Donchian Breakout

The Donchian channel goes back to Richard Donchian, the futures pioneer who popularized N-period breakout rules decades before they became famous — the 1980s Turtle Trading experiment taught Richard Dennis’s novices a Donchian-style breakout, not the other way around. It enters when price breaks above or below the 20-period channel — a signal that a new trend may be establishing.

In the original 2026-04 backtest the SHORT side appeared profitable in both bull and bear samples. A fresh out-of-sample re-test on 2026-06-28 — same product engine, 18-cell sweep with a 5-axis adversarial kill — found this regime-robustness did not hold: the apparent edge was a bear-beta artifact, not a directional skill that survives unseen data. The backtest numbers below are reproducible, but we no longer present this as a live directional edge.

1. Channel calculation — highest high and lowest low over the past 20 bars (look-ahead safe: uses bars up to but not including the signal bar)
2. LONG signal — current close breaks above the 20-bar high (upward momentum)
3. SHORT signal — current close breaks below the 20-bar low (downward momentum)

The indicator is two-sided; this preset is not. Steps 2 and 3 describe what a Donchian channel can signal. What we curated, measured and publish here is the short leg only — the engine’s default for donchian-breakout is short, and every number below comes from short-side runs. (Corrected 2026-09-06: this file’s frontmatter used to say direction: both, which disagreed with the preset, the backend registry and its own results section.)

4. Entry — at the open of the next bar after the break
5. Exit — TP 10% / SL 8%

The 20-period channel captures significant price movements that break out of consolidation zones. When price violates a multi-week high or low, it often signals the beginning of a directional move rather than random noise. The SHORT side looked stronger in the original sample because crypto assets frequently exhibit sharp, fast sell-offs — channel breaks to the downside tend to have stronger follow-through than upside breaks in a market prone to liquidation cascades.

That thesis is plausible, but the fresh OOS re-test (2026-06-28) showed the measured SHORT profit was driven by a falling market (short positions making money simply because price went down) rather than by a regime-independent breakout edge. Strip out that bear-beta and the standalone edge does not survive — so treat the thesis as unconfirmed.

• Failed fresh out-of-sample regime-robustness (2026-06-28). The earlier “profitable in both regimes” claim did not survive re-testing — the edge read as bear-beta, not a directional skill. Status downgraded from verified to testing for this reason, then subsequently retired. Two owners, on purpose: this file’s frontmatter status owns the membership (verified / testing / killed / shelved — the fact that it was killed stays), and src/config/simulator-presets.ts owns the verdict (verdict3) that the strategy badges and the public API read.
• Lower profit factor in bear markets (1.06) — thin edge that would likely disappear under transaction costs in high-frequency trading.
• LONG direction not separately validated.
• Classic trend-following: suffers during choppy, range-bound markets.
• Not live-tracked on OKX. Backtest only — “verified” here means the backtest is reproducible, not that the live directional edge is confirmed.