Election Day

• Election Night Reporting: 528,777
• Cast Vote Records: 528,776
• Ballot Images: 152,209

Recount

• Election Night Reporting: 527,925
• Cast Vote Records: 527,926
• Ballot Images: 510,202

Aberrations

• Double counted ballots in original count: 550
• Double counted ballots in recount: 3,930
• Ballots not appearing in recount: 5,812
• Ballots added to the recount: 1,581
• Test Ballots added to the recount: 250+

General Observations

Fulton County's performance in the ballot counting process was incredibly problematic, presenting the most challenging case among all the counties in Georgia.

In most counties, our analysis typically takes a few days, and even in more complex cases like Gwinnett and DeKalb, it might stretch to a few weeks. However, Fulton County was an entirely different story. The data was so disorganized and inconsistent that it took us over eight months to conduct a thorough analysis. We repeatedly started and restarted our review, struggling to make sense of the chaotic information. Eventually, to ensure the reliability of our findings, I had my analysis partner, Joseph Marolda, independently audit my analysis. After almost a year of analysis, we combined our results to present this report.

Original Count - Double Counted Ballots

In the original count, we identified 550 double counted ballots. These ranged from what appeared to be random single-ballot doublings, likely due to paper jams, to entire batches being scanned twice. Surprisingly, about half of these double-scanned documents were QR-coded (Ballot Marking Device) ballots. We had to rely on the ordering of the ballots and patterns to decipher the duplication of these.

Original Count - Missing Ballot Records

More alarming, however, was the discovery of an astonishing 5,812 ballot records present in the original count but missing in the recount data. Entire batches were nowhere to be found in either the recount ballots or the cast vote records.

Adding to the perplexity, we observed some ballot tabulators from which random ballots seemed to have been arbitrarily removed, without any discernible pattern. It was as if someone had rifled through the ballots, randomly discarding some. There was no clear pattern in terms of the candidate selection or any other identifiable reasoning behind these removals, leaving us with more questions than answers about Fulton County's handling of the election.

12 of the 5,812 Missing Ballots from the Recount

Recount - Double Counted Ballots

In Fulton County's recount, the analysis revealed considerable discrepancies between the original count and the recount, particularly in terms of double/triple counted and stray ballots. A striking finding was the substantial number of ballots that were scanned twice in the recount, with a total of 3,930 ballots either double or triple counted.

10 of the 3,930 double scanned ballots in the Recount

Our analysis, especially the double counted Hand Marked Paper Ballots (HMPB), is believed to be quite precise, benefiting from the availability of most double counted ballot images for examination. We were able to visually examine nearly every double counted ballot image.

The double counted Ballot Marking Device (BMD) ballots relied on the order and grouping of the double counted ballots. Most of the BMD ballots were double counted in batches, unlike the HMPB. So, we were able to easily find most of the double counted ballots. Since a new BMD ballot can be mistaken for a missing BMD ballot, there are probably more double counted BMD ballots that we were not able to count.

Recount - New Ballots Added

The recount process in Fulton County brought to light the addition of 1,581 ballots that were not part of the original count. This significant number of new ballots was striking, as it exceeded the total number of discrepancies identified across all other counties combined. Among these were a considerable number of test ballots included in the recount. Our examination of these new ballots did not reveal any markings typically used to indicate duplications, leaving their origin unclear.

10 of the 1,581 Ballots added to the Recount

Identifying stray Hand Marked Paper Ballots (HMPB) was a more straightforward task. Our method involved comparing ballot images from both the original count and the recount that shared identical voting patterns. We then methodically eliminated the matched ballot images through visual inspection, one by one, until only the new, unmatched ballots remained. This approach, while effective, was time-consuming, particularly in cases where a voting pattern recurred more than 40 times. This painstaking process highlighted the intricate nature of ballot analysis in ensuring the accuracy and integrity of the election results.

Analyzing stray Ballot Marking Device (BMD) ballots in our study presented numerous challenges. A primary difficulty was identifying which specific ballot was missing from certain batches in the tabulators. This task was complex due to the inherent ambiguity, making it tough to pinpoint the exact source of the discrepancies accurately.

