Hi everyone,
I know there are already excellent Go engines and apps out there, and I’m not trying to compete with them. This started as a personal learning project: I wanted to understand the reinforcement learning and Monte Carlo Tree Search approach behind AlphaGo Zero by actually training my own network from scratch, and see how it behaves on a 9x9 board.
The result is iGo Coach, a free Android app with two parts:
- Ray-zeroGo: my self-trained opponent, with levels 1–10
- A move-probability overlay: a KataGo-derived network shows where strong moves are likely, so you can learn while you play
It’s far from perfect, and I’m sure experienced players will find its weaknesses quickly. That’s exactly what I’d love to hear about:
- Odd moves or blind spots the engine has
- Whether the levels feel well balanced
- Anything confusing in the app itself
And a small challenge: can anyone beat level 10? If you do, I’d really appreciate a screenshot or the game record, so I can study what went wrong.
Google Play: https://play.google.com/store/apps/details?id=com.naomijpn.go_coach
Thanks for reading, and thanks to this community for all the resources that helped me learn along the way.