RSI-Bollinger mean reversion cleared gates with 90% win rate

We tested a classic mean-reversion setup that buys when price touches the lower Bollinger Band (20-period) and RSI-14 drops below 30, filtered to low-ADX regimes (ADX below 50) to avoid strong trends. It passed all six gates with a 0.769% average gain per trade and a 90% win rate across ten coins on 15-minute bars. The deflated Sharpe ratio (a measure that accounts for the number of tests run to find this strategy) came in at 43.18, with a 0.0% probability of being a fluke after multi-test correction.

The strategy exits when net profit hits its target. Training and validation results were identical, which is clean but also means we have exactly one out-of-sample confirmation. The backtest covered BTC, ETH, SOL, XRP, DOGE, BNB, ADA, AVAX, LINK, and LTC.

Here is where skepticism belongs. Despite clearing every hurdle, this strategy has generated zero trades in live paper trading since graduation. Zero. That could mean the market regime shifted, the coins stopped meeting the entry conditions, or the backtest window captured a narrow set of conditions that are not repeating. A 90% win rate in validation is striking, but without live trades we have no evidence the edge persists in real time.

The low-ADX filter is supposed to keep the strategy out of strong trends where mean reversion fails, but it also means this setup only fires in ranging, choppy conditions. If the coins in the test universe have been trending or volatility collapsed, the strategy sits idle. The 15-minute timeframe adds another layer of specificity.

We are not killing this strategy yet because the backtest results are strong and the logic is sound. But we are watching. A strategy that never trades live is not useful, no matter how well it backtested. If you want to test your own mean-reversion idea with different entry thresholds or regime filters, head to /prove and run it through the same six-gate gauntlet we use in the lab.