Skip to main content
Track Donchian breakout
6 / 6
Search the archiveLibrary
Library

2014issue C0256-60

Walk-forward complexity in Donchian breakout systems

A Donchian breakout can be written as one entry-and-exit procedure and then extended with a volatility gate and an ATR multiple stop. The archive workflow judges those added factors by how results change from the fit sample to held-out data.

  • A two-length Donchian breakout can be expressed as one procedure: enter on a longer-window high or low break and exit on a shorter-window opposite-channel stop.
  • The same procedure can add an average-true-range volatility gate and an ATR multiple stop without changing the channel-break entry logic.
  • A four-parameter version turns the volatility coefficient and stop multiple on; setting both to zero recovers the basic two-parameter system.
  • Walk-forward analysis compares the simpler model with successively more factors by measuring how results change from the fit sample to held-out data.
Entries in this reading3 entries
Current trackDonchian breakout6 readingsSystem optimization85 readingsWalk-forward analysis95 readings

One procedure for entry and exit

A two-length Donchian channel can be expressed as one procedure. The system enters on a longer-window high or low break and exits on a shorter-window opposite-channel stop.

Extra factors that leave the entry unchanged

The same procedure can be extended with an average-true-range volatility gate and an ATR multiple stop without changing the channel-break entry logic.

A four-parameter version is obtained from the two-parameter version by turning the volatility coefficient and stop multiple on. Setting both to zero recovers the basic system.

Fit sample versus held-out data

Walk-forward analysis compares a simpler model with successively more factors by measuring how results change from the fit sample to held-out data.

System optimization over the two channel lengths can be run as a grid of length pairs, then inspected on a later segment that was not used for the grid.

Out-of-sample Donchian stops on the Dow

After lookback lengths were chosen on earlier bars, the same stop-and-reverse Donchian bands are drawn on later daily Dow prices. A trader should judge the extra rules by whether those bands still frame the held-out advance rather than by the fit-window curve. Index levels were read from the NeuroShell price scale.
After lookback lengths were chosen on earlier bars, the same stop-and-reverse Donchian bands are drawn on later daily Dow prices. A trader should judge the extra rules by whether those bands still frame the held-out advance rather than by the fit-window curve. Index levels were read from the NeuroShell price scale.Dow Jones Industrial Average (continuous) · Daily bars · 2012-01-01T00:00:00.000Z to 2013-11-30T00:00:00.000Z

The chart legend uses optimized lengths of a lagged 8-day high and a lagged 21-day low, not the article’s default 40/15. The only printed time label is 2013; other months are spaced from that year marker across the daily window. Y readings are approximate to the nearest 50 points.

Educational research material, not investment advice. Historical source context does not establish present-day performance.
6 of 6 in the Donchian breakout track
1987Track finished · Next track: Fixed contract sizingVolatility-layered mechanical system with fixed contracts42 readings
All readings on this track · 6 readings
  1. 1989Evaluating channel breakout with intraday and close-only stops
  2. 1990Delayed channel breakout stop construction
  3. 2000Fixed-horizon walk-forward tests of first-print breakouts
  4. 2013Z-score value filter on reversals and Donchian entries
  5. 2014Length fit versus later window for a two-length Donchian breakout
  6. 2014Walk-forward complexity in Donchian breakout systems
By fieldSystem design117 readingsBy year201489 readings
All 6 readings tagged Donchian breakout
Also on Donchian breakout5 readings
2014Length fit versus later window for a two-length Donchian breakout2000Fixed-horizon walk-forward tests of first-print breakouts1989Evaluating channel breakout with intraday and close-only stops1990Delayed channel breakout stop construction2013Z-score value filter on reversals and Donchian entries