Volatility-Adjusted ETF Momentum

[WITH CODE] Investigating the performance of traditional momentum and volatility-adjusted momentum

Reviewed and updated 20 July 2026

Hello!

Today’s post is a deep dive into an underappreciated yet highly practical refinement in systematic investing: volatility-adjusted momentum. While traditional momentum strategies rank assets based on raw returns over a past period, volatility-adjusted momentum modifies this signal by dividing returns by the volatility over the same lookback window.

In other words, it favors high-return assets that achieved those returns with lower volatility. In this post, I explore how this adjustment performs in practice across various lookback and holding periods, using a 32-ETF universe. Backtested return-to-volatility ratios rise above 1 in the medium- and long-horizon tests.

Importantly, this is not a pure replication of an academic paper. Instead, it builds on the ideas presented in the November 2024 SSRN working paper "Shades of Momentum: Alternative Momentum Metrics and their Dissipation in Indian Equities" by Rajan Raju (highlighted in this Recent Academic Research post). The author finds that volatility-adjusted momentum delivers the highest Sharpe ratio among the four signals in the paper’s Indian-equity long-short factor comparison. These findings influenced how I tested and interpreted the strategies presented here.

This post is for paid subscribers only, and it includes:

• A clear explanation of both momentum types

• Backtested results across multiple lookback and holding periods

• Strategy comparisons, with insights on when each works best

• Research companion

Let’s get into it.

---

Paper Summary

In the working paper "Shades of Momentum," Raju identifies a core issue with traditional momentum: it rewards raw return without penalizing risk. As a result, assets with high returns but unstable paths may be selected into the portfolio. By contrast, volatility-adjusted momentum ranks assets by their return-to-volatility ratio over a fixed lookback window. The ETF analysis below reports a trailing return-to-volatility ratio that assumes a zero cash rate, not an excess-return Sharpe ratio.

Why might this adjustment work better, especially at shorter horizons? Two reasons:

1. Short-term price movements are noisy. Some strong performers may be due to transitory spikes, not persistent trends.

2. Volatility is often autocorrelated. Assets with high recent volatility tend to remain volatile, increasing the likelihood of reversion or instability.

The paper uses a 12-month formation period that excludes the most recent month, then studies what happens as holding periods extend from 3 to 12 months. In its Indian-equity sample, volatility-adjusted momentum produces stronger risk-adjusted results than the academic variant. That is related to, but not the same design as, the ETF tests below.

The dissipation analysis in Shades of Momentum is particularly useful. It asks how momentum exposure changes as the holding period lengthens, with exposure to the traditional momentum factor (WML) as a central measure. The ETF exercise instead varies both the lookback and holding period, so its results should be treated as a separate historical implementation rather than a replication.

This naturally leads to the question: does this result hold up using U.S. equity, foreign equity, commodity, and bond ETFs?

---

Methodology

To answer that question, I built and tested two strategies:

1. Traditional Momentum: Ranks assets by total return over a trailing lookback window.

2. Volatility-Adjusted Momentum: Ranks assets by return divided by realized volatility over the same lookback window.

Dataset: I used adjusted daily closing prices for 32 ETFs from January 2010 through April 2025. The displayed results use adjusted-close observations retrieved from Yahoo Finance on 19 July 2026; Tiingo adjusted closes are the supported reader-access route and may differ. The ETFs cover US equities, foreign equities, U.S. equity sectors, commodities, and bonds. May 2025 is excluded because the month was incomplete.

Lookback Periods Tested:

• 3 months

• 6 months

• 12 months

Holding Periods Tested:

• 1 month

• 3 months

• 6 months

The portfolio consists of the top 10 ranked ETFs by each signal. Portfolios are rebalanced at the end of each holding (rebalance) period. The analysis uses non-overlapping holding-period returns and adjusted closes, and estimates volatility from monthly returns rather than price levels. No transaction costs or slippage are included, and returns are not leveraged. No benchmark or cash return is included. The reported return-to-volatility ratio annualizes arithmetic return and volatility with a 0% cash rate; it is not an excess-return Sharpe ratio.

---

Short-Term Performance (3-Month Lookback, 1-Month Holding)