The thesis
The news-sentiment strategy runs a deliberately legible rule: buy names showing 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 (PG, KO, WMT, COST), and industrials and energy (CAT, HON, XOM, CVX). The attraction is interpretability — every trade should trace back to a readable catalyst rather than an opaque factor.
Recent activity
The strategy is live, with executed buys running back to late May 2026 (AAPL, MSFT, BAC, UNH, GOOGL). The most telling stretch is the last week. The scheduled run on 2026-09-23 was the first in over a week to execute anything: it sold 35 shares of BAC at $55.96 and bought 4 shares of NVDA at $225.34, lifting cash from $990.06 to $2,036.71 against a portfolio total of $10,143.16.
That BAC round trip is instructive. The position opened on 2026-06-12 at $55.93 and closed three months later at $55.96 — essentially flat. Meanwhile, the runs from 09-16 through 09-22 executed nothing and rejected every candidate. With the book near-fully invested at $990.06 cash, new signals simply had nowhere to go. The engine was generating ideas it could not fund.
Backtest and validation
The numbers demand caution. Over a 451-day backtest the strategy returned just 0.19% (CAGR 0.1%), on a Sharpe of 0.74, with a shallow 0.05% maximum drawdown — but on only 2 trades and a 0% win rate. Fees were negligible ($2, no FX cost).
Validation did not pass, and it is worth understanding why. Across four walk-forward folds, only one — fold 4, spanning 2025-12-16 to 2026-05-29 — produced any trades at all. Those are the same 2 trades, the same 0.19% return, and a respectable fold Sharpe of 1.5. Folds 1 through 3 generated zero signals over roughly 16 months of history. So the entire track record rests on a single fold and a two-trade sample. The probabilistic Sharpe (0.923) reads well, but the deflated Sharpe (0.551), which penalises for the 6 trials run, is only marginal.
Verdict
The strengths are genuine: a transparent thesis, tiny drawdown, near-zero costs, and one out-of-sample fold that behaved well. The risks are larger. Three of four folds fired no trades, the win rate is zero, and a Sharpe built on two trades is closer to anecdote than evidence. The live cash constraint compounds this by throttling the very signals the strategy is supposed to act on. Until the rule demonstrates it can generate — and fund — a meaningful trade count, treat news-sentiment as a promising sketch, not a proven edge.