The thesis
News-sentiment runs a simple, intuitive idea: buy names on positive recent news sentiment and exit when sentiment turns negative. It operates over a 24-stock universe of US 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 (XOM, CVX, CAT, HON). It is now flagged live.
The appeal is obvious — sentiment is a real, tradeable signal and the rule is easy to reason about. The catch, as the data shows, is that the signal fires rarely.
Recent activity
The recent trade log is buy-only: NVDA (9 @ $213.49), V (4 @ $410.29) and PFE (66 @ $29.92) on 29 Oct, KO (21 @ $95.13) on 3 Nov, and AAPL (5 @ $348.52) on 10 Nov. No exits have printed, which is why the headline win rate reads 0 — nothing has closed round-trip yet.
Scheduled runs tell a quieter story. Across 4–11 Nov, most days executed zero trades, and portfolio total value drifted down from $9,987.62 (4 Nov) to $9,659 (11 Nov) — a soft ~3% slide driven by held positions, not fresh risk-taking. Cash sat idle near $2,387 for a week before the AAPL buy drew it down to $626.
Backtest and validation
The backtest covers 451 days and ends at $10,018.55 — a +0.19% total return, roughly 0.1% CAGR. Sharpe is 0.74, max drawdown a slim 0.05%, turnover 36.53%, with just two trades and $2 in fees. That is a tiny, benign result: the strategy mostly stayed flat.
Validation did not pass, and the fold detail explains why. Of four walk-forward folds, only fold 4 (16 Dec 2025 – 29 May 2026) traded at all — 2 trades, +0.19%, Sharpe 1.5. Folds 1–3 produced zero trades and zero return. So the entire backtest performance comes from a single window; the other three-quarters of the test period generated no signal. PSR is a flattering 0.923, but the deflated Sharpe ratio — which penalises the 6 trials run — falls to 0.551, barely clearing the 0.5 line.
Verdict and risks
Strengths: the thesis is sound, drawdown is negligible, fees are trivial, and the one window that did trade showed a healthy OOS Sharpe of 1.5.
Risks, and they dominate: this is an evidence-starved strategy. Two trades across 451 backtest days, three dead folds, and a failed validation mean we have almost no statistical basis to trust it — the positive result could be a single lucky window. The buy-only live record with no exits also means the "exit on negative sentiment" half of the thesis is untested in practice.
Going live on a strategy that failed validation is a deliberate bet that the signal is real but rare. Fair enough as an experiment — but it should be watched on a short leash, judged on whether exits actually trigger, and sized small until the trade count grows enough to say anything with confidence.