For two weeks the bot looked healthy. It ran 24/7, connected to every exchange, logged hundreds of thousands of price detections, and sent its morning reports on time. It had never executed a single profitable trade. When I finally asked it the only question that mattered — show me a profitable trade — the answer was silence.
The cause was a single number: a trading fee set 7.5× too high, so every opportunity was calculated as unprofitable and silently declined. Three other blockers sat behind it. Any one of them alone would have prevented execution.
The bot was working exactly as designed. It was designed wrong, and nothing in it knew. That's the lesson that defined everything after: running is not the same as working. Since that day, no milestone here is allowed to be called complete until it can show a profitable trade — or, more often now, show the data proving why it correctly didn't.
Bot activity — not financial advice.