The research question
The foundational study documents return continuation across 58 liquid futures and forwards. Later long-history work tests related rules across more than a century, which makes this one of the broader evidence bases in systematic investing.
The useful question is not whether a chart once went up. It is whether the hypothesis has a clear economic mechanism, appears outside one hand-picked sample, can be reconstructed with information that was available at the time, and remains plausible after realistic implementation frictions.
Minimum viable rule
Take the sign of each market's trailing excess return and scale exposure inversely to recent volatility.
Universe: Liquid equity-index, government-bond, currency and commodity futures. A credible test fixes the universe, timestamps, missing-data policy and rebalance convention before model selection. Results should be shown both gross and net of an explicit cost model.
Implementation audit
A research-grade implementation defines the continuous futures series, roll convention, lookback, volatility estimator, rebalance schedule and portfolio risk target before observing results. Multiple markets matter more than a cosmetic search for the best lookback.
At minimum, an audit should report turnover, worst peak-to-trough loss, recovery time, exposure concentration and sensitivity to neighboring parameter choices. A strategy that works only at one exact lookback or threshold deserves a lower level of confidence.
How the idea can fail
Sideways markets can create repeated whipsaws. Leverage, futures rolls, gap risk and rising correlations can turn a smooth backtest into a materially rougher live result.
Failure conditions belong in the strategy definition. They provide a disciplined reason to investigate or stop, instead of changing the story after a loss.
Primary source
Time Series Momentum
Moskowitz, Ooi & Pedersen (2012), Journal of Financial Economics 104(2), 228–250.
DOI record
This page is an independent educational synthesis. It does not reproduce the paper and does not claim that published historical results are currently achievable.