Three good grades in a row feel like a licence. They grant nothing at all.

The loop

Grade, relief, confidence, size — and none of the arrows in that sequence are authorised.

The sequence is familiar enough to be uncomfortable. A green grade lands and is read as confirmation that the system is healthy, which it is. Confirmation reduces the felt risk of the next deployment. Reduced felt risk makes the upper end of whatever range is available more attractive, and the range has some discretion in it — which pairs, how many trades, whether to take the marginal setup. Nothing here is an explicit decision to increase risk, and by the third green week the deployment has quietly moved. Meanwhile the gate has not moved, because the gate reads drawdown from peak and expectancy grades are not an input to it.

Why the grade cannot carry authority

Expectancy and capital state answer different questions, and only one of them constrains size.

The grade describes whether the process is worth running. The gate describes what the account can currently absorb. Those are independent, and the independence is not an oversight — they can and frequently do disagree. A book can hold healthy expectancy while sitting in a deep drawdown, because expectancy is a per-trade property and drawdown is a path property. If a strong grade could lift the tier ceiling, then exactly the situation that most requires containment would be the situation in which containment gets relaxed. The architecture forbids the link so that the disagreement resolves in the direction of the capital state every time.

The reverse is also true

A red grade does not directly revoke anything either, and that surprises people more.

The boundary is symmetric, which is the part operators tend to miss. A red week does not itself cut the tier or shrink the pool; the gate reads drawdown and will do whatever drawdown warrants, which may be nothing at all if the account is still near its peak. What red does is instruct the operator's posture under the risk doctrine and feed the review that asks whether something structural is happening. If the grade could revoke authority directly, the system would have two independent brakes reading two different signals, and the resulting deployment figure would be the product of whichever fired hardest rather than the output of a single stated authority.

• Sizing authority has one source, and expectancy grades are not it.
• Red informs posture and triggers diagnosis; it does not compute a pool.
• Two brakes reading different signals produce a deployment nobody can explain.

What the grade is genuinely for

It answers whether the system is still worth deploying, which is a prior question rather than a sizing one.

None of this makes the grade weak. Deciding whether the process remains worth running at all is the more fundamental question, and nothing else in the stack asks it — the gate would happily authorise full deployment of a strategy whose edge had disappeared, because drawdown says nothing about expectancy until the losses have already arrived. The scorecard is the early instrument, and its authority is real in its own domain: it can tell an operator that continuing to deploy this process is no longer justified by the evidence, which is a considerably stronger statement than any adjustment to a tier.

How the boundary is held

The defence is that the deployment figure is computed somewhere else and simply read.

Boundaries maintained by remembering to respect them do not survive good months. This one holds because the authorised pool and the per-trade figure are produced by a different surface, from drawdown and cycle state, and arrive at the deployment moment as numbers rather than as suggestions. There is no field in which a strong grade can be entered to adjust them. That structural gap is what makes the boundary durable, and it is the general pattern throughout the system: the questions with defensible mechanical answers are computed and read, so that confidence has nowhere to attach itself.

The key idea

Good news is the most common route to unauthorised risk, precisely because nobody guards against it.

Operators anticipate that bad stretches will tempt them into revenge sizing, and most build defences against it. Almost nobody builds a defence against a good stretch, because the good stretch does not feel like a risk event — it feels like the system working, which it is. The separation of grading from sizing is that defence. It says that a healthy month is welcome, informative, and entirely irrelevant to how much of the account the next cycle may commit.