This note focuses on one key parameter in the design of a trend-following strategy: the speed or lookback window of its underlying trend- signals. We assess the return sensitivity of a generic trend-following strategy to different model speeds. We report our findings for the exceptional past year 2022, as well as for every calendar year since 2000 in general.

For that purpose, we simulate the returns of a generic trend-following model over a continuous sequence of lookback windows ranging from a few days up to five years. We demonstrate that over the past 23 years an equivalent lookback window of around one calendar quarter (63 trading days) did allow to best capture the trend inefficiencies across the most liquid financial asset classes (equities, bonds, short rates, and currencies) and commodity markets. Moreover, such “medium- term” lookback period not only produced the strongest risk-adjusted returns for the last year and over the long term.

It also provided, on average, the strongest portfolio diversification benefits in times of falling equity markets or rising interest rates without sacrificing – in contrast to shorter-term trend models – the attractive upside potential in less challenging equity and interest rate regimes.

In other words, a medium-term trend-following strategy has provided superior “smart diversification” characteristics against a variety of risk factors, such as equity market risk, interest rate and inflation risks, but also commodity, currency and volatility risks.

We also note that the trend-following industry apparently sought to capture faster trends until the mid-2000s, and likely operated with a longer average lookback from 2007 onward. We conclude that the trend-following industry as an aggregate has apparently operated with lookback parameters that are very close to the in-sample optimum over the past decade.