Regime Detection for Factor Rotation: A Practical Guide

Introduction

Factor strategies that perform well in one market regime often fail in another. Momentum dominates in trending markets, while value shines in mean-reverting environments. Regime-aware factor rotation can dramatically improve risk-adjusted returns.

What Are Market Regimes?

Market regimes are persistent states characterized by distinct statistical properties:

• Volatility regimes: High vs low volatility
• Trend regimes: Trending vs range-bound
• Macro regimes: Growth vs recession, inflationary vs deflationary
• Liquidity regimes: Abundant vs constrained liquidity

Regime Detection Methods

1. Statistical Methods

Hidden Markov Models (HMM)

• Model the market as a finite state machine
• States transition with estimated probabilities
• Emissions are observable market variables

Advantages: Handles uncertainty, estimates transition probabilities Disadvantages: Requires tuning, sensitive to initialization

Example: Use yield curve shape (2-year vs 10-year rates) as emission to classify inversion vs normalization regimes.

2. Economic Indicators

Leading Indicators:

• Yield curve spread
• Credit spreads
• PMI readings
• Housing starts

Coincident Indicators:

• GDP growth
• Industrial production
• Retail sales

Lagging Indicators:

• Unemployment rate
• CPI inflation
• Corporate earnings

3. Market-Based Signals

Volatility-Based:

• VIX levels and changes
• Realized volatility
• Cross-sectional dispersion

Trend-Based:

• Moving average cross-overs
• Trend strength metrics
• Market breadth

Correlation-Based:

• Asset class correlations
• Factor correlations
• Cross-asset dispersion

Factor Rotation Framework

Step 1: Define Factor Universe

Common factors:

• Value: Book-to-market, earnings yield
• Momentum: Price momentum, earnings revision
• Quality: Profitability, debt ratios
• Low Volatility: Beta, idiosyncratic risk
• Size: Market cap, liquidity

Step 2: Identify Regime-Specific Factor Performance

For each regime, determine which factors historically outperformed:

Regime	Dominant Factors	Underperforming Factors
High Volatility	Quality, Low Volatility	Value, Momentum
Low Volatility	Momentum, Value	Low Volatility
Inversion	Momentum, Quality	Value
Normalization	Value, Low Volatility	Momentum

Step 3: Construct Regime-Conditioned Portfolio

1. Detect current regime
2. Allocate to dominant factors for that regime
3. Hedge or reduce exposure to underperforming factors
4. Rebalance on regime change (monthly/quarterly)

Practical Implementation

Hidden Markov Model for Yield Curve Regimes

from hmmlearn import hmm
import numpy as np

# Prepare data: yield curve spread (10-year - 2-year)
spread = ten_year - two_year

# Fit HMM with 2 states
model = hmm.GaussianHMM(n_components=2, covariance_type="diag")
model.fit(spread.reshape(-1, 1))

# Predict regime
regime = model.predict(spread[-1].reshape(1, -1))[0]

Simple Threshold-Based Regime Classification

def classify_regime(yield_curve_spread):
    """Classify regime based on yield curve spread."""
    if yield_curve_spread < 0:
        return "inversion"
    elif yield_curve_spread < 50:  # bps
        return "flat"
    else:
        return "normalization"

Factor Rotation Logic

def factor_rotation(regime):
    """Return factor weights based on regime."""
    if regime == "inversion":
        return {"momentum": 0.4, "quality": 0.4, "value": 0.1, "low_vol": 0.1}
    elif regime == "normalization":
        return {"momentum": 0.1, "quality": 0.2, "value": 0.4, "low_vol": 0.3}
    else:
        return {"momentum": 0.25, "quality": 0.25, "value": 0.25, "low_vol": 0.25}

Evaluation and Validation

Backtesting Considerations

• Walk-forward validation: Train regime model on historical data, test on forward period
• Regime lag: Regime detection has lag; account for signal delay
• Transaction costs: Frequent rebalancing increases costs
• Regime stability: Evaluate how often regimes change

Performance Metrics

• Sharpe ratio: Risk-adjusted returns
• Regime-specific returns: Performance in each regime
• Drawdown: Maximum drawdown during regime transitions
• Turnover: Trading frequency and costs

Common Pitfalls

Overfitting Regime Definitions

Too many regimes = overfitting. Start with 2-3 regimes:

• Volatility regimes (high/low)
• Trend regimes (trending/range)
• Macro regimes (expansion/recession)

Ignoring Transition Costs

Regime changes often coincide with market stress:

• Higher transaction costs during transitions
• Wider bid-ask spreads
• Reduced liquidity

Look-Ahead Bias

Ensure regime detection doesn’t use future information:

• Use only data available at decision time
• Point-in-time data for economic indicators
• Realistic signal delays

Related Reading

• Yield Curve Inversion as a Regime Classifier for Equity Factor Rotation — HMM-based regime detection with factor rotation
• Earnings Revision Momentum Decay in the Post-2023 Regime — regime-aware factor implementation
• Five Backtesting Pitfalls That Fake Your Sharpe — validation methodology

Further Reading

Upcoming articles will cover:

• Multivariate regime detection
• Dynamic factor models
• Regime-aware risk management