Vicious AI Mod (AI Tweaks) V1.1 (Now works properly)

[Meow:3]

Vicious AI Mod

At its base this mod has only one goal: to significantly improve the capabilities of the AI in SMAC

Files are in the attachment in this post.

Origins
Originally inspired by a civfannatic post and the fact I couldn't find any (major) AI improvement mod, this mod makes the AI much more capable. Even if you prefer multiplayer this is still especially suited for comp stomps and also has several features of interest that might be interesting to those that favour lan/online competitive play.

Originally I wanted to see how fast I could win even on the highest difficulties, playing every trick in the book. Unsurprisingly I found that with enough supply crawlers, city sprawl, probe magic, diplomacy exploits and project sniping this could easily be secured within the first 150 turns. And I've even read stories of players transcending within 100 turns. (On huge maps, but still)

After growing sadly frustrated with the AI's lack of ability to counter me I set forth to apply some changes that would tip the edge ever slightly in the AI's favour.

Even with a few priority tweaks I found the AI being able to overcome some of the familiar tactics and keep up with my tech level and even population on Librarian. But I wanted to go further. I wanted to the lowest to be the equivalent of vanilla Transcend and the highest difficulties to be truly vicious.

The AI as it stands in Vanilla
After observing how the AI reacted to a variety of playstyles I took a step back and simply observed dozens of AI only matches to see how the AI played. I analysed how it used its units, colonisers, probe teams, formers, crawlers, etc. I looked into what it researched, how it used that research, how it dealt with the other AI players, how it build its cities and most of all how it would come on top. I also checked the faction files and the rule set, especially what each AI type tends to focus on. Along the way I applied more tweaks to see how the AI would deal with them, and even played some matches in between using different playstyles and factions to see how it affected human players. An end date of MY 2250 was chosen for each match as by then major issues tend to quickly became apparent.

I learned many things, which while seemingly self-evident aren't entirely clear unless you observe the match in editor mode.  I'll list them below:

