Regime-Optimal Trend Following

Is there a way to enhance time series momentum by considering both trend regime and expected risk-adjusted performance during each state? In his March 2026 paper entitled "Rethinking Trend Following: Optimal Regime-Dependent Allocation", Valeriy Zakamulin describes and tests his optimal regime-dependent allocation (OPT) strategy, which:

1. Determines the trend following regime based on either 2-regime (Bull/Bear) or 4-regime (Bull/Correction/Bear/Rebound) models.
2. Assigns asset exposures that historically maximize gross Sharpe ratio during each regime, with 100% long for the regime with the highest estimated Sharpe ratio and exposures scaled down across other regimes according to their relative signal-to-noise ratios. Regimes with sufficiently negative estimates may reach 100% short, but regimes with weak/very noisy signals have weights close to zero.

Backtests employ data:

• Since July 1926 for value-weighted U.S. equity indexes for the overall market, the largest and smallest fifths of stocks, and the fifths of stocks with the highest and lowest book-to-market ratios.
• Since January 1975 (1977 for Canada) for value-weighted developed market equity indexes of 14 other countries.
• For robustness, since July 1963 for 18 U.S. stock portfolios formed on multiple factors/firm characteristics.

Sharpe ratio estimates derive from expanding windows of historical training data initially through 1968 for the long U.S. samples, December 2003 for the other country samples and 1997 for the U.S. factor/firm characteristics portfolios. Using the specified datasets through December 2025, he finds that: