The Thesis
PanicFade v4 makes a specific, testable bet: markets overreact to bad news. When a large cap gets hit by a negative headline (sentiment below -0.4) while it is already technically oversold — RSI14 under 35 and trading below its 20-day moving average — the strategy buys and sizes into the panic. It looks for the overreaction to fade, exiting when momentum normalises (RSI14 back above 50 or price reclaiming the SMA20). Crucially, it does not assume every panic is noise: a fresh low on continued negativity trips a hard stop, treating that as a genuine fundamental break rather than a dip to buy.
The design is grounded in recent sentiment-asymmetry literature (arXiv:2509.11970 and corroborating work in JRFM), which is a healthier starting point than curve-fitting. The universe is deliberately conservative — 24 liquid names like AAPL, JPM, UNH, and WMT — which limits the falling-knife risk inherent to any contrarian approach.
Recent Activity
The live paper book has been quiet but active. Over the last week the strategy executed buys in UNH, NKE, and HD, and trimmed WMT, with a single fill on the 23 Sep run. That WMT round-trip — bought near $103 in August, sold at $107.48 — is exactly the reversion the thesis targets. Total portfolio value sits at $9,911, drifting modestly below the $10,000 mark across recent sessions. That is worth stating plainly: live, the strategy is currently a touch underwater, even as its history looks strong.
Validation and Performance
On paper the record is solid. The full backtest returns 37.93% over 1,233 days (CAGR 6.79%), a 62.88% win rate across 530 trades, and a contained 9.3% max drawdown. More importantly, the 4-fold walk-forward validation passed cleanly: all four folds were positive, out-of-sample return was 21.7% at a Sharpe of 2.44, and the probabilistic Sharpe ratio of 0.985 is reassuring given 27 trials tested.
The Risks
Balance demands scepticism. The full-sample Sharpe is only 0.96 — respectable, not spectacular — and the deflated Sharpe of 0.56 is honest about the multiple-testing haircut. The out-of-sample glow leans heavily on fold 4 (21.7% return, Sharpe 2.44), while folds 1 and 2 were nearly flat (1.6% and 3.88%). That is recency-weighted comfort. Turnover of 5,807% is enormous; the strategy lives or dies on execution costs and slippage that a paper account understates. And the CAGR, once annualised, is a middling 6.79% — reversion strategies pay for their smoothness in foregone upside.
Verdict
PanicFade v4 is a well-constructed, literature-backed contrarian engine with disciplined stops and clean validation. But a soft full-sample Sharpe, punishing turnover, and a live book sitting just under par mean it should be watched, not celebrated — the coming weeks of live reversion will tell us more than any backtest.