1. We have added a Gift Upgrades feature that allows you to gift an account upgrade to another member, just in time for the holiday season. You can see the gift option when going to the Account Upgrades screen, or on any user profile screen.
    Dismiss Notice

AI much too eager to pillage

Discussion in 'Civ4 - Strategy & Tips' started by morchuflex, Mar 13, 2006.

  1. morchuflex

    morchuflex Emperor

    Joined:
    Feb 19, 2004
    Messages:
    1,385
    Location:
    Paris
    Hello.

    Does anyone but me think the AI puts way too much emphasis on pillaging? This was already the case in previous Civs.
    In a recent game, the Chinese declared war on me. I was terrified when I realized that not only did they outtech me (I had only archers and axemen, they had horse archers, swordsmen and catapults), but they also outnumbered me at least 5 to 1 (being boxxed in a small area, I guess they had nothing to do but build troops...).
    But instead of launching a full-scale attack, they spread their forces very thin and sent them in small groups or even one by one to pillage my cottages. This allowed me to regroup and take their horse archers one by one with only a few spearmen. After about 30 turns of war, the cost of the war amounted to 5 tiles pillaged and one spear killed on my side, and... 20 horse archers killed on their side, plus various other units.
    And a few turns later, I captured Beijing. Go figure.

    Of course, pillaging is a good way to cripple your opponent, but when you're powerful enough to crush your enemy, you should do so immediately, shouldn't you?
     
  2. Smokey McDope

    Smokey McDope Warlord

    Joined:
    Feb 9, 2006
    Messages:
    134
    To answer your question, yes. But, the AI will always pillage everything it can in your borders. Can't seem to grasp the concept that if it takes the city, those will be his tiles. Oh well, makes for some easy counterattacking like you said.
     
  3. Kurbads

    Kurbads Chieftain

    Joined:
    Jan 31, 2006
    Messages:
    50
    Location:
    Riga
    If AI was smart in general there would be no players above Noble, and the ones in Noble would calculate moves in excel spreadheet to have victory chances, not to say that less people would like the game in general.

    Is the AI stuped because its currently no programmable or too expensive laborwise to program it to be smart, or game gesigners make it a bit stuped on purpose to keep interest in game - the question to dispute.

    Basically if AI would be avarage human level smart, plus its natural 100% knowledge of game itself, it would be imposible to win. Thats I quess is why in all games so far AI is stuped.

    so live with it and enjoy it! this is a geme, the aim is You to have good time, and feeling smarter than somebody else - AI this time - is core reason we are all so addicted here.
     
  4. Roland Johansen

    Roland Johansen Deity

    Joined:
    Apr 29, 2003
    Messages:
    4,292
    Location:
    the Netherlands
    I agree with the original poster that the AI pillages too much. It leaves its troops very vulnerable and pillaging is not that useful, especially when the one that is being pillaged still has a large area behind the battle zone that isn't being pillaged. The loss of a tile improvement of the enemy isn't worth the same as the loss of a unit. And a pillaging unit generally dies.

    If the AI would use small mixed groups for pillaging duties, then it would work better. Those small mixed groups would not be slaughtered so easily. Also the AI should focus on capturing cities if it can do so. It shouldn't pillage tiles of a city that it is going to capture.

    And to answer the last poster (Kurbads). I don't think it is possible to programm an AI that can beat the best players at noble difficulty level. The AI will still need bonusses to beat the best humans until we develop an AI like in the Terminator films.
     
  5. Sisiutil

    Sisiutil All Leader Challenger

    Joined:
    Feb 19, 2006
    Messages:
    6,899
    Location:
    Pacific Northwest
    The AI seems addicted to pillaging the way many Civ veterans seem addicted to Wonders.

    I suspect the AI has been programmed to overestimate how hard it is to take a city in Civ IV (admittedly, it's not easy, especially with a hefty culture defense bonus--catapults and their later kin are essential). I also think the AI has been programmed to overestimate the effects of pillaging. (I saw one tip on the opening screen urging me to pillage towns, which is sheer lunacy if you plan to capture the nearest city.) And I also suspect the AI has been programmed to avoid the "Stacks of Doom" that are vulnerable to collateral damage but essential to capturing cities.