Problems in vanilla SMAC/SMAX

First of all; The AI has terrible SE priorities. Someone like Zak - who has a research bonus - should focus on getting more income (economy) or increasing his industry so he can put his tech advantage to use. Instead, the AI will - per design - focus exclusively on research, foregoing useful SE choices and thus acquire a lot of techs it will never end up using. This is the case with pretty much every faction in vanilla, resulting in overal lackluster performance.
The AI is unable to deal with SE and faction penalties.  When you are attacked as a human player you might consider putting the research on the backburner and switching to Power, Police State, and even Fundamentalism so you can put up an effective assault or defend against probes. The AI will generally never do this and suffers as a result. AI Zak is painfully vulnerable to probes and will seldom do anything to counter them. AI Deidre will continue to spam Very Green laser rovers even though she's better suited to psi combat. And Morgan might - assuming he's left alone - sport a loaded wallet, he will still succumb quickly to even the lightest assaults and is crippled by his pre-hab pop limitations. In short, the AI is unable to work around its SE penalties like human players can, further disadvantaging it.
The AI picks useless SE choices. Does Yang really need more police, especially when he starts off with Police State and barely ever gets his cities above pop 5 even during mid-game on Transcendent?  (Yet still goes on to build recreation commons...) The AI is set to gladly pick choices that do not suit it. Deidre doesn't really need 5 Planet (She barely uses the psi combat bonus outside of early game anyway, assuming she survives), and Morgan could better focus on efficiency and even morale/support. It's not surprising that in general only 2 factions really make it out on top under control of the AI; the Hive and the Peacekeepers. One tends to blob early, saving it self for the long run, while the other has no clear disadvantages off the bat and has perhaps the only sane SE priority of all: Growth (which nicely complements its talent bonus and higher pop ceiling). Ideological flavour is nice, but the rigidity and inefficiency of the AI's SE priorities prevent it from reaching its true potential.
The AI spams unescorted colonisers. This is another issue which cripples AI growth. It's not that the AI doesn't try to expand, it's just that it usually loses well over half its colonisers to native life before it has even plotted down its second or third city. And it's not beyond sending several colonisers into mindworm patrolled fungal blooms either. Unsurprisingly this is hardly conductive of maintaining a turn advantage.
Spore launchers can bog down the AI. A single spore launcher can keep the AI busy for at least a dozen or so turns. Either by destroying improvements (which are then futily rebuild) or by sieging AI cities, forcing the units to heal, which will then refuse to attack the spore launcher even if it is standing right next to the base. I've seen matches where an entire faction could be deadlocked for over 50 turns (or until I'd remove it) due to a single hatchling spore launcher.
Probe sprawl. Sometimes the AI uses probes effectively, but most of the time it will gladly spend a dozen or so probes on a single infiltration, even when you have the Hunter Seeker Algorithm. It will also use up all of its movement points even if it ends up next to a base (as stated in another thread). Sadly, I can't really do anything about this besides just making it less likely for the AI to spam probes. Higher probe ratings tend to correspond with probe sprawl, and so one of the things I did was remove Miriam's probe bonus. If the AI does use probes I've made sure (to an extend) that it will have enough energy to make it do something useful (like steal a tech or bribe the base).
Constant and inefficient retooling. I've seen the AI being unable to make up its mind many times. While constant retooling is bad, the worst offender is the inefficient retooling. The AI retooling an insolvent project 1 turn from completion into the production of a supply crawler, a recreation commons, a former or even a simple scout is unfortunately not that rare.
AI will allow humans to hoard projects. I've had several matches where the first secret project would only be build near the end of the 23rd century. Eventually I figured out why the AI tends to forego projects so often: It simply doesn't have enough bases. That, and they are actually too expensive. Though I appear to have found somewhat of a remedy to that...
AI will not HURRY effectively. I've noticed that occasionally the AI will hurry something, but it often isn't something of priority, resulting either in wasted energy or energy piling up, though the latter isn't necessarily that bad (see Probe sprawl). In short though, HURRY hugely favours human players.
Poor energy economy. The vanilla AI suffers from consistent lack of funds and poor handling of the energy economy (due to aforementioned reasons). In general it is more suited to a mineral economy, with its energy surplus best reserved for probe teams.
AI is terrible at diplomacy. Foregoing the infamous city exchange exploit, the AI diplomacy suffers from numerous other issues. One exploit (useful to builder games) is to simply ignore communication if you're at or near "Unsurpassed" power. Even if you have but a few scout patrol garrisons this will usually be enough to hold off an enemy assault till it's too late. Isolationism is splendid.
Extreme bonuses and penalties tend to favour humans. +2 Support is actually more powerful (usually) in human hands than in the hands of the AI. Though certain bonuses in combination with certain SE focusses can produce synergy.
Certain terraform options are barely pursued. Mainly boreholes and forests. The latter which is usually a (builder) nobrainer in multiplayer (and for obvious reasons) is something the AI rarely picks even on the highest difficulties, and it suffers as a result.
Certain factions will barely ever use terraformers even on Transcend. A somewhat better known and palpable problem. I noticed this especially with the Spartans, but also Deidre and Yang. After some testing I think I even know what's causing this: The ecology tech is researched too late. By the time it comes around the AI will be focusing on combat or base facilities. I also noticed that the formers it does have are ill protected and die just as often as its colonisers. Thus severely compromising its ability to use one of the key elements of the game to its advantage.
AI focuses on the "wrong" improvements. While it poorly manages its energy economy, it will still gladly spam solar collectors. Though I noticed that eventually it will adjust the imbalance assuming there are enough formers. I've read that one user here is analysing the exe which might reveal a bit more on the AI's terraform choices. From the looks of it tries to maintain a certain "ratio" relative to its nutrient output.
AI will not switch SE even when there are no penalties. I noticed this issue if the AI has a preference focus in the same category. Deidre for example will refuse to switch to planned even if it's the only economy available. Even when it has no penalties and/or if simple has severe penalties. It's one of many quirks that make "fixing" the AI a difficult task (Without resorting to giving unplayed factions massive bonuses).
AI will spam armoured troop foils on higher difficulties. This became especially apparent on higher difficulties. At one point in a particular match the most powerful faction (Gaians) had 222 silksteel armoured foils.
AI will try to use drop pods in sea tiles. This appears to mostly happen when there are a lot of sea bases.

