Thesis
The donchian-breakout strategy runs one of the oldest ideas in systematic trading: buy 20-day highs, exit on 20-day lows. It is pure trend following — no forecasting, no mean reversion, just a mechanical rule that lets winners run and cuts positions when momentum rolls over. It trades a universe of 24 large-cap US names (AAPL, MSFT, NVDA, JPM, V, MA, XOM, and peers), and it is currently live.
The appeal is honesty. There are no hidden parameters to overfit beyond the 20-day window, and the exit logic is symmetric with the entry. The cost is character: breakout systems are structurally low-hit-rate, and this one is no exception.
Backtest performance
Over 451 days the strategy returned 6.95%, ending at $10,695 on a $10,000 base — a CAGR of just 3.83%. The texture behind that number is classic trend following:
- Win rate: 38.46% across 108 trades. Roughly six in ten trades lose money; the strategy relies on the winners being large enough to pay for them.
- Sharpe: 0.34 — thin risk-adjusted reward.
- Max drawdown: 21.73% — a real gut-check for a return this modest.
- Turnover: 2,082% — very high churn, though fees stayed contained at $108 total with no FX cost.
A 21.7% drawdown to earn 6.95% is an uncomfortable ratio. The engine works, but the pain-to-gain trade is steep.
Validation
This is where the picture gets interesting. Across four walk-forward folds, three were positive:
| Fold | Window | Return | Sharpe |
|---|---|---|---|
| 1 | Aug'24–Jan'25 | +0.73% | 0.21 |
| 2 | Jan'25–Jul'25 | -7.46% | -1.07 |
| 3 | Jul'25–Dec'25 | +14.09% | 2.72 |
| 4 | Dec'25–May'26 | +11.35% | 1.86 |
The most recent out-of-sample fold returned 11.35% at a Sharpe of 1.86 — genuinely strong, and the strategy has clearly performed better lately than its full-history Sharpe of 0.34 suggests. Fold 2's -7.46% drag is what pulls the aggregate down.
Despite this, validation did not pass. The deflated Sharpe ratio (DSR) sits at just 0.198 — after accounting for six trials, the evidence of genuine skill is weak. The probabilistic Sharpe (PSR) of 0.674 is respectable but short of conviction. In plain terms: the recent good folds may be trend-following's cyclical tailwind rather than durable edge.
Recent live activity
Live execution has been quiet. The 22 September run fired 4 trades — buying WMT and AAPL, selling V and MA — but the surrounding sessions on the 21st, 24th, 25th, and 28th executed nothing, logging one or two rejected orders each. The paper account total has drifted between $9,906 and $9,994, sitting just under the $10,000 starting line with $3,513 in cash.
Verdict
Donchian-breakout is behaving exactly as trend following should: low win rate, deep drawdowns, and streaky fold-to-fold results with a recent hot patch. The strong latest folds are encouraging, but the failed validation gate and a 0.198 DSR are the honest signal — keep it live and observed, but do not scale it on the strength of one good year.