The thesis
PanicFade v4 is a contrarian mean-reversion strategy: it buys oversold large caps caught in a negativity spiral and waits for the rebound. The textbook signal pairs a sentiment shock (score below −0.4) with technical capitulation — RSI14 under 35 and price below its 20-day average — then sizes into the fall, exiting when RSI reclaims 50 or price retakes the SMA20, with a hard stop on fresh lows read as a genuine fundamental break. It leans on recent q-fin work (arXiv:2509.11970) on asymmetric overreaction to negative headlines.
One honest caveat up front: the development history shows the sentiment gate was structurally unreachable in backtest — historical per-symbol news coverage was too sparse, so candidates defaulted to neutral and no trade fired. v4 therefore runs on price-defined panic, with sentiment as an amplifier rather than a gate. The validated engine is closer to a pure RSI/SMA reversion system than the sentiment framing implies.
Recent activity
Trading resumed from a clean $10,000 base on 29 October (zero trades that day), then three executions on both 30 Oct and 2 Nov. The tape shows fast rotation: GOOGL (5 sh) and CAT (2 sh) were bought on 30 Oct and sold two sessions later on 2 Nov, with proceeds and fresh cash rotated into NVDA and BAC. That fits a short-horizon reversion mandate — though it also means paying the spread repeatedly.
So far the live book sits at $9,903.68, fractionally underwater versus the reset. Three sessions is far too little to judge; it is simply the honest starting point.
Backtest and validation
On paper the record is strong. Over 1,233 days the strategy returned 37.93% (final equity $13,793), with a 62.9% win rate across 530 trades and a contained 9.3% max drawdown. Validation passed cleanly: four walk-forward folds, all positive, out-of-sample return of 21.7%, and a Reviewer sign-off at risk 0.35.
The caveats matter as much as the headline. Annualised, 37.93% is a 6.79% CAGR and a Sharpe of 0.96 — respectable, not spectacular. Fold returns climb almost too neatly (1.6% → 3.9% → 10.1% → 21.7%), with the eye-catching 2.44 Sharpe concentrated in the most recent, most favourable window. And while the Probabilistic Sharpe is a near-certain 0.985, the Deflated Sharpe — which penalises the 27 configurations tried — is a far soberer 0.56. Turnover near 5,800% also makes real-world fee and slippage assumptions load-bearing; the backtest models only a flat $530 in fees and zero FX cost.
Verdict
A disciplined, well-reviewed reversion engine with a clean validation record — but one whose edge is modest once deflated, whose best numbers sit in the friendliest fold, and whose live track record is three days old. Worth watching; too early to trust.