The thesis
momentum-code runs a simple, legible idea: buy the top positive movers in its universe, capped per position. The universe is 24 large-cap US names spanning tech (AAPL, MSFT, NVDA, GOOGL), financials (JPM, BAC, V, MA), healthcare (JNJ, UNH, PFE, ABBV), staples (PG, KO, WMT, COST), and industrials/energy (CAT, HON, XOM, CVX). No exotic factors, no leverage — just chase strength and size positions modestly. The strategy is flagged live.
Backtest and validation
The headline backtest is respectable: 19.76% total return over 451 days, finishing at $11,976 on a notional $10k, for a 10.6% CAGR. Fees were trivial ($5 total, no FX cost), and turnover ran at 92.74% — active but not frantic.
The walk-forward picture is where it gets interesting. All four folds were positive (15.38%, 3.27%, 12.18%, 13.71%), with fold Sharpes of 2.47, 0.46, 2.31, and 1.93. The most recent out-of-sample window returned 13.71% at a 1.93 Sharpe. On paper, this is exactly the consistency you want from a momentum rule.
And yet validation.passed is false. The tension sits in the risk-adjusted and multiple-testing metrics. The full-sample Sharpe is only 0.65 — far below the individual folds, which implies the drawdowns stitched together across the whole period hurt more than any single fold suggests. Max drawdown on the full run was 20.48%, deeper than the worst single fold (16.55%). Most tellingly, with 6 trials evaluated, the deflated Sharpe ratio (DSR 0.338) lands well below the probabilistic Sharpe (PSR 0.811). The gate is right to be skeptical: strong folds plus a modest aggregate Sharpe and a non-trivial trial count is a classic overfitting-risk signature.
One caveat worth flagging honestly: the reported win rate is 0% across 5 trades. With positions likely still open and unrealized, that figure reads as an artifact of how closed trades are counted, not evidence that every trade lost — but it means the win-rate metric carries no information here.
Recent activity
The live book tells a quieter, more awkward story. The last executed trades were on 31 May – 1 June 2026: buys in MSFT, HON, BAC, NVDA, and XOM. Since then — nothing. Every scheduled run from 28 September through 5 October logged 0 executed, 1–3 rejected, with cash pinned at exactly $608.79 the entire time. Total portfolio value has drifted between roughly $10,344 and $10,534.
The read is operational, not strategic: with only ~$609 of deployable cash against a per-position cap, the strategy keeps generating orders it cannot fill, so they bounce. The engine is alive; the capital is stuck.
Verdict
momentum-code has a genuinely encouraging out-of-sample profile and a thesis that's easy to trust. But the failed gate — driven by a soft aggregate Sharpe and a low deflated Sharpe — is a warranted brake, and the stalled live book shows the rule needs rebalancing logic or fresh cash to actually trade. Promising signal, unfinished plumbing.