01 / Inputs
Multi-source evidence
- Macroeconomic factors
- Market microstructure and capital-market context
- Semantic and news evidence
- On-chain and off-chain activity
- Liquidity and capital flows
Research direction · Market intelligence
Markets change.
Questions evolve.
Investigating how a framework for uncertainty could support market-state analysis, risk research and explainable decision support. The transfer is a research question—not an assumed result.
Planning horizon: 2027, subject to data readiness, implementation and validation. This page describes proposed work, not a live trading system or a validated market product.
The research approach
Observe. Map.
Evaluate. Review.
A regime shift is a meaningful change in system behaviour. The challenge is to distinguish such changes from noise, unstable correlations and explanations that only work in hindsight.
Observe
Define a time-aware view of macroeconomic, market and contextual evidence, with traceable sources and availability dates.
Map
Examine interactions, disagreement, missing confirmation and candidate changes in structure.
Evaluate
Compare hypotheses against historical baselines, held-out periods, scenarios and counter-explanations.
Review
Assess uncertainty and domain-specific risks before treating a result as decision-relevant. Waiting remains a legitimate outcome.
A proposed domain translation
The data changes.
The method must earn its place.
LUMA does not carry benchmark-specific predictions into financial markets. It investigates whether selected analytical principles remain useful when the entities, data and evaluation conditions change.
02 / Methods to investigate
Structure under uncertainty
- Interaction and multi-axis mapping
- Negative-space and regime hypotheses
- ENMF state and experience memory
- Historical and walk-forward evaluation
- Simulation and decision gates
03 / Intended research outputs
Explainable states
- Candidate market regimes
- Cross-layer risk and disagreement maps
- Scenario distributions
- Inspectable candidate universes
- Wait or no-action states
This is a conceptual research mapping. Each proposed output requires its own implementation and validation.
Public mechanism mapping
What could transfer.
What still needs testing.
The established public overview is organised below by research task. It identifies candidate mechanisms without publishing proprietary formulas, weights, thresholds or execution logic.
01 · Observe — build a multi-source view
| Analytical mechanism | Proposed market research use |
|---|---|
| Macro and micro aggregation | Combine macroeconomic conditions with asset behaviour, market microstructure and local liquidity. |
| Consensus and disagreement | Identify where evidence layers agree, conflict or share a common source rather than treating model consensus as proof. |
02 · Map — examine relationships and missing confirmation
| Analytical mechanism | Proposed market research use |
|---|---|
| Pairs, triads and higher-order interactions | Examine relationships among assets, sectors, factors and market variables beyond pairwise correlation. |
| Convergence probes | Test whether apparent alignment across factors adds information beyond correlated inputs. |
| Multi-axis analysis | Keep momentum, stability, liquidity, volatility, sentiment and structural risk distinguishable. |
| Negative-space analysis | Investigate missing confirmation, absent participation, liquidity gaps and silent divergence against an explicit expected baseline. |
| Candidate-space compression | Study how a broad asset or scenario universe can be reduced while recording selection losses and uncertainty. |
03 · Learn — compare states with history and later outcomes
| Analytical mechanism | Proposed market research use |
|---|---|
| ENMF memory | Record prior states, transitions, interactions, failures and the conditions under which observations held. |
| Historical backtesting | Evaluate past behaviour with realistic data availability, costs and known failure periods. |
| Walk-forward validation | Fix each model using only information available before the next evaluation interval; assess performance on held-out subsequent periods. |
| Emerging pattern hypotheses | Test whether newly observed relationships persist out of sample rather than assuming early detection. |
| Post-outcome evaluation | Compare earlier expectations with later observations and preserve discrepancies in research memory. |
04 · Review — control the transition from evidence to action
| Analytical mechanism | Proposed market research use |
|---|---|
| Cross-axis risk assessment | Examine credibility and sensitivity across distinct risk perspectives. |
| Regime-change detection | Test candidate transitions among expansion, contraction, stress, instability and recovery. |
| Monte Carlo simulation | Explore model-conditional outcome distributions, tail risks, stress scenarios and portfolio sensitivity. |
| Adaptive search | Compare alternative analytical policies, parameter settings and decision paths in controlled experiments. |
| Branching scenario exploration | Investigate possible decision sequences under uncertainty rather than treating a single path as certain. |
| Decision memory and gating | Record the conditions under which evidence is accepted, deferred or rejected; review risk before any proposed action. |
| GPU-assisted simulation | Evaluate the practicality of larger scenario and stress-test workloads where the method warrants the computation. |
Validation before application
Plausibility
is not enough.
A credible market translation needs point-in-time data, appropriate baselines, historical and sequential evaluation, and explicit treatment of costs, liquidity and changing market conditions. Domain-specific failure modes cannot be resolved by a successful synthetic example.
- Historical testing: include adverse conditions and failure periods, not only favourable windows.
- Walk-forward evaluation: keep model selection separate from subsequent held-out observations.
- Regime-specific review: inspect where a method works, fails or becomes inconclusive.
- Post-outcome learning: preserve errors and limits as well as positive findings.
No market advantage or dependable trading return is established by the existence of this research agenda.
The intended sequence
A controlled
development path.
The 2027 planning horizon remains subject to evidence, capacity and domain-specific review. Each stage can produce a revision or a decision not to proceed.
Define the data contract
Establish sources, permissions, point-in-time availability and the proposed evaluation question.
Map the market state
Translate entities, relationships, regimes and memory roles into domain-specific representations.
Test the hypotheses
Run historical and sequential experiments, independent checks and relevant baselines.
Assess scenarios and risks
Study model sensitivity, stress conditions and conflicts between evidence sources.
Evaluate a research interface
Consider exposing inspectable states and uncertainty only after the supporting methods and boundaries are established.
Evidence & scope
A research direction.
Not a financial promise.
LUMA Quant does not present this work as automated asset management, a hedge fund or a guarantee of market returns. Proposed outputs are for research, risk analysis and decision support, not investment advice.
The Trust Center links the 19 October product-status record and the separate August source-code baseline. These records do not validate a market product or establish a production audit.
Read the public whitepaper Explore the broader research agenda