Thesis
News-sentiment is a simple, legible idea: buy names on positive recent news sentiment and exit on negative. It runs over a 24-name large-cap universe spanning tech (AAPL, MSFT, GOOGL, NVDA), financials (JPM, BAC, V, MA), healthcare (JNJ, UNH, PFE, ABBV), and consumer staples (PG, KO, WMT, COST). That breadth is sensible — sentiment signals are noisiest on illiquid or thinly-covered names, and this basket is neither.
Recent activity
The strategy is live on a paper book that currently totals about $10,359 against $990.06 in cash — so it is close to fully invested. Its executed trades are all buys: AAPL and MSFT in late May/early June, BAC and UNH in mid-June, and GOOGL (5 shares at $360.40) on 1 July. Since then, the tape has gone quiet. Scheduled runs from 19–26 August report 0 executed every session, with a handful of rejected orders (one or two per run on the 19th–24th). In other words, the book has been holding, not trading, for nearly two months. That is consistent with a sentiment model that simply hasn't seen a strong enough signal to act — but it also means the live sample is tiny.
Backtest and validation
The backtest is where balance matters most. Over 451 days the strategy produced a 0.19% total return (CAGR ~0.1%, final equity $10,018.55) on just 2 trades, with a Sharpe of 0.74, a shallow 0.05% max drawdown, 36.53% turnover, and $2 in fees. On a risk-adjusted basis those numbers look tidy — low drawdown, positive Sharpe, negligible cost drag.
The walk-forward validation tells a harder story. Across 4 folds, only 1 was positive, and — critically — only fold 4 traded at all (Dec 2025–May 2026, 2 trades, fold Sharpe 1.5). Folds 1 through 3 returned exactly zero because the model never fired. The out-of-sample figures (0.19% return, Sharpe 1.5, PSR 0.923, DSR 0.551 across 6 trials) are entirely carried by that single active window. The validation gate's verdict is unambiguous: passed: false.
Verdict
The strengths are real but modest: a clean thesis, a liquid universe, low realized drawdown, and healthy deflated-Sharpe statistics within its one active fold. The risks are larger. A 0% win rate on 2 closed trades, three of four folds with no activity, and a live book that hasn't executed since July all point to the same conclusion — the signal is too sparse to trust yet. This is a strategy that has barely traded, not one that has been tested. Until it generates a meaningfully larger sample, treat the passing OOS Sharpe as a hypothesis, not a track record, and keep it on the watchlist rather than the allocation sheet.