The thesis
News-sentiment runs a simple, legible premise: buy names on positive recent news sentiment and exit when the tone turns negative. It trades a 24-stock large-cap universe spanning tech, financials, healthcare, staples and energy — AAPL, MSFT, NVDA, JPM, JNJ, XOM and the like. The appeal is interpretability: each position should map to a readable catalyst rather than an opaque factor.
Recent activity
The book is live and sat at $10,510 total with $2,036 in cash as of the 2026-10-15 run — modestly ahead of its $10,000 start. Activity has been thin lately: the six most recent scheduled runs (Oct 8–15) executed nothing, and three of them logged a single rejected order. The last real movement was a 2026-09-23 rotation — selling 35 shares of BAC at $55.96 (bought at $55.93 in June, essentially flat) and opening a 4-share NVDA position at $225.34. Earlier executed buys include AAPL, MSFT, UNH and GOOGL between May and July. In short, the strategy acts rarely, and recently not at all.
Backtest and validation
Over a 451-day backtest the strategy returned +0.19% (final equity $10,018.55, ~0.1% CAGR) with a 0.74 Sharpe, a shallow 0.05% max drawdown and just $2 in fees. The headline problem is sample size: that entire result comes from only 2 trades. Walk-forward validation did not pass. Of four folds, only one (Dec 2025–May 2026) produced any trades at all; the other three sat flat with zero activity. The out-of-sample figures look flattering in isolation (0.19% return, 1.5 Sharpe), and the probabilistic metrics are respectable — PSR 0.923, DSR 0.551 after deflating for 6 trials — but all of it rests on that single active fold.
Strengths and risks
Strengths: the logic is transparent, drawdowns have been tiny, turnover is low (36.5%) and costs are negligible. This is not a strategy that bleeds out on fees or churn.
Risks dominate the picture, though. The signal fires so infrequently that there is almost no evidence it generalises across regimes — three of four validation windows never triggered. A reported 0% win rate alongside a positive return hints at a measurement quirk (likely open positions excluded) worth a closer look. And a +0.19% backtest edge is economically indistinguishable from noise once you account for the thin trade count.
Bottom line: a clean, cheap, well-behaved idea that has not yet earned conviction. Until news-sentiment generates enough trades to be judged across more than one window, treat the live book's small gain as encouraging but unproven.