There are some aspects it does tend to get right though. Besides city sprawl (under the right conditions) I also found that aggressive AI's are the most effective. Not surprising in a hindsight if you take into account the Hive's near universal dominance. But what is surprising is that even AI's that aren't geared towards war (originally, such as Deidre, Lal, Zak and especially Morgan) become a lot more effective when set to aggressive. Setting the AIs to embrace more peaceful overtures has the opposite effect and generally correlates negatively both with individual score and total score (all factions combined), though peaceful factions usual enjoy a slight tech lead if they focus on economy or research. Mostly Lal and Zak, but surprisingly occasionally Morgan if he survives on a private and sufficiently large continent.

Goals:
Promote social engineering synergy. Make sure the different choices and the SE focus, together with other faction bonuses complement the AI.
Focus bonuses and attributes around the AI's ability to use them effectively. Remove the penalties that the AI cannot effectively work around or are crippling to the AI's lack of tactical and social engineering flexibility.
Make sure that the bonuses are most conductive of the AI's style of play while still making each faction interesting to human players.
Give every AI a chance to dominate (Including Morgan and late-game Deidre) a match, not just the Hive or the Peacekeepers.
Prevent - as much as possible - "easy" builder victories that make use of "cheesy" mechanics such as diplomacy exploits, crawler spam or project hoarding/sniping.
Prevent the human player(s) from sniping/hoarding all the (early) projects. Especially game winning (faction) combos like the virtual world for the University, cloning vats for the Hive and Peacekeepers or planetary transit for Morgan. Make sure that all early wonders are spread amongst the other AI players and the human player is generally left with those that aren't particularly or immediately useful unless an active effort is made and other elements suffer as a result (eco/forming/military). (Or at the very least prevent the usual combos)
Encourage "blobbing"; A more effective AI aggressively gobbling up all its neighbours and becoming something truly monstrous.
Force ruthless AI expansion; Make sure very little land (and ocean) remains unclaimed by ~MY 2250 and even earlier. Encourage AI city sprawl usually reserved for high tier players. Do not allow builders to build undisturbed.
Nerf HURRY and other mechanics that primarily favour human players. Including boreholes and forests.
Prevent the AI from wasting/losing most of its colonisers and formers early game.
Decrease cost of units favoured or essential to the AI.
Tweak maintenance costs to counter the AI's often poor energy economy.
Give the AI a way to counter early spore launchers.
Prevent AI from falling to constant bribery.
Balance some of the early techs, SE options and buildings so the AI ends up with a more capable economy.


Results:
I've ran dozens of test games, both with me playing and on "auto"/AI-only. While initially I was prepared to be satisfied with subtle improvements, the results were... spectacular. Factions that normally wouldn't hurt a fly are now - frothing at the mouth - aggressive, sprawling until there is practically no free land or ocean by MY 2250. And with bases at your doorstep, appeasement and peaceful construction are not an option. (As a test I had only Morgan set to peaceful, and he still ended up being swarmed even if he was on his own private continent minding his business).
Morgan outsprawls vanilla Hive by a wide margin, AIs will have easily researched most of the tech tree before MY 2300 even on Transcend (old Librarian), early ascends on huge maps are the norm now and I've even seen multiple planet busters fly within 150 turns. (Again, standard map)
Deidre will gladly despoil Planet now and Mirian will occasionally beat you to the supercollider and the theory of everything.

