We all hit that wall where hardcoding build-order responses just stops scaling. The bot needs to start looking at the data and making its own calls.
Sharkbot’s creator sharknice is walking through how he connected his bot to a .NET decision tree model to pick builds based on win probability. He feeds unit counts, building timings, and match results into the library.
You will see a practical example of how to hook your own match data into a basic Machine Learning library.