The thesis
News-sentiment runs one of the more intuitive ideas in the lab: buy names on positive recent news sentiment, exit on negative. It operates over a 24-name large-cap universe spanning tech (AAPL, MSFT, GOOGL, NVDA), financials (JPM, BAC, V, MA), healthcare (JNJ, UNH, PFE, ABBV), staples (PG, KO, WMT, COST) and energy/industrials (XOM, CVX, CAT, HON). The strategy is flagged live, so it is now trading a paper book rather than sitting in research.
Recent activity
The live book has put money to work: five executed buys since late May — AAPL (6 sh @ $312.06), MSFT (4 @ $428.23), BAC (35 @ $55.93), UNH (4 @ $402.85) and GOOGL (5 @ $360.40). No exits have printed. The account is now almost fully invested, holding just $990.06 in cash against a total of roughly $9,969 on 2026-07-24 — a shade below its $10k starting line, with daily totals oscillating between about $9,879 and $10,175 over the past week as the market moves the existing positions.
More telling is what hasn't happened. The last six scheduled runs (2026-07-17 through 07-24) each report 0 executed, 0 rejected. The signal has simply gone quiet — no fresh sentiment strong enough to trip a buy, and no held name negative enough to force a sell.
Backtest and validation
Here is where balance matters. The full backtest returned +0.19% over 451 days (final equity $10,018.55, CAGR ~0.1%), with a Sharpe of 0.74 and a genuinely tiny max drawdown of 0.05%. Turnover was modest at 36.5% and fees were negligible ($2). Those are calm, capital-preserving numbers.
But the trade count is the headline: just 2 trades in the entire backtest, and a reported win rate of 0% despite the flat-positive return. Two observations cannot support any conclusion about edge.
The walk-forward validation makes this explicit — and it did not pass. Of four folds, only fold 4 (2025-12-16 → 2026-05-29) traded at all, contributing all the activity: 2 trades, +0.19%, a Sharpe of 1.5 and a 0.05% drawdown. Folds 1–3 were completely dormant, so only 1 of 4 folds was positive. The probabilistic Sharpe (PSR 0.923) looks strong, but the deflated Sharpe (DSR 0.551), which penalizes the 6 trials run, is far more sober.
The verdict
News-sentiment behaves like a disciplined strategy that refuses to force trades — a virtue that keeps drawdowns near zero. The problem is that discipline shades into inactivity: three of four validation windows and six straight live runs produced nothing. Until the signal fires often enough to build a statistically meaningful record, treat the encouraging PSR and lone strong fold as promise, not proof. The gate is right to hold it back.