Skip to content
Menu
Finabase Learn

Strategies · Trend and momentum

Donchian channel breakout: buying new highs, tested on SPY

The Donchian channel breakout buys a new high and sells a new low. Exact rules, a 33-year SPY backtest with costs, and why every sale since 2009 came before a higher buy.

On this page
  1. The idea
  2. The rules
  3. What the backtest shows
  4. Where it fails: the losing periods
  5. Compared with its neighbours
  6. How Finabase does it
  7. Frequently asked questions
  8. Summary
  9. Check your understanding

A channel breakout is trend following with no averages at all. The rule watches two numbers: the highest price of the last so many periods and the lowest price of a shorter stretch. It buys when the price sets a new high and sells when it sets a new low. The band between the two lines is called a Donchian channel, after Richard Donchian, a pioneer of trend following in commodity futures who started one of the first public futures funds in 1949.

The rule became famous in 1983, when the trader Richard Dennis taught a 20-day version to a group of beginners known as the Turtles. This lesson states a breakout rule exactly and tests it on 33 years of SPY, the oldest S&P 500 exchange-traded fund. On this market it kept the worst fall small and gave up a great deal of return. Since 2009, every single sale was followed by a buy at a higher price.

The idea

A market in a lasting uptrend keeps making new highs. A market in a lasting decline keeps making new lows. A breakout rule takes each as evidence that a trend exists and joins it.

The entry and the exit use different lengths on purpose. The entry asks for a new high over a long stretch, which is hard to reach by accident. The exit uses a shorter stretch, so the rule leaves before a decline has erased a full year of gains. The gap between the two lines is where the rule does nothing: it holds what it has.

Compare this with a moving average rule. An average blends every price in its window. A channel looks at one price only, the extreme. That makes the breakout rule simple to check by eye on any chart, and it makes the rule jumpy when a single month sets a new low and the market turns straight back.

The rules

Checked once a month, on the last trading day:

  1. Find the highest of the last 12 month-end closes, including this one, and the lowest of the last 6.
  2. If you hold cash and today's close is that 12-month high, buy the fund.
  3. If you hold the fund and today's close is that 6-month low, sell to cash.
  4. Otherwise keep what you have.

The classic daily versions follow the same pattern with other lengths. The Turtles' short system bought a 20-day high and sold a 10-day low; their long system bought a 55-day high and sold a 20-day low. They traded futures, went short on new lows as well as long on new highs, and sized each position by its volatility. The test below is long or cash only, on one stock fund.

SPY month-end close with its 12-month high and 6-month low, 1994–2026 (log scale) $50 $100 $200 $500 SPY month-end close 12-month high (buy line) 6-month low (sell line) 1995 2000 2005 2010 2015 2020 2025
SPY month-end close with its 12-month high (buy line) and 6-month low (sell line), 1994 to 2026, log scale. Shaded months are in cash.

A worked example

At the end of January 2008 SPY closed at $137.37. That was the lowest month-end close of the last six, so the rule sold. The 12-month high at that point was $154.65, a long way up.

The rule then waited for a new 12-month high, and in a falling market the 12-month high is far above the price. In July 2009 SPY closed at $98.81 while the 12-month high was still $128.79. In August it closed at $102.46 against a high of $115.99. Only at the end of September 2009, when the pre-crash months had rolled out of the window, did the close of $105.59 equal the 12-month high. The rule bought there, 23.1% below its sale.

Notice the delay. SPY's lowest month-end close was in February 2009. The rule bought seven months later. A breakout entry after a deep fall is slow, because the old high has to age out.

What the backtest shows

The test, stated in full:

  • Data. SPY month-end closes from February 1993 to October 2026. The first 12 months build the first channel, so the test runs from January 1994: 394 months.
  • Trading. The signal and the trade both use the month-end close.
  • Costs. 0.10% of the amount on every buy and every sell.
  • Left out. Dividends, interest on cash and taxes. Buy and hold collects every dividend; the rule collects them about two-thirds of the time and would have earned interest on cash the rest. Short selling, which the original systems used, is also left out.
SPY monthly, 1994–2026: 12/6-month channel breakout against buy and hold, 0.10% per trade, no dividends $1 $2 $5 $10 Growth of $1 Buy and hold 12/6-month channel 0% −20% −40% −60% Fall from the peak 1995 2000 2005 2010 2015 2020 2025
SPY monthly, 1994 to 2026: the 12-month and 6-month channel breakout against buy and hold, 0.10% per trade, dividends not included. Top: growth of $1 (log scale). Bottom: the fall from the highest value so far.
Measure 12/6-month channel Buy and hold
Total return 657.5% 1,467.2%
Return per year (CAGR) 6.4% 8.8%
Deepest fall from a peak −15.4% −52.2%
Longest time below a peak 64 months 80 months
Time invested 68% 100%
Round-trip trades 17 1
Trades with a gain 76% n/a

One dollar became $7.58 under the rule and $15.67 held throughout. The rule gave up 2.4 points a year, which compounds to half the final amount, and in return its worst fall was less than a third of buy and hold's. The 13 winning trades gained 20.5% on average and the four losing trades lost 4.0%.

The two halves of the period

Period 12/6-month channel Buy and hold
January 1994 – February 2009 5.5% a year 2.9% a year
February 2009 – October 2026 7.1% a year 14.1% a year

Before the 2009 low the rule was ahead, because it sold in November 2000 at $132.28 and bought back in June 2003 at $97.63, then sold in January 2008 and bought back in September 2009. After 2009 it earned half of what holding earned.

Where it fails: the losing periods

A breakout rule loses when a decline is deep enough to set a 6-month low and short enough to be over soon after. The rule sells near the bottom and then has to wait for a new 12-month high to get back in, which by definition is above every price of the last year.

Sold Bought back Result
June 2010, $103.22 December 2010, $125.75 Bought back 21.8% higher
July 2011, $129.33 February 2012, $137.02 Bought back 5.9% higher
August 2015, $197.67 July 2016, $217.12 Bought back 9.8% higher
October 2018, $270.63 April 2019, $294.02 Bought back 8.6% higher
February 2020, $296.26 July 2020, $326.52 Bought back 10.2% higher
April 2022, $412.00 April 2023, $415.93 Bought back 1.0% higher
March 2025, $559.39 June 2025, $617.85 Bought back 10.5% higher
March 2026, $650.34 April 2026, $718.66 Bought back 10.5% higher

Eight sales since 2009, eight buys at a higher price. Even 2022, a year in which SPY fell 24.8% from peak to trough on month-end closes, ended with the rule buying back slightly above its sale: it was out for the fall and out for the whole recovery too.

The rule's own deepest fall came in 1998. SPY dropped sharply in August 1998, the rule sold at $96.00, and it bought back in November at $116.13, 21.0% higher.

Other ways it fails or misleads:

  • The re-entry is the weak part. Selling a 6-month low avoided the middle of both big bear markets. Waiting for a 12-month high to return cost months of recovery every time.
  • It was built for many markets, not one. Donchian and the Turtles spread the rule across dozens of futures, expecting most trades to lose and a few trends in a few markets to pay for them. One stock index offers few such trends.
  • Daily versions trade far more. On the two years of daily bars available here, the 20-day and 10-day rule made 15 trades on SPY and gained 2.8% while holding gained 20.7%. Two years is too short to judge it, but it shows how often a 10-day low is followed by a recovery.
  • Taxes. Each sale at a gain is taxable in that year in a US taxable account (IRS Topic 409).

Neighbouring settings

Buy at the high of Sell at the low of Return per year Deepest fall Round trips
12 months 3 months 3.4% −22.7% 38
12 months 6 months 6.4% −15.4% 17
12 months 12 months 7.9% −24.9% 9
9 months 4 months 5.6% −20.4% 27
6 months 3 months 4.2% −22.1% 40
6 months 6 months 7.1% −18.5% 19

The table runs from 3.4% to 7.9% a year, a much wider spread than neighbouring moving-average lengths produce. The pattern is clear: the shorter the exit, the more often the rule is shaken out, and the worse it does. That sensitivity is a warning in itself. A rule whose result depends this much on one setting has not been shown to work; it has been shown to depend on a choice that could only be made well with hindsight.

Other markets

The same 12-month and 6-month rule on other funds, each from the first month its data allows:

Fund From Rule, per year Held, per year Rule's deepest fall Held
QQQ (Nasdaq-100) 2000 6.2% 7.2% −30.4% −81.1%
IWM (small companies) 2001 3.0% 7.0% −25.2% −53.5%
EFA (developed markets) 2002 3.5% 4.5% −34.1% −59.7%
EEM (emerging markets) 2004 3.9% 5.7% −35.0% −61.9%
GLD (gold) 2005 6.6% 9.9% −28.0% −42.9%
TLT (long Treasuries, price) 2003 0.5% −0.1% −22.7% −54.5%

The rule trailed on return in every fund except the bond fund, whose price went nowhere, and it cut the deepest fall by a third or more in every one. The TLT row uses the fund's price only; the interest it paid, which is most of a bond fund's return, is not in the data.

Compared with its neighbours

Rule Return per year Deepest fall Round trips Time invested
Buy and hold 8.8% −52.2% 1 100%
Price above its 10-month average 7.8% −23.8% 25 76%
12/6-month channel breakout 6.4% −15.4% 17 68%

On this fund the channel rule was the more cautious of the two trend rules: less time invested, a smaller worst fall and a lower return. The moving average gets back in as soon as the price recovers above its average. The channel waits for a full new high.

How Finabase does it

Finabase has no Donchian channel study and no tool that tells you when to buy or sell. You can still check the rule's inputs and try related signals.

  • Chart. A five-year chart shows the highs and lows directly, and the chart's drawing tools can mark the last 12-month high and 6-month low. The Breakouts and false breakouts: what the data says about each lesson covers how breakouts look on a chart and how often they fail.
  • Backtest. The Backtest tool trades the signals of the indicators and candlestick patterns it offers on one symbol. It does not include a channel breakout, each signal buys a fixed $10,000, and it charges no trading cost, so its results are not comparable with the table above.
  • Strategy. The Strategy tool replays if/then allocation rules over two years. Two years of a monthly breakout rule is one or two trades, which shows the behaviour and nothing more.

Frequently asked questions

Does the Donchian breakout strategy work on stocks?

On SPY in this test it reduced the worst fall from 52.2% to 15.4% and reduced the return from 8.8% to 6.4% a year. Whether that counts as working depends on which of the two numbers matters more to the person asking. It did not beat buying and holding on return in any of the six stock and gold funds tested.

What were the Turtle trading rules?

They were two breakout systems taught by Richard Dennis and William Eckhardt in 1983: buy a 20-day high and exit at a 10-day low, or buy a 55-day high and exit at a 20-day low, long and short, across many futures markets, with position sizes set by volatility. Curtis Faith, one of the Turtles, describes them in Way of the Turtle.

Why is the exit shorter than the entry?

A shorter exit gives back less of an open gain before selling. The cost is more exits that turn out to be false alarms. In this test the 3-month exit made 38 round trips and earned 3.4% a year, while the 12-month exit made 9 and earned 7.9%.

Is a new 52-week high a good time to buy?

A new high is what a lasting uptrend looks like, so trend rules buy there. It is not a forecast. In this test 13 of 17 purchases at a 12-month high ended with a gain, and the rule still earned less than holding throughout, because of what it missed while waiting in cash.

Summary

A Donchian channel breakout buys when the price sets a new high over a long window and sells when it sets a new low over a shorter one. On 33 years of SPY month-end closes, a 12-month entry and 6-month exit returned 6.4% a year against 8.8% for buying and holding and cut the deepest fall from 52.2% to 15.4%. It avoided the middle of the 2000 and 2008 bear markets. Since 2009 each of its eight sales was followed by a buy at a higher price, and its result swings widely with the exit length. The rule was designed to be spread across many markets with a few big trends; one stock index in a long bull market is close to its worst case.

Check your understanding

4 questions on this lesson. Each answer comes with the reason, and nothing is scored.

  1. You hold the fund. This month-end close is $98. The lowest of the last six month-end closes is $95 and the highest of the last twelve is $110. What does the 12/6 rule do?

    1. Sells, because the price is well below the 12-month high
    2. Nothing: the close is neither a 12-month high nor a 6-month low
    3. Buys more
    Show the answer

    B. Nothing: the close is neither a 12-month high nor a 6-month low Between the two lines the rule keeps what it has. It sells only when the close itself is the lowest of the last six month-end closes.

  2. After selling in January 2008, the rule did not buy until September 2009, seven months after the lowest month-end close. Why so late?

    1. The 12-month high stayed far above the price until the pre-crash months dropped out of the window
    2. Trading costs delayed the purchase
    3. The rule waits a fixed seven months
    Show the answer

    A. The 12-month high stayed far above the price until the pre-crash months dropped out of the window A buy needs a close equal to the highest of the last twelve. After a deep fall, that high belongs to the months before the fall, so it takes up to a year to come down within reach.

  3. Since 2009 the rule sold SPY eight times and bought back higher all eight times. What kind of market does that describe?

    1. Long declines that lasted years
    2. Falls deep enough to set a 6-month low that then reversed within months
    3. A market with no falls at all
    Show the answer

    B. Falls deep enough to set a 6-month low that then reversed within months Each fall triggered the exit and then ended. The rule was built for declines that keep going, as in 2000 to 2002 and 2008, and those did not occur after 2009.

  4. With a 12-month entry, exits of 3, 6 and 12 months returned 3.4%, 6.4% and 7.9% a year. What is the sound reading?

    1. The 12-month exit is the correct one to use
    2. The result depends heavily on one setting, so the rule is fragile on this market
    3. Shorter exits are safer
    Show the answer

    B. The result depends heavily on one setting, so the rule is fragile on this market A spread of more than four points a year from one parameter means the outcome was driven by a choice only hindsight could make. The 3-month exit also had a deeper worst fall than the 6-month one, so shorter was not safer.

Sources and further reading

  1. Richard D. Donchian, “High Finance in Copper,” Financial Analysts Journal 16(6), 1960, pp. 133–142 — an early published trend-following method by the originator of the channel rule
  2. Curtis M. Faith, Way of the Turtle (McGraw-Hill, 2007) — the 20-day and 55-day breakout rules taught by Richard Dennis and William Eckhardt in 1983
  3. IRS — Topic no. 409, Capital gains and losses
  4. Finabase chart data — monthly bars for SPY, QQQ, IWM, EFA, EEM, GLD and TLT to October 2026, weekly bars for five years and daily bars for two; rules and results computed by Finabase