The thesis
momentum-code runs a deliberately simple idea: buy the top positive movers across a 24-name large-cap universe — the usual mega-cap tech, financials, staples and energy anchors like AAPL, NVDA, JPM, XOM and WMT — with a cap per position to avoid concentration. It is a classic momentum tilt dressed in guardrails. No regime filters, no shorting, just lean into what is already rising and size it sensibly.
Backtest performance
Over 451 days the strategy turned a simulated $10,000 into $11,976.21, a total return of 19.76% and a CAGR of 10.6%. That is a respectable number, and it came cheaply: just $5 in fees and zero FX cost across only five trades, with turnover of 92.74%.
The risk side is where the enthusiasm cools. The Sharpe ratio is 0.65 — modest for an equity strategy — and the maximum drawdown reached 20.48%, which actually exceeds the annualised return. An investor would have had to stomach a loss larger than a typical year's gain to collect that gain. One data point deserves a flag: the reported win rate is 0 across five trades. Since the recorded trades are all buys with no sales, this looks like an artefact of unrealised positions rather than five outright losers, but it warrants reconciliation before anyone leans on it.
Validation
This is the most interesting part of the picture, and it cuts both ways. The walk-forward split produced four folds, all four positive (+15.38%, +3.27%, +12.18%, +13.71%), with out-of-sample return of 13.71% and a notably strong OOS Sharpe of 1.93. The Probabilistic Sharpe Ratio sits at a healthy 0.811.
And yet validation passed: false. The likely culprit is the Deflated Sharpe Ratio of just 0.338 against six trials — once you penalise for multiple testing, the edge no longer clears the bar with confidence. That is exactly the kind of overfitting discipline the platform is supposed to enforce, and it is working as intended: consistently positive folds are encouraging, but they are not the same as a statistically deflated-and-survived edge.
Recent activity
Here lies the real concern. The last executed trades were a cluster of buys at the end of May and start of June 2026 — XOM, NVDA, BAC, HON and MSFT. Since then, every scheduled run from 23 September through 30 September has executed zero trades and rejected one to three each, with cash pinned at $608.79 and the live account hovering around $10,300–$10,400. The strategy is fully invested and cash-constrained: it keeps trying to buy new movers and getting bounced for want of capital.
The verdict
momentum-code is a profitable, low-cost idea with reassuringly consistent fold behaviour — but it is not a finished product. The validation gate failed on deflated significance, the drawdown is uncomfortably large relative to return, and the live book has effectively frozen, unable to rotate into new momentum without freeing cash. The fix is less about the thesis and more about plumbing: position sizing and a sell-side rule so the strategy can actually act on its own signals.