The thesis
momentum-code runs a simple, well-worn idea: buy the top positive movers in its 24-name large-cap universe, capped per position to avoid concentration. The universe is defensible and liquid — mega-cap tech (AAPL, MSFT, NVDA), financials (JPM, V, MA), staples (PG, KO, WMT), and energy (XOM, CVX). There is nothing exotic here, which is a point in its favor: momentum on liquid large caps is one of the most durable published anomalies, and a capped, rules-based version is hard to overfit by accident.
Backtest and validation
The headline numbers are attractive. Over 451 days the strategy returned 19.76% (CAGR 10.6%), ending at $11,976 on a $10k base, with a Sharpe of 0.65 and a max drawdown of 20.48%. Turnover was a modest 92.74% and fees were trivial ($5), so the return is not being eaten by friction.
Walk-forward is where it earns real credibility — and also where the caution flag goes up. All four out-of-sample folds were positive (15.38%, 3.27%, 12.18%, 13.71%), with the most recent fold posting a 1.93 Sharpe on a shallow 6.32% drawdown. Consistency across non-overlapping windows is exactly what you want to see; a strategy that only works in one regime would show a losing fold.
Despite that, our validation gate marked it as failed. The tell is in the deflated statistics: the Probabilistic Sharpe Ratio is a healthy 0.811, but the Deflated Sharpe Ratio is just 0.338 across 6 trials. Once you penalize the observed Sharpe for the number of variants tested, the edge is no longer statistically convincing. This is the gate doing its job — four good folds are encouraging, but not the same as a robust, multiple-testing-adjusted result.
One oddity worth flagging: the reported win rate is 0% on only 5 trades, which is inconsistent with a positive return. That almost certainly reflects open positions with no closed round-trips scored, not five losing bets — but it is a reminder that the trade sample is thin and the point estimates are noisy.
Recent activity
Here is the more pressing problem. The strategy is nominally live, but its last six scheduled runs — from 2026-08-03 through 2026-08-10 — all report 0 executed, with 1–3 orders rejected each day. The account is sitting on just $608.79 in cash against a ~$10.8k total, so new buy orders are almost certainly being rejected for insufficient funds. The last real fills were back on 2026-05-31 and 2026-06-01 (MSFT, HON, BAC, NVDA, XOM). In other words, the engine deployed its capital, went fully invested, and has been unable to act since.
Verdict
momentum-code has a sound thesis and genuinely encouraging walk-forward consistency, but two things temper the enthusiasm: the deflated statistics failed the gate, and the live account is effectively frozen for lack of deployable cash. Before treating the backtest as a green light, it needs a cash/sizing fix so it can rebalance, and more trades to firm up a sample that is currently too small to trust.