The thesis
news-sentiment is about as legible as a strategy gets: buy names carrying positive recent news sentiment, exit when sentiment turns negative. It runs over a 24-stock large-cap universe spanning mega-cap tech (AAPL, MSFT, NVDA), financials (JPM, V, MA), healthcare (JNJ, PFE, ABBV) and staples (KO, PG, WMT). Sensible, diversified, and easy to explain to a sceptic — which is exactly why it deserves hard scrutiny.
Recent activity
The live book is modest and quiet. Its opening trades on 29 Oct put it into V, NVDA and PFE, followed by KO on 3 Nov and a 5-share AAPL buy on 10 Nov. Since then the scheduler has mostly come up empty: five of the last six runs executed zero orders, the lone exception being that 10 Nov fill. The headline total has drifted from $9,726 (6 Nov) to $9,575 (13 Nov) — below the $10,000 line it started from — with roughly $626 parked in cash. Nothing alarming, but nothing working yet either.
Backtest and validation
Here enthusiasm should cool. Over 451 days the backtest produced a +0.19% total return, a 0.74 Sharpe, and a trivially small 0.05% max drawdown — but it did so on just two trades. A strategy that trades twice in roughly eighteen months is not being tested; it is barely being observed.
Walk-forward validation confirms the problem and is flagged as failed. Of four folds, three generated zero trades and zero return; the entire result comes from fold 4 (Dec 2025 – May 2026), which logged a 1.5 Sharpe on its two trades. One positive fold out of four is not a track record.
The statistics echo the caution. The Probabilistic Sharpe Ratio is a healthy 0.923, but the Deflated Sharpe Ratio — which penalises the six trials run — falls to 0.551, hovering near the coin-flip line. The reported win rate is 0%, though on two trades that figure carries no information.
Verdict
The strengths are real but narrow: a clear, interpretable thesis, near-zero drawdown, low turnover in practice, and negligible costs ($2 in fees, no FX drag). The risks dominate. The signal fires far too rarely to trust — three of four folds never traded at all — and every favourable number rests on a two-trade sample inside a single window. Failed validation plus a borderline DSR means this strategy has not earned incremental capital.
The right next step is diagnostic: either the sentiment gate is too strict, or the data feed is too sparse to generate signal. Until it trades often enough to be measured, treat news-sentiment as a hypothesis, not a strategy.