Putting the 'role' back in role-playing games since 2002.
Donate to Codex
Good Old Games
  • Welcome to rpgcodex.net, a site dedicated to discussing computer based role-playing games in a free and open fashion. We're less strict than other forums, but please refer to the rules.

    "This message is awaiting moderator approval": All new users must pass through our moderation queue before they will be able to post normally. Until your account has "passed" your posts will only be visible to yourself (and moderators) until they are approved. Give us a week to get around to approving / deleting / ignoring your mundane opinion on crap before hassling us about it. Once you have passed the moderation period (think of it as a test), you will be able to post normally, just like all the other retards.

SMAC 444 - AI patch updated [p]

kyrub

Augur
Joined
Aug 13, 2009
Messages
347
Thanks for feedback. It's nice to see the patch is being used.


SilentCat
I will look into the headquarters issue.

Pelvis Knot
- I had Gaian winners in one of my spectator (all AI) game. The Ai that starts on Monsoon jungle is usually the toughest opponent.
- Empty capitals were an issue in Vanilla as well. One of the main cause seems to be the AI's "obelisk obsession". Once it finds an obelisk, all units are dispatched for the levelling up. It's really stupid - I haven't seen the source as yet. The only point when AI behavior was changed in this respect in the patch, are recycling tanks. They have become a real top priority, which sometimes (rarely, in fact) make the AI playing a bit of vabanque game. I find it correct, though.
- Probes. Exactly, that's weakest point now. I have done my best to enhance Rover-chassis units, but forgot that Probes use it too. Side-effect is massive, it really weakens the AI. Will be corrected in next version, that's sure.
 

Pelvis Knot

Cipher
Joined
Nov 19, 2010
Messages
885
Thanks for feedback. It's nice to see the patch is being used.


SilentCat
I will look into the headquarters issue.

Pelvis Knot
- I had Gaian winners in one of my spectator (all AI) game. The Ai that starts on Monsoon jungle is usually the toughest opponent.
- Empty capitals were an issue in Vanilla as well. One of the main cause seems to be the AI's "obelisk obsession". Once it finds an obelisk, all units are dispatched for the levelling up. It's really stupid - I haven't seen the source as yet. The only point when AI behavior was changed in this respect in the patch, are recycling tanks. They have become a real top priority, which sometimes (rarely, in fact) make the AI playing a bit of vabanque game. I find it correct, though.
- Probes. Exactly, that's weakest point now. I have done my best to enhance Rover-chassis units, but forgot that Probes use it too. Side-effect is massive, it really weakens the AI. Will be corrected in next version, that's sure.


It's great to see you continue working on the patch.

As for Gaia, I played 4 games on Transcend, and they got thrashed twice by Hive, once by Believers. The fourth game was dominated by University, Gaia was second but very close to Uni, I was Sparta, alone on a continent. When everybody else was destroyed by University, they turned on Gaia and just took like 1 base/turn.

It seems to me Gaians can't stand on their own when someone goes all in on them.
 

weirwood

Educated
Joined
Apr 6, 2008
Messages
56
ARID
1) arid + flat = FOR
2) arid + rolling + 2k = FOR
3) arid + rolling + 3k = F / S
4) arid + rocky = M / R


MOIST
1) moist + flat + 1k = FOR
1) moist + flat + 2k = F / S
3) moist + rolling = F / S
4) moist + rocky + 2k = M / R
5) moist + rockx + 3k = S

RAINY
1) flat = S / F
2) rolling = S / F
3) rocky +1k = M / R
4) rocky + 2k = S

------------------------------

MINERAL
1) arid + flat or rolling = FOR
2) arid + rocky = M / R
3) moist + flat / rolling = F / S
5) moist rocky = M/R
6) rainy flat / rolling = S / F
7) rainy rocky = M / R


NUTRI
1) arid + rolling + 3k = F / S
2) any other arid = FOR
3) moist + flat + 2k = FOR
4) moist + flat + 3k = F / S
5) moist + rocky = FOR
6) moist + rolling = S / F
7) rainy flat / rolling = F / S
8) rainy rocky = M / R


ENERGY
1) flat + arid / rolling = FOR
2) flat + moist + 3k = F / S
3) flat + moist + 2k = FOR
4) rocky + arid / moist = FOR (M /R *)
5) moist + rolling = F / S
6) rainy + rolling = F / S
7) rainy + rocky = S only

Are you still using that table? I've been looking at your patch a little closer, and I think there's room for some simple improvements. (Your work is greatly appreciated, that much goes without saying!)

