The thesis
Donchian-breakout is about as orthodox as trend following gets: buy a 20-day high, exit on a 20-day low. The logic is that sustained breakouts tend to persist, so you pay to enter strength and let the market tell you when a move is over. It trades a 24-name large-cap universe — the usual megacap tech, financials, staples and energy (AAPL, MSFT, NVDA, JPM, V, PG, WMT, XOM and friends) — and it's a built-in, hand-written strategy that the reviewer cleared at risk 0.
Recent activity
The strategy went to work on 2026-10-29, executing four buys — ABBV, MA, HD and WMT — and then spent the following week almost entirely idle. The scheduled runs on 10-30, 11-02 and 11-03 each executed zero trades and rejected one or two candidates. Cash has sat pinned at $959.13 throughout, which tells you the book is effectively fully invested; there's little dry powder to act on new signals, which likely explains the steady stream of rejections. Total equity has drifted in a narrow band — $9,911 up to $10,170 and back to $10,015 as of 2026-11-03 — so it is roughly flat to marginally positive since deployment.
Backtest and validation
Over 451 days the backtest returned 6.95% (final equity $10,695, ~3.83% CAGR) on 108 trades. That's an honest, unspectacular number, and the supporting stats are a mixed bag:
- Sharpe 0.34 — thin risk-adjusted reward.
- Win rate 38.46% — expected for trend following, where you take many small losses to catch a few large winners.
- Max drawdown 21.73% — the real sting, and more than three times the full-period return.
- Turnover 2,082% — very high churn, though fees stayed modest at $108 and FX cost was zero.
Cross-validation is where I'd pump the brakes. The strategy did not pass. Across four folds it was positive in three, and folds 3 and 4 were genuinely good (+14.09% at Sharpe 2.72, then +11.35% at Sharpe 1.86). But fold 2 lost 7.46% at a Sharpe of -1.07, and the deflated Sharpe ratio — which penalises for the six trials run — comes in at just 0.198, with PSR at 0.674. In plain terms: once you discount for how many variants were tested, the evidence that this edge is real rather than lucky is weak.
Verdict
Donchian-breakout is doing exactly what its design promises — and that's both the strength and the warning. The recent out-of-sample folds are encouraging and the low win rate is a feature, not a bug. But a 0.34 Sharpe, a 22% drawdown, one badly negative fold, and a sub-0.2 deflated Sharpe mean this is not a strategy I'd size up with conviction. It earns its place as a live, low-risk trend-following baseline; it has not yet earned trust as an alpha engine. I'll be watching whether folds 3 and 4 were the start of a durable regime or simply a favourable stretch.