Structural Diagnostic Engine · The Entity Universe
Four primitives, three composites, one arithmetic rule.
Entities
7
Primitive branches
4
Composites
3
Layer 01 — The arithmetic rule
Never average a ratio.
Composite efficiency is composite net PnL divided by composite risk deployed. It is not the average of the branch ratios, and the difference is not cosmetic. Averaging ratios gives a light week the same weight as a heavy one, so a branch that deployed almost nothing and returned a flattering percentage can drag the composite upward while contributing nothing to the account. The same applies across time: cumulative efficiency is the sum of PnL over the sum of risk, never the average of the weekly figures.
- Cumulative efficiency: sum of net PnL divided by sum of risk deployed.
- Composite efficiency: composite PnL divided by composite risk, not a blend of branch ratios.
- Averaging weekly ratios lets low-risk weeks distort the capital-weighted truth.
Layer 02 — Drawdown is different
The cumulative column is a running worst, not a running mean.
Drawdown is path-dependent, so its cumulative anchor is the running minimum since inception rather than an average of the weekly readings. Averaging drawdown produces a number that describes neither the typical week nor the worst damage, and it will understate exactly the event the metric exists to record. Current drawdown remains the live gate input; the cumulative column is structural damage awareness, and the two are read for different purposes.
How MARS uses this
The Weekly Scorecard converts branch hit probabilities into EV and tags each week by threshold. The rolling line is what MARS actually trusts: single-week readings are treated as noise until the rolling window confirms direction. RED weeks trigger doctrine - reduced aggression and review - rather than negotiation.
How it benefits you
You learn to distinguish a bad week from a broken system. Variance stops triggering rule changes, and genuine expectancy decay gets caught while it is still one line on a chart instead of a hole in the account. The GREEN/YELLOW/RED language also makes weekly review fast and unambiguous.
Weekly EV tagged GREEN / YELLOW / RED against expectancy thresholds, with the rolling-EV line separating persistent edge from one lucky week.
Layer 03 — Why Overflow is separated
The composite that excludes it is the one that reads the plan.
Overflow represents supplemental opportunity flow, and its volume and cost profile can distort blended expectancy in ways that say more about opportunism than about the core system. Keeping a composite that excludes it means the operator can always ask what the plan is doing independently of what the extra flow is doing, which is the difference between diagnosing the strategy and diagnosing the calendar.
Layer 04 — Consistency as a checklist item
All seven entities must exist consistently across every pipeline tab.
The quality checklist opens with this requirement, followed by confirming that value and cumulative columns are paired correctly for each entity. A missing entity in one pipeline is not a gap that announces itself: the cockpit still routes, the lookups still resolve, and the readout still reads confidently, having quietly omitted one branch from a composite that claims to include it.
Reference
The seven entities
| Entity | Type | What it isolates |
|---|---|---|
| Normal | Primitive | Static and normal-regime execution; the baseline quality and stable-edge read |
| Trend Partial | Primitive | Trend execution with partial monetization; steady trend participation |
| Trend No-Partial | Primitive | Fat-tail and trailing behaviour; higher variance, higher acceleration potential |
| Overflow | Primitive | Supplemental higher-timeframe and overflow flow; volume and costs can distort expectancy |
| Normal + Trend | Composite | Core plan behaviour, deliberately excluding Overflow |
| Trend Blend | Composite | The trend engine on its own, both trend branches together |
| All Blended | Composite | The total system view, all four primitives |
Edge cases & failure modes
Where it can mislead
- !A composite recomputed from averaged branch ratios will diverge from the capital-weighted truth, silently and in the flattering direction.
- !Class labels that drift between tabs break the lookup keys, and a broken key produces a wrong meaning rather than an error.
The governing idea
Connected inside MARS
This module doesn't work alone.
Go deeper
Operator briefs on this territory.
Deep dive — 01
Seven entities: what the SDE is actually allowed to diagnose.
Four primitive branches, three composite blends. A closed universe, because the two kinds answer different questions.
Read the full brief →
Deep dive — 02
Never average a ratio: the rule that keeps composites honest.
Composite RAER is composite PnL over composite risk. Average the branch ratios instead and the number inverts.
Read the full brief →
Deep dive — 03
Three composites, three questions — and why one would not do.
All Blended contains everything, which is exactly why it explains nothing. The other two are defined by what they leave out.
Read the full brief →
Every module ships in the complete MARS package.
One price. Eleven workbooks, three TradingView indicators, and the full manual library — $497.