The thesis
Dual-momentum is a simple, well-worn idea executed on a clean large-cap universe of 24 names (AAPL, MSFT, NVDA, JPM, V, JNJ, XOM and peers). The rule: hold the strongest trending names ranked by 60-day return, and exit on a trend break. No stock-picking narrative, no earnings calls — just price. That discipline is a feature, not a bug: it keeps the logic auditable and the overfitting surface small.
Headline numbers
Over a 451-day backtest the strategy returned 23.5% (final equity $12,349.63 from $10,000), a 12.52% CAGR, with a Sharpe of 0.95 and a max drawdown of 15.67%. Respectable, but two figures deserve a closer look. The win rate is just 28.79% across 136 trades — entirely expected for momentum, where a minority of large winners carries the book, but it means the equity curve lives or dies on a few positions. And turnover of 2,638% is high; the $136 in fees is modest on paper, yet real-world slippage at that churn is the first thing I'd stress-test.
Validation: the gate says no
This is where balance matters. Our walk-forward validation did not pass. Three of four folds were positive, and the recent ones look excellent — Fold 3 returned +25.59% (Sharpe 3.32) and Fold 4 +13.34% (Sharpe 2.15). The out-of-sample Sharpe of 2.15 and a Probabilistic Sharpe Ratio of 0.893 are genuinely encouraging.
But Fold 2 (Jan–Jul 2025) lost 7.31% with a 17.15% drawdown — the deepest in the sample — exposing the classic momentum failure mode: trend reversals whipsaw the book. More tellingly, the Deflated Sharpe Ratio is 0.476. After adjusting for the 6 trials behind this configuration, our confidence that the edge is real falls below the bar. The raw Sharpe looks fine; the trial-adjusted one does not. That gap is exactly what the gate exists to catch.
Recent activity: parked in cash
The strategy has been quiet — deliberately so. Every scheduled run from 8 to 15 October executed zero trades, holding a flat $10,000 in cash. With no name clearing the trend threshold, the exit-on-break logic has kept it out of the market entirely for over a week. That is the model behaving correctly, not a malfunction, but it also means current live performance reflects conviction-by-abstention rather than active skill.
The verdict
Dual-momentum has an honest thesis and a plausible out-of-sample signal, but it carries a failed validation flag, a thin win rate, and heavy turnover. I'd treat it as a promising candidate under watch — not one to size up — until a fresh fold either lifts the DSR above threshold or confirms Fold 2 was an outlier rather than a preview.