Thesis and Design
News-sentiment follows a straightforward signal: enter long positions when recent news flow for a stock turns positive, exit when sentiment flips negative. The strategy covers a 24-name universe of large-cap U.S. equities spanning technology, financials, healthcare, consumer staples, and energy — a deliberately diversified sandbox that reduces single-sector dependency.
The appeal is intuitive. News-driven momentum is one of the better-documented short-term anomalies in equity markets, and anchoring entries and exits to a clean binary signal keeps the logic legible and auditable.
Recent Activity
Over the past two weeks the strategy has been largely in a holding pattern. Daily scheduled runs on June 26, 29, 30, July 2, and July 3 all reported zero executions. The one exception was July 1, when the system added five shares of GOOGL at $360.40 — the most recent entry in a book that now holds GOOGL, UNH (4 shares @ $402.85), BAC (35 shares @ $55.93), MSFT (4 shares @ $428.23), and AAPL (6 shares @ $312.06). Portfolio equity sat at $9,961.80 as of the July 3 run against $990.06 in cash, implying most capital is deployed.
The infrequent firing is itself a data point. Either the sentiment pipeline is not finding sufficient conviction across the watchlist on most days, or entry filters are tighter than they appear in the thesis description. Either way, execution cadence is low.
Backtest and Validation Performance
The 451-day backtest produced a total return of +0.19% on a $10,000 starting book, reaching $10,018.55 in final equity. The headline Sharpe of 0.74 is modest but positive; max drawdown of 0.05% is exceptionally contained. Total fees of $2 and zero FX cost reflect the low churn — only 2 completed round-trip trades despite a reported portfolio turnover of 36.5%.
The win rate of 0 warrants a plain explanation: both backtest trades appear to be open positions with no recorded exits, so the closed-trade win rate cannot yet be computed. This is not a signal of consistent losses; it is simply a consequence of the strategy rarely selling.
Validation results are the most textured part of the picture:
| Metric | Value |
|---|---|
| Folds positive | 1 / 4 |
| OOS return | +0.19% |
| OOS Sharpe | 1.50 |
| PSR | 0.923 |
| DSR | 0.551 |
| Validation passed | No |
The out-of-sample Sharpe of 1.50 — drawn entirely from Fold 4 (December 2025 – May 2026) — is genuinely strong. The Probabilistic Sharpe Ratio of 0.923 says there is a 92% chance the true Sharpe is positive, which is encouraging. However, Folds 1 through 3 recorded zero trades each, meaning the strategy was statistically invisible for the first three quarters of the test window. The Deflated Sharpe Ratio drops to 0.551 once multiple trials are accounted for, and with only one active fold the validation framework correctly withheld a pass.
Strengths and Risks
Strengths: Minimal drawdown, sensible large-cap universe, clean signal logic, and a credibly high Sharpe in the one period where it actually traded.
Risks: The core concern is sparse activation. A strategy that does not trade is not generating alpha — it is simply sitting in cash or holding stale positions while the market moves. If the sentiment data pipeline misses signals during low-news regimes, the strategy may routinely underperform even a passive index. The DSR near 0.55 also warns against over-reading the Fold 4 result given the number of parameter trials explored.
The live book provides the cleanest forward test. How the open AAPL, MSFT, BAC, UNH, and GOOGL positions are ultimately closed — and whether the strategy resumes firing regularly — will determine whether Fold 4 was the start of a pattern or an anomaly.