Futures Trend Following with Time Series Momentum and Continuous Forecasts
Summary
This study compares two systematic futures trend approaches across indices, bonds, currencies, and commodities: a one-year time-series momentum rule and a continuous-forecast framework inspired by systematic trading methods. The first takes long or short exposure based on trailing returns and scales positions by volatility. The second turns moving-average differences and Donchian breakouts into risk-adjusted forecasts, rescales and caps them, then combines rules with equal or bootstrapped weights. Position sizes are adjusted for instrument volatility, transaction costs, and a portfolio-level volatility limit.
The author reports stronger risk-adjusted performance before June 2010 than afterward, with both approaches weakening in the later sample. The study also describes a portfolio adjustment during the 2008–09 crisis and discusses diversification and crisis behavior as possible reasons trend following can remain useful. These findings are historical and depend on the sample and data construction. The continuous futures series’ stitching method is unclear, carry forecasts are omitted, and the bootstrapped optimization uses a limited number of blocks; the author proposes longer data and walk-forward analysis as follow-up work.
Key ideas
- Time-series momentum uses trailing annual returns to set the next month’s direction and scales exposure by volatility.
- Continuous forecasts convert moving-average and Donchian signals into capped, risk-adjusted estimates rather than binary entries.
- Forecast rules can be combined with equal weights or weights estimated from bootstrapped in-sample blocks.
- Instrument-level sizing and portfolio-level volatility scaling are used to manage total risk.
- The reported strategy performance weakens after June 2010, and data stitching, omitted carry signals, and limited bootstrap samples constrain the conclusions.
Tags
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.