Reply #26: Think of it as a calibration test

[mgr]

The logic of Ohio's recount procedure, which probably assumes that there is no fraud, but faulty counting, so you would test the machines. So the requirement is first, run a test deck through to see that everything is running properly, e.g. that it can read and tally. Then, manually count a random 3% of the ballots, run the same ballots through the machine. If there is a discrepancy, it would follow that there is a discrepancy throughout the rest of the ballots, so you have to manually recount those. If there is no discrepancy, then run the rest of the ballots through the machine for the recount.

With Fraud, the entire prescription should go out the window. Now you are dealing with possibly a cluster effect, or a programmed miscounting in the tabulators. With the first, you are not likely to sample a cluster of stuffed or altered ballots, and the second would be undetectable if you have an algorithm that does not kick in until a critical number of ballots are read.


Mike