The thesis
news-sentiment runs a clean, interpretable idea: buy names carrying positive recent news flow, and exit when sentiment turns negative. It operates over a 24-stock large-cap universe — the mega-cap tech, financials, healthcare, staples and energy you would expect (AAPL, MSFT, NVDA, JPM, UNH, XOM and friends). The appeal is transparency: you can reason about why a position opened or closed without untangling a black box.
Recent activity
The strategy is live, but lately it has been quiet to the point of dormancy. Across the scheduled runs from 25 September to 2 October, it executed nothing — a steady run of "0 executed" sessions, with only a few rejected orders (two on 25 September, one on 28 September). Cash has sat unchanged at $2,036.71, and total book value has drifted in a narrow band between roughly $10,155 and $10,275.
The last genuine trades landed on 23 September: a 4-share NVDA buy at $225.34 and a 35-share BAC sell at $55.96. That BAC lot, opened on 12 June at $55.93, closed essentially flat — three cents a share — a tidy illustration of a sentiment exit that neither helped nor hurt. Earlier summer buys in MSFT, UNH, GOOGL and AAPL round out an otherwise thin trade log.
Backtest and validation
Here is where balance matters. The headline backtest reads benign: +0.19% total return over 451 days, a 0.74 Sharpe, and a trivial 0.05% max drawdown on just $2 in fees. But those figures rest on only two trades, and the win rate is zero.
The walk-forward validation did not pass. Of four folds, only one was positive — and folds one through three recorded no trades at all. Every bit of performance traces to fold four (Dec 2025 – May 2026): +0.19%, Sharpe 1.5, two trades. A probabilistic Sharpe of 0.92 sounds reassuring until you read the deflated Sharpe: 0.55 after adjusting for six trials, barely above a coin flip.
Verdict
The real issue isn't that news-sentiment loses money — it's that it barely acts. The signal almost never fires, so there is no sample to trust and no edge to measure.
- Strength: it is disciplined and cheap. Minimal fees, near-zero drawdown, no reckless churn despite a 36.5% turnover figure.
- Risk: with two lifetime trades and three empty folds, every performance number is statistical noise, and the failed validation says so plainly.
Until the entry criteria loosen enough to produce a meaningful trade count, this strategy should be filed as unproven — not safe. A flat equity curve from inactivity is not the same thing as a managed one.