Thesis
momentum-code runs a deceptively simple playbook: each session it scans a 24-name universe of household large-caps — technology, financials, healthcare, consumer staples, energy, and industrials — and buys whatever led the day's gainers, subject to per-position caps. The idea is to harvest short-term price momentum in names liquid enough to enter and exit without slippage friction eating the edge.
Backtest Snapshot
Over 451 days of paper trading the strategy compounded to +19.76% total return (10.6% annualised CAGR) on a starting notional that closed at $11,976. Five trades were executed with $5 in total fees and zero FX cost — consistent with a concentrated, low-turnover approach despite the headline 92.74% turnover figure (which reflects position resets, not churn for its own sake).
The Sharpe ratio sits at 0.65 across the full period, and the worst drawdown reached −20.48% — a meaningful peak-to-trough that will test any allocator's conviction.
Walk-Forward Validation: Four Folds, Mixed Signals
The four-fold walk-forward tells a more nuanced story:
| Fold | Period | Return | Sharpe | Max DD |
|---|---|---|---|---|
| 1 | Aug 2024 – Jan 2025 | +15.38% | 2.47 | −6.12% |
| 2 | Jan 2025 – Jul 2025 | +3.27% | 0.46 | −16.55% |
| 3 | Jul 2025 – Dec 2025 | +12.18% | 2.31 | −5.21% |
| 4 | Dec 2025 – May 2026 | +13.71% | 1.93 | −6.32% |
All four folds finished positive — an encouraging baseline. But the dispersion is wide. Fold 2 stands out: a 16.55% drawdown paired with a 0.46 Sharpe signals that momentum strategies can chop badly when the market rotates or trends break. Folds 1, 3, and 4 look genuinely strong, with risk-adjusted returns most discretionary managers would accept.
The out-of-sample (OOS) return of +13.71% and OOS Sharpe of 1.93 are the headline positives — the most recent held-out period is the cleanest proof the edge hasn't fully decayed.
Why Validation Failed
Despite a clean OOS result, the strategy did not pass the automated validation gate. The probabilistic Sharpe ratio (PSR) of 0.811 and deflated Sharpe ratio (DSR) of 0.338 reflect that six trials were run against this parameter set. Each additional trial inflates the multiple-testing penalty: what looks like a 0.65 in-sample Sharpe deflates sharply when adjusted for the search. The DSR of 0.338 is the gating number to watch — it implies a meaningful probability the observed performance is a statistical artefact of the trial count rather than a durable edge.
Recent Activity: Patience Mode
The past week has been conspicuously quiet. Daily scheduled runs from June 24 through July 1 each logged zero executions — between one and three orders were rejected per session. Portfolio value has drifted in a narrow range ($9,401–$9,507) while $608.79 sits in cash. The last executed buys came on June 1 (XOM at $148.69, NVDA at $224.20) and May 31 (BAC, HON, MSFT), all rated "executed."
The rejection pattern suggests the daily movers aren't clearing the strategy's internal filters — either no candidate clears the minimum move threshold or position-cap constraints are binding.
Verdict
Strengths: Positive returns in all four walk-forward folds, a strong OOS Sharpe, and a diversified large-cap universe that limits idiosyncratic blow-ups.
Risks: A 20%+ max drawdown is real money in a sharp rotation (Fold 2 proved it). The DSR failure flag is not cosmetic — the strategy needs more out-of-sample track record before a live allocation is warranted. A zero win-rate metric in the backtest summary also warrants deeper investigation; it likely reflects a data artefact in how closed positions are counted, but it should be confirmed.
momentum-code is a promising first-generation agent. Letting it accumulate another quarter of live paper-trade data before revisiting the gate seems prudent.