The thesis
PanicFade v4 bets that negative headline shocks get overreacted to and revert toward fundamentals. It buys names hit by sentiment below −0.4 that are already technically oversold (RSI14 < 35 and below their SMA20), and sizes into the fall rather than waiting for a bottom. Positions exit on reversion — RSI14 back above 50, or price reclaiming the SMA20 — while a hard stop on fresh lows under continued negativity treats persistent weakness as a genuine fundamental break rather than a panic to fade.
The logic is grounded in arXiv:2509.11970 (q-fin, 2025) on sentiment asymmetries and structural calibration, corroborated by stock-level investor-sentiment work in JRFM (2025) and the broader negative-media overreaction literature. The tradable universe is 24 US large caps across tech, financials, healthcare, staples and industrials.
Recent activity
The strategy is patient. Five of its six most recent scheduled runs logged zero executions; the live paper book has hovered around $9,800 (total $9,839.17 on 28 Oct, cash $6,977.94). The last fills came on 23 October: 1 share of CAT at $688.79 and 3 of GOOGL at $321.67. Earlier in the month it rotated actively — selling NKE, HD, MCD and UNH, while buying BAC, MCD and UNH back. That cadence fits the thesis: wait for qualifying panic, then act. Worth flagging, though, that the live book is roughly flat-to-slightly-soft over this window, a notable contrast to the backtest headline below.
Backtest and validation
Over 1,233 days the backtest returned 37.93% (final equity $13,793), a 6.79% CAGR, Sharpe 0.96, max drawdown 9.3%, and a 62.88% win rate across 530 trades. Validation passed: all four walk-forward folds were positive and improving — 1.6%, 3.88%, 10.14%, then 21.7% — with an out-of-sample Sharpe of 2.44 in the latest fold. The probabilistic Sharpe ratio is 0.985 and the deflated Sharpe ratio, after accounting for 27 trials, is 0.56.
Strengths
- Every fold positive, including fold 1 (Sep 2021–Dec 2022), which survived a hostile bear regime with only a 9.3% drawdown.
- Shallow drawdowns and a high win rate — the downside profile is well-behaved for a strategy that deliberately catches falling knives.
- DSR of 0.56 stays positive after deflating for multiple trials, so this is not an obvious data-mined artefact.
Risks
- Turnover of 5,807% is frantic. The $530 in modelled fees is a real drag on a small base, and slippage on panic names is not captured here.
- Modest risk-adjusted returns: a 6.79% CAGR and 0.96 Sharpe are unremarkable, and much of the headline leans on fold 4 — the most recent, highest-Sharpe window — which recency can flatter.
- The live-vs-backtest gap: a near-flat paper book against a +37.9% backtest is a small sample, but it deserves watching.
- Signal fragility: sentiment inputs are sensitive to data gaps and single-name headline risk. A "fundamental break" that keeps falling is the known failure mode, and the hard stop only partly contains it.
PanicFade v4 is a disciplined, honestly validated contrarian engine — but one whose live proof is still being written.