First of all, I assume what you can easily do is tell a former what it should do on any given square it's on. I guess setting priorities on which squares to TF first isn't yet possible?
Can you make certain improvements dependent on specific techs? For example, there's little benefit to constructing mines before the resource limits are lifted, i.e. Ecological Engineering is researched.

As it is now, the AI builds lots of mines, but it hardly ever works them, even with EE. Rivers and resource boni on the same square seem to help though. Rocky/solar doesn't see much use either, but it rarely gets built. Could you make the AI use 'level terrain'? That's not really efficient, but if you can't assign a low priority to rocky terrain in general, leveling it and building other improvements might be preferable to building a mine that never gets worked.

Second, any progress with advanced terraforming? Something that would be straightforward in that case would be placing farms and condensers on nutrient specials. Aquifiers are also great, especially far inland (or at 2+k elevation, if that's easier).

Even if all that's not possible, I think one key area of improvement is prioritizing forests over farms/solar some more. F/S take 8 turns to construct, a forest takes 4, and can spread on its own. That's the best bang for the buck the AI can get. Plus, once tree farms and hybrid forests come into play, forests come out ahead, especially since the AI doesn't use advanced terraforming well. The tree farm and hybrid forest have some very nice additional perks, so they're a good investment for any faction.

Here's where I would change F/S to FOR: (apart from rocky terrain)

1) moist + flat + <3k
2) rainy + flat + <2k
3) moist + rolling + 0k

We'll have to see how the AI deals with that.

As for the specials: Anything rocky should get a mine and road. As specials are one of the few things that can convince the AI to use its mines, you should take advantage of that.

Moist/rainy nutrient specials should get F/S, as they'll benefit from soil enrichers/condensers, if the AI ever gets around to placing those. I'm pretty sure it won't build condensers on pre-existing solars. At 0k elevation, you could try making it not build those on nut specials.

For anything else, I don't think there's a need to make special exceptions.
 

kyrub

Augur
Joined
Aug 13, 2009
Messages
347
First of all, I assume what you can easily do is tell a former what it should do on any given square it's on. I guess setting priorities on which squares to TF first isn't yet possible?
Oh, surely it is possible, but how? The original AI has this and many more aspects. The AI complexity is its problem or, rather, my problem. I tried to stress resource boni more but it somehow did not work. It does not work properly in the original game, so who should be surprised by it... Many times, you pull some lever up imagining what it might lead to, but instead you get some other response, because some other emergency hidden mechanism (you obviously had no idea about) kicks in. Wish I knew how this can be done.

Can you make certain improvements dependent on specific techs? For example, there's little benefit to constructing mines before the resource limits are lifted, i.e. Ecological Engineering is researched.
This is definitely in the original terraforming AI. I cannot remember if I let it alive in my mechanism. Must have another look.

Could you make the AI use 'level terrain'? That's not really efficient, but if you can't assign a low priority to rocky terrain in general, leveling it and building other improvements might be preferable to building a mine that never gets worked.

Second, any progress with advanced terraforming? Something that would be straightforward in that case would be placing farms and condensers on nutrient specials. Aquifiers are also great, especially far inland (or at 2+k elevation, if that's easier).
I am grateful for your concrete suggestions, very good. You may think about more and putting here, aquifers at 2k+, should work perfectly and is generally easy to do. My problem with advanced terraforming is that the function is very complex and embroiled in some sets of rules which I do not understand one little bit. I believe this has a lot to do with scenario settings and multiplayer. I did not play them, I am unfamiliar with the special rules and I am lost in their quantity. Hell, if original Alpha Centauri weren't multiplayer, I'd have a lot more progress on AI front. It makes the code understanding foolish.

In general, I would love to see here a set of rules for advanced terraforming similar to the table I put above. Can you make it? That would help me.
Example:
Improvement: Borehole
conditions
a) necessary: City has at least 6 size
b) City has at least +2 nutrients output
c) flat terrain unworked tile present
d) City has no other borehole
etc...
then odds are XYZ to start a borehole.



