Hey Mandala players,
We've been training an AI to play Mandala using reinforcement learning (AlphaZero-style). It's trained on 40,000+ self-play games and has learned real strategy — field building, color coverage, mountain timing, defensive play.
We're looking for experienced Mandala players to try it out and give us feedback. It's completely free, runs in your browser, no download or account needed:
https://gamergym.app
The AI uses Monte Carlo Tree Search with a neural network, so it actually evaluates positions and plans ahead rather than following simple heuristics. We'd love to hear how it compares to your BGA opponents.
A few things we're especially curious about:
- Does the AI feel like a real opponent?
- Any obvious strategic mistakes?
- Is the difficulty level right, or should it be harder/easier?
We also have a Lost Cities AI on the same site if you want to try both.
Thanks for any feedback!
We've been training an AI to play Mandala using reinforcement learning (AlphaZero-style). It's trained on 40,000+ self-play games and has learned real strategy — field building, color coverage, mountain timing, defensive play.
We're looking for experienced Mandala players to try it out and give us feedback. It's completely free, runs in your browser, no download or account needed:
https://gamergym.app
The AI uses Monte Carlo Tree Search with a neural network, so it actually evaluates positions and plans ahead rather than following simple heuristics. We'd love to hear how it compares to your BGA opponents.
A few things we're especially curious about:
- Does the AI feel like a real opponent?
- Any obvious strategic mistakes?
- Is the difficulty level right, or should it be harder/easier?
We also have a Lost Cities AI on the same site if you want to try both.
Thanks for any feedback!