Has anyone seen/heard of/performed any research into artificial intelligence for Axis & Allies? I’ve heard that the CD-ROM game had an extremely slow AI, but that’s just second hand, and I can’t seem to find any research being done on the actual game itself.
I’m particularly interested in learning agents and optimal path searches, although anything in terms of using computers to optimize or find solutions to common A&A questions would be ideal.
On a related note:
The AI on the CD-Rom is pretty bad. It gets into loops, sometimes having countries buy nothing but subs for 10 turns, or making a completely illogical move. I don’t play it, I play myself rather than the Computer.
Since Hasbro has not released the A&A source code or .ini files, we really have no way of tampering with it. Even with an incredibly fast computer, some AI turns can seem like an eternity (and sometimes they cause the game to crash), especially if you haven’t switched on Quick AI which I suggest everybody should do. Even then the Allies and Axis AI plays noticeably bad, especially in moving urgently needed units to the front instead of keeping them bottled necked in ICs
With that said, the AI does play a decent USSR and Germany game, with preprogrammed moves tacked on. Island nations like UK and Japan are okay, though the transportation system is unreliable at best. And don’t make me get in USA. I’ve seen first time players have more skill with what to do that country instead of stacking with infantry and RM in W/E USA.
Okay, thanks for the info. We’re actually more interested in whether anyone has explored the use of learning networks or searches to develop/prove optimal methods of A&A play.
This is actually for a research project. So far we’re working on it from the ground up, everything from abstracting the game-state to a graph to being able to generate new strategies. We’d love to hear from anyone who’s conducted similar work/research in the field, to see if we can improve on it.
Of particular note is whether anyone’s done any long-term simulation of gameplay strategies to test out the ideas that people in this group popularly espouse. I’ve only read a handful of existing strategies, but some of them seem to contradict each other, and while all of them seem valid, surely some statistical analysis could bring the real truth to bear?
So I guess the question is: Has anyone conducted or considered methods for conducting this sort of analysis?
Thanks,
Bruce
Heh hey you guys should get together an Axis and Allies 2, and have them market it. Then MAYBE I can find it in the stores! (I’d put a madface thing here, but not sure how…)
“No Bastard ever won a war by dying for his country; he won it by making the other poor dumb Bastard die for his country.”
[ This Message was edited by: Chris on 2002-05-25 12:35 ]
Ha. Still waiting for Axis and Allies: Global War (or last time I heard it) to come out. It was suppose to include, the original Axis and Axis, A&A: Europe, A&A: Pacific, along with an expanding unit and scenario editor. Of course, chances are it probably never got past the initial planning stages. As for the AI strats, the Ai for each country is limited to only a few basic moves, most of which I bet are preprogrammed into the AI. It’s not until the mid-late game where you see any semblance of Runtime AI and how horribly choppy it is. In my opinion, the best way to test the AI is through experimenting with different opening moves and then using the save/load option to see if the AI changes any of its tactics.
I think Hasbro is shifting it’s gaming funds more toward bringing back more old Avalon Hill games and introducing new ones, than improving new games. Maybe an independant company will offer to make a game.
Let hope so. :smile: Though I’m not exactly sure how you would secure the rights in order to do so. The license will probably have to be bided on or with a royalty fee.
