The Thesis
News-sentiment is a simple, intuitive idea: buy names on positive recent news sentiment and exit when sentiment turns negative. It runs across a 24-name large-cap universe spanning tech, financials, healthcare, staples, and energy — AAPL, MSFT, NVDA, JPM, JNJ, XOM and the like. It's the kind of thesis that reads well and is notoriously hard to execute, because sentiment signals are noisy and often already priced in by the time news breaks.
Recent Activity
The live record is mixed. Between late May and early July 2026 the strategy executed five buys — AAPL, MSFT, BAC, UNH, and GOOGL — deploying most of its book (cash now sits at just $990.06). Since then, activity has stalled: every scheduled run from 14–21 August executed zero trades, with one or two orders rejected each day. Total portfolio value has drifted down from $10,403.17 on 14 August to $10,227.33 on 21 August — a modest but consistent bleed while the strategy sits fully invested and unable to act on new signals.
Backtest & Validation
Here the picture gets thinner. The backtest shows a +0.19% total return over 451 days (final equity $10,018.55), a Sharpe of 0.74, a tiny 0.05% max drawdown, and turnover of 36.5% — all off just 2 trades with a 0% win rate. A 0.10% CAGR is effectively flat.
Most tellingly, walk-forward validation failed. Of four folds, only one (Dec 2025–May 2026) traded at all — and that single fold accounts for the entire 0.19% return and its flattering 1.5 out-of-sample Sharpe. Folds 1–3 produced zero trades and zero return. The probabilistic Sharpe ratio looks strong at 0.923, but the deflated Sharpe (0.551), which penalises for the 6 trials run, tells the honest story.
Strengths and Risks
Strengths: capital discipline is evident — drawdowns are negligible and the strategy doesn't overtrade. When it did engage in the one active fold, the risk-adjusted result was respectable.
Risks: the sample is far too small to trust. Two backtest trades and a single productive validation fold is not evidence of edge — it's a coin flip that landed heads once. The recurring live rejections suggest the signal rarely clears its own thresholds, leaving the book static and exposed to drift. A strategy that trades this seldom cannot compound, and its "win" may be luck rather than repeatable skill.
Verdict
News-sentiment remains a promising idea in search of a working signal. Until it demonstrates activity across multiple validation folds — and resolves why live orders keep getting rejected — its live status should be read as an experiment, not a conviction.