Methodology

We select US mega-cap equities from the S&P 100, filtered for sufficient options liquidity. Price, volume, and fundamental data sourced from standard market data providers with daily reconciliation checks.

Multiple cross-sectional factors are computed per stock spanning momentum, mean-reversion, fundamental quality, analyst sentiment, and volatility metrics. Each factor undergoes information coefficient (IC) testing; weak factors are pruned. Remaining factors are winsorized and normalized.

Multiple machine learning models are trained on the factor matrix. Model weights are determined dynamically based on out-of-sample performance and adjusted for the current VIX regime. Shrinkage is applied to blended estimates to reduce overconfidence.

Large-scale Monte Carlo simulation per stock generates model-implied volatility. This is compared against market-quoted implied volatility to compute a volatility edge metric that informs options strategy selection.

Multiple options strategies are evaluated per stock and re-priced using an evidence-based spread model trained on real options chain data. Strategies that appear profitable before execution costs but are unprofitable after are eliminated.

Position sizing uses a drawdown-adjusted methodology with sector diversification constraints. Options allocation is capped based on conviction tier. The result is a fully specified portfolio with conviction-tiered overlays.