have in Wisconsin, is a single, highly skewed, reporting event which was large enough to flip the election. It was the most democrat vote and simultaneously the largest vote batch reported and it was done when the magnitude of the votes needed to flip the election were a known quantity. Also, this report doesn’t meet reasonable criteria at 0.41% probability, to be anywhere near the rest of the votes.

Not quite done yet. In addition, I looked at the direction of votes based on time. We do know that mail-in votes likely came in later but what does the vote magnitude look like over time.

The next graph is total vote ratio Biden/Trump. Above the blue line is Biden winning the total vote count. You can see that in our huge single vote reporting event flipped the election.

After most of the ballots were counted, 5 separate pro-Biden vote entries of significant magnitude were made. This is visible in the graph below:

From about 6am GMT onward, the graph above shows 5 large size voting events for Biden and none strongly for Trump. The last 3 large events were all very strongly for Biden. The dot’s close to the 0 line are large batches of likely absentee votes which went about 50/50 to the candidates.

With my days of time spent in this dataset, I have found a large number of crazy anomalies. Combined with the reports from witnesses and the endless video’s of fraud happening, I’m completely certain that massive election fraud not only existed but did flip this election to Biden. Feel free to draw your own conclusions though.

There is so much more you can find in this single state’s data. If we could get access to the time based reporting of each county with full resolution, a lot of the shenanigans could be quickly and more importantly, accurately identified.

7 Responses to “More ‘no-evidence’ to Ponder”

  1. amac78 said

    Jeff, I think you’ll find this pseudonymous author is thinking along the same lines that you are. vote_pattern_analysis posted Anomalies in Vote Counts and Their Effects on Election 2020 on 11/24/20 on their new Substack site.

    Executive Summary: In the early hours of November 4th, 2020, Democratic candidate Joe Biden received several major “vote spikes” that substantially — and decisively — improved his electoral position in Michigan, Wisconsin, and Georgia. Much skepticism and uncertainty surrounds these “vote spikes.” … While data analysis cannot on its own demonstrate fraud or systemic issues, it can point us to statistically anomalous cases that invite further scrutiny.
    This is one such case… we rely only on publicly available data from the New York Times to identify and analyze statistical anomalies in key states. Looking at 8,954 individual vote updates…, we discover a remarkably consistent mathematical property: there is a clear inverse relationship between difference in candidates’ vote counts and and the ratio of the vote counts… Nearly every vote update, across states of all sizes and political leanings follow this statistical pattern. A very small number, however, are especially aberrant. Of the seven vote updates which follow the pattern the least, four individual vote updates — two in Michigan, one in Wisconsin, and one in Georgia — were particularly anomalous and influential with respect to this property, and all occurred within the same five hour window.

    The four vote updates in question are:
    1. An update in Michigan…
    2. An update in Wisconsin…
    3. A vote update in Georgia…
    4. An update in Michigan listed as of 3:50AM Eastern Time on November 4th, 2020, which shows 54,497 votes for Joe Biden and 4,718 votes for Donald Trump.

    – – – End Executive Summary – – –

    3:50 AM ET would have been 08:50 GMT, I think — or was MI still on Daylight time?

    vote_pattern_analysis transform the voting data in ways that are non-intuitive to me (i.e. I had to read his explanation carefully to understand them). But they make a lot of sense, and allow him or her to draw stronger conclusions.

    This approach reminded me of some of the work that you and SMcI did on temperature time series, back in the day.

    • Jeff Id said

      Amac78, thanks for this. I do have plots and data for all states and i agree with this other author. There is enough difference in the battleground state votes that I was able to list them using stats only in the next post after this one.

      I believe that pennsylvania should be on his list too. I will check out the link tomorrow.

  2. Pouncer said

    Do I understand correctly that the NY Times data set records reports about counts rather than counts themselves? Not sure how to express this, but I consider the counts to be internal to the jurisdiction, while a report is some stringer or wanna-be journalist watching numbers post on the local system, them copying them into a national standardized system. Transcribed numbers, hours into the official process.

    If — If and Only If — the concept of transcription and copying is correct, then copy errors will creep in. The county counters post 510 and the stringer types in, and reports to NY, “5010”. The count is 15 and the stringer reports 51. That sort of thing. I’d also expect some sort of corrective reports — ideally they’d back out an error entirely then post the correct numbers. -5010, +510. But some guys take it upon themselves to grab a calculator and post a difference. -4500 . Some will then bury the correctional difference in with the next, positive, record:

    Actual Reported Needed correction
    510 5010 -4500
    100 -4400 stringer’s confusion is complete

    In a worse case the needed corrections are not SO large that they show up in the report record as negative counts. Suppose the morning count is 150 and the report shows 510 needing correction of -360. If the afternoon update count is 400 the stringer might net the correction to the count and report 40. Now there would two offsetting, nearly invisible, positive value, discrepancies between the count and the report. That evening the (or another) stringer sees that the afternoon count was 40 and the report was 400, so calculates what he believes to be a correction of +360, which he then adds and invisibly embeds to the evening count of maybe 500 leading to a report of 860.