    I base these suppositions on what I've read of the designers' intentions. The programmers seem to have programmed the AI to play they way they THINK the game SHOULD be played, at least in this regard, rather than how experience has shown the game IS played.

    All of this means the AI is less than effective at warfare than a human. The AI just seems too risk-averse; or to put it crudely, the AI has no balls.

    Of course, I say that, and tremble at the thought of the Firaxis programmers taking me at my word...
     
  6. senwiz

    senwiz Prince

    Joined:
    Feb 27, 2006
    Messages:
    304
    Location:
    Appleton, WI
    The AI was programmed once, over a preiod of months. Human players have the ability to read through the tactics at a given time, where the computer can't make changes to the their strategy at any time. It's the reason why comptuers will always bow to humans. It's not that we outthink computers, because they think much faster than we do. It's that our brain has the ability to see a move or two in advance and determine the best change of strategy to make at the time. And we do it ALL the time. :)
     
  7. Wreck

    Wreck Prince

    Joined:
    Jan 4, 2006
    Messages:
    331
    The problem here is that AI is hard. There are many hazy concepts that are very difficult to make concrete enough for a program to use. (One of them is "mass" in the military sense, which is what we're talking about here.) Human being intuitively grasp many such hard notions, and don't even know they're hard. But try to program "mass" into a computer!

    I do think the AI could be better, and I've sort of been hoping to play with it when/if that becomes possible. However I'd guess that will not be possible to fundamentally increase the I of the AI; rather one might be able to program in a few more useful heuristics, such as: don't attack an enemy city if there are more defending units there than there are attackers, or, don't attack if the odds of winning are less than 50%, without twice as many attackers as defenders. That sort of thing.
     
  8. DynamicSpirit

    DynamicSpirit Fear him of the pink tie Moderator GOTM Staff

    Joined:
    Dec 23, 2005
    Messages:
    5,622
    Location:
    London, UK
    Hmm, judging by the number of times I've seen the AI throw a stack of 4-6 units at a city that it'd need at least twice as many units to stand any chance whatsoever of taking, I'm not so sure about that :)

    There is a really annoying thing about pillaging though - you don't get a choice of what to pillage. So often I end up pillaging the improvements (which I know I'm going to have to rebuild in 6-7 turns time) because you have to pillage them before you can remove the road, which I need pillaged to stop the AI circling round me or quickly bringing in reinforcements.

    Actually thinking about it I wouldn't be surprised if Firaxis did that deliberately to stop people from just always pillaging every road in site, and to make the decision of whether to pillage more complex.

    I do tend to draw the line at pillaging towns to get to the roads though. ;)

    I suspect there's a simpler explanation: Conducting warfare intelligently requires you to see the map in 2 dimensions and actually think in those two dimensions. That's something that comes very naturally to our brains but it's incredibly difficult to code up in computer programs. I bet accounts for a lot of the silliness you see in the AI strategy.
     
  9. Fetch

    Fetch When in doubt, reboot.

    Joined:
    Jan 26, 2006
    Messages:
    615
    Location:
    Statesboro, GA
    see in 2D? What 2D are you talking about. I don't follow.
     
  10. DynamicSpirit

    DynamicSpirit Fear him of the pink tie Moderator GOTM Staff

    Joined:
    Dec 23, 2005
    Messages:
    5,622
    Location:
    London, UK
    I mean the fact that when you look at the map (which is a 2-dimensional map) you immediately get an overall picture of what's happening. You just naturally see the whole picture.

    Computer programs don't work like that. They work more by examining individual bits of data at a time then moving on to the next bit of data. This is all hugely oversimplified, but imagine something like: Instead of seeing the whole map in one go, you were allowed to see one square. After examining that square for a moment, you're allowed to look at a different square of your choice. Then you can look at another one, and so on, until you've decided how to move your units. How easy do you think it'd be to plan out your military strategy doing that? That's probably closer to how the AI has to work. That's very roughly what I mean when I say you can visualize the map in 2d, and the computer can't.
     
