Donchian Channel Breakout
Long on new N-day highs, short on new N-day lows — the original Turtle Trading entry rule.
The Intuition
Donchian Channel Breakout is the direct ancestor of modern trend-following. Richard Donchian pioneered systematic commodity trading in the 1950s–70s using channel rules. The strategy that became famous is the "Turtle System," taught by Richard Dennis and William Eckhardt to their "Turtle Traders" in 1983: buy on a new 20-day high, sell short on a new 20-day low.
The insight is elegant: a new N-day high means the asset is at its highest price in N days. Every seller who bought in the past N days is now sitting on a loss; they may accelerate selling (creating downward pressure) or be squeezed out if the price continues higher. Every buyer in the past N days is in profit and becoming more conviction buyers. A new high thus reflects a genuine balance-of-power shift — momentum.
The Turtles proved the system worked across diversified commodity futures: crude oil, corn, gold, currencies, treasury bonds. Their success (several Turtles became very wealthy traders) generated intense interest in systematic trend-following, which eventually became the backbone of the CTA (Commodity Trading Advisor) industry, now managing hundreds of billions in assets.
Key assumptions: (1) New price extremes are meaningful signals, not noise. (2) The channel captures the dominant trend duration — 20 days suits medium-term swings; 55 days (the Turtles' "System 2") suits longer-term trends. (3) The asset is liquid enough for position entry and exit at the channel boundaries without significant market impact. (4) Risk is managed through position sizing and trailing stops (the original Turtle system used ATR-based stops not implemented in this basic version).
The strategy suffers in range-bound markets: the asset repeatedly breaks above the 20-day high then falls back, generating a series of losing trades. The Turtles countered this with two systems (20-day and 55-day channels) and diversification across uncorrelated markets. The worst-case scenario is a market in an extended range with brief false breakouts — exactly the environment that characterised many commodity markets in the mid-2000s as carry-driven range trading dominated.
The Math
Read this as a compact model summary: what the signal sees, what it ignores, and where fragility can creep in.
UpperCh(t) = max(High[t-n : t])
LowerCh(t) = min(Low[t-n : t])
Signal(t) = +1 if Close(t) >= UpperCh(t) [new n-day high]
= -1 if Close(t) <= LowerCh(t) [new n-day low]
Parameters
| Parameter | Type | Default | Description |
|---|---|---|---|
| window | int | 20 | Lookback period for channel calculation (days) |
Source Code
def run(ticker: str, start: str, end: str, **params) -> dict:
window = int(params.get("window", 20))
df = fetch_ohlcv(ticker, start, end)
# Exclude the current bar from the channel so a close is compared only
# against previously completed highs/lows. This avoids lookahead and matches
# the public methodology copy. Positions persist between breakout events.
highest_high = df["High"].rolling(window).max().shift(1)
lowest_low = df["Low"].rolling(window).min().shift(1)
signal = pd.Series(pd.NA, index=df.index, dtype="Float64")
signal[df["Close"] > highest_high] = 1.0
signal[df["Close"] < lowest_low] = -1.0
pos = signal.ffill().fillna(0.0).astype(float)
return run_backtest(df, pos, ticker=ticker, start=start, end=end,
strategy=METADATA["slug"], params={"window": window})
Further Reading
- Faith, C. (2003). Way of the Turtle. McGraw-Hill.
- Covel, M. (2009). The Complete TurtleTrader. Harper Business.
- Donchian, R. (1960). High Finance in Copper. Financial Analysts Journal, 16(6), 133–142.
When It Works / When It Fails
- Sustained trending markets with momentum persistence
- Volatility-expansion phases with new high + range expansion
- Cross-asset diversified futures portfolios
- Active institutional CTA trend-following environments
- Range-bound markets with repeated false breakouts
- Choppy sideways equity indices with no follow-through
- Single-stock equity without a regime filter
- High-turnover environments where round-trip costs compound