1. We have added the ability to collapse/expand forum categories and widgets on forum home.
    Dismiss Notice
  2. All Civ avatars are brought back and available for selection in the Avatar Gallery! There are 945 avatars total.
    Dismiss Notice
  3. To make the site more secure, we have installed SSL certificates and enabled HTTPS for both the main site and forums.
    Dismiss Notice
  4. Civ6 is released! Order now! (Amazon US | Amazon UK | Amazon CA | Amazon DE | Amazon FR)
    Dismiss Notice
  5. Dismiss Notice
  6. Forum account upgrades are available for ad-free browsing.
    Dismiss Notice

Your AI Opponent - by Ision

Discussion in 'Civ3 Strategy Articles' started by Ision, Jul 3, 2004.

  1. Ision

    Ision Master

    Joined:
    Mar 8, 2003
    Messages:
    452
    What follows is my attempt to help players in choosing and understanding their AI opponents. In CIV III, THE single most important factor for any AIs overall performance is the value of their start location and the immediate vicinity. That said, given a sufficient number of games there are certain trends that begin to bear themselves out. All things being about equal (start location, river access, resource availability, ect) the AIs will perform differently. There are 2 primary reasons for this; their traits and their Unique Unit – the former being far more important on average. Unlike the human player, AI play is wholly scripted and therefore incapable of making the proper adjustments to changing situations. As such, certain traits lend themselves better to the scripted play of the AI than others, and thus produce the ‘on average’ better performance of some AI civs over others. Let me state from the onset, that the observations that follow are just that – ‘observations’ – and not hard and true outcomes to be expected in any particular game.

    The Best of the Best (they rarely disappoint):

    Agricultural Civs – the AI will typically begin building a settler with no regard to whether their food/growth output is sufficient enough for the settler to actually be produced within the number of turns the pop-growth/shield output dictates. An AI city waiting innumerable turns without actually building anything is a common site – its simply waiting for the food/pop-growth ratio to catch up to the point where it can finally crank out a settler. The Agri trait mitigates this flaw by giving the AI the extra food/growth bonus that allows for more rapid settler production than other AIs. The net result is a trend for Agri Civs to outperform other civs in expansion – and thus establish themselves more dominantly early.

    Scientific Civs – while Scientific may rate low among human players, it certainly helps the AI. The scientific trait gives the AI that slight little boost that keeps it ahead of the other AIs in research – that slight boost is later accentuated by the faster production of half-priced libraries. That ‘freebie’ tech at the beginning of each age given by scientific is far more beneficial to the AI than the human, and often cements the scientific AIs position of dominance over the other AIs upon entering the Industrial Age. Great Wonders are also far more beneficial to the AI than the human, thus the higher chance of a SGL helps the AI as well. The net result is that these AIs will typically have the tech lead and the consequent overall gold lead that brokering these techs gives them. Lastly, there are an unusually high number of the better UUs possessed by the 9 scientific AIs (Hoplite, Immortal, Enkidu, Droman, Sipahi, Cossack, Bab-Bowman), which tends to give them yet another advantage.

    Industrious Civs – in previous versions of CIV III (prior to C3C) this trait stood alone in a category by itself. So, profound was the advantage of this trait for both AI and human player alike! This is no longer the case; the toning down of the worker speed from a 100% to 50% advantage has brought industrious back to earth. Nevertheless, the trait still helps the AI greatly. The bottom line is this – the AIs never-ever build anywhere near enough workers. As a consequence their reliance on the few workers that they do build is huge. The industrious civs faster workers go a long way towards addressing the AIs unwillingness to build more workers. Further, Industrious civs that lose 1 or 2 early on are not as devastated as other civs. The higher shield output of the typical Industrious AI will manifest itself throughout the entire game. Lastly, a few of the Industrious civs actually have the ‘build often’ checked for workers – this may also result in greater overall worker numbers.

    The Runner Ups (boon or bust)

    Expansionist Civs – typically come out of the starting gate as good or better than other traits. Translating their goody hut/ barbarian advantage into early tech leads and quite often, faster city expansion. A sprawling Expansionist AI by the mid-Middle Ages is common. Unfortunately, from my experience, in spite of their larger size, Expansionist AIs tend to have no staying power and are slowly picked apart from the mid-game on. The factors here are 3-fold; a typically higher than an average aggression level resulting in more wars and thus slowing their infrastructure building, a higher than average number of ultra-early Golden Ages that boost their early age but is for the most part wasted, and lastly the fact that 5 of the 8 expansionist civs are paired with traits that the AI plays poorly. On average outstanding in the first 6 rounds of the boxing match – but without an early knockout (quick overrun of multiple neighbors) - it gets winded fast and craps out at the end.

