Maricopa Hoax: The Ninjas made up the numbers

On Friday, September 24, the political world held their breath as the Cyber Ninjas and other pro-Trump contractors delivered their report on the accuracy of Arizona’s 2020 presidential election. Surprisingly, and somewhat comforting, the Ninjas said that Trump lost 261 votes. Not surprisingly, the Cyber Ninjas affirmed that Joe Biden held onto his 45,000-vote lead in Maricopa County and gained 99 votes.

The next day, the Washington Post quoted Senator Fann, “This [the fact that the audit matches Maricopa’s official machine count] is the most important and encouraging finding of the audit. This finding therefore addresses the sharpest concerns about the integrity of the certified results in the 2020 general election.” 

We held a darker view: Were Senator Fann and the country victims of an epic hoax?

Buried in Randall Pullen’s (Arizona Senate’s co-liasion)  report of the Senate’s machine count were 17 pages of innocuous-looking tables with no description of the column headings.  Anyone with 15 seconds of instruction, and a calculator can come close to our findings in less than 15 minutes.

Herein, our report provides evidence developed by analyzing Pullen’s report. Starting on page 20, he provides an extract from the Ninja’s 695-page report entitled “AZ Audit Consolidation Worksheet 7/10/2021” (hereafter, “Consolidated Worksheet”), which shows:

  • A difference of 15,692 ballots when the Ninja’s hand count of ballots is subtracted from the Senate’s machine count. In short, with this enormous discrepancy, any discussion of vote counts – including Biden’s 99 vote gain and Trump’s 261 vote loss – is meaningless.
  • The difference between our published counts from July 12 and the Senate’s machine count of ballots shows a difference of five ballots. In short, as far back as July 12, we predicted the ballot count per box which now serves as an independent audit of the Senate’s machine count of ballots by box.
  • Our 27-page report and accompanying spreadsheet fully describes how we developed our conclusions solely from documents provided by Pullen.

This is our story.

While everyone was breathing relief that the Ninjas exercise – don’t call it an audit – upheld Maricopa County’s official results, our antennas were already up because three pesky facts have continued to bother us.

  1. The Ninjas did not count the same set of ballots. The Cyber Ninjas said their results tracked the county’s certified results even though they decided to count about 27,000 ballots that Maricopa Elections did not count and ignore about the same number of ballots Maricopa did count. When ballots are damaged or fail tabulation, a two-person bi-partisan team copies the voter’s choices onto a clean ballot in a process called “duplication.”  The Cyber Ninjas, claiming there were fewer original ballots than duplicates, counted the original ballots and ignored the duplicates. Curious.
  • Senator Fann is hiding something. The Senate continues to fight the release of public data on grounds that Arizona’s courts have decisively ruled against. Included in the withheld documents are batch and box-level ballot counts, just like those contained in Pullen’s report on the Senate’s machine count. More on this below.
  • A curious anomaly. Garrett Archer, a reporter at Phoenix’s ABC15 News Affiliate, noticed that the ballot counts for the Presidential and the U.S. Senate contests did not match. The Ninjas excuse? “NOTE: Vote totals for the presidential and senatorial elections mismatch slightly primarily due to small differences (emphasis ours) in hand counts among the 2.1M million ballots.”

This last item – a mismatch in the ballot counts on contests that appeared on the same ballot – piqued our interest because such a mismatch should be impossible.

Now, with Pullen’s report, we had something meaty we could dig into.

In the words of Seth Meyers, “It’s time for a closer look.”

First, a few stats:

Table 1

Our focus was the 17-page extract at the end of Randall Pullen’s report mentioned above. That extract is from the above mentioned Consolidated Worksheet. The worksheet in Pullen’s report is the only detailed data released so far by the Ninjas and is one of the documents we have sought since our August 11 public records request.

