The thesis
Donchian-breakout is trend following in its purest form: buy the 20-day high, sell the 20-day low. The logic is that a fresh breakout marks the start of a durable move, so the strategy pays up to join momentum and cuts positions that fail. It runs a diversified 24-name large-cap universe — mega-cap tech (AAPL, MSFT, NVDA), financials (JPM, V, MA), staples (KO, PG, WMT) and industrials (CAT, HON) — which spreads breakout signals across sectors rather than concentrating in one theme.
Backtest performance
The headline numbers are honest-to-form trend following. Over 451 days the strategy returned 6.95% (final equity $10,695 on a $10k base, ~3.83% CAGR), with a Sharpe of 0.34 and a steep 21.73% max drawdown. The 38.46% win rate across 108 trades is a feature, not a bug: trend systems lose small on frequent false breakouts and rely on a few large winners to carry the year. What is harder to love is turnover above 2,000% — a lot of churn, which is why fees ($108 on the run) matter and why whipsaw in range-bound tape is the core risk.
Validation: promising folds, failing verdict
Walk-forward validation returned passed: false, and it is worth understanding why. Three of four out-of-sample folds were positive, and the two most recent folds were genuinely strong — +14.09% (Sharpe 2.72) and +11.35% (Sharpe 1.86). Fold 2, however, bled -7.46% with a -1.07 Sharpe during what was clearly a choppy stretch.
The deeper caution comes from the risk-adjusted screens. Against 6 trials, the deflated Sharpe ratio (DSR) is just 0.198 and the probabilistic Sharpe (PSR) is 0.674 — meaning once you discount for how many variants were tested, the full-sample edge is not statistically convincing. Recent out-of-sample strength is real, but it may reflect a favourable trending regime rather than a robust, all-weather signal.
Recent live activity
Live behaviour has gone quiet. The last executed trades were in early June — buys in KO, UNH, ABBV and AAPL, sells trimming CAT and MSFT. Every scheduled run from July 23–30 logged 0 executions and 1–2 rejections, with cash pinned at $1,760.89 and total equity drifting between roughly $9,960 and $10,405. That pattern is consistent with a breakout system finding no fresh 20-day highs it can fund — the market simply is not handing it entries.
Verdict
Donchian-breakout is doing exactly what it says: patient in the absence of breakouts, powerful when trends run (see folds 3 and 4). The risks are equally clear — a low win rate, a 20%+ drawdown, heavy turnover, and a validation verdict that says the edge is not yet proven after multiple-testing adjustment. We would keep it live in paper trading as a regime-dependent diversifier, but the failed DSR gate means it is not ready to be trusted with conviction sizing.