    Religious Civs – are hard to measure. The lower the difficulty level the greater the value of the AI being religious. The reason is simple; prior to Monarch level the AI does NOT receive any bonus in reducing anarchy turns during government transitions. Changing governments is a favorite AI past time, doing so FAR too often through the course of a game. 3 to 5 government transitions in lower level games can greatly ****** non-religious AI power. On the other hand, at the upper levels religious AIs are only slightly (if at all) faster in recovering from anarchy. The AI is hampered by the fact that it will NOT use the luxury slider to compensate for any happiness issues. The result is a higher than necessary number of ‘entertainers’ in their most productive cities. Religious AIs tend to be happier AIs due to the cheaper temples and cathedrals. As a consequence it has been my experience that IF they survive, they tend to develop a more solid infrastructure from the mid-game on and become more formidable in the late game. AI performance with religious tends to run the extreme ends from terrific to horrendous.

    Commercial Civs – like religious civs tend to run the extreme ends of performance as well, but even more so. The AI cannot use this trait to its best advantage except by sheer accident. It has no concept of the advantage of a huge commercial/communist empire with its lower OCN. A commercial AI rarely translates its inherent gold advantage out properly, far too often the commercial AI acts like 19-year-old, pro-athlete with his first million-dollar check – he wastes it. However, the trait works as a terrific amplifier – making those AIs with the ‘best of the best’ traits that much more formidable, while doing little to nothing for the ‘artificially ignorant’ traits. Take a little test; run 10 continents games against the Iroquois or Greece – then run 10 more against India or England with the exact same starting positions - enough said.


    Artificial Ignorance (subtlety is NOT an AI strength)

    Seafaring – is a trait that in the hands of the human is as deadly as any. Unfortunately in the hands of the AI – the results are staggering underperformance. The Seafaring advantage of an extra movement point at sea is wasted by an AI that will NEVER attempt a suicide run to find another land mass, an AI that WILL build a good sized fleet and then retreat nearly every unit back to port that loses even a single hit point, an AI that may field a huge ground force – and yet invades with a mere fraction of the force required for successful amphibious operations. Outside of arch maps – seafaring makes the AI appear to be ignorant to the point of the ridiculous. The subtle play as well as the far range strategic planning required to make seafaring work are simply beyond the AIs programming.

    Militaristic – is seafaring’s ugly twin sister. The ability to generate 2 or 3 early great leaders by a human player may result in a game breaking moment! For an AI it means 2 GLs used to rush a temple and a library, with the 3rd used for an army loaded with 1 defensive unit and 1 offensive unit doing garrison duty in a small town! Yes, yes – from time to time the AI will actually field an Army and may even accidentally use it well – but those times are as rare as a white rhino. The early archer rush spring-boarded by those cheaper barracks, the pillaging strat, the artillery red-line strat, selective elite unit use – all of these are all beyond the AIs ability, and while this effects all the AIs militaristic or not – it especially effects the performance of the AIs whose traits is supposedly geared to benefit from war. Here again is a trait that requires subtlety and far range planning to make-work – both are beyond the AI.

    The AI & UUs

    The AI loves to escort units and to bring a great number of defensive units with any offensive force. The result is a much higher than average use of defensive units in offensive roles as compared to a human player – with defensive units having a greater value for the AI than the human. Secondly, the AI is big on having a heavy number of garrison units in each and every one of its cities – the result is once again a greater focus on defensive units and a greater value to the AI than the human. The AI handles offensive units very poorly due to the ‘piecemeal’ method in which it attacks. This lack of ‘concentration of forces’ gives the AIs with the defensive UUs an edge over the offensive AI UUs. The last category are the ‘specialty’ UUs, those requiring subtle play, planning, or an indirect approach to maximize effectiveness: in this the AI is an abject failure – completely wasting a UU like the Hwach’a (the AI has no concept of proper artillery use), and making terrible use of UUs like the Jaguar, Chasqui, Beserk and Man O’ War.

