How We Build the Forecast
A three-stage pipeline, from raw options data to calibrated surface forecasts. Designed by practitioners who spent careers at sell-side vol desks and quantitative asset managers.
Three-Stage Model Ensemble
Each stage addresses a specific failure mode in conventional vol forecasting. No single model handles all three; the ensemble is the architecture.
Where the Signals Come From
Six primary signal categories, each contributing to the surface forecast through the model ensemble.
End-of-day and intraday options chain snapshots across listed markets. Strike and expiry coverage from 1-week to 2-year tenors where liquidity permits.
Net directional positioning from large-lot and dealer flow data. Put/call ratios, skew-adjusted positioning, and open interest changes parsed for directional signal.
Multi-frequency realized vol estimates (5-min, 30-min, close-to-close) over multiple lookback windows. Fed into HAR-family models for the realized vol baseline.
Systematic tracking of macro data release outcomes vs. consensus forecasts. Positive or negative surprise streaks in key indicators have documented effects on vol regimes.
Structured parsing of financial news volume and sentiment scores by topic category. Used as a second-order signal for the sentiment blend, not the primary surface driver.
Rolling correlation matrices across asset class pairs with regime-break detection. Provides context for adjusting single-asset surface forecasts based on cross-market dynamics.
Walk-Forward Validation Protocol
All model configurations are evaluated on out-of-sample walk-forward data, not in-sample fits. No parameter tuning on test periods.
Research note: Metafide is a research and analytics platform. Validation statistics are derived from internal back-tests on historical out-of-sample data. They are provided for informational purposes only and do not constitute a guarantee of future forecast accuracy. Metafide is not a registered investment adviser or broker-dealer. Outputs are for analytical and informational use by institutional professionals only.
Built by Vol Practitioners
12+ years building risk systems and vol research infrastructure for institutional desks in quantitative finance.
Derivatives research background spanning sell-side and buy-side. Specialist in options pricing theory and cross-asset signal development.
Real-time financial data pipeline engineering. ML infrastructure for quantitative systems and low-latency market data processing.
Read the Research Behind the Models
Our public research articles detail the specific techniques behind the platform. No paywalls.