@Trevan Currently, the AI simulates the current turn to the end to determine what it should purchase based on estimated moves/battles. It doesn't simulate future turns due to how quickly the search tree explodes and how long it takes to run. You would need an incredible amount of optimization and fairly accurate prediction of moves/battles to slim the search tree to achieve that.
@ff03k64 I'm talking about my new AI. Hopefully, it'll run fast enough so that multiple turns can be tested. Also, hopefully it will be good enough to actually play against
@Trevan Some AI turns already take a couple of minutes. This would only make that worse wouldn't it?
As part of my new AI, I want to be able to simulate multiple rounds of play. This way, the AI can randomly select player options, diplomacy, etc and then it can play with those selections and see how beneficial they are.
Because of triggers and such, I think I'll need to actually run the game somehow. I've been going through the code to see how possible it is but I'm wondering if this has been done before? Is there some example that I can follow to accomplish this?
In addition, because of the randomness of dice and the fact that it will cause the search tree to be really huge, I want to insert my own battle phase that performs battles in a deterministic manner. Has that been done before?