Thesis
The news-sentiment strategy runs a simple, legible idea: buy names showing positive recent news sentiment and exit when sentiment turns negative. It operates across a 24-name large-cap universe spanning tech (AAPL, MSFT, GOOGL, NVDA), financials (JPM, BAC, V, MA), healthcare (JNJ, UNH, PFE, ABBV), and consumer/industrial staples. The universe choice is sensible — liquid, well-covered stocks are exactly where a sentiment signal has enough news flow to fire cleanly.
Recent Activity
The live paper book currently sits at roughly $10,281 total with only $990.06 in cash, so it is effectively fully invested. Positions were built earlier in the summer through a cluster of executed buys — AAPL (6 @ $312.06), MSFT (4 @ $428.23), BAC (35 @ $55.93), UNH (4 @ $402.85), and GOOGL (5 @ $360.40) between late May and early July.
Since then, activity has stalled. Scheduled runs from Sept 11–18 report 0 executed and (mostly) 1 rejected order each day. Equity has drifted between about $10,210 and $10,377 with no new fills. The repeated single rejection alongside near-zero cash strongly suggests the strategy keeps surfacing a candidate it cannot fund — a sizing or liquidity constraint, not a lack of signal.
Backtest & Validation
Here the picture gets uncomfortable. The full backtest over 451 days returned just +0.19% (CAGR ~0.1%) — essentially flat — on a Sharpe of 0.74, a maximum drawdown of only 0.05%, and a mere 2 trades with a 0% win rate. Turnover was 36.5% and total fees just $2.
Walk-forward validation did not pass. Of four folds, only one was positive — and it carries the entire result: folds 1–3 produced zero trades and zero return, while fold 4 (Dec 2025–May 2026) delivered the whole +0.19% with a punchy out-of-sample Sharpe of 1.5 on two trades. PSR is high at 0.923, but the deflated Sharpe (DSR) is 0.551 across 6 trials — barely above a coin flip once you account for selection.
Strengths
- Risk discipline is excellent: a 0.05% max drawdown and low turnover mean the strategy is not bleeding capital.
- The one active fold looks good: a 1.5 OOS Sharpe is genuinely strong.
- Cheap and legible: minimal fees, no FX cost, and a thesis anyone can audit.
Risks
- The evidence base is tiny. Two trades cannot support a statistical conclusion; a 0% win rate on that sample is noise, not signal.
- Three of four folds never traded, so the signal fires far too rarely to trust across regimes.
- DSR near 0.5 flags real overfitting risk once multiple trials are counted.
- Stuck fully invested with a recurring rejected order suggests a plumbing issue throttling live execution.
Verdict
News-sentiment is well-behaved and conceptually sound, but its edge is unproven. It should stay in observation, not scale — the priority is getting the signal to fire more often and resolving the daily rejections before reading anything into performance.