The thesis
Donchian-breakout is about as classical as systematic trading gets: buy a 20-day high, exit on a 20-day low, and let the trend do the work. It runs across a 24-name universe of US large caps — the usual megacap tech (AAPL, MSFT, GOOGL, NVDA), financials (JPM, V, MA), staples (PG, KO, WMT, COST) and healthcare (JNJ, UNH, ABBV). The logic is hand-written and was approved by the reviewer at risk 0 back in May. No black box here, which is a virtue: every entry and exit is explainable.
Recent activity
The scheduled run on 2026-10-29 executed 4 buys and rejected 1, leaving cash at $959.13 and a total book of $9,911.60. The fills were ABBV (8 @ $289.91), MA (3 @ $636.60), HD (7 @ $328.57) and WMT (20 @ $121.17) — breakouts spread across healthcare, payments, retail and staples rather than crowding into one theme. With the book just below its $10k Season 2 starting line, the strategy is effectively flat-to-slightly-down out of the gate, which is unremarkable for a freshly reset trend follower waiting for trends to form.
Backtest performance
Over 451 days the strategy returned 6.95% (final equity $10,695.11), a 3.83% CAGR. The shape is pure trend-following: a 38.46% win rate across 108 trades, meaning most positions are small losers clipped at the 20-day low, with a minority of winners carrying the result. The Sharpe of 0.34 is weak, and a 21.73% max drawdown is a real stomach test for a single-digit return. Turnover of 2,083% is high but fees stayed trivial ($108 total, no FX cost) given the domestic universe.
Validation — the sticking point
This is where the picture gets interesting. Walk-forward testing across 4 folds was 3-for-4 positive: fold 3 (+14.09%, Sharpe 2.72) and fold 4 (+11.35%, Sharpe 1.86) were genuinely strong, fold 1 was marginally positive, and fold 2 was ugly (−7.46%, Sharpe −1.07). The most recent out-of-sample window actually looks the best the strategy has — encouraging.
Yet the gate reads passed: false. The reason is in the deflated numbers: a PSR of 0.674 is mediocre, and the DSR of just 0.198 — which penalises for the 6 trials run — says we cannot confidently distinguish this edge from luck once selection bias is accounted for. The full-sample Sharpe of 0.34 simply isn't robust enough to survive that haircut.
Verdict
Strengths: transparent logic, clean recent folds, and a diversified, liquid universe. Risks: a low Sharpe, a painful drawdown, and a validation score that honestly flags possible overfitting. The system correctly declined to promote it. Live trading here is best read as continued observation — the thesis is sound, but it needs to prove the strong recent folds weren't the market handing it a favourable regime.