The thesis
Dual-momentum is about as classic as trend-following gets: hold the strongest trending names by 60-day return, and exit when the trend breaks. It runs across a 24-name universe of US large caps spanning tech (AAPL, MSFT, NVDA), financials (JPM, V, MA), staples (PG, KO, WMT) and energy (XOM, CVX). The idea is simple and well-documented in the literature — ride winners, cut losers — which is a point in its favour. Simple theses are easier to trust and harder to overfit by accident.
The backtest
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 with a 15.67% max drawdown. Those are respectable, if unspectacular, numbers.
Two details deserve a flag. First, the win rate is just 28.79% across 136 trades: this is a strategy that loses more often than it wins and relies on a handful of large winners to carry the result — textbook momentum behaviour, but it demands discipline and tolerance for long losing streaks. Second, turnover is a heavy 2,638%, meaning the book churns many times over. Fees totalled only $136 in the backtest, but in a live, fee-realistic setting that churn is a genuine headwind.
Walk-forward validation
This is where enthusiasm needs tempering: validation did not pass. Across four folds, three were positive and one — the January–July 2025 window — lost 7.31% with a -1.05 Sharpe and a 17.15% drawdown. The strongest fold (July–December 2025) returned 25.59% at a 3.32 Sharpe, which flatters the average.
The statistics tell the cautionary story. The Probabilistic Sharpe Ratio is a healthy 0.893, but the Deflated Sharpe Ratio — which penalises for the six trials run — sits at just 0.476. In plain terms, once you account for how many variants were tested, there is less than even odds that the true Sharpe is positive. The out-of-sample return of 13.34% is real but materially below the full-sample 23.5%, the usual signature of some in-sample optimism.
Recent activity
Live behaviour is the most striking part of this update. The last five scheduled runs — 29 October through 4 November — all report the same line: 0 executed, 0 rejected, cash $10,000, total $10,000. The strategy is fully in cash and has done nothing for a week. The reviewer approved it back in May as a built-in hand-written strategy at risk level 0, so this is not a block — it is simply the trend filter finding no qualifying names. Whether that is prudent patience or a signal the universe is currently trendless is worth watching.
Verdict
Dual-momentum has a sound, legible thesis and a backtest that looks good at a glance. But the failed validation, the overfitting-flavoured DSR, the brutal fold-2 drawdown, and the low win rate all argue for caution. Right now it is risking nothing — which protects capital but also means it is contributing no live evidence either way. Keep it live, keep it small, and judge it on out-of-sample behaviour, not the headline 23.5%.