The thesis
news-sentiment is a long-only, news-driven rotation: buy names with positive recent news sentiment, exit when sentiment turns negative. It trades a 24-name large-cap US universe spanning tech (AAPL, MSFT, GOOGL, NVDA), financials (JPM, BAC, V, MA), healthcare (JNJ, UNH, PFE, ABBV), staples (PG, KO, WMT, COST), and a handful of energy and industrial names. The strategy is currently flagged live.
Recent activity
The last executed trades landed on 2026-09-23: a buy of 4 NVDA at $225.34 and a sell of 35 BAC at $55.96. Since then, the six scheduled runs from 1–8 October produced zero executions. Five of those runs logged no rejections; the 8 October run logged a single rejection. Cash has sat unchanged at $2,036.71 throughout.
Meanwhile the portfolio total drifted from $10,154.67 (1 Oct) up to $10,387.83 (7 Oct) before easing back to $10,317.89 (8 Oct). Because cash is static and nothing new has been bought or sold, those swings are mark-to-market moves on existing holdings — NVDA, GOOGL, UNH, MSFT, AAPL — not fresh conviction from the sentiment engine.
Backtest & validation
Over a 451-day backtest the numbers are modest: total return 0.19%, CAGR roughly 0.1%, Sharpe 0.74, max drawdown just 0.05%, turnover 36.53%, and only 2 trades with a 0% win rate. Final equity was $10,018.55 on a $10,000 base, with $2 in fees and no FX cost.
The walk-forward tells the real story. Across four folds, only fold 4 (2025-12-16 → 2026-05-29) traded at all — 2 trades, +0.19%, Sharpe 1.5. Folds 1 through 3 produced zero trades and zero return. The entire track record therefore rests on one window. PSR of 0.923 looks encouraging, but DSR of 0.551 across 6 trials is middling, and the overall verdict is a FAIL: only 1 of 4 folds was positive.
Strengths and risks
Strengths. The out-of-sample return is positive, fold 4 posted a respectable Sharpe of 1.5, and the 0.05% max drawdown means capital preservation has been excellent. The strategy does not appear to take on reckless risk.
Risks. The sample is tiny — two trades is not an edge, it's an anecdote. Three of four folds never fired, the win rate is 0%, and a 0.19% return over 451 days is statistically indistinguishable from flat. The signal fires rarely, leaving the book mostly idle or simply holding, which is exactly what the quiet October run log shows.
Verdict: a sensible mechanism with an unproven edge. Before it earns conviction, news-sentiment needs materially more signal activity and a longer live track record — right now it is passing the validation gate's risk check but failing its evidence check.