The Consolidated Worksheet enumerates the results of five methods the Senate and the Ninjas used to audit their ballot count as shown in rows 2-6 in the table below. In row 1, is a reference to our July 12, 2021 published report which maps Maricopa’s official canvass to storage boxes. We use this report to audit the AZ Senate’s machine count. As shown in row 2, the difference is 5 ballots

Table 2

* The Senate’s machine count matches the official results (CVR) within 5 ballots. Since the machine count was so close, remaining discrepancies are computed from the Senate’s own counts. In short, the Senate’s counts audit the Ninja’s hand count.

** Discrepancies expressed as absolute values to prevent plusses and minuses from cancelling out.

*** Discrepancies expressed as simple differences to show excess ballots counted.

Below is a timeline of our attempts to provide information enabling the Ninjas to check their results against the official results. We did this, not out of any belief that what they were doing was valid because we do not. Rather, we wanted to show that Senator Fann and the Ninjas we were prepared to hold them accountable for spreading disinformation.

We believe our worst fears have happened – the entire exercise in hand counting ballots on lazy Susans for two months, was a hoax.


Late May, 2021: we knew that the Ninjas could not produce an audit because they could not compare their results (i.e., ballot counts and vote counts by storage box) against the official results. That is because Maricopa’s reports its results by precinct and voting method (i.e., Early Voting, Election Day, and Provisional). The Ninjas counted by box.

May – June: We developed a “Rosetta Stone” to link the official results to each of 10,341 batches and 1,634 boxes. On June 6, we published the ballot and vote counts for two boxes of ballots and challenged Senator Fann to direct the Ninjas to confirm their counts against ours. The Arizona Republic’s front page of their print edition read, “Put Up or Shut Up.”

June 30: We believe that by the end of June the Ninjas knew their ballot counts were substantially different from Maricopa’s.  Furthermore, we believe that someone convinced Senator Fann that the official results were wrong and a separate count would support the Ninja’s hand count and if so, a high-speed machine count of the ballots would confirm that Maricopa’s ballot count was wrong. 

July 12: We sent an open letter and spreadsheet to Senator Fann, Kory Langhofer, and Ken Bennett with the ballot counts for 1,634 boxes and the list of boxes that should be counted. A few days prior to Sept. 24, I was told by the Senate’s attorney for the “audit” that the list of boxes was “extremely helpful” to the Ninjas. We now believe it helped them determine which boxes to say they did not count.

July 13: Senator Fann states that their ballot counts did not match the county’s ballot count and that they would initiate a machine count.

July 15: Senators Fann and Peterson conduct a Senate hearing where Doug Logan reveals (skip to 1:23) that his team is confused as to which boxes to count. That is because Maricopa County transferred 1,691 boxes and, of those, 1,634 boxes contained exactly the 2,089,563 ballots listed in the official Canvass. Furthermore, our analysis of the vote counts for every candidate across every contest also matched the official results perfectly.

July 19: Ken Bennett gives us ballot counts for six boxes – five boxes match our published counts exactly. In the sixth box we found a transcription error. The Senate staff reported 1,379 ballots against our published count of 1,397 ballots.

July 21: Ken Bennett provides us with additional 18 box-level ballot counts which had minor differences in four boxes.

July 22: We provided the Arizona Republic with the data that the machine ballot counts had closely matched our prediction of the ballot counts in 24 boxes at an accuracy level of 99.94%.

July 23: Ken Bennett is refused entrance into the Wesley Bolin Building where the Senate’s machine count is underway. It is our view that Bennett has always been interested in a proper audit and gave us the ballot counts out of frustration that the Ninjas had refused to share any information with him. It turns out, we did confirm the machine counting process capable of producing accurate counts.

August 11: We file a public record request for detailed ballot and vote counts organized by batch and box.

Sept 16: We update our July 12 spreadsheet with vote counts for third-party candidate Jo Jorgensen which gives the Ninjas an additional check on their vote counts. Votes for Biden and Jorgensen are not revealed to enable a valid audit where two independent results could be compared.

