It has been a month since I wrote about Making Money with AI. Since that post, my AI trading bots have gone through constant iteration, and a lot has changed. Making AI actually work for you comes down to exactly that: continuous refinement.
Iteration and Progress
For the rapid scalper bot, I eliminated two brokers and all sixteen old strategies. Three new, tweaked strategies are now running. After four days, results look promising. I need at least three consecutive weeks of forward testing to reliably find profitable days, times, and strategies.
Right now, the sample size is too small for the automatic slicer calculator to work its magic. I admit the initial setup was overengineered. When AI writes code, scope creep and overengineering can take over easily. Still, nothing is truly lost. Finding ways that do not work is progress itself.
Crypto Trader Updates
The crypto trader also saw progress. My goal was to forward test all strategies, then eliminate any without at least a 60% win rate and a profit factor over 1. Losses are expected during this phase; that is the point of forward testing. Over time, the clear winners will emerge, and the losers will be cut.
The crypto platform itself did not receive a major overhaul, just a strip-down and some bug fixes here and there to tighten things up.
Building a Dedicated Backtester
Backtesting became crucial. I built a dedicated lab for it. This project kept me awake and pulled me away from the gym. Without AI, this would have taken months, even years, to build. The core backtester components were embedded in my trading platforms. I just extracted and centralized them into a specialized app.
Now, any new strategies for either platform can be submitted to this new backtester app. This streamlines the entire development cycle.
Progress often appears as a relentless cycle of testing and refinement, even when it feels like a step back.
Reflection
This month taught me that consistency in iteration beats perfection. There is always more to test and refine, and always another trap waiting for overconfidence. The work continues.