Thesis
The news-sentiment strategy runs a simple, intuitive premise: buy names showing positive recent news sentiment, and exit when sentiment turns negative. It operates over a 24-symbol universe of U.S. large caps spanning tech (AAPL, MSFT, GOOGL, NVDA), financials (JPM, BAC, V, MA), healthcare (JNJ, UNH, PFE, ABBV), staples (PG, KO, WMT, COST), and industrials/energy (CAT, HON, XOM, CVX). It is currently classified as live.
The appeal is obvious — sentiment is fast-moving and, in theory, front-runs price. The hard part is turning that signal into enough clean, repeatable trades to trust it. On that count, the evidence is thin.
Recent Activity
Live behaviour over the past week has been almost entirely passive. Scheduled runs from September 2 through September 9 executed zero new trades, with one or two orders rejected on most days. Cash has sat flat at $990.06 — roughly 90% of the book is already invested — which is the likely reason fresh orders keep bouncing. Portfolio total drifted from $10,525 down to $10,277 across the week, reflecting mark-to-market on existing positions rather than any active decision.
Those positions came from a burst of executed buys between late May and early July 2026: AAPL (6 @ $312.06), MSFT (4 @ $428.23), BAC (35 @ $55.93), UNH (4 @ $402.85), and GOOGL (5 @ $360.40). No sells appear in the recent record — the strategy is holding, not rotating.
Backtest and Validation
Here the picture gets sobering. The backtest covers 451 days and ends at a final equity of $10,018.55 — a 0.19% total return and a 0.1% CAGR. Sharpe is 0.74, max drawdown a negligible 0.05%, turnover 36.53%, with fees of just $2 and no FX cost. Critically, the entire run rests on 2 trades and a 0% win rate, meaning the modest gain is unrealized rather than booked from closed winners.
Validation did not pass. Across 4 walk-forward folds, only one was positive: fold 4 (2025-12-16 → 2026-05-29) produced the full 0.19% return and a strong 1.5 out-of-sample Sharpe on its 2 trades. Folds 1 through 3 generated no trades at all. So every number this strategy can point to comes from a single five-month window.
The statistics are split. The probabilistic Sharpe ratio (PSR 0.923) looks encouraging, but after deflating for 6 trials the deflated Sharpe (DSR 0.551) sits barely above the 0.5 coin-flip line — a warning that the apparent edge may not survive multiple-testing correction.
Verdict
news-sentiment is a clean idea with a coherent universe and disciplined, low-cost execution. But two trades in one fold is not a track record — it is a data point. With three empty folds, a failed validation gate, and a DSR hovering at 0.55, the responsible read is unproven. I'd want to see the signal fire consistently across multiple regimes, and cash freed up so rejected orders stop masking whether the thesis actually works, before treating this as anything more than a live experiment.