Market State Detector
Interactive Historical Episode Comparison
Compare regime classifications from three major market episodes. Each widget shows the exact state label, 60-day history timeline, and verification data that institutional teams would have received on that date.
Rules-based, not ML. Zero model drift.
Same thresholds for 1990-2025 backtests and live delivery. Historical validation summaries are published state-by-state in the validation documentation. Pilot dataset covers 2012–T-30 days.
Three Major Crisis Episodes
Hover over \"Rule Details & Metrics\" to see rule details, rule version, and historical performance metrics for each classification.
COVID-19 Global Crisis
Crisis - Feb 24, 2020
April 2025 Tariff Shock
Crisis - Apr 4, 2025
Volmageddon
Shock - Feb 5, 2018
How the Market Risk Classifier Works
Deterministic Rules
Six-state system (Stable, Whipsaw, Shifting, Elevated, Shock, Crisis) using fixed thresholds. Same rules for backtests (1990-2025) and live delivery. No ML drift, no black boxes.
// Every classification ties to a specific rule ID + rule version.
High Salience
Typically fewer than 20 risk alerts per year across all states. Crisis is ultra-rare: ~0.7 episodes/year. Designed for signal-to-noise efficiency.
// Critical states fire only during genuinely abnormal conditions.
Pre-Market Context
Classifications delivered by 07:30 AM ET on U.S. trading days (pre-market). Gives risk committees advance context for scenario planning and positioning decisions.
// Historical timing characteristics are documented in validation.
State Classification Logic
MRC applies deterministic thresholds to exogenous environmental indices(NOAA/NASA geomagnetic data, solar activity) together with market structure inputs (SPX, VIX). SPX/VIX are used within the rules and as independent event definitions in validation, detecting regime shifts that traditional price-only models often miss.
CRISIS (Critical)
Historical record (1990-2025): 13 activations, 12 hits, 1 false positive.
Lead time: see validation documentation.
Shock (Severe)
Historical record (1990-2025): See validation documentation for episode summaries and precision.
Median same-day detection. Flags acute expansion.
Shifting (Inflection)
Historical record (1990-2025): See validation documentation for transition summaries and precision.
Median lead: 2 business days. ~4.7 episodes/year.
Stable (Baseline)
Operational baseline: Active ~70.7% of trading days (1990-2025).
Absence of stress signals. Real-time confirmation.
COMPLIANCE NOTICE: Metrics are historical research findings (1990-2025 backtests). Descriptive, not predictive. No investment advice or guarantees.
Pilot API Access
REST API (Pilot Dataset)
Query single-date or range classifications over the pilot dataset (2012–T-30 days). Each response includes the state label, rule ID, and rule version for auditability.
GET https://api.mindforge.tech/api/msd/v1/state?date=2020-02-24
Authorization: Bearer YOUR_EVAL_JWT
Response:
{
"data": {
"date": "2020-02-24",
"state": "systemic_stress",
"derived_from_rule": "systemic_stress_state",
"rule_version": "v6.0.0"
}
}// API key provided with pilot signup. Rate limit: 100 req/hour.
Embeddable Tile (React)
This demo uses the RiskSignalTile component. Pass any historical date to render specific classifications in your own dashboard or research application.
import { RiskSignalTile } from '@mindforge/risk-tile';
// Show COVID crisis classification
<RiskSignalTile date="2020-02-24" />
// Show August 2024 volatility
<RiskSignalTile date="2024-08-02" />
// Default: T-30 days (pilot window)
<RiskSignalTile />Ready to Evaluate?
Download 12+ years of historical classifications (daily_states.csv + episodes.csv) and test MRC against your own risk events. 60-day pilot access includes API key.
// Pilot dataset covers 2012–T-30 days.
Historical backtested episode examples (including April 2025). Research use only. Not investment advice. Past performance ≠ future results.