SparkTrade High-Vol Equity Signals (L/S, VIX ≥ 25)

## **SparkTrade High-Vol Equity Signals (Long-Short, VIX ≥ 25)**

Adaptive long-short scores for any “fast-market” moment—panic **or** euphoria.

## **Short Description**

Daily 0-100 prediction ranks on **2 000+** U.S. equities. This model was trained exclusively on regimes where VIX ≥ 25, high-volatility periods to capture outsized moves. Orthogonal to Fama-French (R² ≈ 0.10).

## **Detailed Description**

*SparkTrade High-Vol Equity Signals* is a machine-learning data feed that pinpoints which U.S.–listed stocks are most likely to surge (long) or plunge (short) whenever the market’s fear-or-euphoria gauge—the Cboe VIX—closes at or above 25. Unlike broad “all-weather” models that dilute signal quality when volatility spikes, this sleeve is **purpose-built** for fast markets. It therefore functions as a convex alpha overlay that complements both quant and discretionary portfolios.

## What makes it different 

1. **Regime-Specific Training** – Models ingest only those history slices where VIX ≥ 25, meaning the algorithms learn behaviors unique to stressed and momentum-charged tapes, not average conditions. 

2. **Multi-Factor Feature Set** – Over two decades of daily price, volume, fundamentals, micro-structure metrics, options-implied skew, ETF flow, and real-time sentiment are distilled into engineered factors before entering a gradient-boosted ensemble capped with an LSTM meta-learner. 

3. **Orthogonality** – Weighted-least-squares regression against the Fama-French 3-factor model returns an adjusted R² of roughly 0.10, indicating that 90 % of the sleeve’s variance is unexplained by classic market, size, or value betas. For allocators this means the data adds novel, diversifying return streams rather than recycling crowded factors.  ### Coverage & history 

**Universe** – 2 000+ U.S. equities with ≥ $300 M float-adjusted cap and consistent liquidity.

**Historical depth** – **January 1999 → present** (T-1 close).

**Delivery** – Secure Share file lands in Snowflake by **20:00 ET** every trading day (≈ 4 hours after the bell).

## Data dictionary

NAME TYPE DESCRIPTION

DATE DATE Date of prediction

TICKER STRING Stock ticker symbol (e.g., AAPL)

LONG_INDEX_SCORE INTEGER 0–100 scaled score (higher = stronger long)

SHORT_INDEX_SCORE INTEGER 0–100 scaled score (higher = stronger short)

LONG_RAW_SCORE FLOAT 0 to 1.0 unscaled signal before index conversion

SHORT_RAW_SCORE FLOAT 0 to 1.0 unscaled signal before index conversion

VOLATILITY_REGIME STRING "low", "high", or "auto" – or for ETF model "all weather"

MODEL_NAME STRING Source model name (e.g., "Low Vol", "AutoSwitch")

MODEL_VERSION STRING Version or release ID (for audit/debug)

**Update Frequency**

Trading days, file posted **by 20:00 ET**