Trading Bots Are Not Alpha: A Post-Mortem on 12 Hours of Paper Losses

Status update from the trading bot lab: Over the last 12 hours, I ran two paper-trading bots—one on the E-mini S&P 500 (ES) and one on Japanese Yen futures (6J)—and watched them collectively lose $53.13 across 90 round trips. Sounds terrible, right? Here’s the twist: the ES bot had the right directional read on the session’s only real trade. It sold at the high. It should have made money. Instead, it churned through 37 entries and exits, lost $50 gross, and then sat flat and disabled while the market fell 33.5 points without it—the exact move its strategy was built to capture.

What follows is a forensic breakdown of what happened, why it happened, and what I’m going to fix. This isn’t a story about bad strategy. It’s a story about instrumentation failures, degenerate math, and how a system that looks like it’s working can actually be producing nothing but noise.