Contradiction Engine · Why Contradictions Matter

A single metric cannot be cross-examined.

Classes per axis

Matrix cells

Baseline risk rail

5.5%

The key idea

In practice

Each closed trade logs its worst adverse excursion and best favorable excursion. The lab plots them into quadrants and pairs the picture with capture efficiency, giveback, and fee drag - then the EV Driver Diagnostics rank which repair (entry location, coefficient, trail logic, cost avoidance) would move net expectancy most.

Every trade plotted by pain taken (MAE) versus opportunity generated (MFE). The four quadrants diagnose entry precision and exit management at a glance.

Two lenses, one capital

Efficiency asks what the risk returned. Profit quality asks how the profit was made.

Efficiency is net result over risk deployed: a capital-conversion ratio. Profit quality is a profit factor read against a risk baseline: a durability ratio. A system can convert capital cleanly while producing thin, fragile profits, and it can produce a robust profit factor while wasting deployment to get there. Neither reading contains the other, which is precisely why their disagreement is informative rather than contradictory.

The fixed rail

A constant baseline is what makes months comparable.

Profit quality is normalised two ways in parallel. Against a fixed baseline risk it produces a reading that can be compared across any period, because the denominator never moves. Against live average risk it produces a reading that shows whether current deployment is improving or diluting efficiency. The two are kept as separate outputs rather than merged, and the gap between them is itself an audit: a large divergence means live deployment has drifted well away from the baseline the classifications were calibrated on.

Classes, not floating values

A matrix key looks like Stable_Strong, never Stable_26.363636.

The matrix is keyed on class labels rather than raw numbers, and the workbook's failure-mode list names the alternative explicitly as a defect. Floating-point keys break lookups, and a broken lookup in an interpretation layer does not error out visibly. It returns the wrong meaning, in the right formatting, with full confidence. Standardising the class vocabulary across every tab is a maintenance rule for the same reason.

The matrix is not rescaled week to week; it is a stable ontology, not a moving target.
Final interpretation is base meaning plus risk context plus expectancy context plus hit-rate context.
Companion rails add context to a class; they never redefine the underlying metric.

Silence as signal

A long stretch of pure agreement earns its own audit.

Engines that agree for many consecutive months are either watching a genuinely stable machine or measuring the same thing twice. Prolonged unanimity is therefore a prompt to verify that the two inputs are actually independent, because a confirmation layer that cannot disagree has quietly stopped being a check and become an echo. This is the same discipline that keeps the interpretation layer separate from the pipelines: merged, they would eventually confirm themselves.

Reference

The base matrix: efficiency class down, profit-quality class across

Efficiency	Weak	Stable	Strong	Elite
Weak	Collapse	Fragile	Distorted	Illusion
Stable	Underperforming	Balanced	Productive	Overextended
Strong	Inefficient Profit	Efficient	High Performance	Aggressive
Elite	Misallocated	Precision	Institutional	Peak

Before you go deeper

Connected inside MARS

This module doesn't work alone.

Connected — 01

Contradiction Engine (full page)

Open system page →

Connected — 02

The Five States

Open system page →

Connected — 03

Operator Playbooks

Open system page →

Go deeper

Operator briefs on this territory.

Deep dive — 01

One number can only be believed or doubted. Two can be tested.

A lone metric can only be accepted or doubted. A pair that should agree makes a claim that can fail.

Read the full brief →

Deep dive — 02

Composite-only: the scope decision that costs the engine its aim.

Three composites, no branches. The engine can prove a contradiction exists and cannot say where it lives.

Read the full brief →

Deep dive — 03

The three cases that built the engine: profitable, and getting worse.

Profitable while drawdown expands. Strong profit factor with inefficient risk. Growth with slowing acceleration.

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.

Get MARS Pro — $497 Explore the Platform