The equity peak only ever moves upward, and it never resets.

The mechanism

One rule in each direction, and they are not symmetric.

Upward, the peak follows equity immediately: a new closing high becomes the new anchor the moment it is realised, with no confirmation period and no requirement that it be held. Downward, it does nothing at all. Equity can fall for months and the anchor stays where it was, which means the measured drawdown grows and the account descends the gate ladder accordingly. There is no mechanism anywhere in the system for lowering the peak through trading — only a deliberate capital-movement adjustment can change the reference, and that is a documented act rather than an automatic one. The asymmetry is the whole design: the anchor is generous about recognising progress and immovable about forgetting it.

The consequence nobody likes

A give-back after a strong run reads as a real drawdown, because it is one.

The uncomfortable case is an account that runs up substantially and then returns part of the gain. Measured from the start of the period the account is well ahead; measured from the peak it is in a drawdown, possibly a deep one, and the gate compresses accordingly. Operators experience this as being punished for a good stretch, and the objection is understandable and wrong. The capital that was given back was real capital — it existed, it was the account's, and it is gone. Recovery from where equity now sits requires the same arithmetic as recovery from a loss of identical size that was never preceded by a gain. The gate is not scoring the period; it is locating the account, and the account is exactly as far below its high as the number says.

Why not a rolling anchor

A peak that forgets would relax the ladder during exactly the wrong periods.

The obvious alternative is to measure from a rolling high — the best equity of the last quarter, say — which would let old peaks expire and prevent a single exceptional high from governing indefinitely. It is worth seeing why that is worse. A rolling anchor decays fastest during long declines, because the old highs roll out of the window and the reference falls toward current equity. So a slow, extended erosion would progressively reduce its own measured drawdown, and the account would climb back up the gate ladder while continuing to lose money. The failure mode the duration doctrine already identifies — a slow bleed evading every depth alarm — would be built directly into the anchor rather than merely being hard to detect.

What resets it

The reference changes for capital movements and for nothing else.

There is exactly one legitimate reason to move the anchor other than making a new high, and it is not performance-related. A deposit or withdrawal changes equity without any trading having occurred, which corrupts the comparison in both directions — a withdrawal manufactures drawdown out of nothing, and a deposit erases genuine drawdown by lifting equity toward an anchor the trading did not earn. In those cases the reference has to be reconsidered deliberately, and the gate override clause exists precisely so a documented divergence can be recorded while the arithmetic is unrepresentative. What is not available is resetting the peak because the run that produced it feels like it belonged to a different regime, or because a new year has started, or because the drawdown has become inconvenient to operate under.

What the ratchet buys

An anchor that cannot be argued with is what makes the gate a lookup.

The design earns its discomfort in one specific way. Because the anchor has no discretion in it — one rule up, no rule down — the drawdown figure is the same number for anyone computing it from the same account history, at any time, with no interpretation available. That is what allows the gate to be a table lookup with zero parameters at the moment of decision, and what makes the capital state something the operator reads rather than assesses. Every softening that has ever been proposed for the anchor introduces a judgement, and a judgement at the anchor propagates into every gate, tier, pool and per-trade figure downstream. The ratchet is uncomfortable in exactly the periods when a softer rule would be most welcome, which is the evidence that it is doing its job.

• Instant upward, immovable downward — the asymmetry is the design, not an oversight.
• A rolling anchor would relax the ladder fastest during a long slow decline.
• Capital movements are the only legitimate reason to reconsider the reference.

The key idea

Measuring from your best moment is the only reference that does not drift.

Any other anchor — a rolling window, a period start, a moving average of equity — has a parameter in it, and a parameter is something that can be chosen after the fact by whoever finds the resulting number more comfortable. The all-time high has no parameter. It is a fact about the account's history that requires no decisions to compute and admits no versions. That property is worth more than the arguable unfairness of being measured against a peak made in favourable conditions, because the alternative is a capital-state reading that can be quietly renegotiated at precisely the moments it most needs to hold.