The thesis
Donchian-breakout is about as classical as systematic trading gets: buy when price makes a new 20-day high, exit when it makes a new 20-day low. It runs long-only across a 24-name large-cap universe spanning tech, financials, healthcare, staples and energy — AAPL, MSFT, NVDA, JPM, V, JNJ, XOM, CAT and others. The logic is pure trend-following, and it carries the behavioural signature of that family: it expects to be wrong often and right big.
Recent activity
The strategy is live and reviewer-approved (risk score 0, flagged as a built-in hand-written strategy). Its most active day was the 29 October run, which executed four buys — ABBV (8 @ $289.91), MA (3 @ $636.60), HD (7 @ $328.57) and WMT (20 @ $121.17) — alongside one rejection. The two sessions since have been quieter: 30 October executed nothing and rejected one order, and 2 November executed nothing and rejected two. Account value has drifted up across those runs, from $9,911.60 on the 29th to $10,004.35 and then $10,170.78 on the 2nd, with cash steady at $959.13. So the capital put to work on the 29th is, for now, paying off.
What the numbers say
The headline backtest covers 451 days: a 6.95% total return (final equity $10,695.11), a 3.83% CAGR, and 108 trades. The weak spots are visible. The Sharpe is a thin 0.34, max drawdown reached 21.73%, and the win rate is just 38.46% — low, but normal for breakout systems that rely on a few large winners. Turnover of 2,083% is heavy, though fees stayed modest at $108 and FX cost was zero.
Validation: a narrow failure
This is where it gets interesting. The strategy did not pass validation, yet the walk-forward picture is not disastrous. Across four folds, three were positive: +0.73%, +14.09% and +11.35%, with a single ugly fold of −7.46% (Sharpe −1.07) in early-to-mid 2025 driving most of the full-period drawdown. The two most recent folds are genuinely strong — Sharpes of 2.72 and 1.86 — and out-of-sample return was 11.35% at a 1.86 Sharpe.
The gate failed on robustness statistics, not raw returns. The deflated Sharpe ratio sits at just 0.198, and with six trials in the search, the probabilistic Sharpe of 0.674 isn't enough to clear the overfitting bar. In plain terms: the recent out-of-sample strength is encouraging, but once you penalise for multiple trials and the strategy's own return variance, there isn't yet statistical confidence that the edge is real rather than lucky.
Verdict
Strengths: a transparent, well-understood mechanism, improving recent folds, and live equity trending up. Risks: a sub-0.4 full-period Sharpe, a 22% drawdown, a failed validation gate, and the ever-present breakout trap of whipsaw in range-bound regimes. It earns its live slot on momentum and clarity, not on proven edge — one to watch, not yet to trust.