01 / Research Protocol
A market state must become measurable before it can become a claim.
LDRG studies global futures and crypto market microstructure as a sequence of observable liquidity states. The work begins with raw and reconstructed order-flow evidence, then asks whether a behavior is unusual, whether it repeats, whether it is linked to measurable outcomes, and whether it remains stable after controls.
The research is not built around a single theory. It is a multi-hypothesis process that tests market microstructure behavior, price-action structure, mathematical features, volatility behavior, cross-market relationships, event context, and risk assumptions against the same underlying data record.
Research starts at the behavioral layer.
Price is treated as an output, not the full evidence. LDRG studies the events and conditions that produce the visible move: order additions, cancellations, modifications, trades, queue behavior, depth, spread, replenishment, retreat, absorption, imbalance, volatility, session structure, and cross-market pressure.
A useful footprint is rarely one isolated event. It is a configuration of behavior through time. The research question is whether that configuration has a repeatable relationship with a future reaction, a failure condition, or a change in liquidity regime.
Public Boundary
- Public research explains process, vocabulary, and validation discipline.
- Proprietary thresholds, feature weights, formulas, scoring logic, and market-specific parameters are not published.
- Public pages do not provide trade instructions, targets, stops, or account-specific advice.
02 / Multi-Hypothesis Discovery
Candidate discovery starts with many competing explanations.
LDRG expects most ideas to fail. That is part of the process. A theory can come from microstructure, trader behavior, quantitative statistics, price-action observation, cross-asset reaction, event analysis, or model-discovered anomaly clusters. It becomes researchable only when it is translated into observable variables.
Depth withdrawal, queue instability, cancellation imbalance, replenishment, absorption, sweep rejection, trapped aggression, and liquidity vacuum behavior.
Failed breakout, compression and expansion, session open behavior, range rejection, prior liquidity-zone reaction, and continuation or reversal structure.
Imbalance ratios, volatility expansion, acceleration, entropy, clustering, lead-lag behavior, regime change, and correlation breakdown.
Cross-market pressure, macro calendar distance, scheduled event windows, delayed reactions, crypto liquidation/funding pressure, and abnormal pre-event market behavior.
The formal distinction matters: a theory is an idea worth testing, a feature is the measurable version of that idea, an anomaly candidate is unusual behavior found in the data, a signature candidate is a recurring anomaly with measurable outcomes, and a validated signature is a candidate that survives defined tests.
03 / Candidate Pipeline
Discovery is broad. Promotion is strict.
LDRG does not promote a candidate because it looks familiar on a chart or appears to work in a small sample. Every candidate must move through a staged process that separates observation, measurement, validation, live monitoring, and eventual system translation.
- 01Theory
An idea from microstructure, price behavior, quantitative features, context, or model discovery.
- 02Feature Set
The measurable variables needed to test the idea against the order-flow record.
- 03Anomaly Candidate
Unusual behavior that appears in the data and deserves structured review.
- 04Signature Candidate
A recurring anomaly with measurable outcomes and defined conditions.
- 05Historical Test
Review against reconstructed data, labels, regimes, and failure cases.
- 06Out-Of-Sample Review
Challenge outside the discovery period to reduce overfit interpretation.
- 07Live Shadow
Observe in live conditions before production-grade treatment.
- 08Library Status
Assign active, conditional, degraded, suspended, or research-only status.
From candidate to signature.
A candidate becomes stronger only if its definition is clear, its inputs are observable, its detection logic is repeatable, and its future reaction distribution can be measured over defined horizons. The result may be directional movement, continuation, reversal, volatility expansion, liquidity replenishment, liquidity withdrawal, delayed reaction, or no reaction.
A failed candidate is not automatically wasted. Failure cases help define where the structure breaks: missing confirmation, unstable queue behavior, insufficient liquidity, excessive slippage, incompatible regime, seasonal distortion, event contamination, or overfit discovery.
Acceptance Requirements
- Clear observable definition.
- Defined reaction horizon and outcome labels.
- Sufficient historical sample and failure review.
- Out-of-sample and walk-forward evidence.
- Known regime, session, and liquidity dependencies.
- Realistic execution and risk assumptions.
04 / Validation Standard
Backtests are only one layer of evidence.
Historical testing is necessary, but not sufficient. LDRG treats validation as an evidence stack: data integrity, reconstruction, leakage control, sample discipline, regime analysis, multi-market review, live observation, cost assumptions, and ongoing monitoring.
Candidate behavior is tested against reconstructed order-flow data, future outcome labels, session context, and known market conditions.
Candidate definitions are challenged outside the discovery window so the research does not simply memorize one period of market behavior.
Definitions are reviewed through time to identify drift, decay, regime sensitivity, and whether the structure survives changing conditions.
Promising candidates can be observed live before being treated as production-grade intelligence or exposed as an official API flag.
When relevant, the same candidate is tested across related global futures and crypto markets to see whether it generalizes, confirms, or remains market-specific.
The research records where a candidate appeared but did not produce the expected reaction, because boundaries are part of the signature definition.
05 / Signature Governance
A signature is a living research object.
Market microstructure changes. Liquidity varies by season, roll period, volatility regime, participation, event density, and cross-market conditions. LDRG therefore treats every accepted signature as versioned, monitored, and conditional on evidence.
The signature has sufficient validation evidence and can be flagged by approved systems under its defined conditions.
The signature remains useful only in specified regimes, sessions, instruments, volatility states, or cross-market contexts.
The measured reaction distribution has weakened, drifted, or become too unstable for normal active treatment.
The signature remains in the research library, but active public or partner-facing use is paused until evidence supports reactivation.
If a signature stops working, LDRG does not erase it. The research asks whether the change is temporary, seasonal, event-related, liquidity-regime dependent, market-specific, or evidence of permanent decay. A suspended signature may later return to reactivation watch if the behavior reappears under identifiable conditions.
06 / Translation To Systems
Research becomes intelligence only after evidence justifies it.
Validated research can be packaged into ORLOS for detection, classification, scoring, and probability-based interpretation. It can later support approved API or SDK access, professional research tools, and controlled terminal workflows such as OFTT.
The objective is not to publish trade calls. The objective is to identify which market states deserve to be measured, tested, monitored, and, only after sufficient evidence, converted into structured liquidity intelligence.