RSI + SMA 200 + ADX Range Finder: 0.85% Per Trade, 87% Wins

We tested a mean-reversion strategy that buys when RSI(14) drops below 35 while price stays above the 200-period moving average, but only when ADX is below 30 to avoid strong trends. It passed all six gates with 0.846% average profit per trade and an 87.5% win rate in both walk-forward training and validation. The deflated Sharpe ratio came in at 4.76, which would normally suggest robust statistical significance.

The strategy hunts for oversold dips in ranging markets. The RSI threshold catches short-term weakness, the SMA 200 filter keeps us on the right side of long-term structure, and the ADX regime filter is the key: it only enters when ADX is below 30, meaning we sit out strong directional moves where mean-reversion typically gets crushed. The exit uses a net profit target, locking in gains when the bounce materializes.

Here's where you should stay skeptical. The backtest shows a 100.0% probability of being a fluke after multi-test correction, which directly contradicts the strong deflated Sharpe ratio. This is a known data artifact we're investigating, but it means the statistical confidence metric is unreliable for this gene. More importantly, this strategy has logged zero forward trades in live paper testing. Zero. That means we have no out-of-sample evidence yet that it fires in current market conditions. The backtest covered BTC, ETH, SOL, and XRP on 15-minute bars, so the opportunity set is narrow. If these coins shift into sustained trends or volatility compresses, the entry conditions may simply stop triggering.

The 87.5% win rate is appealing, but it comes with the usual mean-reversion trade-off: many small wins and occasional large losses when the bounce doesn't come. The regime filter should reduce those blowups, but we won't know until we see forward trades. This one graduated on strong backtest metrics, but it's still waiting for its first live signal.

Check the full survivor registry at /survivors to compare this against other graduates, or head to /prove to test your own regime-filtered ideas with our fee-aware engine.