Donchian channels

Definition

Donchian channels (Richard Donchian’s construction) plot the highest high and lowest low of the last N bars — the rolling range extremes. A close or trade beyond the channel is the canonical breakout event: the N-bar high/low break that classic trend systems (including the 1980s Turtle rules) used as entries and exits. Its simplicity is the point — the channel IS the range; the break IS the signal.

How it works / structure

• Formula: upper_t = max(H over last N), lower_t = min(L over last N); midline = their average.
• Parameters (engine-executable): entry window (20 and 55 are the Turtle conventions), exit window (typically shorter — e.g. enter on 55-bar break, exit on 20-bar opposite break), trade-through vs close-through trigger, and ATR-scaled initial stops (indicator-atr — the historical pairing).
• Relation to platform concepts: the 52-week range position (range_52w_pos_pct) is a normalized Donchian reading at N = 252; “new 52-week high” = upper-channel touch.
• Symmetric system: channel systems are naturally long/ short symmetric — the reason they transplanted so cleanly to futures (strategy-futures-trend-following).

When it applies

Trend-system entries and trailing exits (its native habitat); breakout definitions with zero construction ambiguity (strategy-breakout boundary of choice for replays — no fitted smoothing, one parameter); range-extreme flags for event studies (new-high/new-low behavior).

Risk profile & failure modes

• False-break economics: the channel’s edges are exactly where resting stops cluster; whipsaw costs in ranges are the system’s known loss engine — the win distribution is few large trends against many small failures.
• One-parameter overfitting is still overfitting: scanning N until the replay shines is snooping with fewer knobs.
• Late by construction: an N-bar-high entry surrenders the move’s first N bars; channel systems buy confirmation with performance.
• Gap-through breaks: gaps beyond the channel fill entries at prices far worse than the boundary (ms-slippage-friction).

Evidence & limits

Channel breaks were among the rule families with predictive content in Brock-Lakonishok-LeBaron (1992), subject to the snooping and cost caveats of that literature; the century-scale trend evidence (Hurst-Ooi-Pedersen 2017) covers the systems channel entries historically implemented. The channel itself is descriptive arithmetic; the systems built on it inherit trend- following’s evidence and its whipsaw regimes.

Falsifiable-thesis examples

Illustrations only, not signals:

• “X, printing a 55-day-high break, will reach +3 ATR before breaking its 20-day low” — falsified by which comes first.
• “A 55/20 Donchian system on 20 liquid futures at 10% vol target finishes the next 12 replay months positive” — falsified by the replay P&L.

Cross-references

• Systems built on it: strategy-breakout, strategy-futures-trend-following
• Historical pairing: indicator-atr (stops/sizing), mgmt-stop-loss (trailing exits)
• Platform binding: range_52w_pos_pct
• Method caveats: lens-technical