The thesis
News-sentiment is one of our built-in, hand-written strategies. The rule is deliberately simple: buy names on positive recent news sentiment, and exit when sentiment turns 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 and consumer (PG, KO, WMT, COST, MCD, NKE, HD) and energy/industrials (XOM, CVX, CAT, HON, DIS). The reviewer approved it with a risk score of 0 — a vote of confidence on safety, not on returns.
Recent activity
The strategy went live and fired on 2026-10-29, opening three positions: 66 shares of PFE at $29.92, 9 of NVDA at $213.49, and 4 of V at $410.29. Since then it has gone quiet — the scheduled runs on 2026-10-30 and 2026-11-02 both executed zero trades. Total paper equity sat at $9,987.10 on Nov 2, fractionally below the $10,000 start, with $4,405.24 in cash. In other words, it is mostly in observation mode, waiting for sentiment signals that have not yet arrived.
Backtest and validation
Here is where candour matters. Over 451 days the backtest produced just 2 trades, a total return of 0.19% (final equity $10,018.55), and a CAGR of 0.1%. Sharpe was 0.74, max drawdown a trivial 0.05%, turnover 36.53%, and total fees $2. The win rate is reported as 0.
Cross-validation is more revealing — and it did not pass. Of 4 folds, only 1 was positive. The first three folds (Aug 2024 through Dec 2025) generated no trades at all; the strategy simply never triggered. All of the measured performance comes from fold 4 (Dec 2025–May 2026), which returned 0.19% on a Sharpe of 1.5 off 2 trades. The headline PSR of 0.923 looks reassuring, but the deflated Sharpe ratio (DSR) of 0.551 — after adjusting for 6 trials — is the more honest number, and it is modest.
Strengths and risks
Strengths: capital preservation is excellent. Drawdown is negligible, turnover and fees are low, and the strategy doesn't force trades when it has no conviction. That discipline is genuinely valuable in a portfolio of noisier agents.
Risks: the evidence base is thin to the point of fragility. Two trades across 451 days, with all performance concentrated in a single five-month window, is a textbook small-sample problem — the OOS Sharpe of 1.5 rests on almost nothing. Three silent folds suggest the signal rarely fires, which makes robustness impossible to judge. A failed validation plus a 0% win rate means we cannot yet claim an edge.
Verdict: news-sentiment earns its live slot on safety, not skill. It needs more signal activations and a clean validation pass before we treat its numbers as anything other than provisional.