The bands' authority comes from when they were made.

The problem with mid-drawdown reasoning

Every input degrades at once, and none of them announce it.

The characteristic feature of a bad stretch is not that the operator reasons poorly but that the reasoning feels normal. Losses are recent and vivid; the prior recovery is abstract and distant. The sample being generalised from is the last three weeks. Each individual loss has an explanation, so the aggregate seems explicable. And any proposed rule change arrives with the sincere conviction that it addresses a structural problem rather than a painful one. Nothing in that state flags itself as unreliable, which is why the answer cannot be better in-the-moment judgement. It has to be an artifact produced elsewhere.

What the timestamp guarantees

Not correctness — independence.

The bands are not authoritative because they are right about the future; they are estimates with error bands like everything else in the stack. Their authority is narrower and more useful: whatever they say, they did not say it because of how the current drawdown feels. They were computed before it started, from a sample that does not include it, by a process the operator's present state could not influence. In a situation where every other input has been contaminated by the same source, an uncontaminated one does not need to be excellent to be the best available. It needs only to be independent, and the date is the proof.

The move it blocks

This time is different is unfalsifiable in the moment and checkable afterwards.

The specific argument the artifact defeats is the claim that the current drawdown belongs to a different category than the modelled ones — that conditions have changed, the edge has shifted, the distribution no longer applies. That claim may occasionally be true. What makes it dangerous is that it is always available, feels identical whether true or false, and conveniently licenses whatever the operator already wanted to do. The bands do not refute it. They relocate it: the claim becomes a hypothesis with an evidence threshold and a proper route, rather than a premise that can be acted on immediately from inside the state that generated it.

The condition on the guarantee

Revision must happen in calm or the property is destroyed.

The timestamp only protects while the artifact stays untouched during the period it is measuring. A band regenerated mid-drawdown — for defensible-sounding reasons, with real evidence — is no longer independent of the drawdown, and the operator has quietly converted their ruler into a reflection. This is why the refresh standard fires on approved model changes and on a fixed cadence, never in response to a live period that felt unrepresentative. The rule looks like bureaucracy and is the mechanism that keeps the artifact's one valuable property intact.

The generalisation

The same logic runs through the whole rail.

Pre-commitment authored in calm and consulted under pressure is not a percentile-band trick; it is the shape of the entire system. The gate ladder decides how much to de-risk before any drawdown exists. The checklist decides what blocks a trade before the trade is tempting. The response map fixes what a placement authorises before the placement occurs. In each case the artifact's value comes from the same source — it was written by a version of the operator who had nothing at stake in the answer, and the current version is executing rather than deciding.

• The bands do not need to be right to be useful. They need to be independent.
• Regeneration during the period being measured destroys the property that made them worth consulting.
• Consult and execute; revise later, on cadence, in calm.

The key idea

Build the instrument before you need it, then do not touch it while you do.

A definition of normal pain is worth exactly as much as its independence from the pain. That is why the timestamp matters more than the sophistication of the model behind it, and why the discipline around regeneration is stricter than it looks. The bands are a message from a calmer operator to a compromised one, and the only thing that can invalidate the message is the compromised operator editing it. Read it, execute it, and leave the revision to the version of yourself who is not currently underwater.