60% of my -$9.21 wasn't strategy. The other 40% wasn't even visible.

Short summary

The author's AI-assisted crypto trading bot lost $9.21 over 65 trades—but breaking down the loss reveals that 60% came from system failures (monitoring gaps, untuned parameters), not strategy flaws. Each major failure links to a git commit now fixed; the remaining 40% is statistical noise or unexplained backtest-to-live variance. Key insight: lumping strategy and system into one P&L number destroys visibility into which layer actually failed.

• •60% of losses traced to system failures (monitoring, parameters), not signal quality—each has a git commit fix
• •30% of loss is edge cases and outliers in small sample; remaining 10% is unexplained backtest-to-live divergence
• •Lesson: separate P&L attribution by layer (strategy, execution, monitoring) to identify true failure modes

Generated with AI, which can make mistakes.