Election Day

• Election Night Reporting: 15,277
• Cast Vote Records: 15,277
• Ballot Images: (not available)

Recount

• Election Night Reporting: 15,277
• Cast Vote Records: 15,327
• Ballot Images: (not available)

Aberrations

• Double counted ballots in original count: 1
• Double counted ballots in the recount: 4
• Ballots not appearing in recount: 3
• Filler Ballots added to the recount: 13
• Suspicious Ballots added to the recount: 4
• Test Ballots added to the recount: 37

General Observations

There are 55 new ballots appearing in the recount that were not in the original count. It appears that some test ballots and filler ballots were added to the count.

During Coffee County's Risk-Limiting Audit (RLA) hand audit, it seems that the ballots were organized by the presidential candidate. This organization method led to an unusual grouping of blank, write-in, and Jorgensen ballots. Some of the issues observed could be linked to this sorting approach.

Double counted ballots

A single computer generated ballot was double scanned in the original count

A batch of 4 records were rescanned a second time during the recount in reverse order

Suspicious Added ballots

Four new absentee ballots were added to the recount. There were two ballots for each candidate, with exactly the same unique voting pattern, and they don't seem to be naturally added. Both the Trump and Biden votes were added two the same two precincts.

Presidential Filler Ballots

One notable observation was a set of 11 sequential absentee ballots in the recount, marked only for the presidential choice. These 'filler' ballots, which we've previously seen used to compensate for missing or duplicate ballots in initial counts, were particularly curious in this context. The original count in Coffee County did not exhibit any significant issues, raising the question of why such filler ballots would appear in the recount vote record.

Senatorial Test Ballots

Two batches later in the count, we encountered a notable set of QR coded ballots that stood out for their unique characteristics. These newly added ballots were easily identifiable as they exclusively featured entries for the second senatorial race. This specific type of ballot, marked only for the second senator, wasn't unique to this situation; similar Senatorial Test Ballots were also identified in the recount processes of Fulton, Paulding, and Cherokee counties.

Just a few batches after encountering the Senatorial Test Ballots, we stumbled upon another curious find - eight new QR coded ballots all casting votes for Jorgensen in the presidential race. This discovery was reminiscent of a pattern we had observed in Cherokee and Fulton as well as three other counties, where numerous test ballots featuring Jorgensen as the presidential choice had been added. The similarity in these instances raises the question: could these eight Jorgensen ballots in this batch also be test ballots that were mistakenly added to the count?

Another intriguing discovery in the newly identified images was a group of brand new federal QR coded ballot images. While these did not exhibit the same level of sequential order as previous groups of ballots, the ordering process used in the Risk-Limiting Audit (RLA) likely influenced their placement. Despite the RLA's impact, it was noticeable that these ballots tended to be somewhat sequential relative to each other. This observation adds an interesting dimension to the analysis, suggesting a pattern in how these new federal ballots were organized and integrated into the existing ballot sets.

Five additional ballots were included in the recount that appeared mostly valid. However, two of these ballots each had eight write-in votes, which suggests they might be a different type of write-in ballot. Unfortunately, I couldn't confirm this as I didn't have access to the images of these ballots. Therefore, out of the 55 new ballots discovered, at least 50 are confirmed as test ballots.

Imbalance from Official Count

While I'm not concerned about the original count minor imbalance, the fact that the official count didn't increase with the introduction of these 50 new ballots seems to confirm my analysis. These new ballots don't seem to be included in the final count, so did an election official in Coffee County realize that these test ballots were added, and remove them from the official reporting? That is actually quite doubtful since these ballots were most likely added during the RLA count and the ballots were mixed into various other batches.

Plus, there are 4 double-scanned ballots also included in the final cast vote record. So the official results should be at least 10 votes higher than what is reported from the data, although the final total is indeed correct.