Given that the ballots appeared identical, the usual method of visually comparing ballot images to identify the outlier was rendered impractical. The identical appearance of identical voting patterns posed a significant challenge to the standard approach of distinguishing one ballot from another based on unique visual characteristics. Without distinctive features to guide the comparison, finding the "odd man out" among these identical ballots became an unfeasible task in our analysis.

To tackle this issue, we relied heavily on intuition and a method of analytical grouping. This approach required us to meticulously comb through the data to make educated guesses about which tabulators were most likely responsible for the missing ballots. It was a delicate balance of leveraging data-driven insights and relying on experienced intuition, especially considering the complexity and the lack of clear indicators in the tabulator data.

This aspect of our analysis underscored the complexities involved in handling BMD ballots. Unlike hand-marked paper ballots, which typically have unique identifiers like individual marking styles, BMD ballots lack such distinct traces. This absence of unique markers significantly increases the difficulty in accurately tracking and accounting for each ballot, a challenge amplified in a large and intricate system like that of Fulton County.

Therefore, our analysis in Fulton County went beyond simple data examination. It involved an in-depth exploration of the tabulation process, blending technology and human judgment. We had to delve into the patterns and behaviors within the tabulation data to draw conclusions and hypothesize the most plausible scenarios for the missing ballots.

Recount - New Ballot Source

Although many of the new ballots come from various random precincts, I do want to note some precincts that have the lion's share of new ballots added.

• Hand Marked Ballots = 602
• Ballot Marking Device Ballots = 979

• All new ballots: 1,582 ballots (1050 Biden/ 315 Trump)
• Precinct 10D: 80 ballots (69 Biden/5 Trump)
• Precinct EP04A: 78 ballots (67 Biden/5 Trump)
• Precinct FA01B: 333 ballots (283 Biden/45 Trump)

When I compared the ballot images against the Cast Vote Record, I was surprised to see that many of these ballot images were removed. The ballots from precincts EP04A, AP01B and AP07B are not in the cast vote record data, which accounts for 128 ballots not included in the final count.

Deleted Ballot Images - Original Count

One significant issue we faced was that Fulton County had deleted the ballot images before our analysis could begin. This deletion resulted in a considerable gap: over 374,467 ballot images recorded in the cast vote record were absent from the ballot images. We were left primarily with absentee hand marked paper ballots (HMPB) to examine, as nearly all the Ballot Marking Device (BMD) ballot images were deleted.

Thankfully, it's the HMPB ballots whose images are most crucial for identifying the various errors in the reading process and recognizing the unique marks each voter makes. For BMD ballots, there is no way to conclusively match a ballot in a recount since they lack unique, identifiable features. Despite this limitation, our ability to search for stray and double counted ballots was not significantly hindered.

Deleted Ballot Images - Recount

In the recount, 17,724 ballots were cast but the ballot images were deleted. The ballot images that were deleted seem intentionally selected and not a random computer error as was the official reasoning.

• Tabulator 801 / Batches 113-119
• Tabulator 802 / Batches 81-86
• Tabulator 803 / Batches 2-7,20-23,26,45-46,70,81-88
• Tabulator 804 / Batches 73-75,85-88,124-126
• Tabulator 816 / Batches 53-97

A large number of the double-scanned ballots originate from Tabulator 816, accounting for 2,038 of the doubled-scanned ballots. Notably, in Tabulator 816, Batches 53-58, where a total of 854 ballots were duplicated, happens to align with the start of the deleted ballot images in batches 53-97.

Another source of duplicated ballots is Tabulator 802/Batch 80, which had 162 double-counted ballots. The very next batch (#81) is when the ballot images were deleted.

The selective deletion of only certain batches is remarkable, while others were left untouched. This pattern doesn't seem like a random system error, but rather like it was done intentionally. If this was a USB memory error or files copied incorrectly, they would be more likely contiguous vs. the random batch selection that we are seeing. Each batch was complete with all the images per batch accounted for.

