• 85
  • 1
  • 25
  • 93
  • 20
  • 37
  • 4
  • 31
  • 23
  • 4
  • 1
  • 12
  • 9
  • 7
  • 1
  • 17
  • 15
  • 10
  • 3
  • 25

Recent Posts

  • @kindwind It would seem that training against random games would be every inefficient compared to training against the AI that TripleA comes with.TripleA is not Go, random play should be very bad.

    read more
  • @rogercooper right now I am trying to pin down a good signal from the tree search. I finally stopped get draws. Axis were winning like 25% but the game engine was wrong. I think i have the game engine pinned down. will some training tonight to see if I can get a signal. If I have to I will add turns. have to see how it plays out.

    read more
  • @kindwind Rather than a mercy rule, I would suggest a fixed turn limit, where the game is adjudicated based upon VC's. For the boardgame ports, a 10 turn limit would be sufficient.

    Keep in mind is possible for a game to reach a stalemate.

    read more
  • @kindwind If you want to try a neural net AI for TripleA, you would need to modify TripleA to support repeated self-play. For example of programs that support repeated self-play checkout Ludii & Chess Arena.

    Supporting repeated self-play would be project of moderate difficulty in Java. (TripleA already supports self-play). You would need to learn Java (Vibe coding doesn't cut it).. Repeated self-play would also be useful for balancing mods.

    I don't know how well a neural net would play TripleA. The number of possible moves in Go is a most 361, far fewer than the number of possible moves in a TripleA game. An effort to get a neural net to play a 2-player Risk variant did not go well.

    One advantage is that TripleA already has heuristic AI's that could be used to train a neural net.

    read more