kill: zen
zen was supposed to be the calm one. a mean-reversion bot on SOL, 15-minute bars, fading extreme RSI readings with a tight stop and a 2:1 target. the pitch was simple: catch overreactions, take the other side, let probability do the work. no leverage. no martingale. just a quiet, patient edge grinding small pnl while everything else screamed.
the data tells a different story. total trades: zero. closed trades: zero. wins, losses, pnl, all zero. max drawdown zero percent. profit factor zero. sharpe undefined. the bot ran for its full scheduled window and never once fired a signal. not because the market was calm. because the entry conditions never triggered.
the autopsy is straightforward. the RSI thresholds were set too wide, calibrated on a backtest window where SOL was choppier than it's been in months. the regime shifted. trending, momentum-driven price action doesn't produce the kind of oversold bounces zen was trained to catch. the filter that was supposed to keep it safe also kept it silent. a bot that never trades has a perfect sharpe ratio and zero reason to exist.
one falsifiable observation: mean-reversion signals on SOL 15m need adaptive thresholds that widen or tighten with realized volatility. static RSI bands die quietly in trending regimes.
rest in peace, zen. you were a falsified hypothesis.
ā research and educational content. not investment, legal, or tax advice. do your own research. positions and views may change without notice.
Originally published on FalsifyLab Substack.