Here's where I would change F/S to FOR: (apart from rocky terrain)

1) moist + flat + <3k
2) rainy + flat + <2k
3) moist + rolling + 0k
Very good suggestions 1) and 2), I disagree with the 3). It's easy to repair and it will improve the AI.
 

weirwood

Educated
Joined
Apr 6, 2008
Messages
56
Oh, surely it is possible, but how? The original AI has this and many more aspects. The AI complexity is its problem or, rather, my problem. I tried to stress resource boni more but it somehow did not work. It does not work properly in the original game, so who should be surprised by it... Many times, you pull some lever up imagining what it might lead to, but instead you get some other response, because some other emergency hidden mechanism (you obviously had no idea about) kicks in. Wish I knew how this can be done.

I'm not certain about this, might it be possible that the AI is so eager to plant mines all over the place because you also explicitly told it to build roads there? You could try encouraging it to place roads on resource specials, and see what happens.

If watched a few more games since I wrote the first post, here's what I've noticed:

1) Mines are definitely a problem as is. The AI builds lots of them, but they rarely ever get worked. This amounts to a lot of wasted former time.

I'm inclined to recommend you change it that the AI either leaves ordinary rocky squares unworked, or that it uses 'terraform level' on them and builds other improvements. If a rocky square is on level terrain, it would also a good spot for a borehole later on.

The exception are mineral specials - a mineral mine will produce 0/7/0, that's big enough to convince the AI to work them. There is one borderline case where it might make sense (for the AI) to build a mine in rolling terrain; if it is rainy and has a mineral special, a farm + mine + road will boost it to 2/5/0. However, a forest or depending on elevation, a farm + solar would get you nearly as much, for less terraforming investment. I thought I'd mention it because the AI can tend to get a little mineral-starved with its focus on farm + solar.

For nutrient and energy specials, or minerals on flat terrain, it's better to build something else. for nutrients, leveling the terrain is best. It might be possible to leave a rocky/energy square for a borehole, if it's on level terrain.

2) The AI does build condensers. However, with your rules for farms + solars, it won't build them on farms anymore. This is a problem. Absent pop-booming, condenser farms are probably the best way to boost the AI's growth and mid-game power.

3a) It's rarely used by human players, but remember that 'logging' a forest by building something else there will add 5 minerals to the nearest base. For some squares, it might be worthwhile to simply have the AI plant forests at first, and then construct other improvements as the restrictions get lifted and/or advanced terraforming becomes available.
3b) Forests are great on rivers. Before restrictions are lifted, you can't beat getting 1/2/2 for only 4 turns of terraforming. They also seem to spread fast along rivers.

3c) I believe the original AI would increase planting forests once it had tree farms/hybrid forests. In tends to happen quite late for the AI, but even so, you should add a check that once a base has a TF, forests take precedent over farm/solar at <2k elevation.

4) On reconsideration, I think farm/solars aren't bad at 2k+ elevation. If you can get the AI to intersperse them with an occasional Echelon Mirror, they should become quite solid.

I am grateful for your concrete suggestions, very good. You may think about more and putting here, aquifers at 2k+, should work perfectly and is generally easy to do. My problem with advanced terraforming is that the function is very complex and embroiled in some sets of rules which I do not understand one little bit. I believe this has a lot to do with scenario settings and multiplayer. I did not play them, I am unfamiliar with the special rules and I am lost in their quantity. Hell, if original Alpha Centauri weren't multiplayer, I'd have a lot more progress on AI front. It makes the code understanding foolish.

In general, I would love to see here a set of rules for advanced terraforming similar to the table I put above. Can you make it? That would help me.

Can you go into a little more detail what conditions you can easily check for? Buildings present in the city, i.e. tree farms? Eco-damage? Can you make a check what terraformings/conditions are present in adjacent tiles? IIRC, the AI can 'cheat' in the base game by building boreholes next to each other. It would be good to fix that, and this would probably also be necessary so it can't drill aquifiers next to existing rivers.

