Thesis
News-sentiment runs a simple, intuitive premise: buy names on positive recent news sentiment and exit when sentiment turns negative. It hunts across a 24-name large-cap universe spanning tech (AAPL, MSFT, GOOGL, NVDA), financials (JPM, BAC, V, MA), healthcare (JNJ, UNH, PFE, ABBV), and staples/industrials (PG, KO, WMT, COST, CAT, XOM). The strategy is currently live in paper trading.
Recent Activity
The live book has been quiet. The last five executed orders were all buys — 6 AAPL at $312.06 (May 31), 4 MSFT at $428.23 (Jun 3), 35 BAC at $55.93 (Jun 12), 4 UNH at $402.85 (Jun 18), and 5 GOOGL at $360.40 (Jul 1). Since then, every scheduled run from July 23 through July 30 reported 0 executed, 0 rejected. Cash sits at $990.06 with total equity drifting up from $9,879.36 (Jul 23) to $10,348.98 (Jul 30) — gains driven by mark-to-market on existing holdings, not new activity.
Notably, we've seen no exits yet. The book is nearly fully invested and holding, which means the "sell on negative sentiment" half of the thesis remains largely untested in the live tape.
Backtest & Validation
Here the picture turns sober. Over 451 backtested days the strategy returned just +0.19% (CAGR ~0.1%), on only 2 trades, with a 0% win rate, a Sharpe of 0.74, and a tightly contained 5% max drawdown. Turnover was low at ~36.5% and fees negligible ($2). The low drawdown is genuinely attractive, but it's the flip side of a signal that almost never trades.
Walk-forward validation did not pass. Of four folds, only one was positive — fold 4 (Dec 2025–May 2026) carried the entire result, posting +0.19% and a 1.5 Sharpe on 2 trades. Folds 1 through 3 recorded zero trades and zero return. In other words, three-quarters of the out-of-sample history produced no signal at all. The deflated Sharpe (DSR 0.551) and probabilistic Sharpe (PSR 0.923) look superficially decent, but they rest on a single active window across six trials — a thin foundation.
Strengths and Risks
Strengths: disciplined, low-turnover, and capital-preserving. A 5% max drawdown and near-zero fees mean the strategy rarely hurts itself, and the one active fold hinted at respectable risk-adjusted returns.
Risks: the signal fires far too infrequently to draw conclusions from. A 0% win rate on 2 trades, three empty validation folds, and a failed validation gate all point to a sample size too small to be meaningful. The all-buy, no-exit live posture also leaves the strategy's downside behavior unproven.
Verdict
News-sentiment is intriguing but unproven. Until the signal fires more often — and until we see it actually exit on deteriorating sentiment — this belongs firmly in the observation bucket, not the conviction bucket.