Operator brief · 157

Why the SDE, Regime, and Contradiction engines live inside CP3.

The key idea

The roster

What CP3 actually contains.

The flagship is a single ~4.7MB workbook holding the Journal, the Weekly Summary, Gate/Brake state, the EV Scorecard, the Branch Risk Engine, and Capital Dynamics — plus the housed SDE, Regime, and Contradiction engines. Around it, and deliberately outside it, sit the Weekly Trading Scorecard as the weekly EV cockpit, the Advanced EV Analytics Lab for deep expectancy R&D, the MAE/MFE Execution Efficiency Lab, and the Variant Attribution Module. The line between inside and outside is not arbitrary, and reading it correctly explains most of the ecosystem's shape.

FigureDivergence between two copies of the same evidence

Schematic. A housed engine reads the live journal, so its view never differs from CP3's. A separate workbook works from a copy, and that copy drifts further from the truth with every row entered since the last sync — then snaps back, having been wrong the whole interval.

The reason

An engine that reads the journal must read the same journal.

If the SDE lived in a separate workbook, it would need its own copy of the evidence — and the moment there are two copies there are two versions of the truth, differing by whatever was entered since the last sync. A structural read computed against a stale snapshot is not merely slightly out of date; it can disagree with the EV Scorecard about the same week, and the operator is then adjudicating between two components that were supposed to be describing one system. Housing removes the possibility rather than managing it.

Read, don't recompute

The housed engines consume CP3's outputs rather than deriving their own.

Co-location is paired with a discipline that matters as much: the housed engines read the values CP3 already computed instead of recalculating them from raw rows. This is why the SDE's EV figure always matches the EV Scorecard's — not because two calculations happen to agree, but because there is only one calculation and the SDE is looking at its result. The engines add interpretation on top of settled numbers. They do not produce a competing arithmetic, which is what would happen if each derived its own inputs from the journal independently.

The cost of the flagship

The cost is indivisibility, and it is knowingly accepted.

Consolidation has consequences the design does not pretend away, and it is worth being precise about which ones are real. File weight is not among them — CP3 is roughly 4,700 KB, which is unremarkable for a workbook of its formula density. The genuine costs are structural. Version control is coarse, because everything moves together: a fix to the Gate/Brake logic reissues Capital Dynamics whether or not it changed. A corruption event threatens more at once, so the backup discipline has to be stricter than a smaller file would demand. And the workbook cannot be partially shared — handing someone the Journal means handing them the EV Scorecard and the housed engines too, which forecloses the option of releasing one component to a reviewer. These are paid for the single benefit of eliminating divergence between components that must agree, which the architecture treats as worth more than the convenience of separable files.

Why the labs stayed out

External components are the ones permitted to be out of date.

The four external modules share a property: none of them needs to agree with CP3 at the instant of computation. The MAE/MFE Lab analyses closed trades from a settled period. The Variant Attribution Module decomposes a completed record. The Advanced EV Analytics Lab is explicitly R&D and must be sandboxed rather than integrated. The Weekly Scorecard operates on manually-entered branch probabilities and answers its own weekly question. Each can safely work from a snapshot, which means each can safely be a separate file — and separate files are cheaper, faster, and independently versionable.

The reading order it implies

Inside first, outside second — and never the reverse.

The architecture's boundary produces a practical sequence for review. CP3's own outputs establish what happened and what state the system is in; the housed engines then say what that state means structurally. Only after both are settled do the external labs add their diagnosis, because each of them is interpreting a record that CP3 has already finalized. An operator who opens the MAE/MFE Lab before reconciling the journal is analysing a period that has not finished being defined, and any finding produced there will need re-checking once it has.

Connected inside MARS

Compliance Panel 3 →

Structural Diagnostic Engine →

Regime Classification Engine →

Module Ecosystem →

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