When the inputs disagree, the tie is broken toward the lower-risk reading. Always.

What ordering leaves open

A hierarchy settles precedence and not every disagreement is about precedence.

The authority stack answers a specific question: when two layers of different standing disagree, the higher one governs. The gate outranks structural permission, structural permission outranks tactical evidence, and so on down. That resolves a large class of conflicts cleanly. It does not resolve the cases where the disagreeing readings sit at the same level — the efficiency modifier pointing one way while the profit-factor modifier points the other, or acceleration supporting a promotion the structural drawdown read suppresses. Both are modifiers, neither outranks the other, and the ordering has nothing further to say. Something has to decide, and if nothing in the system does, the decision falls to the operator in the moment, which is the condition the whole panel exists to avoid.

Why downward

The rule follows from an asymmetry in what being wrong costs.

The direction is not arbitrary and it is not temperamental. Choosing the lower-risk reading when the evidence is genuinely ambiguous means that in the cases where the optimistic reading was correct, the account deployed less than it might have and gave up some upside — bounded, measurable, and recoverable from. In the cases where the pessimistic reading was correct, the account deployed less into a deteriorating condition, which is the outcome the entire governance layer is built to produce. The two errors are not equally priced, so a symmetric tie-break would be the wrong rule even if it felt more even-handed. The stated purpose says as much: the caps exist to prevent expectancy from being over-deployed during fragile conditions, and fragile conditions are precisely where readings tend to disagree.

The alternative

An unresolved conflict does not stay unresolved — it gets resolved by mood.

Every tie has to break somewhere. A panel that produced no answer when two modifiers disagreed would simply hand the case back to the operator, who would then resolve it using whatever considerations were available in the moment — how the last cycle went, how the current positions feel, how much the account is behind where it expected to be. Those considerations are not neutral and they do not average out, because they correlate: an operator is most inclined to prefer the optimistic reading after a losing stretch, which is exactly when the pessimistic one is more likely to be right. A fixed directional rule eliminates the category. The conflict is still real, the evidence is still ambiguous, and the answer is settled in advance and identically in a good month and a bad one.

What it does not authorise

The rule breaks ties; it does not make the panel a pessimism machine.

Two limits are worth stating because the rule is easy to over-extend. First, it applies to conflicts, not to readings — a set of modifiers that agree favourably is not a conflict, and the rule has no business suppressing a promotion that the evidence supports coherently. Applied indiscriminately it would collapse the tier to the bottom of the ladder permanently, which would make the whole modifier apparatus decorative. Second, it operates within the resolution, not on top of it. It is a tie-break inside the arithmetic, not a licence for the operator to prefer the lower reading whenever they are uncertain about something. Operator caution has its own sanctioned instrument — the manual tier cap — and routing it through a doctrine sentence instead would put a discretionary judgement somewhere the log cannot see it.

Reading a conflict

A disagreement is a finding, and the tie-break should not end the enquiry.

Because the rule resolves the deployment question immediately, there is a temptation to treat a conflict as handled once the tier is set. It is worth a second look, because disagreement between same-level readings is itself informative. Efficiency strong while profit quality is weak is not a random collision — it is the under-monetisation signature the contradiction engine names, and it points at exits rather than at entries. Acceleration positive while structural drawdown is expanding describes a system growing while its damage profile worsens. In each case the tie-break has correctly produced a conservative tier, and the conflict is separately worth carrying into the weekly review. The rule protects the cycle; the pattern of what keeps conflicting protects the quarter.

• Precedence first. Only when standing is equal does direction decide.
• The tie-break is downward because the two errors are not equally priced.
• It resolves the cycle; the recurring shape of the conflict is a separate finding.

The key idea

Determinacy under ambiguity is worth more than being right on average.

A tie-break rule cannot be correct in the sense of picking the better reading, because if one reading were reliably better the case would not be a tie. What it can be is fixed, so that the same evidence produces the same deployment regardless of when it arrives or who is reading it. That property is what allows the panel's output to be audited later — a reviewer can reconstruct why a cycle resolved as it did without needing to know anything about the operator's state of mind. Choosing a direction and applying it without exception buys determinacy at the cost of some forgone upside, and the governance layer makes that trade knowingly every time.