Loading market research view…
Multi-horizon diagnostics
Tactical, Swing, and Structural analysis with fixed lookback windows, training cutoffs, and retained threshold distributions.
FRACTAL INTELLIGENCE ENGINE
An evidence-first view of market behavior.
A quantitative research platform that examines market behavior across horizons. Explore what is implemented today and what the broader intelligence layer aims to become.
Three horizons.
One inspectable framework.
Compare hourly and daily price-history diagnostics using declared lookbacks and frozen training artifacts. Missing information stays visible throughout the analysis.
Implemented as experimental research. Current classifications are not qualified for analytical use or recommendations.
Horizon MapSYNTHETIC DEMONSTRATION
Swing horizon
30 / 90 / 180-day lookbacks · Daily source bars. Intended holding horizon: 3–15 days.
Normalized trend +0.42Volatility percentile 64thPersistence UnqualifiedLiquidity adapter Unavailable
Illustrative descriptors, not source-engine results. Cross-horizon qualification remains incomplete.Implemented in the source application
Experimental fractal analysis
Classical and modified R/S, DFA, Haar wavelet energy, and surrogate comparisons. No Hurst interpretation or predictive qualification is enabled.
Daily statistical baselines
Saved-close arithmetic diagnostics, including return behavior and realized volatility, with explicit input completeness requirements.
Quote & liquidity evidence
Displayed book depth, bid–ask spread, Coinbase–Kraken quote comparison, bounded sampling, and configured fee scenarios.
Historical evidence
Immutable feature records, original definitions and results, knowledge cutoffs, source manifests, and frozen-fit lineage.
Portfolio research foundations
Holdings import, reconciliation, shared-cash benchmarks and registered historical account replay. Cash, holdings and pending orders persist across joined source-book windows within each fold.
Registered strategy replay
Pre-fit registration, training-only fitting, retained variants and controls, three cost scenarios, and immutable reports reproducible in a new process. Implemented as a local research command.
Chronological return diagnostics
Daily observable account equity with explicit quarterly capital resets. Approximate overfitting and Sharpe diagnostics, common-block bootstrap comparisons, and retained missing or failed cases.
Accepted strategy robustness
Registered lag, date and liquidity perturbations, retained missing cases, daily risk and recovery diagnostics, and an approved lockbox protocol. Phase 3 is complete within its declared research bounds.
Recommendation & risk engineering
Five strategy paths, deterministic sizing, cost gates, correlated-exposure limits, stale-data rejection and immutable replay. Phase 4 engineering is accepted; recommendations and execution remain disabled.
Connected portfolio research
Bounded read-only Coinbase, Binance.US and IBKR workflows feed dated portfolio evidence and a partial combined book. Historical reconciliation, basis, valuation and operational risk inputs remain incomplete.
Accepted research dashboard
Portfolio, source health, recommendation history, performance/risk charts, retained calibration bins and regime timelines. Execution-link review, versioned configuration and audit paths preserve missing evidence.
Scheduled reports & email
Persistent daily, weekly and on-demand reports with retained versions, retries and restart recovery. Microsoft 365 delivery is configured and verified in the local application; blocked data remains visible in reports.
Shadow monitoring & paper portfolios
Immutable observations, missed-cutoff records and bounded simulated fills. Paper journals retain shared cash, quantities, fees and version-bound re-evaluation; scheduled midnight marks support paper risk checks.
Daily research & frozen forecasts
Saved daily closes, first-known timestamps, features and strategy conditions. Forecast populations preserve exclusions and actual freeze time, keeping the reserved latest 12 calendar months out of training.
Cost & reliability calibration workflows
Immutable cost-population registration and source-bound nested calibration retain causal checks, predictions and reliability evidence when sufficient data exists. Missing inputs stay unavailable; measured-cost acceptance is unfinished.
Saved-universe screening
Screens immutable saved crypto and securities evidence with explicit history and cutoff bounds. Missing data is not retrieved or inferred; every new-entry flag remains false.
SOURCE CHECKPOINT · September 25, 2026
Phase 8 is underway. Qualification is still ahead.
