Time-series momentum
MomentumAdvanced
Time-series momentum looks only at an asset's own past return. If that return over a lookback, classically 12 months, is positive, the rule is long. If it is negative, the rule is short or flat. It is the futures version of trend following, and it is not the same thing as ranking stocks against each other.
Not investment advice. Nothing here is a recommendation to buy or sell anything. Backtests and historical studies are not guarantees, and a rule that worked in one sample can fail in the next. Trading can lose money. Links are for education, not endorsements. Do your own research.
- Rules
- On a monthly schedule, compute the excess return over the past 12 months for each futures market, go long if the sign is positive and short if it is negative, and size each position by inverse volatility so one noisy market does not dominate. Hold for about a month and repeat. A simpler long-only cousin is a 10-month moving-average filter on an ETF: hold it when price is above the average, otherwise hold cash or bonds.
- Where AI fits
- The portfolio construction is where automation matters: volatility scaling, correlation caps, and a rule for when many markets flip together. AI can monitor that and flag when the book has become one big risk-on bet. It should not be used to retune the 12-month window every month. An LLM is useful for reading the original paper and turning the stated rule into testable code.
- Markets
- A diversified set of liquid futures and forwards: equity indexes, bonds, currencies, and commodities.
- Timeframes
- Monthly rebalance on daily data. This is not an intraday strategy.
- Risks
- Trends reverse, and a book of correlated positions can draw down together when the macro regime changes. Shorting futures has financing, roll, and gap risk. A single-asset version drops the diversification the research actually used.