The Thesis
Donchian breakout is one of the oldest systematic ideas in trend-following: buy a stock the moment it prints a new 20-day high, and exit when it falls back to a 20-day low. The logic is regime-detection by price action alone — no earnings forecasts, no macro overlays. Applied here to a 24-stock universe of large-cap US names spanning tech, financials, healthcare, consumer staples, energy, and industrials, the strategy stays fully systematic and runs daily after market close.
Backtest in Brief
Over 451 days the strategy compounded $10,000 to roughly $10,695 — a 6.95% total return, or about 3.83% annualised. Those headline numbers are underwhelming on their own. The deeper concern is the shape of the distribution: a Sharpe of 0.34, a win rate of only 38.46% across 108 trades, and a maximum drawdown of 21.73%. Turnover ran at an eye-watering 2,082%, which on a paper portfolio with flat $1-per-trade fees came to $108 in costs — manageable, but worth watching as position sizes scale.
A sub-40% win rate is normal for trend strategies; they make money by letting winners run far longer than they cut losers. The risk is that a 21.7% peak-to-trough drawdown demands real investor conviction during the losing stretches.
Fold-by-Fold: A Tale of Two Regimes
Walk-forward validation divides the backtest into four roughly equal periods, and the contrast is stark:
| Fold | Period | Return | Sharpe | Max DD |
|---|---|---|---|---|
| 1 | Aug 2024 – Jan 2025 | +0.73% | 0.21 | 6.71% |
| 2 | Jan 2025 – Jul 2025 | −7.46% | −1.07 | 15.58% |
| 3 | Jul 2025 – Dec 2025 | +14.09% | 2.72 | 3.14% |
| 4 | Dec 2025 – May 2026 | +11.35% | 1.86 | 6.01% |
Fold 2 is the red flag: a sustained losing period with a Sharpe below −1 and the strategy's heaviest drawdown. Fold 3 and 4, however, are a different story — back-to-back strong quarters with Sharpe ratios that any discretionary manager would be happy with. Three of four folds finished positive.
Why Validation Didn't Pass
Despite the recent momentum, donchian-breakout did not clear Headmars' validation gate. The Probabilistic Sharpe Ratio sits at 0.674 and the Deflated Sharpe Ratio at just 0.198 — the latter adjusting for the number of parameter combinations trialled (six). A DSR below 0.5 suggests there is meaningful probability the observed edge does not persist out of sample. The validation framework is deliberately conservative for good reason: trend strategies are prone to look-ahead and regime-fitting, and a bad fold like Fold 2 can erase months of gains quickly.
Live Activity: One Week In
The strategy went live on 31 May 2026 with a $10,000 allocation and has been steadily deploying capital into breakout signals:
- 31 May: MSFT × 5 @ $450.24
- 2 Jun: AAPL × 7 @ $314.77
- 3 Jun: ABBV × 11 @ $218.75
- 4 Jun: CAT × 2 @ $941.58
As of the 5 June run the portfolio stood at $9,829 with $1,256 in cash — fractionally below starting equity, which is expected given entry costs and the early stage of position building. One signal was rejected on both 4 and 5 June, indicating the position-sizing or risk rules turned away at least one breakout candidate.
Balanced View
Strengths: The out-of-sample Fold 4 result — 11.35% return with a 1.86 Sharpe and only a 6% drawdown — is genuinely encouraging and aligns with the current live-period market environment. The rule set is transparent, parameter-light, and well-understood.
Risks: The failed DSR is a real caution flag, not a technicality. A strategy with six parameter trials, a full-period Sharpe of 0.34, and one deeply negative fold has not proven enough statistical edge to be considered robust. Until live performance accumulates, this one warrants close monitoring rather than expanded allocation.