Each gate row has a written justification, and that is what makes the ladder revisable.

Two layers

The table holds the values; the notes hold the claims the values encode.

A calibration table without commentary presents an operator with a set of numbers and exactly two options: use them or do not. There is no third position, because disagreeing with a number requires knowing what it was trying to express, and a bare figure expresses nothing beyond itself. Attaching a reason to each row changes the object. Recovery compressing its upper tiers first is no longer a fact about the table, it is a claim — that re-damage during healing is the most expensive failure mode — and claims can be examined, tested against the record, and found to hold or not. The values become the conclusion of an argument that is available for inspection rather than the argument itself.

Why recovery compresses from the top

The most instructive note is the one that could plausibly have gone the other way.

The recovery row is worth dwelling on because its shape is not obvious. An account seven to thirteen percent below its peak is not in serious trouble, and a reasonable designer might argue for preserving the upper tiers there so the drawdown can be recovered efficiently — the edge is presumably intact, and smaller deployment lengthens the climb. The note takes the opposite position, and states its reason: re-damage during healing is the costliest failure. The underlying arithmetic is that recovery from a deeper drawdown is disproportionately harder, so an expansion that fails while the account is already down converts a manageable situation into a structural one. Reasonable people could disagree, and because the reason is written the disagreement has something to be about.

What a written reason permits

A calibration change becomes a specific, bounded revision.

The practical dividend arrives when the ladder needs adjusting, which any calibrated system eventually does. Without stated reasons a change is a negotiation over numbers, and the usual outcome is that the whole table gets nudged in whichever direction recent experience suggests. With them, the process is narrower: identify which claim the record contradicts, revise the row that encodes that claim, and leave every other row untouched because their claims were not the ones tested. That containment is what stops calibration from becoming drift. It also means a change carries an explanation into the future — the next reader inherits both the new number and the finding that produced it, rather than a table that has quietly moved.

The alternative

An unexplained cap is one the operator will eventually override.

There is a durable observation running through this system: an operator complies with a constraint they can trace and quietly disregards one they cannot. A tier ceiling that exists because the table says so has no defence at the moment it becomes inconvenient, and the moment always arrives — a strong stretch under a compressed gate, a setup that looks like the best of the quarter, a milestone slipping away. A ceiling that exists because re-damage during healing is the costliest failure is arguing back, with a reason the operator can weigh against their own. It may still lose that argument occasionally. It will lose far fewer of them than a number with nothing behind it.

The auditable property

Written reasons are what let the ladder be tested rather than merely trusted.

The claims are also, in principle, checkable against the system's own record. Whether re-damage during recovery is genuinely more costly than the equivalent damage from a peak is an empirical question, answerable from drawdown history and the benchmark's modelled paths. Whether collapsing toward the bottom in floor states produces better survival than a shallower compression is likewise testable in simulation. None of that is available for a table whose values have no stated purpose, because there is nothing to test — the numbers do not predict anything. Writing the reason down converts each row into a small hypothesis about how the account behaves, which is the only form in which a calibration can improve rather than merely change.

• A bare number can be accepted or rejected; a justified one can be examined.
• Revision stays contained to the row whose claim the evidence contradicts.
• Each note is a testable hypothesis — untested tables cannot improve, only drift.

The key idea

The reasons are the part of the ladder that survives a rebuild.

If the workbook were lost tomorrow, the pool values would be tedious to reconstruct and the reasoning behind them would be nearly impossible to recover from memory — which of them were considered, what each was protecting against, why recovery compresses downward from the top rather than upward from the bottom. The construction notes are the durable artefact and the numbers are their current expression. That ordering is worth keeping in view, because it explains why the notes are not an accessory to the table: a system that records only its conclusions has to rediscover its reasoning every time it changes, and a system that records its reasoning can rebuild its conclusions whenever it needs to.