This all points to a potential connection between ballot duplication and the selective deletion of ballot images. It raises the question of whether someone at Fulton County knew about the double-counted ballots in this tabulator and deliberately deleted images to cover up the error, rather than a mere coincidence.

Write In Ballots Consolidated

During the hand recount in Georgia, it appears that at least 1,100 ballots with write-in candidates for president were segregated from their original batches. This move seems to have been an effort to simplify the hand recount process, which focused solely on the presidential vote. Consequently, ballots that either had 'no candidate' or contained a write-in candidate for president appear to have been removed from their original batches and consolidated in Tabulator 794.

A notable characteristic of these batches is the high level of proper adjudication, with an average rate ranging from 85% to 100% per batch. These batches predominantly feature ballots where the presidential contest included a write-in candidate, and it's commendable that the election officials properly adjudicated these ballots, ensuring every vote was accounted for.

Senatorial Test Ballots Added to the Recount

In the realm of unusual occurrences during the recount, this stands out as especially peculiar. It consisted of 17 sequential ballots where voters seemingly chose only the second senatorial option, with no down-ballot votes marked. This pattern is highly atypical and defies the usual expectations of voting behavior.

It's hard to imagine a scenario where a consecutive line of real voters would each choose only the second senatorial candidate on their ballot, completely ignoring all other races. A more likely explanation for this anomaly could be that these ballots were part of a testing procedure on the ballot machines. This theory is supported by a similar pattern I recall observing in three other counties, where a batch was marked exclusively for a group of senatorial candidates, again without any down-ballot selections.

I was informed about an affidavit from Bridget Thorne, a voting technician for Fulton County. According to Thorne's statement, the testing and calibration process for each ballot marking device (BMD) in each voting precinct involved printing 21 test ballots. The choice of 21 was dictated by the number of candidates in the Loeffler senate race. While these test ballots could have been printed on plain paper to distinguish them from actual voting ballots, they were, in fact, printed on the same heavy cardstock used for live Georgia ballots, making them indistinguishable from the real ones.

Thorne's affidavit revealed a concerning lack of security in the handling of these test ballots. They were reportedly managed in a haphazard and careless manner by Dominion employees, often left unsecured in various locations. Thorne recalled instances where entire batches of these test ballots were misplaced, necessitating reprinting. On realizing the security risk posed by these indistinguishable test ballots, Thorne and her daughter took the initiative to 'spoil' many of them by marking the QR codes or tearing the ballots, spoiling approximately a couple of thousand. However, Thorne expressed concern that thousands of test ballots remained unspoiled and supposedly were collected for shredding, though she could not confirm if the shredding actually took place. You should also notice that all of these senatorial test ballots were for one precinct, 08E.

Blank Test Ballots

In this next group of ballots, the focus remains on selecting senatorial candidates. However, unlike previous instances, the Write-In candidate section was left entirely empty. I wasn't aware this was an option. Leaving the Write-In space blank is a recurring pattern we've seen in these test ballots, and there are numerous other examples that demonstrate this.

RJ's Test Ballots

This next set of ballots exhibited the same pattern of voting for the senate race, and again, there were 17 sequential write-in ballots added to the recount. In this batch, none of the entries were remotely valid, consisting of a few random letters. At least half of the write in ballot names, used the 'R' and 'RJ' initials. This is starting to point to someone's name who was testing the machines.

Notice again how the precinct codes are quite similar and the repetition of the pattern for a group of sequential ballots.

I do want to make a specific and clear note that although RJ's initials are on many sets of test ballots, that doesn't mean that this is the person who did anything wrong. It just means that he was a prolific tester and he was consistent in his work. There is no need to blame this person for anything other than doing a lot of testing.

More Senatorial Test Ballots - John Fortuin

Continuing the pattern where testers are entering the second senate contest, this set of 12 ballots had John Fortuin, the Green Party candidate, as the 2nd senator choice. The President and 1st Senator contests were all blank write ins, with all down ballot races blank.

Here is the voting choices for the ballots. Notice how they are using precinct codes that are very similar and how the patterns repeat over and over on the ballots. I reordered the ballots so that the precincts group together, but they're all sequential in their original order.