7 / 10
Project phases complete
0 / 1
Phase 5 acceptance · still open
0 / 1
Phase 8 acceptance · still open
Counts reflect the source project’s development plan, not investment performance, a release date or a percentage of engineering effort.
Qualification: not started. Phase 8 requires at least 60 qualifying scheduled decisions spanning at least 90 calendar days, alongside all research, accounting, data and risk gates, approved cost tolerance and reviewed material incidents. Monitoring observations, report jobs, synthetic tests and unqualified runs do not count toward that minimum.
Seven phases are accepted: 0–4, 6 and 7. Phase 5 remains open while Phase 8 continues. The accepted dashboard and reporting workflows make partial evidence inspectable; they do not turn incomplete accounting into complete portfolio value, returns or current advice.
Next: establish verified continuous portfolio, performance and risk inputs; qualified historical and market-cycle evidence; calibration matched to registered outer-test outcomes; and measured costs. Engineering fixtures can exercise the pipeline while these remain open. Downloading older prices today cannot recreate what was known at a historical decision cutoff.
From evidence to an auditable outcome
The shadow pipeline binds retained observations, current research, prior frozen forecasts and matching calibrations. Explicit synthetic contexts can exercise the risk engine; operational recommendations and execution remain disabled.
Passing a number is not acceptance
Registered outer-fold replay now combines numerical checks with evidence provenance, settled trade episodes and cost stress. A favorable statistic cannot override missing qualification evidence. Fold-reset diagnostics are not a continuous portfolio record.
Count only eligible decisions
Source-bound counters, missed-slot review and incident coverage preserve why a run does or does not qualify. Historical accounting exceptions remain explicit, including unknown cost basis. Current progress remains 0/60 decisions and 0/90 days.
Accepted
Research, data, validation & risk
Engineering milestones accepted within their declared scope; no strategy is qualified for investment use.
Open
Portfolio accounting & account reads
Coinbase, Binance.US and IBKR reads and partial combined accounting are implemented. Historical import provenance and exception handling have advanced; unknown cost basis stays unknown. Combined reconciliation remains open.
Accepted
Dashboard, history & explainability
Source-bound portfolio and research views, performance/risk charts, execution matching, configuration and audit review.
Accepted
Reports, scheduling & email
Persistent daily/weekly and on-demand reports, retained history, retry/restart recovery and configured Microsoft 365 delivery.
In progress
Shadow/paper operation & calibration
Retained signal/forecast/calibration pipeline, synthetic risk evaluation, registered outer-fold acceptance checks, incident coverage and source-bound qualification counters are implemented. Verified operational inputs and qualified research are still missing. No qualification clock is running.
Preparation only
Production hardening
A synthetic encrypted recovery drill has been exercised. This is preparation, not production restore acceptance or Phase 9 completion. Production hardening and operational acceptance remain future work.
Scheduled shadow collectors read saved evidence; they do not refresh exchange feeds. Account session renewal and expiring permission declarations remain operational limits. The local paper workspace requires retained publications; its implemented simulator does not establish an actual multi-day owner strategy run.
No research family or strategy is qualified. Experimental FMH interpretation, live recommendations and execution remain disabled. The website demo uses synthetic data and has no account connection.
Capital, custody and exposure connect. Portfolio research keeps those relationships visible.
Conceptual artwork · not market dataThe next intelligence layer
Fractal Market Stability Index
Proposed · FMSI would summarize structural stability across horizons. Candidate dimensions include diversity, liquidity, persistence, volatility, and tail risk. No production formula or score exists.
Qualified regime context
Future · Persistence, mean-reversion and liquidity adapters must be qualified before directional state labels become available in the saved-close runtime.
Risk & Opportunity scores
Proposed · Explainable composite summaries require defined formulas, calibrated inputs, and validation. Opportunity would not mean probability of profit.
Public API & broader markets
Future · Commercial authentication, service commitments, data licensing, wider production coverage and public subscriptions are not launched. Local scheduled reporting is implemented; a public reporting service is not available.
SEE THE MARKET IN CONTEXT
A better starting point
for your market research.
Help shape the next generation of multi-horizon intelligence.
Get early access No payment required. Access timing is not yet announced.