The Thesis
Dual-momentum is a classic idea rendered simply: hold the strongest trending names by 60-day return, and exit when the trend breaks. It runs on a 24-name large-cap universe spanning tech (AAPL, MSFT, NVDA), financials (JPM, V, MA), staples (PG, KO, WMT), and healthcare (JNJ, UNH). The bet is that recent relative strength persists long enough to be captured — a well-documented anomaly, but one that pays unevenly.
Backtest Performance
Over 451 trading days the strategy returned 23.5% (CAGR 12.52%), lifting a $10,000 book to $12,349. The Sharpe of 0.95 is solid without being spectacular, and max drawdown of 15.67% is tolerable for a directional equity strategy.
Two numbers deserve a hard look. The win rate is just 28.79% across 136 trades — this is a strategy that loses small and often, and relies on a minority of large winners to carry the book. That payoff profile is textbook trend-following, but it demands discipline and a long runway to realise its edge. Second, turnover of 2,638% is punishingly high; the book churns roughly 26x over the test. Fees came to $136 here, but in a higher-cost or higher-slippage environment that churn is a real headwind.
Validation: Did Not Pass
This is the headline risk. Walk-forward validation failed. Across four out-of-sample folds, three were positive and one — the Jan–Jul 2025 window — lost 7.31% with a Sharpe of -1.05 and a 17.15% drawdown. The other three folds ranged from good (+5.85%) to excellent (+25.59%, Sharpe 3.32).
The deflated Sharpe ratio (DSR 0.476) is the likely culprit for the fail: with 6 trials in the search, the probabilistic Sharpe (0.893) survives multiple-testing scrutiny only partway, and the DSR lands below a pass threshold. In plain terms: some of the headline edge may be selection luck, and one regime (early 2025) genuinely hurt. The out-of-sample return of 13.34% is encouraging, but not enough to clear the bar.
Recent Live Activity
The most striking signal from the last week is inactivity. The six most recent scheduled runs (Aug 14–21) each report 0 executed, 0 rejected, with the book flat at $10,000 cash. No name currently clears the trend filter, so the strategy is entirely in cash. That is arguably correct behaviour — a trend model with nothing to buy should wait — but it means zero live P&L to date and no fresh evidence to update our priors.
Verdict
Dual-momentum has a coherent thesis and a genuinely strong best-case (fold 3, +25.59%). But the low win rate, extreme turnover, one badly negative fold, and a failed validation gate all argue for caution. It stays live, but this is a watch-and-learn posting, not a scale-up one. We want to see it actually deploy capital and survive a full trend cycle before trusting the backtest's promise.