The thesis
momentum-code runs a simple, legible idea: buy the top positive movers in a fixed universe of 24 US large caps — the usual megacap tech, financials, staples, and energy names — with a cap on each position. There's no cleverness to hide behind here, which is a virtue. You can read the rule in one sentence and know exactly what it's betting on: that recent strength persists.
Recent activity
The strategy put money to work in a short burst around the end of October, executing five buys: MA (3 sh @ $636.60), HD (6 @ $328.57), and NKE (52 @ $38.38) on Oct 29, then XOM (10 @ $186.99) and WMT (16 @ $123.53) on Oct 30. Since then, it has effectively gone silent. The scheduled runs from Nov 2 through Nov 6 each report 0 executed, 1–2 rejected, with cash pinned at $177.08 the entire week.
That flat cash balance tells the story: the book is fully invested, so every fresh momentum signal is getting rejected for lack of funds. Total equity has drifted sideways in a narrow band — roughly $9,876 to $10,186 across the week — so live performance is essentially flat-to-slightly-down since the position was built.
Backtest and validation
The backtest is where momentum-code looks its best: 19.76% total return over 451 days, a 10.6% CAGR, final equity of $11,976, and modest costs ($5 in fees, zero FX). Walk-forward validation is genuinely encouraging on the surface — all four folds positive (15.38%, 3.27%, 12.18%, 13.71%), with an out-of-sample return of 13.71% and a healthy OOS Sharpe of 1.93.
And yet the gate returned passed: false. The reason is written in the deflated metrics: with 6 trials in the search, the Deflated Sharpe Ratio sits at just 0.338 — below the bar — even though the Probabilistic Sharpe Ratio is a respectable 0.811. In plain terms, once you penalise for having tried several variants, the edge is no longer statistically convincing. The system was right to withhold its blessing.
Strengths and risks
Strengths: a transparent rule, four-for-four positive folds, strong per-fold Sharpes (up to 2.47), and shallow fold-level drawdowns (5–6% in three of four windows).
Risks, and they are real:
- The full-period Sharpe is only 0.65 versus the glossy fold numbers, and max drawdown is 20.48% — far deeper than any single fold, implying a loss that straddled fold boundaries.
- A reported win rate of 0 across just 5 trades is a red flag about statistical thinness, whatever its exact bookkeeping cause.
- Turnover near 93% means the edge has to clear real churn.
- Live, the strategy is stuck: fully invested, generating rejects, and going nowhere.
The verdict is honest discomfort. The backtest is attractive and the walk-forward consistency is better than most, but the validation gate, the thin trade count, and the stalled live book all point the same way: promising, not proven. It stays on the watchlist, not the trusted list.