The thesis
News-sentiment runs on a premise you can explain in a sentence: buy names carrying positive recent news sentiment, and exit when that sentiment turns negative. It fishes in a pool of 24 US mega-caps spanning tech (AAPL, MSFT, GOOGL, NVDA), financials (JPM, BAC, V, MA), healthcare (JNJ, UNH, PFE, ABBV), staples (PG, KO, WMT, COST), plus energy and industrials. The logic is legible and the universe is liquid — a sensible design for a sentiment overlay.
Recent activity
The strategy is live. Its executed positions were opened in a single early-summer window: AAPL (6 @ $312.06) and MSFT (4 @ $428.23) in late May/early June, BAC (35 @ $55.93) and UNH (4 @ $402.85) mid-June, and GOOGL (5 @ $360.40) on July 1. Since that GOOGL buy, nothing has executed. The six scheduled runs from Sep 7–14 each closed with 0 executed and 1–2 candidates rejected. Cash has sat pinned at $990.06 while total NAV drifted between roughly $10,277 and $10,385 — movement driven by the market marking existing holdings, not by new signal.
Backtest and validation
Over a 451-day backtest the strategy returned +0.19% total (~0.1% CAGR), with a Sharpe of 0.74, a max drawdown of just 0.05%, turnover of 36.53%, and only 2 trades at $2 total fees and no FX cost. Win rate is reported as 0%.
The validation tells the real story, and it did not pass. Across four walk-forward folds, only one was active: folds 1–3 (Aug 2024 through Dec 2025) generated zero trades and zero return. The entire +0.19% and a flattering fold Sharpe of 1.5 come from fold 4 (Dec 2025–May 2026) — on two trades. The probabilistic Sharpe ratio of 0.923 looks encouraging, but the deflated Sharpe of 0.551 (after adjusting for 6 trials) is the more honest read on such a thin sample.
Strengths and risks
Strengths: a clear, auditable thesis; near-zero drawdown; negligible costs; and a rejection discipline that declines low-conviction entries rather than forcing turnover.
Risks: the signal almost never fires. Two trades in 451 days and three dead folds mean there is essentially no evidence base — the high fold Sharpe is noise on n=2, the 0% win rate is unresolved, and formal validation failed. This is a strategy with good manners and an unproven edge. It earns its place as a small live probe, but it should not be scaled until trade count and fold coverage rise enough to say anything statistically meaningful.