Research on Optimized Time Series Momentum Strategy Based on Volatility Weighting - Evidence from S&P 500 Constituent Stocks

Abstract

Time series momentum (TSMOM) is a trend-following strategy that uses each asset's past price trajectory to construct trading signals, exhibiting persistent return premia worldwide. Standard equal-weighted TSMOM does not adjust for cross-sectional volatility differences across stocks, which can cause large drawdowns amid abrupt market swings. However, traditional equal weight time series momentum strategies ignore differences in individual stock volatility, leading to significant pullbacks when the market undergoes drastic reversals. To this end, this study constructed a Mean-Variance Optimization (MVO) time series momentum strategy under volatility target constraints. Based on daily data of S&P 500 index constituent stocks from 2015 to 2025, the model selects the Top 20 stocks with the strongest momentum signal and rebalances portfolios bi-monthly, with a risk aversion coefficient γ =1.2, a fixed annualized volatility target of 18% and strict no-leverage constraints. This article uses equal weight buy and hold portfolios and traditional equal weight time series momentum portfolios as dual benchmarks, introduces CAPM regression model to decompose source of strategy excess returns, and constructs a rolling sample backtesting framework with five years of training data and one year of out-of-sample testing, avoiding parameter overfitting problems. The empirical results indicate that MVO momentum strategy achieves dual optimization of returns and risks after volatility weighted optimization. The annualized return rate, Sharpe ratio, and Calmar ratio have significantly increased. CAPM regression results show that the optimized strategy generates obvious positive abnormal returns, proving that the excess returns of strategy stem from independent trend premiums rather than random market fluctuations.