Sept 21-23: Ee obtain an early draft of the Ninja’s report and publish a report which debunks most of the allegations in Volume III related to voter registration and the “voted file.”

Sept 24: The AZ Senate’s hearing – a mountain of detail and, at the end, an Easter egg. Randall Pullen’s report on the machine count contains 17 pages of detailed ballot counts.

Sept 25-27: we analyze Pullen’s data drawn from the Ninja’s 695-page report.


  1. An enormous discrepancy of 15,692 missing hand counted ballots from 40 boxes  out of out of 1,634, supports our opinion that the Ninja’s hand count of ballots was so far off the Senate’s machine count of ballots that any statements about the vote counts (e.g., that Trump lost 261 votes) are meaningless.
Table 3

2. Our analysis shows that the machine count at the box level agrees with official results; as shown below, the difference in our published box-level ballot counts was off by 5 ballots.

Table 4

3. We believe the Ninjas have profoundly misled the public, Senator Fann and her colleagues. Senator Fann should exercise oversight, order the release of all public data and publicly distance herself from the Ninjas as quickly as possible.

4. Having zero experience in election audits, the Ninjas announcement that they had confirmed, to a high degree of accuracy, the election results of the second largest county in the country is, we believe, laughable.

5. The assertion that Trump had lost 261 votes was, we believe, a “shiny object” designed to convey believability to an otherwise unbelievable hoax.


We are not done; much is left to do. As we stated in this Op-Ed in the Arizona Republic, Senator Fann must release the Ninja’s findings. Now, with the Arizona Supreme Court upholding the decision of the Appellate Court, she must comply – or be held in contempt. Those court decisions require her to turn over lawfully requested public data including our request of the 695-page Consolidated Worksheet as well as a similarly detailed report containing vote counts.

Once we obtain the data we have requested, we will conclude a real audit of the Ninja’s sham process, publish our results and stand ready to defend them vigorously.

The events surrounding the Maricopa Hoax are a shameful episode in America’s history. Each of us feels lucky to be able to serve America, to protect our elections, debunk lies and confront those who would use their power to spread disinformation and doubt about America’s electoral systems – the heart of our democracy.

Election Audits – A Simple Explanation

Larry Moore

September 20, 2021

Recently I posted a blog entry that provided additional data by storage box. I released the vote counts for third-party Presidential candidate Jo Jorgenson for each of 1,634 storage boxes.

To the organizers of the Arizona Senate’s months-long process designed to undermine confidence in the Maricopa’s election, this data signaled that we could effectively audit their numbers.

Several people asked me questions that indicated that they did not have a clear concept of the audit. Having spent the last decade working on this audit methodology now marketed and supported by the company I founded, the Clear Ballot Group, it behooves me to explain it more clearly.

First, a definition. An audit compares two independently produced results based on the same data. It’s like asking two people to add an identical list of numbers.  If the results are the same and they did not talk to one another, we can be reasonably sure the totals are correct.

But, what is a “result?”

In elections, there are two types of results: ballot counts and vote counts.  Like Maricopa’s, all U.S. election departments create ballot counts by precinct and vote counts by candidate and precinct. However, the Senate contractors, the Cyber Ninjas, Inc., did not have the ballots organized by Maricopa’s 743 precincts. Instead, ballots were organized by storage box. Each storage box had an average of 1,274 ballots and 374 distinct precincts. Precincts were comingled in the same storage boxes because 90% of ballot cast were returned during the early voting period. There was no need to physically sort the ballots by precinct because the tabulation software could do that automatically.

Why does this matter? Because the county’s results and the Ninja’s results were not comparable.  Precinct counts cannot be compared to counts by storage box. The Ninjas cannot do an audit. Period.

We, the Audit Guys, linked Maricopa county’s official results to the 1,634 ballot storage boxes. The result: we can make an apples-to-apples comparison between the county’s results and the Ninja’s results.

Here are excerpts from the Audit Template with which we intend to audit the Ninja’s ballot and vote counts.

