MINDFORGE INTELLIGENCE
Market Risk Classifier
THE DATA BUYER
BRIEF
[ Orthogonal Regime Data — Inputs, Outputs, Evaluation ]
SYSTEM CONTEXT
Market Risk Classifier (MRC) : Daily regime classification. Delivered pre-market. Research tool, not trading signal.
INPUT CLASS : Space weather telemetry from NOAA and NASA — the same input class Federal Reserve research (Krivelyova & Robotti, 2003) tied to equity returns.
CORE FINDING
A daily regime feed nobody else has.
Orthogonal to VIX, options flow, sentiment, and macro.
Backtest: 2012–2024 • 92.31% precision (CRISIS, 12/13, backtest) • 2025–2026 out-of-sample • mindforge.tech/terms
INPUTSWHAT THE DATA ISInputs, Processing, Output, Delivery
INPUTS
Space weather telemetry from NOAA and NASA monitoring networks — the same input class Federal Reserve research (Krivelyova & Robotti, 2003, FRB Atlanta WP 2003-5b) tied to equity returns.
ORTHOGONALITY
Structurally uncorrelated to VIX, options flow, sentiment, and macro indicators — a genuinely independent axis of information.
PROCESSING
Rules-based, not ML. Fixed thresholds with rolling 252-day normalization. Walk-forward only.
AUDITABILITY
Versioned rules with full audit trail. No model drift, no retraining, no hindsight tuning.
OUTPUT
5 daily regime states: STABLE, SHIFTING, ELEVATED, SHOCK, CRISIS. Single classification per day. No intraday noise.
HISTORY
5+ years of historical classifications included for evaluation. Extended history (35 years) available under pilot agreement.
DELIVERY
Delivered pre-market every trading day.
CHANNELS
API (REST, JSON) · Email (rich HTML) · CSV (S3, SFTP) · Webhook on state change (<5 min latency)
Cross-era CRISIS 1990–2011 backtest history available at mindforge.tech/market-risk-classifier/systemic-stress-history.
STATESTHE 5 REGIME STATES2012–2024 backtest • 2025–2026 out-of-sample
| State | What it classifies | Precision | Hits/Total | Freq/Yr |
|---|
| CRISIS | Crisis-level regime: COVID, carry unwind, tariff shock | 92.31% | 12/13 | 0.4 |
| SHOCK | VIX expansion without systemic collapse | 85.07% | 57/67 | 1.9 |
| SHIFTING | Regime pivot point: transition between states | 92.75% | 64/69 | 2.0 |
| ELEVATED | Elevated conditions, advisory-level warning | 75.51% | 37/49 | 1.4 |
| STABLE | Benign regime: reduced risk posture appropriate | Meta | — | ~87% of days |
WHY 5 STATES, NOT RISK-ON / RISK-OFF
The system distinguishes a SHOCK (VIX expansion) from CRISIS (crisis-level). Oct 2024 was SHOCK, not CRISIS — no forced liquidation on a temporary expansion. A binary toggle would have over-reacted; the granularity is what makes the feed usable daily.
PROOFLIVE PROOF — IRAN SHOCKMar 2026 • production classifications • not backtested
THE TAPE
STABLE (Feb 27) → SHIFTING (Mar 4) → ELEVATED (Mar 6) → SHIFTING (Mar 13) → SHOCK (Mar 21, eff. Mar 23) → SHIFTING (Mar 25) → RESOLVED (Mar 31). Six transitions, 28 trading days.
CLASSIFIED PRE-MARKET
Every transition delivered before the trading session opened, by 07:30 ET — production classifications, not back-stamped.
PEAK & TROUGH
VIX peaked 31.05 (Mar 27, CBOE close); S&P trough 6,344 (Mar 30). Same deterministic rules validated on 2012–2024 — no overrides.
PROOFRESOLVED AT VIX 31Mar 31 2026 • the orthogonality moment
THE CALL
On Mar 31 MRC classified RESOLVED with the prior-session VIX close at 30.61 — while headlines still ran crisis narratives and price-derived signals said stay defensive.
WHY IT IS POSSIBLE
MRC’s primary inputs are environmental, on timescales independent of price. When those inputs normalized, the system said so — regardless of where VIX sat.
WHAT FOLLOWED
By Apr 8, VIX was back near 21 and the S&P had recovered +254 points. A VIX-reading signal cannot call normalization while VIX is still elevated — by the time it confirms, the move is behind you.
EVALUATIONPRIOR EPISODES & INTEGRATIONFrom canonical records • 60-day pilot terms
10
Trading Days Early
COVID-19 • SS 2020-02-24
Before first s&p 500 circuit breaker
1
Trading Day Early
Carry unwind • SS 2024-08-02
Before flash crash / vix spike to 66
22
Trading Days Early
Tariff episode • VS 2025-03-06
Before tariff-driven vix peak
60-DAY PILOT
Full production feed. 60-day license.
BACKTEST
5+ years historical classifications (extended history under pilot).
SCORING
Pre-defined success criteria.
COMMITMENT
No commitment beyond pilot.
API
REST/JSON, historical + live endpoints.
EMAIL
Rich HTML, daily pre-market.
CSV
Daily drop (S3, SFTP, email).
WEBHOOK
State-change push (<5 min latency).
AUDIT TRAIL
Versioned rules with integrity verification.
EPISODES
Episode catalog with exact dates and values.
METHODS
Walk-forward methodology documentation.
COMPLIANCE
Research-only framing, no advice language.
SAMPLE OUTPUT (API RESPONSE)
{
"date": "2026-03-21",
"state": "volatility_spike",
"previous_state": "turning",
"delivered_at": "2026-03-21T07:28:00-05:00",
"rule_version": "2.0.0"
}ACCESSPILOT
VERIFY IT LIVE
60 Days. Full Access.
INCLUDES
- • Full production feed for 60 days
- • 5+ years historical classifications
- • Pre-defined scoring & success criteria
- • Due diligence documentation package
QUALIFICATION
Institutional 60-day pilot for qualified research teams.
RESTRICTED TO INSTITUTIONAL USE
