The thesis
Dual-momentum is a classic trend-follower: hold the strongest names by 60-day return and exit when the trend breaks. Its universe is 24 US large caps spanning tech, financials, healthcare, staples and energy — AAPL, MSFT, NVDA, JPM, JNJ, XOM and the like. The logic is deliberately simple: let winners run, cut laggards. Strategies like this tend to buy strength and tolerate a low hit rate in exchange for a few large winners.
Backtest performance
Over 451 days the strategy turned $10,000 into $12,349.63 — a 23.5% total return, or roughly 12.5% CAGR. Sharpe came in at 0.95 against a 15.67% max drawdown. Those are defensible numbers, but two figures deserve a closer look. The win rate is only 28.79% across 136 trades: consistent with a trend model that loses small and often while winning big, though it demands patience and tight risk control. More concerning is turnover of 2,638% — the book churns heavily, and while fees here totalled just $136, that level of trading leaves the strategy exposed to slippage and transaction costs in live conditions.
Validation: did not pass
This is where the headline number earns scrutiny. Our walk-forward validation failed. Across 4 folds, 3 were positive, but the spread is wide:
- Fold 1 (Aug 2024–Jan 2025): +5.85%, Sharpe 1.23
- Fold 2 (Jan–Jul 2025): -7.31%, Sharpe -1.05, 17.15% drawdown across 60 trades
- Fold 3 (Jul–Dec 2025): +25.59%, Sharpe 3.32
- Fold 4 (Dec 2025–May 2026): +13.34%, Sharpe 2.15
Fold 2 is the warning sign: a choppy, trendless market whipsawed the model into 60 trades and a double-digit drawdown. The out-of-sample return of 13.34% and the strong OOS Sharpe of 2.15 are encouraging, but the Deflated Sharpe Ratio of 0.476 — after accounting for 6 trials — sits below the bar we require. The Probabilistic Sharpe (0.893) is healthy in isolation; the DSR, which penalises for multiple testing, is what fails it. In plain terms: the edge may be real, but we can't yet rule out that it survived by luck of selection.
Recent activity: flat and waiting
Live, the strategy has done nothing lately. The last six scheduled runs (16–23 October) each report 0 executed, 0 rejected, with cash and total equity both pinned at $10,000. No name is currently clearing its trend filter, so the model is fully in cash. For a trend-follower this is normal — sitting out a directionless tape is a feature, not a bug — but it means there is no live track record to confirm the backtest yet.
Verdict
Dual-momentum has a coherent thesis and a plausible edge, but it is not validated. The low win rate and high turnover are tolerable if the big winners keep showing up; Fold 2 shows what happens when they don't. We'd keep it live and observed, but we would not scale capital into it until the out-of-sample edge clears the deflated-Sharpe gate and live trades start accumulating.