The thesis
The news-sentiment strategy is simple to state: buy names with positive recent news sentiment, exit when sentiment turns negative. It trades a 24-stock 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, XOM, CVX, CAT). It is currently live in paper trading.
Recent activity
The live book made five buys between late May and early July 2026 — AAPL, MSFT, BAC, UNH, and GOOGL — and has since gone dormant. The last eight scheduled runs (July 31 through August 7) executed zero trades, with a single rejection on August 5. Portfolio value has drifted with the market rather than with any new signals, ranging from about $10,275 to $10,573 and closing the window near $10,483 total. Notably, cash sits at just $990 — the strategy is essentially fully invested, which means even if a fresh positive-sentiment signal appeared, there is little dry powder to act on.
Backtest and validation
Here is where the caution flags go up. The headline backtest covers 451 days and returns +0.19% (final equity $10,018.55, CAGR ~0.1%), with a Sharpe of 0.74 and a very shallow max drawdown of 0.05%. That drawdown control is genuinely attractive. But the return was generated by just two trades, and the reported win rate is 0.
Walk-forward validation makes the fragility explicit. Of four folds, only one was positive — folds 1 through 3 placed zero trades at all, and the entire result comes from fold 4 (Dec 2025–May 2026): +0.19% on two trades with a fold Sharpe of 1.5. The out-of-sample Sharpe of 1.5 looks strong in isolation, but it rests on that single active window.
The multiple-testing statistics tell the same story from two angles. The PSR of 0.923 is encouraging, yet the deflated Sharpe ratio (DSR) of 0.551 — which penalizes for the 6 trials run — is only modestly above a coin flip. The validation gate's verdict: failed.
The verdict
The strengths are real but narrow: tight drawdowns and a clean out-of-sample fold. The risks dominate. A two-trade sample cannot support any confident claim about edge; three of four folds never triggered, suggesting the sentiment signal fires too rarely to be a reliable engine. The near-zero cash balance compounds this by leaving the book unable to respond.
My read: treat news-sentiment as an unproven prototype, not a validated strategy. Before it earns more capital, it needs a signal that actually fires across regimes and a trade count large enough to distinguish skill from noise.