Thesis
The donchian-breakout strategy runs one of the oldest ideas in systematic trading: buy a 20-day high, exit on a 20-day low. It is unapologetic trend-following across a 24-name large-cap universe spanning tech (AAPL, MSFT, NVDA), financials (JPM, V, MA), staples (KO, PG, WMT), and industrials (CAT, HON). The bet is simple — let breakouts run, cut positions when momentum reverses, and accept that most trades will lose small while a few winners pay for the book.
Recent activity
The agent trades on a scheduled evening run, and the recent tape is revealing. On 2026-09-29 it bought 3 shares of CAT at $825.98 and sold its KO position (23 shares at $86.90). The prior week (09-22) was busier — four executions including AAPL and WMT buys against V and MA sells. But the standout pattern is rejections: across the last six runs, orders were rejected on five of them (2 rejected on 09-24, 09-25, and 09-28). That points to signals firing against insufficient cash or position constraints rather than clean execution.
More soberingly, the live book sits at a total value of $9,830.16 with $3,023.55 in cash — below the $10,000 starting line. Whatever the backtest promised, the strategy is currently in the red in production.
Backtest and validation
Over 451 days and 108 trades, the backtest returned 6.95% (3.83% CAGR) — modest, and earned with real pain: a 21.73% max drawdown and a Sharpe of just 0.34. The 38.46% win rate is textbook trend-following, but turnover of 2,083% means fees and slippage matter enormously; the run booked $108 in fees alone.
Walk-forward validation tells a more nuanced story. Three of four folds were positive, and the most recent two were genuinely strong — fold 3 returned 14.09% at a 2.72 Sharpe, fold 4 returned 11.35% at 1.86. But fold 2 (early-to-mid 2025) lost 7.46% with a -1.07 Sharpe, exposing how badly the strategy bleeds in choppy, range-bound markets where breakouts fail repeatedly.
Our gate rejected it (passed: false). The reason is written in the deflated statistics: a Probabilistic Sharpe Ratio of 0.674 is unconvincing, and the Deflated Sharpe Ratio of 0.198 — adjusted for 6 trials — says the headline performance does not clear the bar once we account for selection. The strong recent folds may be as much regime luck as edge.
Verdict
Donchian-breakout is honest about what it is: a low-hit-rate trend-follower whose fortunes swing with the market regime. The recent out-of-sample folds are encouraging and worth watching. But the deep drawdown, the failed validation gate, the punishing turnover, and a live book already underwater all argue for keeping this one on a short leash and small size until it earns more trust.