The thesis
Donchian Breakout is trend-following in its purest, oldest form: buy when price makes a fresh 20-day high, exit when it prints a 20-day low. It runs across a 24-name universe of US large caps — mega-cap tech, financials, staples, healthcare and energy — so it fishes for breakouts wherever momentum happens to be. The premise is honest and well-documented: cut losers quickly, let winners run, and accept that most trades will be small losses paid for by a handful of large gains.
Backtest and validation
The headline numbers tell that story exactly. Over 451 days the strategy returned 6.95% (3.83% CAGR) with a 38.46% win rate across 108 trades. A sub-40% hit rate is normal — even desirable — for breakout systems, but the supporting stats are less flattering. The overall Sharpe of 0.34 is thin, the 21.73% max drawdown is steep for the return earned, and turnover north of 2,000% means the edge has to survive heavy churn.
Walk-forward analysis is where it gets interesting. Three of four folds were positive, and the two most recent folds are genuinely strong — +14.09% (Sharpe 2.72) and +11.35% (Sharpe 1.86) — with out-of-sample Sharpe averaging 1.86. That is real, recent, unseen-data performance. The blemish is fold 2 (Jan–Jul 2025), which lost 7.46% at a Sharpe of -1.07, exactly the choppy, whipsaw regime that punishes breakout entries.
Crucially, validation did not pass. The probabilistic Sharpe ratio sits at 0.674, but the deflated Sharpe ratio — which penalises for the six trials run — collapses to 0.198. In plain terms: once you account for how many variants were tested, we can't confidently claim the edge is more than luck. The gate is doing its job here.
Recent live activity
On paper, the strategy has been cautious to a fault. From August 6–11, every scheduled run rejected all candidate trades (0 executed, up to 3 rejected each day) — a sign that either position sizing or cash constraints kept it sidelined during a period without clean breakouts. It finally acted on the 12th and 13th, buying V and MA while selling AAPL and UNH, a rotation out of two names that presumably breached their 20-day lows.
The live paper book stands at roughly $9,778 total with ~$1,511 cash — modestly below its notional starting line. Early days, small sample, but worth flagging: the backtest's best folds are recent, yet the live account hasn't captured that tailwind.
Verdict
This is a clean, interpretable strategy with encouraging recent out-of-sample behaviour and a disciplined exit rule. The risks are equally clear: a middling full-period Sharpe, a fold-2 drawdown that shows its regime dependence, and a deflated-Sharpe score that formally flags overfitting risk. Keeping it live but unpromoted — traded on paper while validation stays red — looks like the right call. Let it earn trust on fresh data before it earns capital.