    The Strongest AIs

    Unlike the human player that has successfully engaged in early wars - the AI does not benefit nearly as much - initially. The reason is that AI vs AI wars are ‘wars of attrition’ – they last far longer than human wars, and take a tremendous toll on the AIs including the winning AI! Having put all their builds and commerce into unit production; it often leaves their infrastructure building well behind that of other AIs, a large trade debt, their military reduced to a fraction of what they need, and commonly they are well behind the other AIs in tech research. So in spite of they’re having the largest total city count, they are far from the strongest civ. What this means for the human player is that your typical method of choosing the AI to go to war with (the - whose the highest AI in total score method) is quite often the wrong approach. You will find more often than not that that nice, and until now, peaceful mid-sized AI 3rd or 4th on the score list – is often the single most dangerous AI in the game. Their infrastructure is deep, their units numerous and up to date, and their ability to form alliances from an unsullied reputation is still strong. Keep this in mind when trying to assess who the most powerful AIs really are. Naturally, you can not lose sight of the fact that that much larger AI, given a long enough period to recuperate – will catch up to and eclipse all the other AIs.


    Conclusion: Let me repeat it one more time: “In CIV III, THE single most important factor for any AIs overall performance is the quality of their start location and the immediate vicinity.” However, my personal experience playing numerous games against every civ has lead me to certain conclusions; the more direct and obvious the benefit of the trait the better the AI will perform on average, the more subtle the trait the worse the AI will perform. The more defensive the UU the better the AI handles it – if the UU requires anything other than a ‘direct’ approach to be used well – the AI will bungle it. Unlike its human counterpart the AI is almost always incapable of ‘coming back from behind.’ For the AIs, the rich get richer and the poor get poorer with each passing turn. While both the AI and human share in the advantages of the ‘REX’ traits like Agri/Ind/Exp they part ways to varying degrees with the other 5 traits. Last but not least, do not forget that the AI plays the game strictly ‘one turn at a time’ re-computing its strategy each and every turn. In essence it has no real ‘strategy’.

    Ision
     
  2. Longasc

    Longasc Chieftain

    Joined:
    Jun 18, 2003
    Messages:
    2,763
    Hi Ision!

    I share your observations, however I miss the aggressiveness as an important factor.

    Germans, Mongols and Celts e.g. - they are very aggressive, you can only see this slider in the editor.

    But they all have a tendency of conquering their neighbours early on or well, produce military units, fail the attack and descend into the third league and are usually of no major importance anymore, weak nations.

    I think it would be good to take this into the article - perhaps a ranking of civ aggressiveness level. Everyone could take a look at this, but it would fit great into your article.

    People with a knack for history probably know that the Romans are likely to go for war in self-defense... ;) but well...

    Here is the list for C3C:

    Aggressiveness Factor:
    5: Germany, Zulu, Mongols
    4: Rome, Russia, Japan, Persia, Aztecs, Vikings, Celts, Arabia, Hittites
    3: Egypt, Greece, Babylon, America, England, Spain, Ottomans, Netherlands, Portugal, Byzantines, Inca, Maya
    2: China, Iroquis, Carthage, Korea, Sumeria
    1: France, India

    It should also be noted that some civs have production priorities regarding military land units, e.g.: Celts, Mongols, Germany, Zulus... the more aggressive ones.

    Favoring military units and high aggressiveness makes for early wars that are either boon or bust for this Civ.
     
  3. Pfeffersack

    Pfeffersack Chieftain

    Joined:
    May 10, 2003
    Messages:
    2,169
    Location:
    Germany
    I agree on the evaluation of the traits, but ironically the civ which has the both worst "AI"-traits, the Vikings, is one the AI plays very well.The Vikings are often one of the superpowers in my games.I think it because of their fantastic UU (despite its offensive kind), they are often sucessful with there middleage wars.

    And as Longasc has already mentioned, the building preferences may have an influence, too.The Vikings for example have only naval and offensive land units - this may fight a bit against the tendence to overbuild defense units and with only two set flags they don't lose their impact.
     
  4. Tranquility

    Tranquility Grand unifier

    Joined:
    Jan 18, 2004
    Messages:
    91
    Location:
    Mascouche, Quebec, Canada
    Intresting post. This is gonna help me alot depecting my enemies.

    Your right about the GL thing. I,ve only seen twice the A.I. use to make a full army. And there we're all civs with very high agression(the zulu and the aztecs).

