Debug Log #3: I Had Claude Code Backtest a Trading Strategy for a Week, and It Found Three Bugs In Its Own Simulation

A developer recently shared their experience using Claude Code to build a backtesting engine for a leveraged ETF trading strategy. The project highlighted the dangers of 'silent' bugs—errors that do not trigger exceptions but produce incorrect, plausible-looking data. The author identified three distinct issues: a misaligned trigger price formula, a flawed reactivation timer that functioned differently than specified, and a failure to account for historical stock splits. These bugs were only discovered after manual verification of individual data points against the original requirements. The author emphasizes that when building automated systems, developers cannot rely solely on aggregate outputs or the absence of stack traces. Instead, they must perform rigorous, granular validation to ensure that the logic implemented matches the intended strategy, as even small discrepancies can lead to significant errors in long-term financial simulations.