It does not classify the market. It classifies the machine.

The confusion

Two things called regime, sharing nothing but the word.

A market-regime detector reads price: volatility state, trend persistence, range compression, correlation structure. Its inputs are the market's, its output describes external conditions, and its use is to decide what kind of environment you are trading into. This engine reads none of that. Its six inputs are monthly gate and brake metrics — pressure, structural health, gate volatility, transition count, persistence, and compression — all of which are properties of the governance system's own behaviour over the month. A month in which the gate transitioned frequently and held its states poorly produces an Instability reading regardless of what any instrument did. The output describes the machine's condition, and the market appears in it only indirectly, as one of several things that might explain why the machine behaved as it did.

Why the distinction is load-bearing

Misreading the label produces confident conclusions about nothing.

An operator who reads Compression as a statement about market conditions will start adjusting instrument selection, timeframe, or setup criteria on the strength of a reading that contains no information about any of those things. Worse, the adjustment will sometimes appear to work, because system stress and difficult markets do frequently coincide — which converts a category error into a reinforced habit. The engine's own inputs make the error easy to spot once named: a Gate Transition Count is a count of how often the system's own capital state changed, and no amount of examining it will reveal anything about price. If the reading were about the market, it would have to differ by instrument, and it does not.

The relationship that does exist

Difficult conditions can cause system stress, but so can behaviour.

There is a real link, and it runs in one direction with substantial noise. A genuinely hostile month can generate drawdown, spike clustering, and gate compression, which will show up as elevated pressure and reduced structural health. But identical readings can be produced by a system trading a normal month badly — poor selectivity, forced branch selection, size taken at the wrong point in the gate ladder. The engine measures the strain and is constitutionally unable to attribute it. That attribution is what the monthly review is for, and it requires other evidence entirely: the checklist record, the branch-level diagnostics, the execution efficiency lab. The regime label opens that enquiry rather than concluding it.

• The same regime label can describe a hard month traded well or an ordinary month traded poorly.
• The engine has no input that could distinguish the two, and does not pretend to.
• Attribution is a review task supported by other modules, not a property of the classification.

What the naming buys

A governance readout deserves the strongest word available.

Given the confusion, the term could have been avoided — system state, monthly condition, governance status. Regime was kept because it carries the right connotation for what the label is meant to do: describe a persistent condition that shapes behaviour until it changes, rather than a momentary reading. A regime is not a measurement; it is a state you are operating within, and that is exactly the posture the engine is asking the operator to adopt toward their own machine. The cost of the word is a confusion that has to be named once. The benefit is that the label carries its intended weight without further explanation.

The key idea

The system is the subject, not the observer.

Nearly every analytical tool a trader touches points outward at the market. This one points back, and the reversal is the whole reason it exists — a machine that measures everything except itself will discover its own deterioration from its equity curve, which is the most expensive possible place to find out.