Lesson T10.2 · 2 of 6 · Strategy, testing and the trading journal
Trend following: the evidence from 137 years of data (1880-2016)
Trend following means buying the markets that have risen over the past few months and shorting the ones that have fallen. An AQR study simulated the strategy across 67 markets from 1880 to 2016 and found it in profit in every decade, after assumed costs, but with falls of up to 25% that lasted a long time.
Terms in this lesson
- Price momentum
- How much the stock has risen or fallen in recent months: the force of the move.
- Futures
- An agreement to buy or sell something on a future date, at a price fixed today.
- Maximum drawdown
- The largest fall in price from a peak to the lowest point after it.
- Sharpe ratio
- How much extra return you got for each unit of risk.
What trend following is (time series momentum)
Definition
What does a trend-following strategy do?
It goes in the direction the price has moved over the past few months.
If oil has risen over the past year, the strategy buys it. If German government bonds have fallen, it shorts them, betting on further falls. Positions are reset at fixed intervals, monthly for example, using the same rule. It is usually done through futures contracts on equity indices, bonds, commodities and currencies.
The lesson on ROC and momentum mentioned the 2012 study by Moskowitz, Ooi and Pedersen: across 58 futures contracts, returns over the previous 1-12 months tended to persist. The obvious question was whether this was just a quirk of the last few decades. A study published in 2017 went looking much further back.
The rest of this lesson is in the subscription
The subscription unlocks every chapter of Intermediate, Trading school, Investor psychology and Advanced, with their exercises and unlimited reviews.
Create a free accountThe AQR study: 137 years, 67 markets
Brian Hurst, Yao Hua Ooi and Lasse Pedersen of the investment firm AQR published “A Century of Evidence on Trend-Following Investing” in The Journal of Portfolio Management in autumn 2017. They rebuilt the strategy month by month, from January 1880 to December 2016.
Watch out
How was the simulation built?
1-, 3- and 12-month signals, with assumed costs and fees.
Each market got a signal from its 1-, 3- and 12-month returns, in equal parts, and the portfolio aimed for volatility of 10% a year. The results are after hypothetical fund fees of 2% a year plus 20% of gains, and after trading costs assumed to be twice the estimate for recent years in 1993-2002 and six times that estimate in 1880-1992. The authors write that they are not claiming the strategy could have been run this way in the 1880s, and that the cost estimates are highly uncertain.
The headline result: after the assumed costs and fees, the simulated strategy made money in every decade. Taken market by market, before costs and fees, its average return was positive in all 67 markets, with an average Sharpe ratio of about 0.4. Its correlation with equities and bonds was low, over the whole period and in each decade.
The weak periods for trend following
In profit every decade doesn’t mean painless. The authors write that the strategy had significant drawdowns, losing up to 25%, over extended periods. Anyone following it would have gone a long time without a new high in their account.
The authors also say the strategy can do badly in very fast sell-offs such as the 1987 stock market crash: a signal worked out over whole months can turn too late. And from late 2008 to mid-2014, the average correlation between markets rose sharply. When markets rise and fall together, there are fewer genuinely different trends for the strategy to bet on.
Why trend following has many small losses and a few big wins
Trend rules cut the trades that go nowhere and hold on to the ones that work. To show the shape of the results, we ran an illustrative simulation for this lesson, not a study: 400 randomly generated price series (1,000 days each, 1.5% daily volatility, no real trend at all, no costs). The rule: buy when the price closes above the highest close of the previous 50 days, with a stop 3 × the daily volatility below entry, raised after every new high.
Share of trades in each result band, in the simulation on random prices: 25.2% lose 1R or more, 21.5% lose between 0.5R and 1R, 16.5% lose less than 0.5R, 18.6% gain up to 1R, 9.2% between 1R and 2R, 4.3% between 2R and 3R, 3.6% between 3R and 5R and 1.2% more than 5R.
Lessons are educational. They are not a recommendation to buy or sell. Investing involves risk and you can lose money. Sources: B. Hurst, Y. H. Ooi, L. H. Pedersen, “A Century of Evidence on Trend-Following Investing”, The Journal of Portfolio Management 44(1), Fall 2017 (January 1880 - December 2016, 67 markets, construction and costs, results by decade and by market, average Sharpe ratio of about 0.4 per market, drawdowns of up to 25%, 8 out of 10 stress periods, implementability in the 1880s, 1987, correlations in 2008-2014); T. Moskowitz, Y. H. Ooi, L. H. Pedersen, “Time Series Momentum”, Journal of Financial Economics, 2012. The distribution of trades is an illustrative simulation made for this lesson, not a published study (400 random series of 1,000 days each, 1.5% daily volatility, no costs).