Operator brief · 09

Structural diagnostics: seeing decay before the equity curve does.

The key idea

The question

Is this noise, or is the machine changing?

Every metric in MARS — EV, drawdown, RAER, RAPF, acceleration — gets the same structural interrogation: its rolling value, its stability, its drift direction, and how many standard deviations the current reading sits from its own history. A −3R week inside a stable distribution is weather. A slow six-week EV drift with tightening stability bands is climate — and climate change in a trading system demands intervention before the equity curve confesses. Distinguishing the two is the single hardest judgement in running a system, and it is the one humans are worst at making unaided.

FigureDrift shows up before the curve does

Schematic. Rolling expectancy declines steadily from window three; the equity curve keeps climbing until window seven.

Branch-level x-ray

Aggregate health can hide a dying branch.

System-level numbers average away the truth. One branch can deteriorate for months while a strong branch carries the blended figures — the operator sees green and learns the problem only when the strong branch has an ordinary cold streak. The SDE diagnoses each branch separately, so the response can be surgical: repair or retire the decaying branch, leave the healthy ones untouched. Blended health is an average of things that do not average meaningfully.

What a z-score is doing here

It converts ‘this feels wrong’ into a number with a threshold.

Every metric is scored against its own history rather than an external standard, which is what makes the diagnosis portable across very different systems. A reading two standard deviations below a metric's own rolling mean is unusual for that system, whether the metric is expectancy, efficiency, or acceleration — no universal benchmark is required. This matters because the alternative is comparison against someone else's numbers, which is where most retail self-assessment goes wrong: a 45% hit rate is excellent in one design and catastrophic in another, and only the system's own distribution can say which.

Drift is a direction, not an event

The signal is the slope of the deterioration, not any single reading.

A metric that falls sharply for one window is usually variance. The same metric declining modestly across six consecutive windows, while its stability band narrows, is a system changing shape — and it will typically produce no dramatic reading at any point along the way. This is precisely why the equity curve confesses late: it aggregates, and aggregation smooths exactly the gradual signal that matters most. The engine watches direction and persistence rather than magnitude, which is what allows a diagnosis while the numbers still look acceptable.

Diagnosis before prescription

The engine names the problem. It does not choose the repair.

A structural finding is deliberately short of a recommendation. Detecting that a branch's expectancy has drifted downward across six windows with a narrowing stability band is a diagnosis; deciding whether to widen its stops, tighten its selection criteria, demote its tier, or retire it entirely is an operator judgement that depends on things the engine cannot see. This boundary is the same one the regime layer observes and it exists for the same reason: an automated system that both detected problems and prescribed fixes would accumulate authority no model has earned. What the diagnostics buy is time and specificity — knowing which branch, which metric, and how long it has been happening. What the operator supplies is the decision about what that means.

The payoff

Early detection is cheap. Late detection is a drawdown.

Structural decay caught at drift stage costs a review and maybe a branch demotion. The same decay caught at the equity curve costs a gate compression, a recovery climb, and weeks of reduced authority. The engine exists to move the detection point forward — from pain to signal — which is the difference between managing a system and being surprised by one.

Connected inside MARS

Structural Diagnostic Engine →

Structural Health →

Contradiction Engine →

Every brief documents the same shipped system.

The complete MARS package — eleven workbooks, three TradingView indicators, the full manual library — $497.

Get MARS Pro — $497 Explore the Platform