← Entropy Atlas

Uniform or fair

Randomized audits & sampling

A sample can make a large population inspectable when selection is genuinely representative.

Evaluation card

What “plenty” means here.

Required property
A documented sampling frame, intended probabilities, and a draw that cannot be quietly steered.
If it fails
Missing populations or biased selection makes the audit confidently answer the wrong question.
Relevant attacker
An audited party predicting which records will be selected, plus unintentional sampling bias.
The Plenty Line
Protect the seed until the population is fixed, publish the method, and retain enough evidence to reproduce the sample.
Why more is not automatically better
A stronger generator cannot repair a biased source list or a sample too small for the inference.

Where it appears

Recognize the pattern.

  • 01risk-limiting election audits
  • 02expense audits
  • 03quality-control sampling

The Plentropy rule

Protect the requirement, then stop.

Protect the seed until the population is fixed, publish the method, and retain enough evidence to reproduce the sample.
Once that claim has comfortable evidence and margin, improve the next limiting factor instead of worshipping a larger entropy number.