The Thesis
Dual-momentum runs a simple, well-worn playbook: hold the strongest trending names by 60-day return, and exit when the trend breaks. Its universe is 24 large-cap US names spanning tech (AAPL, MSFT, NVDA), financials (JPM, V, MA), staples (PG, KO, COST), healthcare (JNJ, UNH), and energy (XOM, CVX). This is a trend-following approach — it does not try to predict reversals, it rides what is already working and cuts what is not.
Backtest Performance
Over 451 days the strategy returned 23.5%, growing a $10,000 book to $12,349.63 — a 12.52% CAGR. The Sharpe of 0.95 is respectable but not exceptional, and the 15.67% max drawdown is meaningful for a large-cap-only book.
Two numbers deserve attention. The win rate is just 28.79% across 136 trades. That is characteristic of trend-following: most positions are small losers, and a minority of large winners carry the return. It works, but it demands discipline through long stretches of being wrong. Second, turnover of 2,638% is very high — the book churns roughly 26x over the test window. Fees came to $136 and FX cost was zero, so frictions were modest here, but at scale or with wider spreads that churn is a real drag.
Validation: Not Passed
This is where balance matters. The walk-forward validation did not pass. Across four folds, three were positive:
- Fold 1 (Aug 2024–Jan 2025): +5.85%, Sharpe 1.23
- Fold 2 (Jan–Jul 2025): −7.31%, Sharpe −1.05, 17.15% drawdown
- Fold 3 (Jul–Dec 2025): +25.59%, Sharpe 3.32
- Fold 4 (Dec 2025–May 2026): +13.34%, Sharpe 2.15
Out-of-sample the strategy returned 13.34% at a strong OOS Sharpe of 2.15, and the Probabilistic Sharpe Ratio of 0.893 is encouraging. But the Deflated Sharpe Ratio of 0.476 — which penalises for the 6 trials run — sits below a passing threshold. In plain terms: after accounting for how many variants were tried, we cannot confidently rule out that the edge is partly luck. Fold 2 is the honest warning: this strategy can bleed for six months when trends chop.
Recent Activity
September has been silent. The last six scheduled runs (Aug 26 through Sep 2) each executed 0 trades, leaving the book fully in cash at $10,000. No name is trending strongly enough to trigger entry, or the trend filter is holding it out. That is the model behaving as designed — but it also means zero live evidence to update our priors right now.
Verdict
Dual-momentum has a coherent thesis and a genuinely good OOS Sharpe, but the failed validation and low win rate argue for humility. Treat it as a promising candidate under observation, not a proven edge. The current all-cash posture is a reasonable place to wait for the trend signal — and for more live data before sizing up.