rlauxe

OneAudit version 4

last changed 9/11/2025

Strategies

We try different betting “strategies”:

• bet99: Set eta0 = reportedMean, always bet 99% of the maximum allowed bet.
• eta0Eps: Set eta0 = upper*(1 - eps), use shrinkTrunk for betting strategy.
• optimalBet: Use betting mart with optimal_comparison betting from SHANGRLA.

Here we compare these 3 strategies along with polling and clca audits, when there are no errors, as a function of margin. The OneAudit has CVRs for 95% of its cards, and 5% are in a single pool without CVRS, but with Card Style Data.

• The bet99 and optimalBet are mostly identical and do best when there are no errors.

Here we compare the same audit types when there are errors, at a fixed 4% margin, as a function of fuzzPct:

• The optimalBet strategy does better even than CLCL once the fuzzPct are bigger than .005.

Compare Audit Types

We will use the optimalBet strategy for OneAudit, and compare Polling, Comparison (CLCA) and OneAudit (with 50%, 75%, 83%, and 90% of cards having CVRs).

When there are no errors in the CVRs, as a function of margin:

• OneAudit looks ok as long as the % pooled data is not too high, and the margin is not too low.

When there are errors (parameterized by fuzzPct, the percent of ballots randomly changed), for fixed margin of 2% and 4%:

• The spread among the OneAudit-cvrPercent audits follow the expectation that higher cvr percents look more like CLCA.
• OneAudit results have similar sensitivities to errors as CLCA.
• IRV (Raire) audits are less likely to have their outcomes altered due to random changes in the ballots.
• Polling audit sample sizes are all but impervious to errors.