Ten regimes, forty sub-labels: naming a month precisely.

The ten states

Four groups, from expansion to crisis.

The regimes sort into four bands. Positive: Expansion and Expansion (Developing) — the states where alpha compounds and where it's beginning to. Neutral: Balanced, Transitional, and Compression — quiet operation, rotation, and constraint respectively. Warning: Fragile Recovery and Instability — the system rebuilding unsteadily, or oscillating without holding. Negative: Degradation, Drawdown Control, and Critical State — structural erosion, active containment, and breakdown. The ladder is deliberately asymmetric toward the downside, with five of ten labels describing degrees of trouble, because the distinctions that matter operationally are mostly distinctions between kinds of difficulty.

How the sub-label is chosen

The most extreme relevant metric names the driver.

Sub-label assignment runs after the main regime is fixed, and it's a priority search rather than a second score. The engine looks only at the metrics relevant to that regime, finds which is most extreme, and assigns the corresponding descriptor by priority order. Instability, for example, checks volatility, then transitions, then persistence, then pressure — severe volatility yields High Volatility, else severe transitions yields High Transition, and so on down the chain. The result is a label that names the month's dominant cause: two Instability months with different drivers demand different diagnostic attention, and 'Instability — High Transition' points at churn while 'Instability — Pressure-Driven' points at capital stress.

The one-label rule

No stacking, no hedging, no partial classifications.

Two of the engine's non-negotiables enforce singularity: one main regime per month, one sub-label per month. The temptation to hedge — 'mostly Balanced but with Compression characteristics' — is exactly what the rule forbids, because a hedged classification is uncomparable with every other month and defeats the compression the engine exists to provide. If the month genuinely sits between states, the composite ladder has already decided which side it falls on, and the sub-label supplies the nuance: Balanced — Mixed says the same thing as the hedge, in a form that can be counted, sequenced, and compared across a year of records. Sub-labels are descriptors, not separate regimes; they refine interpretation without changing the state.

• A month's final label is one string: main regime, em-dash, sub-label. That string is the record.
• The QA layer validates that the assigned sub-label actually belongs to the assigned regime — invalid pairs are flagged, not tolerated.
• Regime sequence across months is where the real signal lives: three Compression months running says more than any single classification.

What it's for

Posture and review priority — advisory, never authority.

The regime feeds the Call to Action layer, which maps regime and sub-label to recommended posture: where to focus the month's review, what to be cautious about, what the system's current character suggests about aggression. And the boundary is stated plainly in the design rules — the Call to Action sheet is advisory. It informs posture and review, but does not override capital-state authority. A Critical State classification doesn't compress deployment; the gate ladder already did that on drawdown arithmetic. An Expansion classification doesn't authorize anything at all. The regime tells you what kind of month you're in so the review asks the right questions, and that's the entirety of its remit.

The key idea

A precise name is worth more than a precise number.

'Composite +3, pressure 41, persistence 68' is data. 'Transitional — Recovery-Driven' is a month you can reason about, compare to last November, and hand to a review with the right questions already implied. The ten-by-four ontology exists to make that translation reliable: fixed states, fixed descriptors, one of each, every month. It's the same compression the gate ladder performs on drawdown and the scorecard performs on EV — turning telemetry into a state an operator can actually hold in their head.