Trend and momentum

Assets that have gone up tend to keep going up, for a while. This pattern has been documented across equities, bonds, currencies and commodities over more than a century of data, and it is the engine behind most of the trend-following strategies in the vault.

Two kinds of momentum

Time-series momentum (trend following) looks at each asset on its own: go long if its own past return is positive, short if negative. Moskowitz, Ooi and Pedersen (2012) documented it across 58 futures markets with a 12-month lookback.

Cross-sectional momentum ranks assets against each other: buy the recent winners, sell the recent losers, and stay roughly market-neutral. Jegadeesh and Titman (1993) found it in US stocks using 3- to 12-month formation periods, typically skipping the most recent month, which tends to reverse.

A minimal time-series momentum rule

For each asset $i$, with $r_{i,t-L:t}$ its return over the lookback $L$ and $\hat\sigma_{i,t}$ its estimated volatility:

The volatility scaling matters as much as the signal: it makes each asset contribute similar risk and keeps the portfolio’s volatility roughly constant. Moving-average crossovers, breakouts of $N$-day highs (the Donchian/“turtle” rule), and the sign of a regression slope are all variations of the same idea.

Why might it work?

• Underreaction. Information diffuses slowly; investors anchor and update gradually.
• Overreaction later. Herding and feedback trading extend trends past fair value.
• Risk transfer. Hedgers pay to offload price risk; trend followers provide the insurance and are paid over time.

How it behaves

Trend following has a characteristic payoff: many small losses in choppy, range-bound markets, and occasional large gains in sustained moves. That makes its returns positively skewed and its performance in crises historically good (“crisis alpha”), because crises tend to be extended trends.

Cross-sectional momentum is different. It is prone to sudden, violent momentum crashes, typically when a beaten-down market rebounds sharply and the short leg (last year’s losers) rallies hardest — 2009 is the textbook example.

Check your understanding

• Why does volatility scaling make a portfolio’s Sharpe more stable even if it doesn’t change the signal?
• Why is trend following positively skewed while mean reversion tends to be negatively skewed?

Then browse trend-following specs and momentum specs in the vault to see how people implement these ideas in Pine Script.