This next sequential set of ballots exhibited a pattern that were far from a typical voting behavior, but extremely similar to the senatorial test ballots. The Green Party candidate 'John Fortuin' was the only valid candidates selected, followed by single/double letter entries mainly marked 'R' or 'RJ' for the top three contests. This is repeating the senatorial test ballots, but using the same candidate for every ballot.

This pattern strongly suggested that these ballots were not genuine votes but rather more test entries made while working with the voting machines.

Jorgensen's Ballots

As we delve deeper into the recount data, a discernible pattern begins to emerge, particularly in the use of 'senator only' and Green Party test ballots. The previous 4 sets of identified test ballots are at least attempting to follow the test procedure of using this senatorial position as an indication that these are tests. The next set of ballots will show a continuation of this pattern in the same tabulator, appearing just a few batches later.

It is crucial for the reader to note the repetitive use of single characters like 'R' and 'RJ' by the operator. This consistent pattern, observed across various batches, sets the stage for understanding the more significant issues that follow.

Initially, the problems encountered, such as the mishandling of ballots, seemed like mere annoyances that would not substantially impact the final count. However, a more alarming situation started to surface with the analysis this batch. 17 sequentially added ballots voted straight Libertarian candidates, with write in votes consisting of a few letters.

The pattern of voting straight Libertarian ballots continues with another sequential list of 10 ballots. The top candidates voting pattern is exactly the same as last batch, but this time, the write in candidates are either left totally blank or single letter write in.

The striking similarity between these ballots and the previously identified fake test ballots suggests they were created by the same technician. Unlike the first set of ballots, which could be dismissed as test data, these ballots present a more serious issue. They were counted as legitimate votes for actual candidates, thereby directly influencing the election results. The protocol for creating test ballots has been violated and nearly impossible to identify ballots created. The fact that these ballots were included in the official count points to a significant lapse in ballot security and validation.

Even More Test Ballots - Joseph Biden & Donald Trump

Now, let's get to the extremely controversial ballots that I've located. Only 30 ballots later in the ballots where the 17 Joe Jorgensen ballots were found with the write in ballots of R and RJ, comes 9 ballots voting for Joe Biden. These ballots are only identified by the write-in candidates in the middle of the ballot. The write-in ballots are all single/double letter words, with 4 of them consisting of the initials of our favorite tester. The other 6-8 ballots were single letter entry.

As you view these batches of ballots, another thing you should note is the continued lack of diversity of the precincts. Each of the precincts listed in the ballots (the code right after the dash) is similar in all of these batches. In this batch, it's precincts that start with a 11 or a 12. You may want to go back and look at the batches over the last few pages and you should see that it's a pattern. (12G, 11P, 11R)

-

In an apparent attempt to not show any voting bias, in the same exact batch where we found the Jorgensen and Biden ballots, we find 10 sequential Donald Trump test ballots added to the recount.

If it wasn't for RJ's consistent write-in names, I would have certainly thought that this was a valid ballot if it wasn't for the fact that 6 of the 10 write-ins were R and RJ. Once again, notice the precincts are all extremely similar in this set of images. (11P/12G/11N/12F)

There is another set of 5 ballots that follow the same pattern, but the write-in for Surveyor is left blank, or with a single letter. Again in this batch, precincts are similar (CP04B, CP07F, CP07D)

These last few ballots were just barely identifiable! Thanks to RJ's write-in votes and the proximity of the test ballots with obvious test ballots could I even identify them. How many more test ballots are included in the count that I couldn't find? If I search for all write-ins with just the letter 'R', I find a total of 77 ballots. Do these ballots indicate that they were tests, or did the voter want all republican entries here instead?

Is the following ballot real or test? There really is no way to tell.

Given the nature of these ballots and their potential impact on the election outcome, it's imperative that this situation triggers a thorough fraud investigation. Going back to the Bridget Thorne affidavit, there are potentially a thousand of test ballots unaccounted for. And without the testers following the testing procedures that were set out, we cannot say whether there are an additional thousand votes included in the election or not.

