The thesis
The idea behind news-sentiment is deliberately simple: buy names carrying positive recent news sentiment across a 24-stock large-cap universe (AAPL, MSFT, GOOGL, NVDA, JPM, and the usual mega-cap suspects), and exit when the tone turns negative. It is a clean, interpretable premise — the kind of rule you can explain in one sentence and audit after the fact. The strategy is currently flagged live.
Recent activity
On the live book, the strategy has been quiet. The last five executed orders were all buys, spanning late May through early July 2026 — AAPL at 312.06, MSFT at 428.23, BAC at 55.93, UNH at 402.85, and GOOGL at 360.40. Since then it has largely sat on its hands: the scheduled runs from 25 August through 1 September executed zero trades each, rejecting one candidate on 28 August and again on 1 September. Cash sits at just $990.06 against a total paper equity of roughly $10,373 on 1 September, so the book is close to fully invested and drifting with its holdings rather than actively rotating.
That is consistent with the thesis — no strong negative sentiment means no forced exits — but it also means the recent P&L is really a story about the underlying mega-caps, not about the signal doing much work.
Backtest and validation
Here the picture gets more sober. Over a 451-day backtest the strategy returned a negligible 0.19% (final equity $10,018.55, CAGR ~0.1%) on just 2 trades, with a shallow 0.05% max drawdown and a Sharpe of 0.74. A Sharpe near 0.74 on an essentially flat return simply reflects how little capital was ever at risk — the engine barely fired.
Formal validation did not pass. Across four walk-forward folds, only one was positive: folds 1–3 (August 2024 through December 2025) recorded zero trades and zero return, and the entire 0.19% came from fold 4, which managed 2 trades and a 1.5 fold Sharpe. So the headline out-of-sample numbers rest on a single active window. The probabilistic Sharpe ratio looks flattering at 0.923, but the deflated Sharpe — which penalises the 6 trials run — drops to 0.551, and the win rate is reported as 0. That combination is the tell.
Verdict
Strengths: the rule is transparent, drawdown has been tiny, and the strategy shows discipline — it does not overtrade, and turnover is a modest 36.5%. Risks: the evidence base is dangerously thin. Three of four validation folds never generated a signal, the whole backtest edge is two trades, and validation failed outright. A strategy that trades this rarely gives us almost no statistical footing to trust it in live conditions.
Right now, news-sentiment reads less like a proven edge and more like a well-behaved placeholder waiting for a sentiment regime loud enough to test it. I would keep it live on a small allocation to gather real signal-firing data, but I would not lean on the backtest as validation until it accumulates far more trades.