Regime Detection: Market States and Transitions

Module: Fintech AI | Difficulty: Advanced

Market Regimes

• Bull: High returns, low volatility
• Bear: Negative returns, high volatility
• Sideways: Low returns, low volatility

Hidden Markov Model

Regime-Dependent Strategy

import numpy as np
from hmmlearn import hmm

class RegimeDetector:
    def __init__(self, n_regimes=3):
        self.model = hmm.GaussianHMM(n_components=n_regimes)
    def fit(self, returns, volatility):
        X = np.column_stack([returns, volatility])
        self.model.fit(X)
    def predict(self, returns, volatility):
        X = np.column_stack([returns, volatility])
        return self.model.predict(X)
    def get_regime_stats(self, returns, regimes):
        stats = {}
        for r in np.unique(regimes):
            mask = regimes == r
            stats[r] = {
                'mean_return': returns[mask].mean(),
                'volatility': returns[mask].std(),
                'frequency': mask.mean()
            }
        return stats

Research Insight: Regime-aware models outperform static models by 30-50% in Sharpe ratio. The key challenge is that regimes are not directly observable and must be inferred. The transition matrix captures the persistence of regimes, which is crucial for strategy timing.