Thesis and Universe
The news-sentiment strategy operates on a straightforward premise: buy when recent news sentiment across a holding turns positive, exit when it turns negative. Its universe spans 24 large-cap U.S. names — tech giants (AAPL, MSFT, NVDA), financials (JPM, V, MA), healthcare (JNJ, UNH), and consumer/energy stalwarts (WMT, XOM, MCD). Coverage is broad by design; the logic being that sentiment signals are more reliable on heavily-followed names with dense news flow.
First Live Steps
The agent deployed on May 31 with a $10,000 allocation and immediately found a signal: 6 shares of AAPL purchased at $312.06, leaving $8,127.64 in cash. That single position represents roughly 19% of the portfolio — a measured, not aggressive, opening move. The reviewer approved deployment at risk level zero, consistent with its classification as a built-in, hand-written strategy with no learned parameters.
Backtest Snapshot
Over 451 backtest days the strategy produced two total trades, a 0.19% total return, and a maximum drawdown of just 0.05%. The Sharpe ratio sits at 0.74 — positive but not compelling in isolation. Annual turnover of 36.53% is moderate for a sentiment strategy, though with only two trades it reflects how rarely the signal actually fires rather than active rotation.
The near-zero win rate figure deserves a note: with only two completed backtest trades and the strategy still in its first live session, there simply isn't enough trade history to compute a meaningful win rate. That's a data-scarcity issue, not evidence of consistent losses.
Validation: Where the Story Gets Complicated
The four-fold walk-forward validation is where news-sentiment shows its weak spots. The strategy did not pass the validation gate.
| Fold | Period | Return | Sharpe | Trades |
|---|---|---|---|---|
| 1 | Aug 2024 – Jan 2025 | 0% | 0 | 0 |
| 2 | Jan 2025 – Jul 2025 | 0% | 0 | 0 |
| 3 | Jul 2025 – Dec 2025 | 0% | 0 | 0 |
| 4 | Dec 2025 – May 2026 | 0.19% | 1.5 | 2 |
Three of four folds produced zero trades and zero return. All of the strategy's backtest activity is concentrated in the most recent five-month window. That single active fold carries an impressive out-of-sample Sharpe of 1.5, but it also means we're drawing conclusions from a sample of two trades across one market regime.
The Probabilistic Sharpe Ratio (PSR) of 0.923 is encouraging — it suggests a reasonable probability the Sharpe is above zero. But the Deflated Sharpe Ratio (DSR) of 0.551 tells a more cautious story: after adjusting for the fact that this strategy was tested across six trials, the odds that the observed Sharpe reflects genuine edge rather than selection bias drop significantly. A DSR below the conventional 0.95 threshold is a yellow flag, not a red one, but it warrants attention.
Strengths and Risks
Strengths: The drawdown profile is exceptionally clean — 0.05% max drawdown over the full backtest period. When the strategy is active, its risk-adjusted performance in fold 4 is genuinely strong. The low-friction cost structure (just $2 in total fees, zero FX cost) means even modest signals aren't eaten by execution overhead.
Risks: Infrequent firing is the central concern. A strategy that generates two signals in 15 months will be slow to validate or invalidate in live trading. If market conditions shift the news-sentiment correlation, months could pass before there's enough signal to detect the change. The DSR also flags multiple-testing risk — the apparent edge may owe something to the search process that produced the current parameterization.
Outlook
News-sentiment is a logical, intuitive strategy on a well-covered universe. The architecture is sound; the question is whether the signal fires reliably enough to accumulate meaningful performance history. The next 90 days of live trading will matter more than the entire backtest: either the strategy starts generating frequent, profitable signals — or the near-dormancy of folds 1–3 reasserts itself, and the team will need to revisit the sentiment threshold or lookback window.