ml4t-regime-awareness
Regime Awareness
Markets alternate between regimes (low/high volatility, trending/mean-reverting, risk-on/risk-off). Regime detection for diagnostics and risk scaling is reliable. Regime detection for market timing is not.
The Problem
Regime-switching models promise to predict when to be in or out of the market. In practice, regime transitions are identified with high confidence only after they have already occurred. A model that correctly labels the March 2020 crash as "crisis" does so 2-4 weeks late, after the drawdown has already happened. Trading on regime predictions produces whipsaw losses and underperforms a regime-conditioned but always-invested approach.
The correct use of regimes is as a conditioning feature: scale risk, adjust position sizes, and evaluate strategy performance per regime - but stay invested.