Original Summary

I’m a scientist by profession, and automated trading has been a hobby of mine for years. I think the reason I enjoy designing trading bots so much is that I naturally approach them the same way I approach experiments in the lab. I don’t really believe a strategy should be built, backtested once, and then immediately trusted with real money. For me, a good strategy should spend a long time in the lab first . You build an idea.. You test it. You collect the data. If it performs well, you don’t immediately declare success. You can clone it into slightly different versions, change one variable at a time, and test those versions against the original. Maybe one improves the entry, another changes the take-profit logic, another changes the stop-loss, and another works better on different pairs... Then the strongest versions survive and become the starting point for the next round of experiments. So the loop becomes: build >> test >> compare >> improve >> test again And it can continue forever. That is actually how I used to use 3Commas several years ago. I had many bots running at the same time, testing different ideas and variations against each other. Eventually I stopped because 3Commas’ pricing for running multiple bots became too expensive for me. The whole point of my setup was to run many experiments in parallel, and the new pricing made that workflow impractical. But I always missed the process. So recently I built my own environment specifically so I could start experimenting this way again... The first thing I did was go back through my old 3Commas tests and recreate the three bots that had performed best for me back then. Today I want to share the first one: TrendFrend . I recreated it and started it again as a forward paper test, with the rules frozen. After roughly two weeks: 14 closed trades 10 wins / 4 losses 71.4% win rate +$15.40 realized PnL +15.4% ROI on max capital 1.77 profit factor -$12.10 max drawdown +$1.10 expectancy per closed trade 11h average holding time I attached both the performance overview and the underlying trades. But I don’t consider these results particularly meaningful yet. 14 trades is still an extremely small experiment. What interests me is what happens if I leave the strategy untouched and allow the data to accumulate. Then, once there is enough evidence, I can start creating controlled variations and test them against the original. That paper-testing environment eventually became LabNarrative , which I built mainly because I wanted this experimental workflow back! I now have all three of my old top-performing bots running again.. I’ll share the other two as well, and over time I want to start cloning the strongest ones into new versions and testing them against their originals. Basically, I want to treat trading strategies the way I would treat experiments in the lab. Does anyone else here develop bots this way — keeping a strong “control” strategy and testing new variations against it

  • 情报分类：综合情报
  • 分类依据：内容未命中明确的垂直分类规则，归入综合情报
  • 信息来源：Reddit · SideProject
  • 发布时间：2026/9/20 20:00:20