In my previous analysis , I explored the structural trade-offs between pure growth equities via QQQ and options-derived income via JEPQ. While static back tests provided valuable insights into performance across historical market phases, I wanted to create a more rigorous quantitative analysis, taking into account the non-stationarity of financial markets. To this end, I developed an unsupervised 3-state Gaussian Hidden Markov Model (HMM) that dynamically identifies market regimes in real-time and allocates capital among QQQ, JEPQ, and Risk-Free Cash. 1. Methodology & Unsupervised State Discovery Feature Engineering & Emission Selection Rather than relying on ad-hoc technical indicators, the HMM relies on two mathematical features observed at daily frequency: 1. QQQ Daily Log Returns: 2. QQQ 5-Day Realized Volatility: These emissions capture both price momentum and local volatility clustering. Model Mechanics & Expect...