Example:

Improvement: Borehole

conditions

a) necessary: City has at least 6 size

b) City has at least +2 nutrients output

c) flat terrain unworked tile present

d) City has no other borehole

etc...

then odds are XYZ to start a borehole.

I can try my hand on condensers for a start. I think those are more important, as they can be available quite early, if the AI gets the Weather Paradigm. (Boreholes are pointless until the restrictions get lifted, even if you could build them earlier.) The big question is how well the AI can handle the eco-damage from advanced terraforming.

Improvement: Condenser

First of all the basic squares

moist/rainy + rolling + 0k
moist/rainy + rolling + nuts

Making rolling a requirement should keep the AI from simply plastering condensers everywhere at ground level. The condenser should bring a moist square up to rainy, so there's no difference between the two. Before soil enrichers, these will provide 4 or 7 nuts each. These are the kinds of squares the AI shouldn't build solars on to leave room for condensers.

Conditions:

a) Eco-damage is 0
b) Base isn't at population limit, if it has a Hab Complex
c) farm built on square
d) limit 2/3 condensers if there's no Hab Complex in base (Morgan/rest). - (This assumes that bases' radius overlaps, i.e. condensers can be in the radius, but worked by another city.)
e) limit 4/6 if there's a Hab Complex. (Morgan/rest) (By the time you get Domes, you shouldn't depend on condensers alone anymore.)

That's my first shot. It's probably a good idea if we take this discussion to the various Alpha Centauri forums, so that anyone who still plays the game a lot can contribute.
 

kyrub

Augur
Joined
Aug 13, 2009
Messages
347
Moving to apolyton for detailed discussion, let's not clutter rpg forum.
 

kyrub

Augur
Joined
Aug 13, 2009
Messages
347
Yeah, I meant Codex, sorry for lazy typing. I meant: If a specific game thread grows into a detailed discussion of AI algorithms, it belongs more to a specialized forum, like apolyton with a SMAC section.

(Had Codex had a special mod section, it could have gone there. But I am probably not a fan of this idea either, I like the crazy melting pot here in a strange way.)
 

Elwro

Arcane
Joined
Dec 29, 2002
Messages
11,746
Location
Krakow, Poland
Divinity: Original Sin Wasteland 2
If you think this doesn't belong into the "Strategy Gaming" section, isn't "Codex Workshop" the best place for this?
 

Baptismbyfire

Liturgist
Joined
Sep 9, 2012
Messages
182
Really appreciate it that you keep working on it. SMAC is still the best 4X there is even after so many years.
 

DakaSha

Arcane
Joined
Dec 4, 2010
Messages
4,792
I posted a new update of the SMAC AI patch, if anyone's interested.
http://alphacentauri2.info/index.php?action=downloads;sa=view;down=122


I corrected the most annoying side effects from the previous game (mass probes and mass sea transports).
I found a major mismatch in AI social engineering. The correction has unknown consequences, hopefully for the better. I added a few fixes there as well, see the link.
this link is sucking my dick

edt: in fact everything i find about this patch gives me broken pages wtf.


edit: had peerblock active :retarded:
 

kyrub

Augur
Joined
Aug 13, 2009
Messages
347
edit: BUGGED - see corrected SMAC_444_n version.
http://dfiles.eu/files/4cxkbh09p

Major update
of SMAC_444 patch,
about 70 assembly hits for a ruthless SMAC AI. I have had this on hard disk for about year, finally got an afternoon to finish it.

The patch does *not* work with SMAX (I am sorry, I had/have no time for it; I have not got even the time to finish this).
The patch includes scient's latest patch, fixing a good number of bugs.

- AI will stand up and fight in air wars. It builds a lot more AAA defenders, it uses planes for other things than bombing formers. You may see it use hit and drop tactic.
- AI will attack with while in "hurry" (with 2/3 strength etc.). It should counterattack from bases more.
- AI sees enemy probes as a primary target and may go long way to destroy them.
- AI uses more intelligent pattern for city spreading. AI colpods will now rarely walk half map somewhere.
- AIs ruthlessly exchanges technologies between themselves, hopefully this makes them competitive until later stages of the game.-
- AI likes buildings a lot more. It is able to rebuild its hedaquarters now. It loves orbital pods, tree farms etc.

http://alphacentauri2.info/index.php?action=downloads;sa=downfile&id=164

This update was untested, only for crashes. I saw just 1 test game, the early signs were good, but who knows.
I'll be glad to hear reactions and feedback.
 
