The Thesis
News-sentiment is a simple, defensible idea: buy names when recent news skews positive, and exit when the tone turns negative. It runs on a 24-name large-cap universe spanning tech (AAPL, MSFT, GOOGL, NVDA), financials (JPM, BAC, V, MA), healthcare (JNJ, UNH, PFE, ABBV), and staples and industrials (PG, KO, WMT, COST, CAT, HON). The premise is sound — sentiment does move price on short horizons — but everything downstream depends on the signal actually firing often enough to matter. That is exactly where this strategy struggles.
Recent Activity
The live paper book is quietly invested rather than active. Executed buys cluster in late May through early July 2026: AAPL (6 @ $312.06), MSFT (4 @ $428.23), BAC (35 @ $55.93), UNH (4 @ $402.85), and GOOGL (5 @ $360.40). Since that GOOGL purchase on July 1, the scheduled daily runs have produced no new executions — the log shows a steady string of "0 executed" days, with a single rejected order on August 5. Cash sits at $990.06, meaning the book is almost fully deployed, and total equity has drifted between roughly $10,275 and $10,573 over the past week, resting near $10,466. That is a modest gain, but it reflects buy-and-hold drift in held positions more than any active sentiment edge.
Backtest and Validation
Here the picture turns cautionary. Over 451 backtested days the strategy returned just 0.19% (final equity $10,018.55, CAGR ~0.1%) on only two trades, with a reported win rate of zero and a Sharpe of 0.74. Drawdown was negligible at 0.05% and fees were trivial — but that is the signature of a system that barely does anything, not one that is well-controlled.
Walk-forward validation did not pass, and the fold detail explains why. Of four folds, three registered zero trades and zero return; only fold 4 (Dec 2025–May 2026) traded at all, delivering the entire 0.19% and a headline out-of-sample Sharpe of 1.5. So the encouraging OOS numbers rest on a single window and two fills. The probabilistic Sharpe ratio looks strong at 0.923, but the deflated Sharpe — which penalizes for the six trials run — falls to 0.551, and only one of four folds was positive. That is a thin, fragile track record dressed in flattering summary stats.
Verdict
Strengths: an intuitive thesis, tight drawdowns, and a live book that is slightly in the green. Risks dominate, though — the signal fires far too rarely to trust, the sample is statistically meaningless, and validation failed on breadth of folds. Before this earns real conviction, the sentiment trigger needs recalibration so it engages across market regimes. Right now it is a good idea waiting for enough evidence to prove it works.