The inclusion of these test ballots in the recount not only raises questions about the integrity of the counting process but also calls for an urgent review of the security protocols and handling of ballots. This scenario highlights the critical need for heightened vigilance and strict adherence to electoral procedures to safeguard the fairness and accuracy of elections.

Intentional Doubling?

As we delve deeper into the recount data from Fulton County, a striking pattern emerges, one that resonates with trends observed in other counties. There is a notable similarity in the number of stray and double counted ballots between the original count and the recount. In the original count, there were 6,362 ballots that are not in the recount, while the recount showed 5,511 new ballots that are not in the original. This similarity aligns with a recurring theme seen across various counties where new ballots were either introduced into the recount or an unusually convenient number of ballots were double counted. The end result of these actions invariably aligns the recount numbers with the original counts, raising questions about their intentionality.

This pattern begs the question: Were the double counted ballots in Fulton County intentionally created to align the recount with the original count? Given that Fulton County had conducted a hand count prior, it's plausible that there was an awareness of the mismatch between the two counts. While it's impossible to definitively conclude that this was a deliberate act based solely on the data, the frequency and consistency of this occurrence across counties suggest that it may not merely be a coincidence.

The data points towards a possibility that these discrepancies were not just random errors but part of a broader pattern. This speculation is not without merit, considering the similar patterns observed in other counties. The repeated occurrence of these anomalies in both the count and recount processes in Fulton, and indeed in other counties, signals a need for a more in-depth investigation to fully understand the motives and methods behind these discrepancies.

Absentee by Mail Ballots

During the comparison of original and recount ballots in Fulton County, a peculiar pattern emerged. Certain ballots exhibited consistently light or dark un-filled circles, a feature independent of the tabulator or scanner used. This consistency suggested that the characteristic was inherent to the ballots themselves, not a result of scanner variability. These ballots were colloquially termed 'fuzzy' ballots, a phenomenon unique to Fulton County, and appeared in a significant number.

A critical realization was the presence of a prominent barcode at the top of these ballots. This barcode was indicative of an on-demand ballot, a contingency method used when a county faces a shortage of standard ballots for advance voting.

Inquiries into the prevalence of On Demand Ballots in Fulton County suggested that approximately 5,000 such ballots were generated. With the latest software update, I was able to identify 3,575 of these On Demand Ballots that were cast in the election.

With the On Demand Ballots now identified, a crucial aspect of our analysis focused on the timing of their scanning. This factor provides insight into the distribution and handling of these ballots in the voting process.

Contrary to expectations, these On Demand Ballots did not exhibit a random distribution among early voting ballots. Instead, they tended to appear in batches, often grouped together, which deviates significantly from the expected norm. Under typical circumstances, if these ballots were mailed, received, and processed like other early votes, they would be randomly intermixed with the rest of the ballots.

A striking example of this pattern was observed in (Tabulator 5162/Batch 387/Ballots 1-75), where 75 consecutive ballots were all On Demand. Such a sequence starkly contradicts the probability of a random mix that would occur through normal postal return processes.

This pattern of grouping was not an isolated incident. In most instances, On Demand Ballots were not randomly dispersed within the batches as one would expect. Instead, they were commonly found clustered together.

The most intriguing aspect emerged when we analyzed the scanning times of these ballots. On Election Day, only a minimal number of On Demand Ballots were processed. The following day, November 4th, saw a slightly higher but still small number. However, at 10 AM on November 5th, two days post-election, a surge occurred with over 1,800 of these ballots being scanned - an anomaly in the context of time analysis.

Upon charting the scan times of all ballots, it became evident that the On Demand Ballots were predominantly scanned towards the end of the entire ballot scanning period. This finding is highly unusual, given that these ballots, by their nature, should have been randomly distributed among the overall ballot pool.

The peculiar timing and grouping patterns of the On Demand Ballots in our analysis raise significant questions about their handling and processing. These ballots' concentrated appearance in batches, and their late scanning times, particularly days after the election, point to an irregularity that deviates from standard electoral procedures. This abnormality warrants further investigation to understand the implications and reasons behind such a distribution pattern.