In short: This is more in line with a human playing at a very high level, and that's assuming you are being left alone all game. I ran a few playtest with the new rules and the AIs on Transcend(Old Librarian renamed) is now able to out-tech and out-grow me and generally able to keep up, and that's while they are in a state of total war. In a way, "Drone" (Replaces citizen) almost is the new (or rather old) Transcend (Though obviously the project, prototype, retool and mind control handicaps are still in place). And "Diety" (new highest difficulty) is some truly vicious stuff.
On average the highest difficulty generally ends up with a total score (All faction combined) of up to and over 5-6x higher than Vanilla transcend (Turn 150) on a standard/Planet map, while Drone difficulty generally approaches Vanilla Transcend's total score.

Compare these maps: Three are Vicious AI Drone, the other three are Vanilla Transcend.



Yep.

EDIT: Lists are difficult

[Meow:3]

Changelog V1.1:
I'd like to start off by giving credit where it is due and say that certain changes were inspired by SNAC mod (It self a multiplayer oriented mod, I reccomend trying it out). Mainly the forest, borehole, Peacekeeper archivist bonus, Believer tech steal, Morgan's Energy Base, colony pod cost and poly encryption changes.

Faction Changes:
The key here was achieving synergy, supporting the AI's hybrid playstyle, while even making the factions more interesting to human players.
I'll begin by stating that all penalties have been removed and all AIs are aggressive now. HURRY cost has been increased by 200% for all factions.

Believers:
Removed Research penalty and Probe bonus.
Gave new +2 Morale bonus instead.
Reduced Support bonus to +1.
Added, Sociological expertise: Free Recreation Commons at every base.
Added, Agreeable interrogators: Can steal tech when conquering base. (Interesting for human players too)
Changed Tech focus from Military to Tech.
Changed AI focus from Probe to Growth.

Gaians:
Removed Morale and Police penalty.
Increased Planet bonus to +2. (Partially offsets morale increase for all factions at start, prevents AI Deidre from running into too much eco damage from her rampant terraforming)
Reduced Efficiency bonus to +1.
Added, Ecological Affinity: +1 Minerals from fungus squares. (Makes for a more unique experience and allows AI Deidre to sprawl more effectively, replaces Nutrient bonus)
Changed AI focus from Growth to Military and Tech.
Changed AI focus from Planet to Industry. (:xD)

Hive:
Removed Economy penalty.
Removed starting perimeter defenses.
Added, Self-Sufficient Bases: Free Recyling Tanks at every base. (Might appear overpowered, but this allows the other AIs to overcome the Hive more easily early game)
Changed tech focus from Military and Wealth to Tech and Growth.
Changes AI focus from Police to Economy.

Morgan:
Removed Support penalty.
Increase Economy bonus to +3.
Removed crippling Pop restriction penalty.
Removed energy bonus.
Added, Economic Development: Free Energy Bank at every base.
Changed Tech focus from wealth to Tech and Growth.
Changed AI priority from Economy to Efficiency.

Peacekeepers:
Removed Efficiency penalty.
Gave new +1 Efficiency bonus.
Starting tech changed to Information Networks.
Added, Planetary Archivists: Gains any technology known by 3 other factions
Changed Tech focus from Tech and Growth to Military.

Spartans:
Removed Industry penalty.
Increased Police bonus to +2.
Added, Organised Militias: Free Perimeter Defense at every base.
Added, Well-Prepared: Free former at start.
Changed Tech focus from Military to Wealth (Together with Tech).
Changed AI focus from Morale to Growth.

University:
Removed Probe penalty.
Increased Research bonus to +3.
Removed crippling (for AI) drone penalty.
Starting tech changed to Biogenetics.
Changed Tech focus from Tech to Wealth and Growth.
Changed AI focus from Research to Industry.

Social Engineering Changes:
All penalties removed.
Frontier gives +1 efficiency.
Simple gives +1 Economy.
Survival gives +1 Morale. Mostly useful to the AI early game as it makes poor efforts to avoid or first strike mindworms and fungal blooms.
Police State changes to 2 Police, 1 Support, 1 Growth.
Fundamentalism has an additional 1 Growth.
Free Market gives +1 efficiency.