Compares official ballot counts to Ninja’s ballot counts and links counts to storage box and pallet.
Compares official vote counts to Ninja’s vote counts and links counts to storage box and pallet.

Here is a link to an expanded version of the above snapshots that explains the audit in greater detail.

Hopefully, you will see the conceptual simplicity of the audit:

List the boxes that should be counted, subtract the Ninja’s results from the official results for ballots and votes, analyze the differences.

I’ve personally participated in over 50 audits. We typically find that ballot control is the hardest thing for an election department do consistently. 

Maricopa’s process, which traces every ballot from the envelope opening process straight through tabulation, is among the very best I’ve ever seen.

What next?

Benny White, J.D.

September 18, 2021

The Senate is scheduled to receive a report of some sort from the Cyber Ninjas on Friday, September 24. This will be another press extravaganza designed to continue to disinform the public about elections generally and the 2020 General Election in Maricopa County. Many claims will be made. Most will probably have no evidence and will be difficult to verify.

We think the Ninjas will say they found more votes for Donald Trump and there are several areas where they continue to have questions because they have not been given all of the information they need from the Maricopa County officials.

I will address a few areas we suspect they will try to tell the public where they found problems.


The Ninjas were confused about ballots that were duplicated; they were not sure which ballots to count or did not trust the duplication process. The Secretary of State reported that instead of counting the ballots that were duplicated and were actually counted in the official results the Ninjas decided to count the original ballots that had some sort of defect that prevented them from going through the vote counting machines.

Here is a table that shows the actual number of ballots that were duplicated during the entire election:

You can see there were a total of 27,457 ballots that were duplicated and included in the official results. These included mail and early ballots that were duplicated by the election boards before they went to the central count room where the ballots were tabulated. The remaining 7,436 ballots were those that were rejected by the tabulators for some reason and had to be duplicated.

There is no mystery here. There are clear, detailed public records available that account for each of these ballots.

Electronic Adjudication of Ballots By Adjudication Boards

The next area the Ninjas will probably complain about is the process of electronic adjudication. Electronic adjudication allows an expedited and accurate process of resolving ambiguous vote marks and determination of write-in vote marks. Maricopa County is the only county in Arizona that employs electronic adjudication, in part due to the size of the voting population and the number of ballots cast in their elections.

Here is a table that accounts for every ballot that underwent electronic adjudication:

This table shows that there were 2,089,563 total ballots counted. We have public records that confirm that each of these ballots were cast by a qualified voter and the daily records of ballot tabulation of Early, Provisional and Election Day ballots exactly matches the official results.

Much was made initially about the high percentage of ballots that required electronic adjudication. There were 235,392 ballots (11.27% of total ballots cast) that were reviewed by adjudication boards. The vast majority were contests (102K) for which no changes were made, the reason being that the oval was identified as being ‘ambiguous’ (an insufficiently filled in oval), and the adjudicator examined the oval and decided that the weak mark was actually vote intent. The weak mark was allowed to stand without changing anything. Finally, 132,139 ballots needed some sort of modification to resolve which whether the write-in mark was for a qualified candidate, which qualified write-in candidate should receive a vote, whether it was really an overvote or undervote.

11,954 of these electronically adjudicated ballots involved the Presidential Elector contest. There may have been other contests that had some issue with vote marks but the table below shows exactly what the issues were, how they were resolved and what the results were:

This data shows us a few very important things. First, the vast majority of these ballots (6,611) were marked for unqualified write-in candidates. Secondly, these adjudications awarded additional votes to all three Presidential candidates with Biden receiving 2,069 to Trump’s 1,516.

There is a detailed record for each one of these ballots, the actions taken and identification of the election board involved and the time the action was taken. In addition to those records that were produced at the time the action was taken there is a complete record of the adjudication action included with every digital image of every ballot no matter whether some ambiguity was resolved or not.

We anticipate that the Ninjas will once again demonstrate that they don’t know what they are doing. They will talk about counts that we will show are wrong. They will talk about various problems they have discovered. Most of these will be about things involving election administration and operation of the election systems which they did not understand when they started and haven’t learned about over the last six months.

They attempted to create a different result that would show Donald Trump won Arizona and they have failed.

The Audit Guys Publish Vote Counts in Preparation for a Real Audit

Larry Moore

September 16, 2021

Today, the Audit Guys provide additional data to Arizona Senate President Karen Fann, her contractor, the Cyber Ninjas, and the media. With this “September 16th spreadsheet“, the Ninjas can easily verify the accuracy of their vote-counting methodology before they release their report to the Senate. The September 16 spreadsheet updates the July 12 spreadsheet, which has enabled the Ninjas to test the accuracy of their ballot-counting methods for the past two months.

To recap:

In mid-May, 2021, we began to develop the capability to provide an independent audit of the Ninja’s unorthodox recount procedures.  That is because the Ninjas seemed unaware that they would not be able to conduct an audit. Why?

An audit compares two independent results based on the same data.  Maricopa counted votes by precinct; the Ninjas counted votes on ballots in storage boxes.  It would have been challenging for the Ninjas to count ballots and votes by precinct; Maricopa has 743 precincts and, on average, storage boxes contain 1,274 ballots and 374 distinct precincts.  

Drawing on public data, we devised a method (see Appendix F of our report) to recast the official results to compare the official vote counts to votes counted by ballot storage box.

On July 12, the Audit Guys sent an email to Senator Fann, Ken Bennett (Senate liaison), and Kory Langhoffer (counsel to the Senate) containing a spreadsheet that listed the 1,634 boxes containing exactly the 2,089,563 ballots reported in the officially certified election results. Doug Logan, CEO of Cyber Ninjas, should welcome this data because, on July 15, he publicly expressed confusion over which ballot storage boxes should be counted out of the 1,691 boxes delivered by Maricopa County in response to the AZ Senate’s subpoena. See Logan’s testimony at 1:23 here.

In addition, the July 12th spreadsheet contained the ballot counts for each of the 1,634 boxes. One day later, on July 13, Senator Fann announced that the Ninja’s hand count of ballots did not match the official ballot count. The difference, although never revealed, was sufficiently significant to justify purchasing two high-speed paper counters at the cost of $30,000 to create an independent ballot count.

What is new in the September 16 spreadsheet? Two additional items which the Ninjas can use to check their results.

First, it contains the count of ballots in 24 boxes which Ken Bennet provided the Audit Guys. This ballot count, performed on high-speed paper counters, matched the official ballot counts at 99.94% level of accuracy. It will be interesting to compare the hand count to the offical results and to the Ninja’s machine count.

Second, the September 16 spreadsheet provides box-level vote count for Jo Jorgensen, the third-party candidate for President.  

With the names of 1,634 boxes to be counted, ballot counts for each box and vote counts for one candidate, it should be relatively easy for the Ninjas to see how close their ballot counts are to the official results.

Once the Senate complies with our public record request, we will compare ballot and vote counts from the Ninjas hand count and machine count to the official results.

Disaffected Voters: Impact on the Presidential race in the Arizona 2020 Election

I am publishing a spreadsheet that contains the backup data to support our contention that Trump lost Arizona because a sufficient number of disaffected Republican-supportive voters did not vote for him in the two most populous counties – Maricopa and Pima Counties.

This spreadsheet provides additional information not contained in our report, Lessons from Maricopa …

For our purposes, disaffected voters vote for a majority of down-ballot candidates of one party but not for that party’s President.

Here’s a link to the supporting spreadsheet.


C2674,822 Disaffected Republican Supportive Voters
C1948,577 voted for Biden
(represents 4.6x the margin of Biden’s statewide victory (10,457)
Source: Cast Vote Records from Maricopa and Pima Counties
obtained under a Public Records Request.

%d bloggers like this: