Donchian channels: the breakout indicator the Turtles got rich on

In 1983, commodities trader Richard Dennis trained a group of complete beginners — the "Turtles" — in a single trend-following system, gave them money, and over five years they collectively made about $175 million. The signal at the core of the system was almost embarrassingly simple: enter on a new N-period high, exit on a new M-period low. The indicator that draws those highs and lows is the Donchian channel.

What it is

A Donchian channel has three lines. The upper band is the highest high of the last N bars. The lower band is the lowest low of the last N bars. The midline is the average of the two.

# Donchian channel with lookback N
upper[t] = max(high[t-N+1 : t+1])    # highest high of last N bars
lower[t] = min(low[t-N+1 : t+1])     # lowest low of last N bars
mid[t]   = (upper[t] + lower[t]) / 2

# Classic Turtle System 1:
#   Entry: long when close breaks above 20-bar upper band
#   Exit:  when close breaks below 10-bar lower band
# System 2:
#   Entry on 55-bar breakout, exit on 20-bar opposite breakout.

There's no smoothing, no period weighting, no parameters except the lookback. The signal is literal: did price just print a level it hasn't seen in the last N bars? If yes, the trend is making a new high (or low) and the strategy joins it.

Why it works (sometimes)

Markets that trend really do print new highs (or lows) more often than chance would predict during the trending phase. A break of the 20-bar high tells you that buyers have absorbed all selling pressure in the last 20 bars and are now pushing into new territory. That's structural information — not just an indicator value.

The cost: false breakouts. In range-bound markets, price oscillates around the channel midline and occasionally pokes through the upper or lower band only to reverse. Each false breakout is a small loss. The Turtle system tolerated this because the rare 4× and 10× trades in true trends covered all the noise.

The Turtle rules, summarised

• System 1: enter on 20-day breakout, exit on 10-day reverse breakout. Skip the next entry if the previous one was a winner (theory: trends pause after fast wins).
• System 2: enter on 55-day breakout, exit on 20-day reverse breakout. Always take the entry, even if the previous trade was a winner.
• Risk per trade: 2% of equity, sized by ATR(20). Add to winning positions in 0.5×ATR increments up to four units.
• Stop loss: 2×ATR(20) from entry.
• Diversify across uncorrelated markets — the system depends on the rare big trend, and trends arrive at different times in different assets.

Rule snippet

strategy:
  name: donchian_breakout
  indicators:
    - { id: don,   kind: Donchian, period: 20 }
    - { id: don10, kind: Donchian, period: 10 }
    - { id: atr,   kind: ATR,      period: 20 }
  rules:
    entry:
      type: AND
      children:
        - { type: gte, left: bar.close, right: don.upper }   # new 20-bar high
        - { type: gt,  left: atr,       right: don.range_q }  # require some volatility
    exit:
      type: lte
      left: bar.close
      right: don10.lower                                       # 10-bar low exit
  risk:
    size_method: atr_target
    target_risk_pct: 2.0
    atr_period: 20
    atr_mult: 2.0
    pyramid:
      enabled: true
      step_atr: 0.5
      max_units: 4

When not to use it

Range-bound, low-volatility markets. The 2017-2018 sideways grind in commodities, the 2022-2023 chop in BTC — these environments shred trend-following systems. Three defences: a volatility filter (only trade when ATR is above its 100-bar median), a regime filter (only trade when ADX is above 22), or accept that some years will be flat-to-down and only the trending years pay.

Next steps

Donchian sits in the broader family of trend-following techniques — see trend-following vs mean-reversion for the trade-offs against the other side, and ADX explained for the regime filter most trend systems add on top.