The thesis
PanicFade v4 bets that markets overreact to bad news. Its rulebook is deliberately narrow: buy a name only when negative sentiment (below -0.4) coincides with a price-defined panic — RSI14 under 35 and trading below its 20-day moving average — then size into the drop. Exits are equally mechanical: take the reversion when RSI14 clears 50 or price reclaims the SMA20, and hard-stop on fresh lows under continued negativity, treating that as a genuine fundamental break rather than noise.
The approach is grounded in recent sentiment-feedback literature (arXiv:2509.11970, 2025, plus corroborating work in JRFM/MDPI), which documents asymmetric overreaction and reversion to negative media sentiment. It is a large-cap-only strategy — a 24-name universe spanning mega-cap tech, financials, healthcare, staples, and energy — which keeps liquidity high and idiosyncratic blow-up risk lower than a small-cap contrarian book would carry.
Validation: the strong point
The walk-forward evidence is the most encouraging part of the file. Across four sequential folds from late 2021 to August 2026, every fold was positive (4/4), with out-of-sample return of 21.7% and an out-of-sample Sharpe of 2.44 in the most recent window. The deflated Sharpe ratio (0.56) and probabilistic Sharpe ratio (0.985) are reported alongside 27 trials — meaning the headline numbers are already discounted for the number of configurations tried, which is exactly the honesty a contrarian strategy needs to survive scrutiny.
That said, read the folds carefully. Performance was strongly back-loaded: fold 1 returned just 1.6% (Sharpe 0.19), and the bulk of the edge appeared in folds 3 and 4. The full-sample CAGR is a modest 6.79% with a 0.96 Sharpe and a 9.3% max drawdown. This is a steady, low-drama profile — not a fireworks strategy — and its recent brilliance may partly reflect a favourable regime rather than a permanent step-change.
Recent activity: quiet by design
Live behaviour has been sparse, which is consistent with the strategy's selectivity. The last six scheduled runs (Aug 28–Sep 4) fired exactly one trade: a 10-share buy of WMT at $103.115 on Aug 28. Every run since executed nothing — no qualifying panics met the sentiment-and-oversold gate. Total account value has drifted in a tight band around $10,000–$10,045, with roughly $8,958 sitting in cash.
That idleness cuts both ways. On one hand, discipline is a feature: a fade strategy that refuses to force trades avoids the classic failure of catching falling knives on thin signals. On the other, a book that is ~90% cash is barely deployed, so live results so far tell us little. The backtest logged 530 trades over 1,233 days with 62.88% win rate; the live sample is a handful of runs and one fill.
The balance
Strengths: a coherent, literature-backed thesis; clean, all-positive walk-forward folds; and DSR/PSR figures that survive multiple-testing correction. Risks: heavy back-loading of returns, a genuinely modest full-sample CAGR, and a live track record too short and too cash-heavy to confirm anything yet. PanicFade v4 has earned its live slot on validation quality — now it needs enough panics to actually trade through.