Unit changes:
Colony pods cost reduced by 50%.
Crawlers cost reduced by 1/6th.
Transport cost increased by 100%.
Crawler cost increased by 60%.
Supply Crawlers moved to Industrial Nanorobotics. No more early game crawlers. (Where they were most effective)
Colony pods, formers, crawlers and transports now use PSI by default. Makes them survive a bit better early game (in combination with the SE changes, as the AI is bad at avoiding native life) and prevents some cheesy late game tactics like walls of armoured foils or crawlers from being effective.
Unconventional warhead (Inc. Planet Busters) base cost reduced to 18. AI will gladly spam planet busters now on higher difficulties.
Conventional warhead cost reduced to 12. Missiles are most effective against the highly elite units mostly favoured by human players, while they are largely ineffective against the AI's spam tactics.

Ability changes:
Polymorphic encryption, super former and fungicide tanks are free now for all units.
Gas pods cannot be fit on choppers now.

Tech changes:
DocLoy is a now a "starting" tech (no prerequisites)
All starting techs are prioritized for all AI types. (Meaning the AI has a better chance of getting vital technologies such as ecology early)

Buildings changes:
Creche, Network Node, Recreation commons and Energy Bank now have no maintenance.
Creche and recycling tanks are cheaper.
Merchant Exchange available for everyone from start.
Virtual World, Ascetic Virtues, Planetary Datalinks and Planetary transit changed to "early" projects.
Virtual World now enabled with Psych.
Planetary Datalinks enabled with InfNet.
Planetary Transit enabled with Mobile. (Doctrine: Mobility)
Ascetic Virtues enabled with DocLoy.
Planetary Energy Grid enabled with Indust.
Command Nexus enabled with Physic.
Voice of Planet base cost reduced to 10. Giving players much less time to prepare for the ascend if they are caught off guard.

Terraform changes:
Boreholes only give 4 minerals and energy. Cost reduced by 4 to compensate slightly.
Forests only give +1 food and +1 mineral initially. Tree and Hybrid farms are now specifically required to run a forest economy.

Text changes:
Dialogue options and content made a bit darker for certain factions to reflect the extremely aggressive AI.
Help entries edited to reflect changes.
Difficulty level names changed to (From Easiest to most Difficult): Drone, Engineer, Thinker, Transcend, Planet, Diety

Other changes:
Retooling is now free for the first 30 minerals.

Known problems:
Deidre refuses to switch to Green. ( )

Future:
I'm not sure about morgan's +3 eco boost as the AI rarely (efficiently) uses hurry. I might change it to +1 IND & +1 ECO instead as minerals are generally more important to the AI. Also with such aggressive AI commerce is a lot less useful.
Change Yang's starting tech. Even with nerfed Police State he still feels a bit overpowered, especially now when he starts off with recyling tanks. I might just switch them around with the Spartan's.
Move more techs around to making it harder for human players to beeline for certain powerful techs.
Early artillery enabled with Physic. I holding back on this for now because early artillery access can lead to artillery sprawl, which negatively affects the AIs offensive capabilities.
Reduce cost of later game wonders and move them earlier into the tech tree, assuring the AI gets an easy chance at those as well. Especially the Universal Translator.

[Meow:3]

Been a few hours but I already found some issues. 

Changelog for V1.1:
Merchant Exchange available for everyone from start.
Planetary Energy Grid available with Indust.
Gaian nutrient bonus in fungus squares removed. Will still gather minerals. (2+ Nutrient made player sprawling too easy)
Crawlers now properly enabled with Industrial Nanorobotics.

If anyone has suggestions let me know.

Download as an attachment below.

[Buster's Uncle]

Is this an .exe mod or what?

[Meow:3]

No, just the text files (Alphax, Helpx, and the 7 faction files).

The idea is to balance the game around the AI to give the player more of a challenge.

