What Is Trend Following, and Does It Still Work?

A rule that held US stocks only after a month-end level above the 10-month average made 9.6% a year from January 1928 to August 2026, before costs. Holding stocks throughout made 10.1%. The rule's worst fall from a peak was 42.9%, against 83.7% for holding.

That rule is trend following in its simplest form: hold what is rising and step aside from what is falling. This guide defines the idea, shows all 73 exits the rule made and checks period by period whether it still works.

Trend following holds an asset while its price trend is up and holds cash while the trend is down. It reacts to prices that have already moved and makes no forecast.

The trend is usually read from a moving average, the average of recent closing prices. A price above its average counts as an uptrend. A golden cross is a related signal built from two averages.

Such rules are trading signals: fixed rules that turn price data into a hold or sell instruction. Momentum investing is a close relative that ranks assets against each other by past return.

The test uses a 10-month moving average. We fixed the length before running the test and did not fit it to this data. Four other lengths appear further down, so the choice can be checked against the alternatives.

The rule changed position 146 times in 98.7 years, which is 73 exits and 73 returns.

The chart shows $1 invested in January 1928 on a logarithmic axis, where each gridline is 10 times the one below. Holding grew to $13,565 and the rule to $8,742. That is 10.12% a year against 9.63%, a gap of 0.5 points.

The rule's edge sits in the falls. From the August 1929 peak, holding lost 83.7% by June 1932, while the rule's worst fall was 42.9%, reaching its low in February 1933.

A drawdown is a fall from a peak to the next low. The 1929 crash produced the deepest in this record.

In 2008 the rule made 1.6% and holding lost 36.7%. In 2022 the rule lost 10.9% and holding lost 19.9%. In 2020 the rule made 19.4% against 24.1%.

Each bar is one exit, in date order: the Treasury bill return minus the market's return while the rule was out, in percentage points. Green bars above the line are exits that helped. The 73 exits lasted from 1 to 34 months, with a median of 2.

15 exits helped and 58 cost money. 7 avoided a market fall of more than 10%.

They started in November 1929, September 1937, July 1969, December 1973, November 2000, May 2002 and January 2008. The first, a 34-month exit, was worth 65.9 points.

The costly exits were mostly brief. All 38 that lasted one or two months cost money. Of the 35 that lasted three months or more, 15 helped.

The largest cost was a two-month exit from March 1933 that missed a 43.2% market gain, a loss of 43.1 points.

The chart splits the record into 11 periods and shows the rule's yearly return minus the yearly return from holding. The rule was ahead in four: 1928 to 1929, the 1930s, the 1970s and the 2000s. In each, the market's worst fall was 33.2% or more.

In the other seven periods the market's worst fall was 29.9% or less. The rule trailed by 1.2 to 5.4 points a year in them.

From 1928 to 1999 the rule made 9.96% a year against 10.72%. From 2000 to August 2026 it made 8.75% against 8.54%, with worst falls of 17.6% and 50.3%. Since 2010 it made 9.48% against 14.28%, with worst falls of 17.6% and 24.8%.

Across the full record there are 1,065 overlapping 10-year windows. The rule came out ahead in 389 of them, or 36.5%. Overlapping windows are not 1,065 separate tests.

A different test is the same rule on 15 exchange-traded funds (ETFs), funds that trade like shares. Nine hold one US sector each: materials (XLB), energy (XLE), financials (XLF), industrials (XLI), technology (XLK), consumer staples (XLP), utilities (XLU), health care (XLV) and consumer discretionary (XLY).

The other six are the S&P 500 ETF (SPY), the Nasdaq-100 ETF (QQQ), the gold ETF (GLD), the long-dated Treasury bond ETF (TLT), the non-US developed-markets ETF (EFA) and the small-company ETF (IWM).

Each fund starts at its first signal, between November 1993 and September 2005. The rule's worst fall was smaller in all 15, by 4.3 to 56.9 points.

Its return beat holding in 6 of the 15. The median difference was 0.81 points a year in holding's favour.

Only 7 exits avoided a fall of more than 10%, which is 7 events in 98 years. The test covers one country's stock market with tax left out.

A backtest tests a rule on past data. Quant investing for beginners covers what such tests can and cannot show.

An illustrative cost of 0.1% per switch lowers the yearly return from 9.63% to 9.47%. At 0.25% it is 9.23%. These are assumed costs, not measured ones.

Other lengths gave other answers. From January 1928, 6, 8, 12 and 14 months gave 9.27%, 8.91%, 9.81% and 9.59% a year, against 9.63% for 10.

Worst falls ran from 42.9% to 53.8%. The 10-month length had the smallest worst fall of the five.

A trend rule acts after the price has moved. By the end of October 1929, the month-end at which the rule first sold, the market was already 23.7% below its end-August level.

The gains come in falling years. In the 24 calendar years when holding lost money, the rule did better in 17. In the 74 years when holding gained, it did better in 5.

Over 98 years the rule roughly halved the worst fall and gave up about half a point a year. Since 2010 it has given up 4.8 points a year for a worst fall of 17.6% against 24.8%.

This explainer is part of the YX Insights Academy. Each one takes a single idea and checks it against real data.

The same approach runs through everything else we publish:

Good places to start on the website:

Trend following is a rule for holding an asset while its price is rising and leaving it when the price turns down. The test uses a 10-month moving average: stocks are held next month if the latest close is above the average of the last 10 month-ends, otherwise Treasury bills.

From January 1928 to August 2026, a 10-month rule made 9.63% a year against 10.12% for holding US stocks, with a worst fall of 42.9% against 83.7%. It trailed holding in 7 of 11 periods and by 4.8 points a year since 2010. One country's record cannot settle the question.

Trend following compares one asset with its own past, for example holding it while its price sits above its 10-month average. Momentum investing compares assets with each other, holding those with the highest past returns. The first needs one price series; the second needs a group of assets to rank.

Over 98.7 years the rule changed position 146 times, which is 73 exits from stocks and 73 returns, about 1.5 changes a year. It held stocks in 72.9% of months. Costs were not measured, but an assumed 0.1% per switch lowered the yearly return from 9.63% to 9.47%.

DISCLAIMER: This article is strictly educational. Any information or analysis in this note is not an offer to sell or the solicitation of an offer to buy any securities. Nothing in this note is intended to be investment advice and nor should it be relied upon to make investment decisions. Any opinions, analyses, or probabilities expressed in this note are those of the author as of the note's date of publication and are subject to change without notice.

A rule that held US stocks only after a month-end level above the 10-month average made 9.6% a year from January 1928 to August 2026, before costs. Holding stocks throughout made 10.1%. The rule's worst fall from a peak was 42.9%, against 83.7% for holding.

That rule is trend following in its simplest form: hold what is rising and step aside from what is falling. This guide defines the idea, shows all 73 exits the rule made and checks period by period whether it still works.

Trend following holds an asset while its price trend is up and holds cash while the trend is down. It reacts to prices that have already moved and makes no forecast.

The trend is usually read from a moving average, the average of recent closing prices. A price above its average counts as an uptrend. A golden cross is a related signal built from two averages.

Such rules are trading signals: fixed rules that turn price data into a hold or sell instruction. Momentum investing is a close relative that ranks assets against each other by past return.

The test uses a 10-month moving average. We fixed the length before running the test and did not fit it to this data. Four other lengths appear further down, so the choice can be checked against the alternatives.

The rule changed position 146 times in 98.7 years, which is 73 exits and 73 returns.

The chart shows $1 invested in January 1928 on a logarithmic axis, where each gridline is 10 times the one below. Holding grew to $13,565 and the rule to $8,742. That is 10.12% a year against 9.63%, a gap of 0.5 points.

The rule's edge sits in the falls. From the August 1929 peak, holding lost 83.7% by June 1932, while the rule's worst fall was 42.9%, reaching its low in February 1933.

A drawdown is a fall from a peak to the next low. The 1929 crash produced the deepest in this record.

In 2008 the rule made 1.6% and holding lost 36.7%. In 2022 the rule lost 10.9% and holding lost 19.9%. In 2020 the rule made 19.4% against 24.1%.

Each bar is one exit, in date order: the Treasury bill return minus the market's return while the rule was out, in percentage points. Green bars above the line are exits that helped. The 73 exits lasted from 1 to 34 months, with a median of 2.

15 exits helped and 58 cost money. 7 avoided a market fall of more than 10%.

They started in November 1929, September 1937, July 1969, December 1973, November 2000, May 2002 and January 2008. The first, a 34-month exit, was worth 65.9 points.

The costly exits were mostly brief. All 38 that lasted one or two months cost money. Of the 35 that lasted three months or more, 15 helped.

The largest cost was a two-month exit from March 1933 that missed a 43.2% market gain, a loss of 43.1 points.

The chart splits the record into 11 periods and shows the rule's yearly return minus the yearly return from holding. The rule was ahead in four: 1928 to 1929, the 1930s, the 1970s and the 2000s. In each, the market's worst fall was 33.2% or more.

In the other seven periods the market's worst fall was 29.9% or less. The rule trailed by 1.2 to 5.4 points a year in them.

From 1928 to 1999 the rule made 9.96% a year against 10.72%. From 2000 to August 2026 it made 8.75% against 8.54%, with worst falls of 17.6% and 50.3%. Since 2010 it made 9.48% against 14.28%, with worst falls of 17.6% and 24.8%.

Across the full record there are 1,065 overlapping 10-year windows. The rule came out ahead in 389 of them, or 36.5%. Overlapping windows are not 1,065 separate tests.

A different test is the same rule on 15 exchange-traded funds (ETFs), funds that trade like shares. Nine hold one US sector each: materials (XLB), energy (XLE), financials (XLF), industrials (XLI), technology (XLK), consumer staples (XLP), utilities (XLU), health care (XLV) and consumer discretionary (XLY).

The other six are the S&P 500 ETF (SPY), the Nasdaq-100 ETF (QQQ), the gold ETF (GLD), the long-dated Treasury bond ETF (TLT), the non-US developed-markets ETF (EFA) and the small-company ETF (IWM).

Each fund starts at its first signal, between November 1993 and September 2005. The rule's worst fall was smaller in all 15, by 4.3 to 56.9 points.

Its return beat holding in 6 of the 15. The median difference was 0.81 points a year in holding's favour.

Only 7 exits avoided a fall of more than 10%, which is 7 events in 98 years. The test covers one country's stock market with tax left out.

A backtest tests a rule on past data. Quant investing for beginners covers what such tests can and cannot show.

An illustrative cost of 0.1% per switch lowers the yearly return from 9.63% to 9.47%. At 0.25% it is 9.23%. These are assumed costs, not measured ones.

Other lengths gave other answers. From January 1928, 6, 8, 12 and 14 months gave 9.27%, 8.91%, 9.81% and 9.59% a year, against 9.63% for 10.

Worst falls ran from 42.9% to 53.8%. The 10-month length had the smallest worst fall of the five.

A trend rule acts after the price has moved. By the end of October 1929, the month-end at which the rule first sold, the market was already 23.7% below its end-August level.

The gains come in falling years. In the 24 calendar years when holding lost money, the rule did better in 17. In the 74 years when holding gained, it did better in 5.

Over 98 years the rule roughly halved the worst fall and gave up about half a point a year. Since 2010 it has given up 4.8 points a year for a worst fall of 17.6% against 24.8%.

This explainer is part of the YX Insights Academy. Each one takes a single idea and checks it against real data.

The same approach runs through everything else we publish:

Good places to start on the website:

Trend following is a rule for holding an asset while its price is rising and leaving it when the price turns down. The test uses a 10-month moving average: stocks are held next month if the latest close is above the average of the last 10 month-ends, otherwise Treasury bills.

From January 1928 to August 2026, a 10-month rule made 9.63% a year against 10.12% for holding US stocks, with a worst fall of 42.9% against 83.7%. It trailed holding in 7 of 11 periods and by 4.8 points a year since 2010. One country's record cannot settle the question.

Trend following compares one asset with its own past, for example holding it while its price sits above its 10-month average. Momentum investing compares assets with each other, holding those with the highest past returns. The first needs one price series; the second needs a group of assets to rank.

Over 98.7 years the rule changed position 146 times, which is 73 exits from stocks and 73 returns, about 1.5 changes a year. It held stocks in 72.9% of months. Costs were not measured, but an assumed 0.1% per switch lowered the yearly return from 9.63% to 9.47%.

DISCLAIMER: This article is strictly educational. Any information or analysis in this note is not an offer to sell or the solicitation of an offer to buy any securities. Nothing in this note is intended to be investment advice and nor should it be relied upon to make investment decisions. Any opinions, analyses, or probabilities expressed in this note are those of the author as of the note's date of publication and are subject to change without notice.