The thesis
News-sentiment runs a simple, legible rule across 24 large-cap U.S. names — Apple, Microsoft, Nvidia, JPMorgan, UnitedHealth and the like: buy when recent news sentiment turns positive, exit when it turns negative. It is the kind of thesis that is easy to reason about and easy to audit, which is a genuine virtue. The strategy is currently flagged live.
Recent activity
The live tape is quiet. The last executed order was a buy of 5 GOOGL at $360.40 on 2026-07-01, following a cluster of executed buys in late May and June (AAPL, MSFT, BAC, UNH). Since then, every scheduled run has come up empty. Across 2026-09-04 through 2026-09-11, six runs each executed zero orders while rejecting one or two candidates apiece. The live account holds cash of $990.06 and a total value drifting between roughly $10,277 and $10,398 as its existing positions mark to market.
That gap between "live and holding" and "live and trading" matters: the signal is either not firing or its candidates are failing risk checks. Neither is fatal, but a strategy that stops acting stops learning.
Backtest and validation
The headline backtest is tidy but slight. Over 451 days it returned 0.19% (final equity $10,018.55), a 0.1% CAGR, with a Sharpe of 0.74, a maximum drawdown of just 0.05%, turnover of 36.5%, and only $2 in fees. The drawdown discipline is the standout number — capital preservation here is excellent.
The problem is sample size. The entire result rests on 2 trades, and the reported win rate is 0. That is not a track record; it is an anecdote.
Walk-forward validation makes the fragility explicit. Of four folds, only one was positive — and it is the only fold that traded at all. Folds one through three (spanning 2024-08 to 2025-12) registered zero trades and zero return; fold four (2025-12 to 2026-05) produced the whole 0.19% at a 1.5 Sharpe. Formal validation did not pass. On the multiple-testing side, the probabilistic Sharpe ratio is reassuring at 0.923, but across 6 trials the deflated Sharpe falls to 0.551, and a strategy that only engages in one window of four cannot claim robustness.
Verdict
Strengths: an interpretable thesis, negligible costs, and remarkably contained drawdown. Risks: a two-trade backtest, a failed validation gate, three of four folds with no activity, and a live account that hasn't executed in over two months.
The honest read is that news-sentiment is under-tested rather than disproven. It needs to trade — a lot more — before its low-drawdown profile means anything. Until the sample grows, treat the numbers as a hypothesis, not a result.