SparkTrade Adaptive Regime-Switch Equity Signals (Long-Short, Auto)

## **SparkTrade Adaptive Regime-Switch Equity Signals (Long-Short, Auto)**

One feed that flips seamlessly between two signal models: (a) low-vol calm-market drift **and** (b) high-vol fast-market chaos based on the current regime (VIX level).

## **Short Description**

Daily 0-100 prediction ranks on **2 000+** U.S. equities. Two gradient-boosted models—one Low-Vol (VIX < 25), one High-Vol (VIX ≥ 25)—and a **transparent VIX-based rule** auto-selects the right engine each day. Designed as a core, all-weather alpha layer. Orthogonal to Fama-French (R² ≈ 0.11).

## **Detailed Description**

*SparkTrade Adaptive Regime-Switch Equity Signals* delivers a single, friction-free data feed that toggles between SparkTrade’s Low-Vol and High-Vol models using a **simple, rule-based switch**: if yesterday’s Cboe VIX closes below 25, the Low-Vol learner publishes; if ≥ 25, the High-Vol learner publishes. PMs get **all-weather coverage** without maintaining two separate pipelines.

## Why SparkTrade Clients Love It:

1. **Automatic context awareness** – No need to forecast volatility regimes in-house; the feed self-adjusts yet remains auditable (VIX rule is transparent and back-testable). 

2. **Dual ensemble stack** – Each regime uses gradient-boosted trees & random forests trained on 20 + years of price/volume, fundamentals, micro-structure metrics, options skew, ETF flow, and real-time sentiment. No RNNs, no black-box sequence nets—just robust supervised learning. 

3. **Orthogonality & diversification** – Weighted-least-squares regression vs the Fama-French 3-factor model shows **adjusted R² ≈ 0.11**, so ~89 % of variance is unexplained by classic factors. Adds **fresh return vectors** to factor stacks and discretionary books alike.

## Coverage & history 

## **Universe**

2 000+ U.S. tickers (≥ $300 M float-adjusted cap, liquidity screens).

## **Historical depth**

**Jan-1999 → present** (T-1 close).

## **Delivery cadence**

Snowflake Secure Share file posts **by 20:00 ET** every trading day.

## 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**