Read on Twitter
I read ASOG's report on their exam in Antrim County. It is illogical nonsense. The "68% error rate" that is getting attention is not 68% miscounted votes. Instead, it is just that ASOG found a log file where 68% of the lines contain the word "error". Don't even know what kind.
Their report is mostly about the log files. They didn't copy data from the tabulators or BMDs (which is difficult). They just copied the Election Management System (EMS) computer and some USBs and CFLASHs with encrypted data.
The report is consistently wrong about laws and federal certification of voting systems. It cites FEC Voting System Standards replaced in 2002. It cites EAC VVSG 1.1 which no voting system uses. It references a HAVA "Safe Harbor" 90-day requirement that is a total fabrication.
EAC certification under VVSG 1.0, which is used, requires a voting system to read 1.5 million consecutive votes accurately during their tests. Dominion and almost all of the other voting systems used in the US have passed the "million-mark test".
But again, it is apples and oranges. The number of errors the system wrote into a log file has nothing to do with the system's ability to accurately translate marks on paper into votes for a candidate.
The report doesn't say what the errors are so we can judge their significance. It doesn't bode well that one of the only ones in the report appears to be misinterpreted. @get_innocuous wrote up a good description. https://twitter.com/get_innocuous/status/1338576587780812800
The EMS designs printed ballots (which contests, which candidates) & creates a matching digital version for each tabulator ("ballot definition file"). That's how a tabulator reads a mark on the paper 3" over and 4" down and assigns a vote to a particular candidate.
The EMS also imports the data that comes back from the tabulators and aggregates them into election results. It's not clear if "errors" are from scanning ballots, operating the EMS (login for example), or another task. Without more info, an "error rate" is meaningless.
The MI SoS says the ballot definition file caused the Antrim miscount. The clerk created and shared Set A with all tabulators, corrected the candidates, and shared Set B with only a few tabulators. Sets A & B translate marks into candidates differently. https://www.michigan.gov/documents/sos/Antrim_Fact_Check_707197_7.pdf
This explanation makes sense. Imagine if I gave 22 people a map to get to 10 buried treasures, but then I moved and rearranged them and only gave a few people an updated map. A bunch would end up in the wrong places.
The exam did not compare the ballot definition files. It says the difference is because votes were changed. In D.7-12, it says votes were changed before the 2nd count. But the 2nd count are the results they prefer! Biden votes were changed to Trump votes? 
I think their alternative theory is: errors were intentional (via a hidden algorithm?) so that votes could be sent to adjudication and secretly changed (by staff in 48 different counties?). That would be quite a conspiracy.
The report has a lot of wrong info about adjudication. The references to "reversed ballots" and "rejection rate" aren't explained and don't have a common meaning. "All reversed ballots are sent to adjudication" is gibberish.
For precinct, in-person voting, the tabulator returns ballots with ambiguous marks back to the voter to fix. A 
or an X are common examples. There is no adjudication unless there is an audit or hand recount. Only ballots with write-in votes get examined by a human.
or an X are common examples. There is no adjudication unless there is an audit or hand recount. Only ballots with write-in votes get examined by a human.
With mail-in ballots counted on a central scanner, the voter is not around to fix any problems. So ambiguous marks get flagged for human review. Every state has different rules about what counts as a vote. Usually it is a bi-partisan review process, not a single trusted person.
Yes, adjudication tools are powerful. They have to be so human reviewers can choose how to count a vote according to state law. Every adjudication is recorded. Dominion adds data to the ballot's AuditMark. The report did not examine digital ballot images or AuditMarks.
The report doesn't say how many ballots were adjudicated or show any were changed. Remember: Antrim & many other counties have hand-marked paper ballots. The MI SoS intends to recount them by hand during an upcoming audit. Adjudication fraud would be dumb crime for a clerk.
The report raises legit concerns about system configs & out of date software. It is a known problem being worked on. They can't be updated because current rules favor stability over being fully patched. Not crazy if a system is physically-secured & offline, but we can do better.
TLDR; The ASOG report frets over errors in log files, flags security that is not best practices, floats a conspiracy theory about adjudication, and gets many basic facts wrong. It presents no evidence of election fraud, vote manipulation, or hacking.
Dominion has a fact-checking page on Antrim. They note, and just told MI lawmakers, that Antrim does not use digital adjudication or ranked choicevVoting (modules with an upcharge). ASOG signed an affidavit without checking? They have the EMS. 
https://www.dominionvoting.com/election-2020-setting-the-record-straight-antrim-county-michigan/
https://www.dominionvoting.com/election-2020-setting-the-record-straight-antrim-county-michigan/
The ASOG report actually says its bogeyman "high error rate" includes both errors and warning messages and "most of the errors" are related to configuration. Their example: "Ballot's size exceeds maximum expected ballot size." (e.g., 8.5" x 14") Oh brother.
