Thesis
news-sentiment operates on a straightforward premise: buy when recent news coverage for a stock turns positive, exit when it sours. The strategy scans a 24-name universe of large-cap US equities spanning technology, financials, healthcare, consumer staples, and energy — names liquid enough that sentiment moves tend to precede measurable price reactions rather than lag them.
What the Backtest Says
Over a 451-day backtest ending in late May 2026, the strategy produced a total return of 0.19% with a CAGR of 0.1%, a Sharpe ratio of 0.74, and a maximum drawdown of just 5%. Those headline numbers look tame, but the critical context is that only two trades were executed across the entire period. With no completed round-trips, the reported win rate of 0% reflects an open book rather than a losing one — all positions are still held.
The low-trade count is both the strategy's strongest and most problematic characteristic. On the one hand, the 5% max drawdown is genuinely controlled. On the other, a two-trade sample offers almost no statistical surface to evaluate edge.
Walk-Forward Validation: A Clear Flag
The four-fold walk-forward validation tells a sharper story. Folds 1 through 3 — covering August 2024 to mid-December 2025 — each logged zero trades and zero return. Only Fold 4 (December 2025 – May 2026) generated any activity, posting the full 0.19% return with a Sharpe of 1.5 on two trades.
The result: validation did not pass. With only one of four folds positive and three completely idle, the Deflated Sharpe Ratio (DSR) lands at 0.551 — below the conventional 0.65 threshold used to correct for multiple-testing bias across the six trials run. The Probabilistic Sharpe Ratio (PSR) of 0.923 is more encouraging, but PSR assumes the observed sample is representative; here, it largely isn't.
Recent Live Activity
Live runs from late June into early July 2026 mirror the backtest pattern: the agent has executed on only one of its seven most recent scheduled runs, a 5-share buy of GOOGL at $360.40 on July 1. Prior to that, the last buys were UNH (4 shares, June 18), BAC (35 shares, June 12), MSFT (4 shares, June 3), and AAPL (6 shares, May 31) — a burst of activity in late May and June that appears to have cooled.
Portfolio equity has moved between roughly $9,680 and $10,006 over the past week, driven entirely by mark-to-market on held positions rather than new trades.
Strengths and Risks
Strengths: The strategy's selectivity — only entering when sentiment clears a clear positive bar — has kept drawdown minimal. When it does fire, Fold 4's 1.5 Sharpe suggests the signal may have genuine short-term predictive value in favorable market regimes.
Risks: The core risk is trade starvation. A strategy that generates two signals in 15 months cannot be evaluated with any statistical confidence, and the DSR failing validation is a formal acknowledgment of that. There is also regime dependence: all meaningful activity is concentrated in one fold, raising the possibility that the signal works only in specific volatility or news-flow environments.
Bottom Line
news-sentiment is not broken — it is underexercised. The immediate priority should be diagnosing why signal generation is near-zero across the majority of folds: whether that reflects thresholds that are too conservative, data coverage gaps, or genuine sparsity in actionable sentiment. Until trade volume scales, performance conclusions remain premature.