How to Prove a Draw Was Fair Afterwards

The frozen entrant list

The single most important piece of evidence is a published list of entrants, timestamped before the draw. Almost every genuine dispute about a draw is about the list rather than the mechanism — somebody entered late, somebody entered twice, somebody who should have been in was not.

Freezing means publishing it somewhere you cannot silently edit: a post with a visible timestamp, a message to a group, a comment on the announcement. The requirement is not cryptographic; it is that a third party can see the list existed in that form before the result was known.

This also protects you. A published list means the person claiming they entered can be answered by pointing at it rather than by an argument about your records.

A stated method, in advance

Say how you will draw before you draw. "I will paste the list into this tool with this seed and take the first name" is a complete method; "I'll pick a winner at random" is not, because it leaves the choice of mechanism open until after you know the entrants.

The reason this matters is subtle and important: an unstated method lets you choose, consciously or otherwise, among several plausible methods until one gives an outcome you like. Stating it removes that freedom and is therefore evidence about your intent, not just about your process.

It also settles the tie-break question in advance. What happens if the winner does not respond, or is ineligible, or turns out to be your cousin, are all questions with obvious answers before the draw and contested ones afterwards.

• Name the tool or method you will use.
• State how many winners and in what order.
• State the no-response rule and the deadline.
• State any eligibility rules — before you know who won.

A reproducible result

The strongest available evidence is a result anyone can recompute. Publish a seed alongside the entrant list, run a seeded draw, and any reader can repeat it and confirm the winner without trusting you at all.

There is one requirement that makes or breaks this: the seed has to be committed to before the draw. A seed published afterwards proves nothing, because you could have tried two hundred seeds and published the one that produced the outcome you wanted.

The full version is commit-and-reveal — publish a hash of the seed beforehand and the seed itself afterwards — which removes the seed-shopping attack while keeping the winner unknown until the draw. It is more ceremony than most giveaways need and it is the only method that survives a genuinely determined sceptic.

Recording the draw

A screen recording or a live stream is weaker evidence than it feels, and it is worth understanding why before relying on it. It shows that a draw happened and produced a result; it does not show that it was the first draw, or that the list on screen was the list that was frozen.

It is still worth doing, because it is cheap and it addresses the intuitive version of the doubt. Most people are not modelling an adversary — they want to see that something happened rather than being told about it.

The way to make a recording genuinely strong is to include the verifiable elements in it: show the published list, show the seed, run the draw. At that point the recording is evidence of the sequence rather than just of the outcome.

What does not help

Describing the randomness in technical terms does not help. Telling people you used a cryptographically secure generator addresses a doubt nobody has — the suspicion is about you, not about the quality of the entropy.

Re-running the draw because the first result looked odd is the single most damaging thing you can do, and it destroys every other piece of evidence you assembled. If the method was stated in advance, the first result is the result.

Announcing the winner without the list is the common failure, and it is a complete failure. A name with no context is exactly as convincing as no announcement at all, which is why the frozen list is first on this page rather than last.

• Technical descriptions of the generator: addresses the wrong doubt.
• Re-running after an unexpected result: destroys everything else.
• Announcing a winner with no published list: no evidence at all.
• Deleting a losing entrant's comment: turns a question into a story.

Proportionate effort

None of this scales down gracefully, so it is worth saying what is proportionate. For a draw among friends, a stated method and a screenshot is plenty. Nobody is going to audit it, and treating it as an adversarial problem is odd.

For a public giveaway with a prize of real value, the frozen list and the stated method are the minimum, and a published seed costs almost nothing on top. That combination answers every reasonable question.

For anything where money changes hands or a regulator might be interested, commit-and-reveal plus a retained record is proportionate, and the general expectations page covers what else tends to be expected.

Frequently asked questions

Tools that use this

Related guides