    I can’t tell you how often these and similar kinds of transcription errors afflict warehouse inventory counts when the COUNT is done at storage location, on paper, but the ACCOUNTING is copied from multiple pages of paper via keyboard into the computer database.

    • amac78 said

      Pouncer, this data set was generated by Edison Research and shared with media subscribers. The NYT is the one who posted the time series.

      Edison says: “In an effort to improve quality, streamline data collection, and expand election coverage in 2018, ABC News, CBS News, CNN and NBC News ended their arrangement with the Associated Press for vote tabulation and now partner with Edison Research for these data.

      For the 2020 General Election, Edison Research provided the NEP [National Election Pool] with a fast and accurate vote count throughout the nation, providing data for all statewide races and all House races.”

      My understanding (which could be off) is that Edison didn’t report raw votes as they came in, but first converted them to a ratio with three decimal places. So for a batch of 621 votes for Biden, 328 for Trump, and 51 for Jorgenson/other, Edison would have reported something like (1000, 0.621, 0.328, 0.051).

      Obviously, for accuracy and for forensic purposes, this gets worse as batch size increases. (On the other hand, I suspect that the NYT regrets opening access to these records in any form, given how it has allowed people to undercut the media’s No Misconduct Anywhere narrative.)

      The above description may be off. While looking for a link to transcription-type errors, I came on this un-referenced assertion: “… [Edison didn’t supply] the exact votes per candidate per batch, rather we have the candidate % of total votes being updated for every batch. When shifts are small, we have no way of determining the actual change as the % given only goes to one decimal place (e.g. 52.2 or .522).” See also here.

      One claim of a large error in the Virginia NYT (= Edison) count is here.

      And, back to the comments in the Substack article I linked above, here is “Sam” recounting how a major error in transcription was handled. (Sam’s source seems to be

      “‘There was a typo in the initial report that went out to the Bureau of Elections. It was caught in a very short amount of time and quickly corrected,’ [said Shiawassee County Elections Clerk Abby Bowen]. ‘There was an extra 0 added on the end when Biden’s votes were keyed in.’

      The county has reported Biden received 15,371 votes (compared to Trump’s 21,354). When the typo was made, it displayed as 153,710 — which would have far exceeded the county’s 55,612 registered voters.

      On Twitter, live updates from Decision Desk HQ showed that between 5:04 a.m. and 5:43 a.m. — as additional votes in Michigan came in for Trump — Biden’s numbers were updated and actually dropped by 110,796.

      Shiawassee County Clerk Caroline Wilson told us she was responsible for the error. ‘Nothing was jeopardized in any fashion,’ she said. ‘It was clearly a human error.’

      Bowen emphasized that ‘everything that gets reported on election night is unofficial’ and that a routine county canvass will be done to further verify vote tallies before they become official.

      ‘This is why we have the checks and balances,” she said. “This is no kind of voter fraud. It was literally just a typo.'”

  3. amac78 said

    Hell, I’ll make one more comment before getting on with real life, riffing from this one.

    Not so long ago, the Democrats and their media arm were convinced of the reality of interference in US elections. That only changed a few months ago. Now the shoe is on the other foot. It ought to be obvious to everybody (but isn’t) that of course US elections are vulnerable to fraud. An honest system has to (1) be transparent, (2) have clear procedures (including procedures for investigating and publicizing deviations), and (3) be auditable, and regularly audited. Whether or not any of the 50 states qualify on all three points, it doesn’t seem that the states in the news meet any of them.

    Now many on the right are identifying anomalous events in the battleground states, which is a good start. And ascribing them all to fraud, which doesn’t necessarily follow. States aren’t uniform in geography or demography. For most, the landscape is the opposite of homogeneous (e.g. those who vote in-person vs. by mail, conservative small towns vs. woke cities).

    It would add a lot of perspective to run many controls for the analyses that highlight anomalies (possible fraud) in battleground states. In particular, non-contested states for 2020, and other elections (though I gather that the Edison/NYT time-series data was unavailable prior to this year).

    At what frequency do aberrant events and results show up in contests where the motive for fraud is low?

    For instance, most red states have blue cities. Some urban precincts trend 90%+ D. For a uniform 40% D voter pool, a large batch of 95% D ballots would be very improbable. But if they were collected from those 90%+ D neighborhoods, much less so.

    OH, IN, IA, MO, KS, MA, TX, UT, CO, and CA are examples of states where the electoral votes weren’t in doubt. Is the prediction that they should lack the fraud telltales that have been seen in MI, WI, GA, and PA? Do they?

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: