A Donchian channel breakout is a buy signal that fires when price closes above the highest high of the prior N sessions, and the classic exit is a close below the lowest low of a shorter window. Run on $SPY, $QQQ, $IWM and $DIA from 2016 to the November 4, 2026 close, the 20-day-entry, 10-day-exit version produced 184 closed trades and won 48.9% of them. Across both the 20-day and 55-day versions, eight tests in all, the channel cut the worst drawdown every time and trailed buy-and-hold on price return every time.
That is the trade-off in one sentence: channel breakouts buy you smaller drawdowns and pay for them with missed upside. The numbers below show exactly how much, and where the channels sit right now.
What a Donchian channel is, and where 20 and 55 come from
The channel is three lines. The upper band is the highest high of the last N sessions, the lower band is the lowest low, and the midline splits the difference. StockCharts ChartSchool credits the indicator to Richard Donchian and notes that its price channel formula leaves out the most recent period, because a breakout would be impossible if today's bar were part of the band it is trying to break.
The two lookbacks that traders still quote come from the Turtle program. The Original Turtle Rules, which describe the system "as taught by Richard Dennis and William Eckhardt to their students in the 1980s," lay out two breakout systems:
- System 1: enter on a break of the 20-day high, exit on a 10-day low. A System 1 signal was skipped if the previous breakout would have been a winner.
- System 2: enter on a break of the 55-day high, exit on a 20-day low. Every System 2 breakout was taken.
The original rules went further. The Turtles traded futures long and short, entered on an intraday tick through the band, added units as a trade worked, and used volatility stops measured in multiples of N, their average-true-range figure. None of that is tested here. This is the bare channel, long only.
The backtest rules used here
- Data: daily OHLC for the four ETFs from November 4, 2016 to November 4, 2026, 2,513 sessions each, from Nasdaq.com historical data.
- Entry: a close above the highest high of the prior N sessions, with today excluded.
- Exit: a close below the lowest low of the prior M sessions, with today excluded.
- Two settings: N = 20 and M = 10 (System 1 without the skip rule), then N = 55 and M = 20 (System 2).
- Fills: at the close of the signal day, one position at a time, cash earning nothing between trades.
- Benchmark: buy-and-hold from the first session with a full lookback to the November 4, 2026 close.
The prices are not dividend-adjusted, so every return here is price-only, for the strategy and the benchmark alike. The test also ignores commissions, taxes and slippage, fills every signal at the close, and covers a single 10-year window in four ETFs that, by definition, survived it.
Results: the 20-day breakout with a 10-day exit
| ETF | Trades | Win rate | Avg win | Avg loss | Time in market | Strategy return | Buy-and-hold | Max drawdown (strategy / hold) |
|---|---|---|---|---|---|---|---|---|
| $SPY | 47 | 46.8% | +4.8% | -1.4% | 58.1% | +92.4% | +247.0% | -9.8% / -34.1% |
| $QQQ | 45 | 51.1% | +7.3% | -2.7% | 55.0% | +167.4% | +493.4% | -17.8% / -35.6% |
| $IWM | 43 | 46.5% | +5.2% | -2.6% | 45.7% | +48.4% | +113.3% | -29.2% / -42.3% |
| $DIA | 49 | 51.0% | +3.4% | -1.7% | 54.2% | +47.1% | +177.3% | -12.4% / -37.1% |
Test window: December 5, 2016 to November 4, 2026. Returns are compounded across closed trades; drawdowns are measured on closing prices.
Read the win rates first. Roughly half the trades lost. The system made money anyway because the average $SPY winner (+4.8%) was about 3.4 times the size of the average loser (-1.4%). That is what a trend-following payoff looks like: many small cuts, a few long rides.
Those few rides carry most of the result. The best $SPY trade entered at $278.20 on April 9, 2020 and exited at $333.21 on September 8, 2020, a gain of 19.8% over 104 sessions. Strip out the three best $SPY trades and the remaining 44 compound to +32.5% instead of +92.4%. Skip the wrong breakout and the edge mostly disappears.
The cost is plain in the return columns. Sitting in cash for roughly 42% to 54% of sessions meant missing a large share of a strong decade, and every one of the four ETFs finished well behind buy-and-hold.
Results: the slower 55-day breakout with a 20-day exit
| ETF | Trades | Win rate | Avg win | Avg loss | Avg hold (sessions) | Strategy return | Buy-and-hold | Max drawdown (strategy / hold) |
|---|---|---|---|---|---|---|---|---|
| $SPY | 26 | 61.5% | +4.0% | -3.3% | 51.7 | +33.1% | +234.4% | -23.2% / -34.1% |
| $QQQ | 21 | 66.7% | +8.5% | -3.7% | 66.7 | +133.8% | +451.1% | -17.5% / -35.6% |
| $IWM | 23 | 47.8% | +6.2% | -4.2% | 45.9 | +13.0% | +107.8% | -28.0% / -42.3% |
| $DIA | 24 | 45.8% | +5.2% | -2.5% | 50.4 | +24.5% | +165.6% | -18.1% / -37.1% |
Test window: January 26, 2017 to November 4, 2026. Time in market ran from 43.0% ($IWM) to 57.0% ($QQQ).
The slower system traded about half as often, 94 closed trades against 184, and won more of them: 55.3% across the four ETFs. It still made less money. The wider 20-day exit gave back more of each gain before it triggered, so the $SPY average loss grew to -3.3% while the average win shrank to +4.0%. A 61.5% win rate on that payoff produced only +33.1%.
That is the clearest lesson in the data: a higher win rate is not the same thing as a better system. The same hit-rate-versus-payoff trade-off shows up when choosing the stop distance for an ATR trailing stop.
Where the channels sit at the November 4, 2026 close
The levels below use the 20 and 55 sessions through November 4, so they are the bands the next close would be measured against.
| ETF | Nov 4 close | 20-day high | 20-day low | 55-day high | 55-day low | 10-day low |
|---|---|---|---|---|---|---|
| $SPY | $766.81 | $779.42 | $750.36 | $781.62 | $749.60 | $750.36 |
| $QQQ | $691.95 | $757.18 | $680.33 | $762.86 | $680.33 | $680.33 |
| $IWM | $284.02 | $286.17 | $269.88 | $303.41 | $269.88 | $273.35 |
| $DIA | $533.11 | $534.28 | $508.66 | $537.73 | $504.70 | $515.01 |
Two readings stand out. $DIA closed at $533.11 on November 4, above its prior 20-day high of $528.45 set on October 22, which is a fresh 20-day breakout under these rules; its 10-day exit line sits at $515.01. $QQQ is the opposite case. It closed 8.6% below its 20-day high of $757.18 and just 1.7% above its 20-day low of $680.33, set on November 2, so it sits near the bottom of its channel. $SPY and $IWM closed inside their channels with no signal on either system.
How traders use the channel without fooling themselves
- Treat it as a risk dial, not a return engine. In this sample the channel's job was cutting drawdowns, not beating the index.
- Expect to lose about half the time. A 46% to 51% win rate on the 20-day system was normal, not broken.
- Never skip the next signal. With the top three trades doing most of the work, cherry-picking entries is the fastest way to lose the edge.
- Add context, not just triggers. A 200-day average filter is one way to separate breakouts in uptrends from breakouts in downtrends, the same logic tested in the golden cross versus death cross test.
More breakdowns like this one live in the Analysis section and on the Traders On Trend homepage.
Bull/Bear Verdict
Bull Case: Where it helps: the rules are mechanical, the exits are defined before entry, and in all eight tests the channel kept the worst drawdown shallower than buy-and-hold, with $SPY at -9.8% versus -34.1% on the 20-day system.
Bear Case: Where it misleads: in a rising decade, half the time in cash cost dearly, and no setting beat buy-and-hold on return. A 61.5% win rate on $SPY still produced only +33.1%, and the results hinge on a handful of trades.
FAQ
What is a Donchian channel breakout?
It is a close above the highest high of the prior N sessions. The lower band, the lowest low of the prior M sessions, is the usual exit for a long position.
What settings did the Turtles use?
According to the Original Turtle Rules, System 1 entered on a 20-day breakout and exited on a 10-day low, and System 2 entered on a 55-day breakout and exited on a 20-day low.
Does the Donchian breakout beat buy-and-hold on ETFs?
Not in this test. From 2016 to 2026, neither the 20/10 nor the 55/20 version beat buy-and-hold on price return for $SPY, $QQQ, $IWM or $DIA, but both had smaller maximum drawdowns in every case.
Why is the win rate so low?
About half the breakouts in this test failed and were cut quickly. The 20-day system made money because the average win was between 1.9 and 3.4 times the average loss, depending on the ETF.