Georgia's 159 counties are conducting a risk-limiting audit of the secretary of state's race, hand counting randomly selected batches of ballots to verify incumbent Brad Raffensperger's victory and that election equipment worked properly.

The audit, required by law, is different from the post-2020 election audit that saw workers spend days hand counting all 5 million ballots cast in that race, one of three separate counts of that contest that saw President Joe Biden narrowly defeat former President Donald Trump. 

On Wednesday, Gabriel Sterling with the secretary of state's office kicked off the audit by having members of his staff roll 20 10-sided dice to create a random seed number that was then put into the state's auditing system, along with a list of every batch of ballots cast in the midterm.

From there, the software uses statistics to identify enough random batches of ballots to be counted so that the state could have confidence that scanners worked correctly and the correct winner won. In this case, the state upped the number of batches to be counted so every elections office could participate.

"Every county will audit at least one batch that is from Election Day or advanced voting, and another batch that is from either absentee or provisional voting," Georgia Elections Director Blake Evans said. "The reason for that is we wanted to make sure that every county is auditing one batch of [ballot-marking device] ballots and one batch of hand-marked paper ballots."

Since this morning, county audit teams have been busily working in pairs to count the number of ballots in those batches and then the number of votes for Raffensperger, Democrat Bee Nguyen and Libertarian Ted Metz.

The "risk limit" aspect of the risk-limiting audit is the largest chance an incorrect election result would not be corrected through the audit process. The margin of the 2020 presidential election was so close that the state opted to hand count every ballot to ensure the outcome was correct, but since the secretary of state's race had a much wider margin, fewer ballots need to be examined.

"So this year, since it's a little over a 9% margin in the contest that we're auditing, according to the early math we're only having to audit what will probably end up being about 5% to 7% of ballots statewide in order to achieve the risk limit," Evans said. "Which is much lower than what the law actually requires: We're setting it at 5%."

In Fulton County, about two dozen teams settled into work shortly after 9 a.m. as interim elections director Nadine Williams surveyed the convention hall and reflected upon the general election. 

"We have a very strong staff and we're very proud of our department, we did an awesome job" she said. "We still know that the runoff is upon us, plus we're doing this RLA today, so we're kind of doing two things at once. But we're going to be able to go ahead and make sure everything is efficient, everything is functioning."

It is important to note that the hand count of the batches may have slightly different totals than the machine counts of the ballot, but it is not evidence of impropriety.

"That's expected because human beings really suck at counting things computers don't," Sterling said. "So for all the conspiracy theorists out there when these don't match exactly, that's expected. That is not showing fraud. That is not showing anything. Just understand that on the front end."

Georgia's post-certification audit is also not designed to change the specific margins or results of the election, and is the final step before the state certifies the results of the midterms. The results of the audit, including copies of the tally sheets of ballots, should be online next week.