The Thesis
The donchian-breakout strategy is about as orthodox as trend following gets: buy when a name prints a 20-day high, exit when it prints a 20-day low. It runs across a 24-name large-cap universe spanning tech, financials, staples, healthcare and energy — AAPL, MSFT, NVDA, JPM, V, JNJ, WMT, XOM and the rest. The logic is deliberately dumb in the best sense: let the market tell you what's trending, ride it, and cut losers at the lower channel. No forecasting, no fundamentals.
Backtest Snapshot
Over 451 days the strategy returned +6.95% (final equity $10,695.11, CAGR 3.83%). The headline numbers read exactly like a trend-follower's: a 38.46% win rate across 108 trades, meaning a minority of winners carry the book. The catch is the ride down — a 21.73% max drawdown against a Sharpe of just 0.34. Turnover is a striking 2,082%, so this is an active system, though fees stayed negligible at $108 and FX cost was zero.
Does It Hold Up Out-of-Sample?
This is where it gets interesting — and where the strategy currently fails our gate (passed: false). The walk-forward across four folds tells a story of feast and famine:
- Fold 1 (Aug 2024–Jan 2025): +0.73%, barely positive
- Fold 2 (Jan–Jul 2025): −7.46%, Sharpe −1.07 — the painful chop
- Fold 3 (Jul–Dec 2025): +14.09%, Sharpe 2.72 — the trend payday
- Fold 4 (Dec 2025–May 2026): +11.35%, Sharpe 1.86
Three of four folds positive, and the most recent out-of-sample stretch posted a healthy +11.35% at Sharpe 1.86. That's encouraging. But the deflated Sharpe ratio (DSR 0.198) — which penalises for the 6 trials we ran searching for this configuration — is the sticking point. A PSR of 0.674 isn't terrible, yet the full-sample Sharpe of 0.34 and that ugly Fold 2 drag keep the risk-adjusted case unconvincing. The gate is right to withhold its blessing.
Recent Activity: Quiet and Fully Deployed
The strategy deployed capital on 2026-10-29, executing buys in ABBV (8 @ $289.91), MA (3 @ $636.60), HD (7 @ $328.57) and WMT (20 @ $121.17). Since then it has gone quiet: every scheduled run from Nov 4 through Nov 11 executed zero trades and rejected one or two candidates, with cash pinned at $959.13. Total equity has drifted in a narrow $10,162–$10,391 band. That's the system working as designed — it's fully invested, no fresh 20-day highs cleared the bar, and nothing breached a lower channel to force an exit.
Verdict
Strengths: a transparent, robust mechanism with genuinely strong recent out-of-sample performance. Risks: a 21.73% drawdown, a sub-coinflip win rate that demands patience through chop like Fold 2, and a deflated Sharpe that flags possible overfitting from the search. It's live and it's green, but it hasn't earned full validation — treat it as promising-on-probation rather than proven.