    Jsut out curiosity, do you play mostly solo civ? Or are ya a muiltiplaya.
     
  5. Ision

    Ision Master

    Joined:
    Mar 8, 2003
    Messages:
    452
    Thanks for your input gentlemen, and for that listing of the specific aggression levels of the Civs.

    One note however, the 'build prefrences' have little overall impact on Civ performance. This particular factor has been grossly over-rated. My comment on build preference for workers by a few Industrious Civs was made only because low worker production is such an enormous AI handicap that even 1 or 2 more during the entire game can often make a great difference. That said, I have had extensive expierence with modded games addressing the build preference issue and have found that the impact was overwhelmingly modest - so much so that the trait & UU factors completely overshadow it as a factor.

    Ision
     
  6. axxessdenied

    axxessdenied Chieftain

    Joined:
    Nov 26, 2002
    Messages:
    12
    Location:
    Waterloo, Ontario
    I always notice Sumeria to be the #1 civ. The combo of agri and science for the AI is just a seriously good combo for some reason.
     
  7. yankees

    yankees Chieftain

    Joined:
    Jan 4, 2004
    Messages:
    48
    the sumers have all the things ision was talking about. scientific and agri and a early defensive uu that is cheap. this article proves many things i notice before. i will use it alot when i play on pangia maps. thanks ison!

    Y
     
  8. nullspace

    nullspace Chieftain

    Joined:
    Aug 20, 2002
    Messages:
    728
    Very nice article. It makes sense that the traits that get used automatically (like Agri and Sci) are best for the AI, but traits like Mil and Sea don't do much for them. Good comments on the UUs too. This info would also be good for modders who are trying to balance civs against other civs.
     
  9. Sparta

    Sparta Chieftain

    Joined:
    Sep 25, 2003
    Messages:
    1,256
    Great article, Ision! Thanks! :goodjob:
     
  10. Drakan

    Drakan Voluntas Omnia Vincit

    Joined:
    Jun 15, 2004
    Messages:
    1,370
    Location:
    Bloomberg
    Ision, once again, a great article.xnks

    "Yes, yes – from time to time the AI will actually field an Army and may even accidentally use it well".

    In C3C? The AI uses them very well in the Scenarios, but after months playing C3C, I haven't seen any AI Armies in the core game yet. As a matter of fact I started a thread on that. An AI Army is a rara avis in the Battlefield, more like a unicorn I'd say in C3C.
     
  11. Ision

    Ision Master

    Joined:
    Mar 8, 2003
    Messages:
    452
    rofl -

    Ision
     
  12. bradleyfeanor

    bradleyfeanor Chieftain

    Joined:
    Jul 25, 2003
    Messages:
    676
    Location:
    St. Louis
    Excellent analysis Ision. I have noticed the advatage of the Agricultural civs, but I never made the same connection for scientific civs. In thinking back on many of my games though, you are certainly correct. Your breakdown of civ traits will help me pick my early victims more wisely in the future.

    Thanks!
     
  13. KayEss

    KayEss Gone to the dogs

    Joined:
    Feb 1, 2004
    Messages:
    227
    Location:
    Bangkok
    Of course the 'strategy' here is the weights that it uses to determine what to execute so it will seem like it is working to plan. This can also make it seem more determined than many human players - especially those who are new to the game.

    It also allows it to abandon its current plans when something drastic happens - I've noticed many human players aren't ready enough to change their plans when the situation changes. They put too much value into sunk costs and not enough into future costs (both real and opportunity). It's all about balance.
     
  14. Bibor

    Bibor Doomsday Machine

    Joined:
    Jun 6, 2004
    Messages:
    2,300
    Location:
    Zagreb, Croatia
    As a matter of fact, I've run an analysis on my Bibor's(TM) Supercomputer(TM).
    It says that there's a greater chance for a unicorn(TM) appearing on battlefield than an AI army.

    :lol: :crazyeye:

    -bibor

    All trademarks are owned by their respeted owners. Bibor's, and Supercomputer are a registered trademark of Bibor's Living Quarters. Unicorn is a registered trademark of Germanic Tribes Fairytales Inc.

    ***

    To ision:

    I have a question about this great article:

    How would a game vs. AI look like if all the civ's (except the human players') would be set to a combination of the first three traits? Would that mean that the actual difficulty of play against them would be much higher (making a sid difficulty game tougher than sid)?
    This would mean that in a "vs. 12 AI game" there would be very few civs that are technologically backward vis a vis their "fat neighbours".

