How Spread Foundry Uses Regime Context

A regime label is a compressed description of recent market behavior. It is not a forecast, and it is not a trade.

That distinction matters in an options workflow. It is easy to let a market state become a hidden rule that deletes candidates, manufactures confidence, or implies that one structure will outperform another. Spread Foundry uses a narrower contract: regime is evidence attached to a decision, not an oracle that makes the decision.

From market history to an evidence pack

FerroWave is MorphIQ Labs' Rust signal processing engine. In the Spread Foundry workflow, it turns trailing, time-ordered market history into a versioned per-symbol evidence pack. The pack summarizes several different questions:

• Trend and regime. Is the recent path directional, range-bound, volatile, or changing character?
• Market posture. Does the combined evidence support normal exposure, smaller exposure, or standing aside?
• Jump and stress state. Is a discontinuity or event still material to the current read?
• Volatility horizon. Which time scales are carrying the observed variation?
• Cross-asset coherence. Is the symbol behaving independently or moving with broader market, sector, or volatility context?
• Readiness and methodology. Was the evidence usable, degraded, or unavailable, and which version produced it?

These are separate observations. A directional label does not make jump risk disappear. A clean symbol-level read does not make broad-market coupling irrelevant. Keeping the evidence typed and versioned prevents a single badge from pretending to summarize everything.

What Spread Foundry does with the evidence

Spread Foundry carries the evidence through a snapshot-scoped workflow:

1. FerroWave computes the market-context pack from trailing history.
2. The pipeline validates and persists the read with quality and methodology.
3. Spread Foundry derives bounded sizing guidance and explicit caution states.
4. Structurally valid spreads are built from option-chain inputs and their own strategy rules.
5. Regime alignment, jump risk, stand-aside state, and sizing guidance are attached to those candidates for review and explanation.
6. The default view can place stronger context ahead of weaker context without hiding the rest of the valid set.

This separation is deliberate. Candidate construction answers, "Is this a valid spread under the structural rules?" Regime context answers, "How does this spread sit inside the market evidence available right now?" Combining those questions too early makes it impossible to tell whether a candidate did not exist or was merely considered a poor fit.

Unavailable is not sideways

Missing evidence is not neutral evidence.

If the history is insufficient or a required read cannot be produced, Spread Foundry marks the context unavailable and stands aside on sizing. It does not substitute a range-bound label, carry an old answer forward without provenance, or let an absent signal look like a normal low-conviction state.

That failure behavior is part of the model contract. The system should explain what it knows, what it does not know, and what changed in the workflow because of that boundary.

Context, not predictive certainty

This design does not claim that a regime label predicts the next return, identifies a winning strategy, or demonstrates trading efficacy. It provides a consistent way to organize observed state before a trader compares structure, volatility, liquidity, scenarios, and risk.

A useful regime layer should therefore be judged on more than whether colored bands look plausible after the fact. It needs causal inputs, explicit ground truth where available, stability away from transitions, detection-latency measurement, and honest failure states. Our public research notes on evaluating market regime detectors and the synthetic benchmark for advanced decomposition methods show how we approach those questions without presenting a product configuration as settled science.

Why the implementation boundary matters

The workflow is a research-to-runtime problem, not just a modeling problem. The mathematical method has to survive typed interfaces, versioned methodology, missing data, deterministic recomputation, persistence, and product explanation. FerroWave provides the reusable signal-processing and finance contracts; Spread Foundry owns the options-specific composition and workflow policy.

That boundary also protects the research. The exact production configuration, calibration data, thresholds, strategy-alignment map, and sizing policy remain proprietary. The public claim is narrower and testable: Spread Foundry uses a versioned FerroWave evidence pack to make market context explicit, keeps candidate construction independent from that context, and fails visibly when the evidence is unavailable.

For the broader workflow, see Options Spread Analytics Should Be a Workflow, Not a Table. For the underlying signal-processing idea, see Wavelet-Domain Regime Detection with the MODWT.