RSI-7 <25 with ADX filter: restored, still watching

We tested a countertrend strategy that buys when RSI-7 drops below 25, filtered to low-ADX (below 30) environments to avoid strong trends, and exits on a 0.5% net profit target. It ran 29 live paper trades with a 0.259% average P&L per trade and a 79.3% win rate. Then we killed it because the backtest validation told us we should have.

The backtest numbers were harsh. Walk-forward validation showed a negative 0.254% average P&L per trade across 1,136 out-of-sample trades, despite an 86.6% win rate. The deflated Sharpe ratio (a measure that penalizes you for running many tests and picking the best looking one) came in at -1.85, and the probability of being a fluke hit 100.0% after multi-test correction. Under our strict audit gates, that is an automatic kill.

We soft-killed it in early May 2026, then restored it the same day when regime analysis showed it had profitable fit in specific market conditions. The strategy now sits in observation status, meaning it continues to accumulate forward-test data but is not available for general selection until we finish integrating regime-aware gating into the forge.

The core tension is clear: the backtest sampled 1,136 trades across BTC, ETH, XRP, and BNB on 5-minute bars and found a consistent small drag. The forward test sampled 29 trades in a narrow time window and showed a small edge. Twenty-nine trades is not enough to override 1,136, but it is enough to justify watching, especially if those 29 fell into a regime the strategy was built for.

This is not a survivor you can trade yet. It is a case study in why we do not let forward results alone overturn backtest discipline, and why regime fit matters. If you are building mean-reversion ideas that rely on low-ADX windows, this gene will stay on your watchlist. If you want strategies you can run today, check the full registry at stratproof.com/survivors.