Last edited:

kyrub

Augur
Joined
Aug 13, 2009
Messages
347
Corrected version SMAC_444_n
http://dfiles.eu/files/4cxkbh09p

- 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
 

Trash

Pointing and laughing.
Joined
Dec 12, 2002
Messages
29,683
Location
About 8 meters beneath sea level.
Corrected version SMAC_444_n
http://dfiles.eu/files/4cxkbh09p

- 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

Your patching already gave me lots of joy when I replayed UFO and MoM and now this. Thanks for the incline you brought!
 

Lyric Suite

Converting to Islam
Joined
Mar 23, 2006
Messages
56,164
The patch does *not* work with SMAX (I am sorry, I had/have no time for it; I have not got even the time to finish this).

The consensus here seems to be that SMAX is shit, so don't sweat over it. Keep focusing on the original.
 

Nevill

Arcane
Joined
Jun 6, 2009
Messages
11,211
Shadorwun: Hong Kong
The patch does *not* work with SMAX (I am sorry, I had/have no time for it; I have not got even the time to finish this).

The consensus here seems to be that SMAX is shit, so don't sweat over it. Keep focusing on the original.
There were several annoying bugs that were fixed only in SMAX. For example, the maintenance cost for buildings on higher difficulty levels was supposed to be lower for AI, but this was mistakingly applied to the player as well. Or the bug when +2 GROWTH provided by a Golden Age combined with +4 gained by other means sometimes did not trigger a population boom in the base - this was critical for factions that could not achieve +6 GROWTH by means other than Golden Age.

While the new content was lackluster, the add-on did have redeeming qualities. Hopefully, Yitzi will be able (and willing) to adapt these new changes for SMAX.
 

Xbalanque

Educated
Joined
Apr 7, 2014
Messages
92
Location
Land of blossoming onion
Never played SMAC and decided to start. Can you recommend any place where I can learn how to play and the basic rules?
Unfortunately, got the computteer when the Civ3 was released so my experience is somewhat limited. Will appreciate all the help I can gget
 

kyrub

Augur
Joined
Aug 13, 2009
Messages
347
Never played SMAC and decided to start. Can you recommend any place where I can learn how to play and the basic rules?
Unfortunately, got the computteer when the Civ3 was released so my experience is somewhat limited. Will appreciate all the help I can gget

If you had some kind of Civ experience then I don't think you need any particular help. The game has a solid internal help system to lead you through.
There are a few mysteries, though, like Free Market effect (special type of unrest) or eco damage mechanics. For these, you may need to consult wiki

There were several annoying bugs that were fixed only in SMAX.
Not true.
SMAC was patched to 4.0 version alongside SMAX release and it is relatively bug free (or, should I say, scient made it so). A few cool landmarks were added as well, like Borehole Cluster and Manifold Nexus. SMAX has done more wrong than right. The worst part is the fact they let AI to "somehow" cope with all the new features, that is amateurish. It is hidden under the fact that the AI of the original is not exactly scintillating, but SMAX made it downright incompetent (Pirates, aliens, Cha Dawn).
 

Xbalanque

Educated
Joined
Apr 7, 2014
Messages
92
Location
Land of blossoming onion
Never played SMAC and decided to start. Can you recommend any place where I can learn how to play and the basic rules?
Unfortunately, got the computteer when the Civ3 was released so my experience is somewhat limited. Will appreciate all the help I can gget

If you had some kind of Civ experience then I don't think you need any particular help. The game has a solid internal help system to lead you through.
There are a few mysteries, though, like Free Market effect (special type of unrest) or eco damage mechanics. For these, you may need to consult wiki

thanks! gonna get to it
 
Unwanted

Xu Fugui

Unwanted
Joined
Apr 15, 2014
Messages
253
Location
香巴拉
Yeah, I may not be able to brofist you yet but your efforts are greatly appreciated, kyrub, it's important that you know that. Thanks to your work I have started to enjoy SMAC even more than before.
 

As an Amazon Associate, rpgcodex.net earns from qualifying purchases.
Back
Top Bottom