The thesis
News-sentiment runs one of the most intuitive ideas in systematic trading: buy names when recent news sentiment turns positive, exit when it turns negative. Its universe is a clean, liquid slice of US large caps — 24 tickers spanning tech (AAPL, MSFT, GOOGL, NVDA), financials (JPM, BAC, V, MA), healthcare (JNJ, UNH, PFE, ABBV), staples (PG, KO, WMT, COST) and industrials/energy (CAT, HON, XOM, CVX). The logic is easy to reason about, which is a genuine strength: when the strategy acts, you can usually explain why.
What the backtest actually shows
Here is where enthusiasm needs to meet the numbers. Over 451 backtest days the strategy returned just 0.19% (final equity $10,018.55), a 0.74 Sharpe, a shallow 0.05% max drawdown, and only 2 trades with a 0% win rate. That is not a track record — it is barely a sample. Two trades cannot support any confident claim about edge, and a 0% win rate alongside a marginally positive return means the result rests on open, marked-to-market positions rather than realized gains.
Validation: it failed, and it matters
The walk-forward validation returns a blunt verdict: passed = false. Of 4 folds, only 1 was positive — and it was the only fold that traded at all. Folds 1 through 3 (August 2024 through December 2025) produced zero trades and zero return; every point of performance comes from fold 4 (Dec 2025–May 2026), which posted +0.19% on a 1.5 out-of-sample Sharpe.
That concentration is the core risk. A signal that stays silent for three consecutive folds and then fires twice in the fourth is not demonstrating a robust, repeatable edge — it is showing us a single lucky window. The headline PSR of 0.923 looks reassuring until you weigh it against the deflated Sharpe of 0.551 across 6 trials and a trade count you can hold up on one hand. DSR is the more honest metric here, and 0.55 is unremarkable.
Live activity: mostly standing still
Live, the picture is quiet. The most recent scheduled runs (Aug 28 – Sep 4) executed 0 trades, with several orders rejected — unsurprising given cash is pinned at $990.06 against a portfolio total hovering around $10,350–$10,525. The book is effectively fully invested from an earlier burst of buys (AAPL, MSFT, BAC, UNH, GOOGL executed late May through early July), and there is little dry powder to act on fresh signals. So the strategy is currently more holder than trader.
Verdict
News-sentiment is a reasonable idea deployed on unreasonably thin evidence. Its strengths are interpretability, a disciplined universe and a benign drawdown profile. Its risks are decisive: a failed validation, performance sourced from one fold, a two-trade sample, and a live book with no cash to express new views. This is a strategy to watch and starve of size — not one to scale — until it trades enough to prove the thesis survives outside a single fortunate window.