    -bibor
     
  15. Ision

    Ision Master

    Joined:
    Mar 8, 2003
    Messages:
    452
    Natuarlly it would be harder on the human - although I do not think it ranks with being 1 full level higher.

    Ision
     
  16. Ision

    Ision Master

    Joined:
    Mar 8, 2003
    Messages:
    452
    I would like to add one more comment on AI performance that I should have included in the original article (I have edited the article to include the below paragraph):

    Unlike the human player that has successfully engaged in early wars - the AI does not benefit nearly as much - initially. The reason is that AI vs AI wars are ‘wars of attrition’ – they last far longer than human wars, and take a tremendous toll on the AIs including the winning AI! Having put all their builds and commerce into unit production; it often leaves their infrastructure building well behind that of other AIs, a large trade debt, their military reduced to a fraction of what they need, and commonly they are well behind the other AIs in tech research. So in spite of they’re having the largest total city count, they are far from the strongest civ. What this means for the human player is that your typical method of choosing the AI to go to war with (the - whose the highest AI in total score method) is quite often the wrong approach. You will find more often than not that that nice, and until now, peaceful mid-sized AI 3rd or 4th on the score list – is often the single most dangerous AI in the game. Their infrastructure is deep, their units numerous and up to date, and their ability to form alliances from an unsullied reputation is still strong. Keep this in mind when trying to assess who the most powerful AIs really are. Naturally, you can not lose sight of the fact that that much larger AI, given a long enough period to recuperate – will catch up to and eclipse all the other AIs.

    Ision
     
  17. Pfeffersack

    Pfeffersack Chieftain

    Joined:
    May 10, 2003
    Messages:
    2,169
    Location:
    Germany
    Good point.It applies in my experience to modern wars, too.Despite the AI is able to fight more effective against rivals behind in tech (in my games, most AI civs are eliminated in the industrial or beginning modern era) because of the greater differences in weapons, it fails in using the conquered lands.In most cases some AIs fight against a single one and the result is a patchwork land.The AI is rarely able to develop those areas; the strongest reason is its fatal leaning towards Fascism.
    AI civs in communism gain more advantage and might get more dangerous after a won inter-AI-war.
     
  18. nullspace

    nullspace Chieftain

    Joined:
    Aug 20, 2002
    Messages:
    728
    Yeah, that is ture. In AI vs AI wars, both sides lose their offensive forces quickly. That's a good time to attack them. On the other hand, and AI that hasn't fought for a long time has had plenty of time to build up offensive units. Since those have just been sitting around in cities, the AI does a good job of upgrading them, too.

    It seems to me that the very aggressive civs don't last long in the game. They declare war often, and spend most of their time building units so their infrastructure is terrible. They might get a large empire in the ancient and middle ages, but most cities will be missing basic improvements. Once the peaceful civs build factories and hospitals, the warmongers will still be working on marketplaces and aqueducts. Their frequent wars also make many enemies, who ally up later and beat them with superior units. I think that Germany is the only aggresive civ that I've ever seen do well through the industrial and modern age.
     
  19. Pfeffersack

    Pfeffersack Chieftain

    Joined:
    May 10, 2003
    Messages:
    2,169
    Location:
    Germany
    :eek:
    My experience is far from being representative, but Germany is in my games the civ the AI plays worst.Most times the are eliminated in the ancient age or middle ages, sometimes they survive until middle of the industrial ages.But apart from the very early beginning, they are never really dangerous.There infrastructure is bad and they usually have lost some wars, which means they have only a few cities.I have never, really never, seen their Panzer UU in action if the AI plays Germany...
    I think its the combination of to high aggresssion level, late UU (and golden age in most cases of course) and the traits.Militaristic supports their warmongering, but it is in general not a strength in the AIs hand.Scientific is nice, but only if a civ makes use of it - here it is complete waste.They rarely make use of the cheaper buildings (because they are at war most of the time), they fall behind in tech (and don't get SGLs) and they rarely get more than one free era tech from it.
     
  20. dgfred

    dgfred Sports Freak

    Joined:
    Apr 21, 2004
    Messages:
    7,256
    Location:
    N.C., USA
    In most of my games Germany is either weak or destroyed by game end :( :confused: . That is unless I am the Germans ;) :hammer: .
     

Share This Page