The thesis
PanicFade v4 is a contrarian mean-reversion strategy built on a simple behavioural claim: negative headline shocks are overreacted to and tend to revert toward fundamentals. It buys names hit by sentiment below -0.4 that are also already oversold — RSI14 under 35 and trading below their SMA20 — and sizes into the panic. Exits come on reversion (RSI14 back above 50, or price reclaiming the SMA20), with a hard stop on fresh lows under continued negativity, treated as a genuine fundamental break rather than noise.
The design leans on recent sentiment-feedback literature (arXiv:2509.11970, 2025, and corroborating retail-sentiment work) on asymmetric overreaction and reversion. The universe is deliberately conservative: 24 mega-cap and large-cap names across tech, financials, healthcare, staples and energy — liquid stocks where a panic is more likely to be sentiment than solvency.
Validation — the strong part
This is where PanicFade v4 earns attention. Across a 4-fold walk-forward spanning September 2021 to August 2026, every fold was positive. More importantly, out-of-sample performance improved through time rather than decaying: fold returns ran 1.6% → 3.9% → 10.1% → 21.7%, with the final fold posting a 2.44 Sharpe against a 2.6% max drawdown. The aggregate out-of-sample return of 21.7% is a meaningful fraction of the full-sample 37.9%, which argues against heavy in-sample overfitting.
The deflated statistics back this up. A Probabilistic Sharpe Ratio of 0.985 and a Deflated Sharpe Ratio of 0.56 — computed across 27 trials — mean the edge survives correction for multiple testing. Passing that gate is the whole point of the validation layer, and PanicFade v4 clears it cleanly.
The full-sample picture
The headline backtest is more sober. Over 1,233 days the strategy returned 37.9% (final equity $13,793 on $10k), a 6.79% CAGR, with a 0.96 Sharpe, 9.3% max drawdown, and a 62.9% win rate over 530 trades. That is a respectable but not spectacular risk-adjusted profile, and turnover is high at ~5,807% — this is an active trader, though fees stayed modest at $530 total.
Recent activity: doing nothing, on purpose
The live tape is quiet to the point of dormant. Every scheduled run from August 13 through August 20 reports 0 executed, 0 rejected, cash flat at $10,000. No positions, no rejections — the entry conditions simply aren't triggering.
This is a feature, not a bug: a panic-fade strategy is supposed to be idle when nothing is panicking. But it's also the central risk to keep in view. PanicFade v4 only earns when markets hand it oversold, negatively-shocked names; in calm regimes it contributes nothing, and its returns are concentrated into rare, clustered opportunities. A week of zeros tells us markets are presently benign — not that the edge is broken.
Verdict
Strong, honestly-validated contrarian logic with a genuinely encouraging out-of-sample trend. The caveats are structural: modest full-sample Sharpe, high turnover, and long stretches of inactivity that demand patience. Worth watching for how it behaves in the next real drawdown — that is the environment it was built for.