[Nexii]

I don't think the AI's play is really that much tied to SEs.  I play with bigger penalties and the AI plays with about the same efficiency.  If you take a game past 2250 you'll usually see the AI hardly puts much in research...its rare to see an AI tech well enough to Ascend.

Aggressive AI helps as it makes the AI expand more (horizontal expansion >> vertical until higher terraforming comes out).  And it makes the AI build units with attack rather than useless 1/X garrisons.  Unfortunately super aggression isn't good play.  Pacting is much better for the commerce...optimally the AI would be more open to pacts including the human player if they are losing.  But it plays the opposite.

The terraforming (or lack thereof) is probably the thing that holds the AI back the most past the early game.  Early game the AI doesn't do quite as bad since unterraformed land is not as big a difference from terraformed.

I don't think Forests at 1/1/0 would ever be planted.  Yang and Morgan look much stronger than the others..

[Dio]

I presently believe, unless future evidence points otherwise, that the poor AI terraforming behavior stems from the fact that it employs many of the default options from terraforming action list (e.g. Farm + Solar + Road or Mine + Farm + Road). I would need to look more into this particular area before I could provide additional information. I also know that the AI only cares about postive ECONOMY bonuses through +4 from Social Models (this provides the maximum energy bonus at each base besides the commerce bonus of +5). It appears the +5 Economy only relates with the Morgan AI.

[Nexii]

The AI does prefer farm/solar and farm/mine over forest.  But even more than that a lot of times a lot of the land doesn't even get terraformed.  It generally doesn't build enough formers and tends to go a bit heavy on frivolous things like extra roads, sensors, bunkers.  So even if you mod to encourage farm/solar the AI still doesn't do great because unterraformed land is really bad. 

[Harpalus]

Looks interesting. Do you have any plans on tweaking the SMACX factions? I doubt the Progenitors need much, but I can't imagine the Cultists in particular doing well as AI compared to your improved starting 7.

[vonbach]

Interesting. I have done something similar with the believers I just gave them techsteal and immunity to probe teams and lowered their research penalty and added a +GROWTH. Now she's even more of a nightmare to deal with. With the Spartans i just gave them -planet.
Any thoughts on the new factions?

[MercantileInterest]

This looks really great. Would you mind my incorporating it into the Binary Dawn mod? Or, better yet, would appreciate your help. I'm looking to create an optional folder full of AI factions, which could be copied into the game when needed.

[vonbach]

I've tried some of these ideas. Like removing the government penalties and some of the facility maintenance costs.
It definitely has an effect on the game and helps the AI.

[realjackryan]

Noob here:


First off THANK YOU SO MUCH FOR THIS.

Secondly, I am trying to find the best mixture of patches / mods etc. Quick question about this, when I overwrite the 9 .txt files, won't this apply to those factions in SMAX as well as SMAC? If not can you (if possible) briefly explain why or why not? I would really enjoy this mod for SMAX, especially if it is an easy tweak I can do myself / help with / share with others.

Additionally, how does this effect other patches IE Yitzi patch, Kyrubs SMAC 444 and Scietns v 1.0?

Finally, and I realize this may not be the place for this, but how do these partches affect each other, and are there a standard set of patches people use (primarily for multiplayer games to get everyone on the same version?)

Thanks a ton for the awesome patch, anyone's help with my questions and ya'lls dedication to the community!

[Buster's Uncle]

MPlayers seem to be going with the latest YitziPatch, v.3.4.

[BFG]

Wow, wish I'd seen this sooner as I was complaining in another thread about the lack of AI challenge.  Not sure how I missed this.

I won't have a chance to actually try this for a while, but it sounds impressive and you've certainly done your homework.  One modest suggestion: I have found that the AI can be quite a bit more effective if they can learn how to use Aero Colonies effectively.  Might be something to look into.  (I also have a pet peeve that Automate Grav Formers / Grav Supplies doesn't work correctly, but that's a different matter...)