The Thesis
The donchian-breakout strategy runs one of the oldest ideas in systematic trading: buy the 20-day high, sell the 20-day low. It is pure trend following — a bet that when a large-cap name breaks out of its recent range, momentum carries it further. The strategy is live and trades a 24-name universe of U.S. blue chips spanning tech (AAPL, MSFT, NVDA), financials (JPM, V, MA), healthcare (JNJ, UNH), staples (KO, PG, WMT), and industrials (CAT, HON).
Recent Activity
The live book has gone quiet. Across the six scheduled runs from July 24 to July 31, the strategy executed zero trades and rejected one to two candidate orders each day, holding cash flat at $1,760.89. Over that stretch the total account value drifted from about $10,406 down to $9,925 — mark-to-market erosion on existing positions rather than any new decisions. The last real executions came in early June: buys in KO, UNH, ABBV, AAPL, CAT, and MSFT, alongside sells in CAT and MSFT. The pattern of repeated rejections suggests the strategy is finding signals it cannot fund or that fail an entry filter — worth investigating, because a trend follower that can't act on breakouts forfeits its edge.
Backtest Performance
Over 451 days the backtest returned 6.95% (3.83% CAGR), ending at $10,695 from a $10,000 start. The character is textbook trend following: a 38.46% win rate across 108 trades, meaning most positions lose small while a minority of runners pay for them. The costs of that style are visible — a 21.73% max drawdown and a punishing 2,083% turnover, though fees stayed negligible at $108. The 0.34 Sharpe is uninspiring; this is a strategy that made money without being comfortable to hold.
Validation: The Hard Part
Here the picture sharpens. Walk-forward validation across four folds failed the deployment gate, and the details explain why the raw returns flatter the strategy. Three of four folds were positive — fold 3 (+14.09%, Sharpe 2.72) and fold 4 (+11.35%, Sharpe 1.86) were genuinely strong — but fold 2 cratered at -7.46% with a -1.07 Sharpe. The out-of-sample stretch actually beat the full backtest (11.35% vs 6.95%), which is encouraging.
The statistical adjustments are what sink it. With six trials searched, the Deflated Sharpe Ratio sits at just 0.198 and the Probabilistic Sharpe Ratio at 0.674 — below the confidence needed to trust the headline Sharpe as real rather than a product of selection. In plain terms: the recent good folds may be regime luck, not durable edge.
Verdict
Donchian breakout is doing exactly what trend following does — surviving on a few big winners, tolerating deep drawdowns, and living or dying by regime. The recent all-rejection runs and the failed validation gate are two separate warning lights: one operational, one statistical. I'd treat this as a promising template that has not yet earned capital conviction. The near-term priorities are diagnosing why live orders keep getting rejected and gathering more out-of-sample evidence before the deflated numbers can be trusted.