Weak audits that can miss a flipped result
Even when the law requires a check of the paper ballots, a weak method — like always recounting a tiny fixed percentage — can easily miss a wrong result in a close race. A smarter method checks however many ballots it takes to be sure.
Explainer video coming soon
A 60-second, plain-language walkthrough of this gap.
What actually happened
Statistician Philip Stark introduced the 'risk-limiting audit' (RLA): instead of a fixed percentage, you hand-count enough randomly-chosen paper ballots to reach a defined statistical confidence that the reported winner really won — checking more ballots when the race is close, fewer when it's a blowout. A flat 1–3% recount has no such guarantee.
The same small handful every time
Picture checking a huge jar of marbles for a few wrong-colored ones by always pulling out the same small handful — no matter how big the jar. If only a few are wrong, you'll probably miss them.
That's a fixed-percentage recount (say, 1–3% of ballots). It looks like a check, but in a close race it has no real guarantee of catching a result that was flipped by a small margin.
A method that actually guarantees something
A risk-limiting audit counts enough randomly-chosen paper ballots to reach a defined statistical confidence that the reported winner truly won — checking more ballots when the race is close and fewer when it's a landslide. That's the difference between "we looked at some" and "we're confident it's right."
How TrustVoting closes it
Voter-verified paper plus ballot-level records are built to support risk-limiting audits, so the strength of the check scales to the closeness of the race.
How TrustVoting closes it
TrustVoting's voter-verified paper and ballot-level records are built to support risk-limiting audits, so the check scales to the closeness of the race and reaches real statistical confidence.