SMAC 444 - AI patch updated [n]

[kyrub]

SMAC 444 - AI patch updated [n]

http://alphacentauri2.info/index.php?action=downloads;sa=view;down=165

Corrected version SMAC_444_n

- corrected bad bug from previious release!
- improved basic terraforming to include more forests (see weirwood's suggestion up in the thread)
- improved sea terraforming, we'll see a lot less sea mines
- AI will now use its mines and boreholes

http://dfiles.eu/files/4cxkbh09p

[Geo]

The Sun Spots event hasn't been touched by you, hasn't it Kyrub? Because, at Librarian difficulty level, it starts at MY 2154. Sounds a bit soon?

[kyrub]

I have done nothing to the events in the game.

(Only data instability in patches may cause non-standard behaviour. But such an instability usually proves itself with crashes to desktop later. I have tested this version with several games and there were no CTDs, it seems clean. --- The version I posted YESTERDAY though had CTDs, so beware... and update.)

[Geo]

I am playing with version n, no worries.
I am impressed with the Gaians sofar though. They already constructed 3 SP's to my one SP, and its MY 2201 now.
OTOH, Yang is very agressive with little to no army. He declared war at me and I could walk in The Hive with a scout patrol after which he immediately sued for peace.

[kyrub]

Re: Yang
There is an issue with AIs walking their units to obelisks. They often leave their bases for the taking, it's outrageous design. I cannot seem to do anything with it, no idea where to find it. SMAC AI code is incredibly complex, it's probably a modified Civ2 AI.

In my test games, Yang was by far the most succesful. Usually, the AI that starts near Monsoon Jungle has best results.

[Yitzi]

Re: Yang
There is an issue with AIs walking their units to obelisks. They often leave their bases for the taking, it's outrageous design. I cannot seem to do anything with it, no idea where to find it. SMAC AI code is incredibly complex, it's probably a modified Civ2 AI.

If you can identify where it sets goals for units, maybe you could just have it test for undefended or underdefended bases and have the nearest defensive unit which isn't already defending a base that needs it set its goal to go to the underdefended base at an extremely high priority.

[kyrub]

If you can identify where it sets goals for units, maybe you could just have it test for undefended or underdefended bases and have the nearest defensive unit which isn't already defending a base that needs it set its goal to go to the underdefended base at an extremely high priority.

If I can move a mountain, I can build a new mountain chain, sure. Just how?

The trick is that the AI does a lot of thing we aren't dreaming about, like, I bet, it already has a send-defender-unit-somewhere function. I think I saw that one, even. But this is not helping much since defenders in AI games are hard to come by, bases are normally underdefended. You find a free defender rarely. I also don't have the flag for "going to monolith for an upgrade" thing, that would be a massive start. Personnally, I'd stop AI sending units to monolith  if only 1 defender in base, case solved.

[Yitzi]

If you can identify where it sets goals for units, maybe you could just have it test for undefended or underdefended bases and have the nearest defensive unit which isn't already defending a base that needs it set its goal to go to the underdefended base at an extremely high priority.

If I can move a mountain, I can build a new mountain chain, sure. Just how?

I would run it with a breakpoint at a unit that it's likely to be moving, and see where it sets orders and backtrack from there.  Unfortunately, AI is probably one of the hardest parts to work with, but that's the best I can think of.