The Regime Intelligence Engine is the six-dimensional composite market-regime classification system inside Narion Execution Research Infrastructure. On its own terms — independent of any specific downstream signal — it answers a structural question: does the market organize into a comparatively small, recurring set of conditions across trend, momentum, volatility, positioning, and price structure, and does knowing the current condition say something meaningful about the condition that follows.
A separate, forthcoming publication will examine whether conditioning Narion's microstructure research on this regime state changes what a given order-flow reading should be taken to mean. This module deliberately does not make that combined claim — it states and evaluates the regime hypothesis on its own.
Consistent with the disclosure approach used across Narion's research modules, the governing hypothesis, system architecture, and the function of each regime-classification engine are disclosed below in the form a research publication reports them. The closed-form scoring formulas, statistical weights, and classification thresholds inside each engine are proprietary research infrastructure and are withheld — see Disclosure & Scope for the exact boundary.
Narion's research program currently runs two independent tracks. Microstructure research — the Institutional Flow Anticipation Engine, IFAE — studies the order-flow information content of trade and order-book activity. Structural regime research, documented in this module, asks a narrower and separate question: is the market's broader condition at any moment reducible to a small, recurring set of structural states, and does knowing the current state say something meaningful about the state that follows.
This module deliberately does not combine the two tracks into a single hypothesis. Whether regime context changes the interpretation of a microstructure reading is a real question Narion intends to answer, but it is reserved for a forthcoming joint publication once each research track has been independently validated to Narion's standard. What follows here is the regime hypothesis, the six independent classification engines that reconstruct market structure into a categorical fingerprint, how those six outputs are fused into a single composite regime read, and the validation discipline used to keep the system's live and research-reconstruction implementations honest against each other.
Stated on its own terms, independent of any specific downstream signal, Narion's regime research bet is that a market's broader condition — its trend, momentum, volatility regime, positioning, and price-structure phase, taken together — does not drift continuously and unpredictably through every one of the states it could theoretically occupy. Instead, it is hypothesized to organize into a comparatively small number of recurring, persistent structural states, such that knowing the current state materially reduces uncertainty about the state the market will occupy a short time later.
This is a deliberately narrower and more falsifiable claim than asserting that regimes predict price outcomes: it only asserts that regimes are structured and persistent, rather than effectively random draws from the theoretical regime space described in Section 05. Rt is not a single observable number — it is a theoretical construct that the Regime Intelligence Engine reconstructs operationally, once per bar, from six independently classified structural dimensions (Section 04), fused into a single composite state (Section 05).
The Regime Intelligence Engine is not a standalone product — it is a component of Narion Execution Research Infrastructure, the research and engineering program that studies how live signal intelligence, structural regime classification, and a research-validated policy universe combine into a deterministic, auditable execution-decision pipeline. Within that pipeline, the engine sits between raw market data and the research-policy lookup layer: it converts a snapshot of market conditions into the composite regime label that the execution-policy research described elsewhere in Narion's Execution Research program uses to decide whether a given microstructure candidate falls inside a research-validated state.
Each of the six regime dimensions is produced by an independent, single-purpose engine reading from a shared market-data snapshot; a seventh component fuses the six outputs into one composite string. Running in fixed dependency order, each stage reads only the fields already available to it — no engine other than the fusion stage depends on another engine's classification output.
The engine exists in two parallel, independently coded forms, run side by side as a matter of research discipline rather than as a single implementation replayed twice. A live path evaluates each of the six engines statelessly, once per bar, directly against streaming market data. A separate offline research-reconstruction pipeline recomputes the same six dimensions in batch, vectorized across the full historical shadow dataset, to generate the research artefacts behind Narion's regime-level validation work — and, for one dimension, to produce the statistical calibration baseline the live engine itself depends on. Because these are two independent implementations of the same governing rules rather than one implementation replayed, Narion treats formula-for-formula reconciliation between them as an ongoing research responsibility (Section 06), not a one-time engineering task.
Each dimension is described below by function — what it measures and the categorical states it can resolve to — rather than by its closed-form scoring formula, statistical weights, or numeric thresholds.
The six dimension outputs are joined, in a fixed column order, into a single pipe-delimited composite regime string — market cycle, trend state, trend strength, expansion state, VWAP bucket, then location state. Any dimension that cannot be classified for a given bar is filled with an explicit unknown marker rather than left null or silently dropped, so a composite regime read is always fully formed.
Because the live and offline paths are independently coded rather than one replaying the other, Narion runs a standing reconciliation practice between them: every classification rule is checked formula-for-formula so that research conclusions drawn from offline-reconstructed history can be trusted to describe the live engine's actual behavior. This is treated as ongoing research quality assurance, in the same spirit as the empirical validation discipline applied to Model_D (see the IFAE Core Research Module).
| Dimension | Reconciliation Status |
|---|---|
| Market Cycle | RECONCILED |
| Trend State | RECONCILED |
| Expansion State | RECONCILED |
| Composite Fusion Rule | RECONCILED |
| Trend Strength | RECONCILED — classification rubric tracked across two source locations as a single-source-of-truth item |
| VWAP Context — bucket | RECONCILED |
| VWAP Context — trend | UNDER RECONCILIATION |
| Structural Location | UNDER RECONCILIATION — a subset of boundary conditions |
The Regime Intelligence Engine reduces raw market data to a six-dimensional categorical fingerprint through six independent classifiers, fused by a simple, fixed-order concatenation rule. Its dual-path design — a live, stateless, per-bar inference engine and an independently coded offline research-reconstruction pipeline — is a deliberate choice for research reproducibility, and the reconciliation status in Section 06 is published as part of that discipline rather than smoothed over.
| Dimension | Status |
|---|---|
| Governing hypothesis | DISCLOSED — regime-conditioning bet (Eq. 1) |
| System architecture | DISCLOSED — six engines, fixed fusion order |
| Engine functions & classification states | DISCLOSED |
| Live / offline reconciliation — majority of dimensions | RECONCILED |
| Live / offline reconciliation — VWAP trend, structural location | OPEN — active engineering item |
| Regime-level performance validation | SEPARATE RESEARCH TRACK — not covered by this module |
The recommended next step is the same discipline Narion applies to Model_D: close the two remaining reconciliation items on their own timeline, and treat any regime-level performance claim as requiring its own dedicated forward-validation study rather than being inferred from classification accuracy alone.
This module states the governing hypothesis, the system's position inside Narion Execution Research Infrastructure, its architecture, and the function of each regime-classification engine because that is what a research publication owes its readers. It withholds the closed-form scoring formulas, statistical weights, calibration mechanics, and numeric classification thresholds because those are the proprietary research output the hypothesis was built to produce.
A full mathematical architecture specification — including every scoring formula, weight, and threshold withheld from this page, and the complete cross-engine reconciliation ledger — is maintained internally and is available to research partners under a research-access agreement.
Full formula-level specification of all six classification engines, the composite fusion rule, and the live/offline reconciliation ledger.
RESEARCH ACCESS ONLYThe companion IFAE research module — governing hypothesis, signal family, and Model_D validation record.
RESEARCH ACCESS ONLYHow live signal intelligence, regime research, and policy combine into an auditable, fail-closed execution-decision pipeline.
RESEARCH ACCESS ONLYResearch briefings for qualified partners and engineering collaborators, including the material withheld from this page, are reviewed within 48 hours.