The thesis
Donchian breakout is textbook trend following: buy when price sets a fresh 20-day high, and exit when it prints a 20-day low. The idea is old, well-documented, and honest about what it is — it doesn't try to predict, it just tries to be along for the ride when a large-cap name in its 24-ticker universe (AAPL, MSFT, NVDA, JPM, XOM and friends) breaks out and keeps going.
That design has a signature that shows up clearly in the data: a 38.46% win rate across 108 backtested trades. Most breakouts fizzle; a minority run. This is expected, not a red flag — trend followers make their money on the fat tail of the winners, not on being right often.
The backtest
Over 451 days the strategy returned 6.95%, ending at $10,695 on a $10k base, for a 3.83% CAGR. The Sharpe of 0.34 is thin, and the 21.73% max drawdown is the number that should give any allocator pause — you are risking more than a fifth of capital for a single-digit annual return. Turnover of 2,083% is heavy but the fee drag stayed modest at $108 total, with no FX cost.
Validation: why it failed
Here the picture gets genuinely interesting. Walk-forward across four folds went 3-of-4 positive, and the two most recent folds are the best of the lot:
- Fold 1: +0.73% (Sharpe 0.21)
- Fold 2: -7.46% (Sharpe -1.07)
- Fold 3: +14.09% (Sharpe 2.72)
- Fold 4: +11.35% (Sharpe 1.86)
The out-of-sample fold returned 11.35% at a 1.86 Sharpe — far better than the full-period 0.34. So why did validation stamp it failed? Because we don't judge on raw returns. The Probabilistic Sharpe Ratio sits at 0.674, but the Deflated Sharpe Ratio — which penalises for the 6 trials run in search of this configuration — collapses to 0.198. Once you account for how many variants were tested, the edge isn't statistically distinguishable from luck. Fold 2's -7.46% is a reminder that the strategy can and did bleed through a full six-month regime.
Recent activity
Live behaviour matches a strategy waiting for a signal that isn't coming. The last six scheduled runs (Sep 8–15) all read 0 executed, 1–2 rejected, with cash pinned at $1,510.88 and total equity drifting between $9,824 and $10,025. The most recent real fills were back in mid-August (buying V and MA, selling AAPL and UNH). Rejections without executions suggest breakout candidates are being blocked — by position sizing or cash constraints — rather than the market simply going quiet.
The verdict
Donchian breakout is behaving exactly as trend following should: patient, low win-rate, occasionally brilliant, occasionally painful. The recent folds are encouraging and the strategy is worth keeping live on paper. But the failed DSR is the honest signal here — the promising out-of-sample numbers may be a product of the search, not a durable edge. Watch the rejection log, and don't let one great fold overwrite a -7.46% one.