Comment Re:OT: How to build an trustable voting machine (Score 1) 188
Auditing an election of, say, 3M voters where one candidate allegedly beat the other 50.5% to 46.5% to 3% for minor candidates need only determine that there's less than a 5% chance that the true election result had the winning candidate with 50%+1 votes to avoid a runoff. With a paper ballot satisfying #1 and #2 and generally accepted statistical analysis, this won't require a recount of nearly the entire pool of votes, only a random sample from each ballot box sufficiently large to rule out the need for a runoff.
If on the other hand the alleged winning amount was exactly 1,500,001 out of 3M votes, or if it was 1,499,499 and the winner wanted a recount to avoid a runoff, a full manual recount would likely be necessary.
This only applies if you have a voting system where it is sufficient to have the most votes. (Like in the US) In Germany for example, the percentage of every party is important, as it determines the number of seats in the parliament.