Okay, remember how I was working on Kampaku, a small game about supply lines?
Holy shit is this gameplay emergent. There's no possible way I can write a decent AI for this game.
I've tried every trick short of neural nets and alife. I can get it to perform well, but there's a clear counter to each, and an experienced player will have no problems outwitting the computer foes I have designed.
This becomes especially true as I tweak the statistics of the combat - giving one side a faster spawn or march rate, another side more powerful fortifications...
The only way I see to get this game to work is to make it strictly multiplayer.
So... hrm... I think I'll need to totally sack the idea and make a new system that doesn't carry a required AI that's only theoretically possible.
The good news? The foundation for the AI - the map scanning and weighting - worked perfectly. I just didn't realize how complex the gameplay would get.
So, a new Kampaku. I have ideas.