Six families of EV truth: how the lab decomposes one number into causes.

The decomposition premise

'EV is weak' is a symptom with at least six possible diseases.

Two systems can print identical mediocre expectancy for completely different reasons: one has a thin edge honestly measured, another has a strong edge taxed by giveback, a third has a real edge that its operator degrades after losses, a fourth has an edge that only exists in one session being averaged against five where it doesn't. Treating those four identically — 'EV is weak, trade less' — fixes at most one of them. The lab's architecture is the answer: a 64-metric engine computes every diagnostic against the all-blended baseline, and each metric belongs to a family that names a cause. The workflow doctrine follows directly: read EV through statistical truth first, then diagnose the family, and only then form an action.

The six families

Each family answers one question and routes to one tab.

Statistical Truth asks whether the edge is credible at all: realized EV, lower bound, probability EV is above zero, trimmed and median EV, outlier reliance. Mechanical Leakage asks whether execution is taxing it: exit drag, capture ratio, fee-adjusted EV, breakeven-rule cost. Behavioral asks whether the operator is: post-loss EV, adherence premium, emotional drag, forced-entry effects. Growth & Capital Efficiency asks whether it scales: risk-weighted versus equal-weighted EV, sizing alpha, Kelly reference, dollar throughput. Conditional Context asks where it lives: branch, session, timeframe, ATR state, pair, stop-width, gate-state slices. And Pip Translation asks whether the R accounting is real: native-unit reconciliation, pip efficiency, divergence counts. Six questions, six tabs, one engine underneath.

The routing habit

Dashboard last — the headline is where reading ends, not begins.

The manual's workflow inverts the natural instinct: do not start on the Dashboard and stop there. The clean sequence runs from source truth outward — refresh the Source_Bridge, set the control-panel lens, confirm the scope is valid against the sample gate, read the Structural Truth Snapshot, inspect the full engine, and only then drill into whichever family tab the weakness routes to. The Dashboard's executive verdict comes last, as compression of a diagnosis already made. The reason is the same one behind every reading protocol in MARS: a headline consumed first becomes a conclusion, and every subsequent tab gets read as confirmation. A headline consumed last is a summary, which is all it ever claimed to be.

The authority boundary

The lab diagnoses opportunity; it cannot grant permission.

The manual draws the line in its own words: if the workbook shows strong EV but the Gate/Brake state is Floor, Deep-Floor, Ground-Floor, or System Lock, the gate wins. The lab can influence review posture; it cannot expand past gate or throttle constraints — and its scenario outputs and what-if insights belong in the experimentation layer until tested, approved, and promoted. This boundary is what makes the lab's aggression safe: precisely because nothing it finds can touch capital directly, it's free to slice ruthlessly, hunt pockets, and stress-test hostile cases. The strongest filtered sample in the workbook is still evidence for a review, never a key to a gate.

• Strong lens + restricted gate = the gate wins, every time, with no appeal to sample quality.
• The correct workflow is not chasing the highest EV pocket — it's identifying durable expectancy and removing leaks.
• A filtered sample that looks too strong is a prompt to check the sample, not to raise deployment.

The key idea

Causal decomposition is what turns analytics into repairs.

The lab's value isn't precision for its own sake — it's that a cause-shaped diagnosis produces an action-shaped conclusion. 'EV weak' suggests nothing; 'EV credible but capture ratio is leaking a third of available MFE' names a repair; 'EV strong except post-loss' names a different one; 'EV only exists in London session' names a third. Six families, one routing question, and the discipline of never acting on the headline: that's the whole method, and it's why the lab sits above the scorecard in the diagnostic stack.