Repeated Batch Report

Fulton County Georgia, 2020 General Election

AuditEngine version:'v2.32.X'

Run Date: 2026-11-11 00:33:14.840049 Copyright (c) Citizens' Oversight

Return to Master Report Index

A batch scanned twice gives images that differ in bytes, ballot ids and pixel values. The votes the audit extracts from each scan stay the same. A vote pattern is the precinct, the style and the extracted votes of a ballot. A balanced pattern is held only by ballots of two batches, the same number in each. A rare twin has one ballot in each batch, a rare quad two in each, and so on. A repeated batch doubles each of its patterns, so every one of them is balanced. Chance links are scattered. A repeated batch puts many links between one pair of batches. A batch scanned whole in one run and split in another links only by its twins.

How the batch pairs are found

Comparing every pair of batches would take too long, so balanced patterns pick the pairs to look at. A batch pair with 3 or more balanced patterns is paired in full. Links counts them, each pattern once. A pair with fewer is also paired in full when every ballot of its smaller batch is linked, and that batch holds 3 or more. Paired in full means every ballot of both batches is offered for pairing, not only the linked ones. Each ballot of one batch pairs with a ballot of the other that has the same vote pattern, each ballot used once. Paired counts those pairs. It does not mean that every ballot pairs.

Paired % is the paired ballots over the ballots of the smaller batch. The paired ballots have the same votes, so their vote totals match exactly. Take out the unpaired ballots of each batch, and the two batches have exactly the same counts. A possible repeated batch has 90% or more of the smaller batch paired.

In CVR says whether the ballots of a batch have CVR records. Where both batches are in the CVR, the same ballots may be counted twice. A match by votes does not prove that two images are one paper. The images do.

Possible repeated batches

Status says whether a person's grades and the batch counts confirm the pair. A sample is a ballot whose vote pattern is held by one ballot in each batch, so its two copies must be one paper. A confirmed pair needs enough samples graded 5 and none graded 3 or less. The two sides must also have about the same ballot count, and nearly all of the larger side paired. Only confirmed pairs change the Reconciliation Report.

Batches with exactly the same ballots

These batches hold the same vote patterns, the same number of each. So they have the same ballot count and the same vote totals. This search checks every batch, with no need for links. Small batches of common patterns, such as one blank ballot each, can match by chance.

Ballots counted twice

208 ballots are in a possible repeated batch where both scans have CVR records. The CVR counts each of these papers twice. The tables give the votes of one copy of each. The audit counts each paper once, so it takes these votes out of its totals. The CVR is the official record and is not changed. So the CVR has these votes more than the audit.

• 05162_00234 and 05162_00235: 98, confirmed
• 05162_00139 and 05162_00147: 99, confirmed
• 05150_00346 and 05162_00389: 11, not confirmed, 05150_00346 and 05162_00389: a sample graded 3 or less; the sides differ by 94 ballots

Only the confirmed pairs are taken out of the audit in the Reconciliation Report.

Votes counted twice, by contest

Votes counted twice, by option

Near misses

These batch pairs were paired in full, but less than 90% of the smaller batch paired. Common patterns within one precinct pair some ballots by chance.

Review app: grade each pair of scans from 1, certain not matched, to 5, distinctive and certain match.

Pairs to inspect

For each possible repeated batch, up to 8 of its samples are shown side by side. Look for the same paper: the shape and place of each hand mark, stamps, tears and handwriting. Pixel darkness changes from scan to scan, so it is no test.

05162_00234 and 05162_00235

8 of 12 samples, 39 links. 98 of 98 ballots paired.

Mark comparison page

batch a	batch b
05162_00234_000002	05162_00235_000061
05162_00234_000006	05162_00235_000065
05162_00234_000008	05162_00235_000067
05162_00234_000010	05162_00235_000069
05162_00234_000013	05162_00235_000072
05162_00234_000018	05162_00235_000077
05162_00234_000021	05162_00235_000080
05162_00234_000022	05162_00235_000081

05162_00139 and 05162_00147

8 of 12 samples, 38 links. 99 of 99 ballots paired.

Mark comparison page

batch a	batch b
05162_00139_000001	05162_00147_000099
05162_00139_000005	05162_00147_000095
05162_00139_000012	05162_00147_000088
05162_00139_000025	05162_00147_000075
05162_00139_000026	05162_00147_000074
05162_00139_000029	05162_00147_000071
05162_00139_000031	05162_00147_000069
05162_00139_000032	05162_00147_000068

05150_00346 and 05162_00389

8 of 11 samples, 9 links. 11 of 11 ballots paired.

Mark comparison page

batch a	batch b
05150_00346_000001	05162_00389_000030
05150_00346_000002	05162_00389_000079
05150_00346_000003	05162_00389_000032
05150_00346_000005	05162_00389_000034
05150_00346_000006	05162_00389_000029
05150_00346_000007	05162_00389_000078
05150_00346_000008	05162_00389_000080
05150_00346_000010	05162_00389_000081