  11. Fetch

    Fetch When in doubt, reboot.

    Joined:
    Jan 26, 2006
    Messages:
    615
    Location:
    Statesboro, GA
    oh. 2d map.
    In war: Dim a 2d array with a boolean flag: blnInteresting would be useful in this. Make the array the same size as the map. Apply a pattern finding algorithm to the interesintg tiles and you'll get an idea of the units. Take a before and after turn shot and you've got troop movements. See what cities are along those lines, and you've got something. Easy? Nope. I hate 2d arrays.

    for i
    for j
    ....
    ....
    next j
    next i

    Yuck!
     
  12. Dominus Belli

    Dominus Belli Chieftain

    Joined:
    Feb 26, 2006
    Messages:
    21
    The AI I've seen doesn't avoid stacks of doom at all. In fact, more often than not when an AI declares war on me, I see a huge stack of units next to one of my cities within a few turns. However, I do agree with the pillaging observation.
     
  13. Ray Patterson

    Ray Patterson the dude is not in

    Joined:
    Nov 15, 2005
    Messages:
    592
    Location:
    Not in 'Nam of course
    That gives me an idea... maybe artificial learning techniques should be applied, just let them battle it out and learn from it. Then again it could indeed be that people don't want superior AI. Would be cool though, maybe some sort of "civilization challenge" could be made (you know, like a million dollars for an AI that beats the current AI at Deity).
     
  14. atreas

    atreas King

    Joined:
    Jan 10, 2006
    Messages:
    799
    Location:
    Greece
    I think the game would be challenging, but I'm not sure if I would play it, if the AI was acting like a human. Than means:

    You could never trust anyone around.
    AI would make some movement (sometimes as a kind of "team") just to prevent you from winning.
    Diplomatic victory? Let's laugh all together loud! Only if you already could have domination!
    After 1000BC there would be no trees around, not even for picture.

    And so many other "small" things. It would just be no real strategy - nothing to be able to count on. If you didn't happen to have a UU in ancient age you wouldn't live to see next day, if you happened to be 2 turns late for BW you could safely reload, and all these funny things.

    But it's very unrealistic to think that, if they wanted to, they wouldn't be able to create an AI that would (easily) beat any man on Noble (i.e. with fair odds). Civ is a finite game, and in all these games you can (more or less easily) develop an AI that beats the best of men. The cost in time and money is another story.
     
  15. Roland Johansen

    Roland Johansen Deity

    Joined:
    Apr 29, 2003
    Messages:
    4,292
    Location:
    the Netherlands
    Ha, ha, very funny. Then why haven't they done this with the very very much easier games like chess and go? (easier in the sense that these games have a more limited set of possibilities) They have been making AI's for these games for decades, but still the human mind is better.

    The number of possibilities in a game of go are greater then the number of molecules in the galaxy. And the number of possibilities in a game of go are very much smaller then in a civ game. But computers are very fast so they should be able to calculate this finite game. Yeah right.:lol:

    Another problem with writing an AI for a game like civ is that there is a concept of 'chance' that plays a big role in the game. This makes programming the AI much more difficult and makes calculating things in the game much more difficult. A lot of decisions that you take in the game are based on a (good) chance that it will work. It's very difficult to translate that into an AI that controls a civilization.
     
  16. lutzj

    lutzj The Last Thing You See

    Joined:
    Mar 6, 2006
    Messages:
    1,693
    Location:
    New England
    wat if each character had its own playing style, like ghengis khan likes to pillage and harass while caeser likes to take cities one by one. or ghandi goes all dimplomacy and bismark gets pissed off real easy
     
  17. sydhe

    sydhe King of Kongs

    Joined:
    Nov 4, 2001
    Messages:
    2,426
    Location:
    Norman Oklahoma
    We don't need Bismarck getting pissed off. That's what we have Montezuma and Isabella for.
     
  18. megalomaniac

    megalomaniac Push the red button!!

    Joined:
    Dec 31, 2005
    Messages:
    64
    Location:
    South Korea
    There are two reasons they dont have AI that can defeat the best of men.
    1. A program created by a human cannot be superior in its logic to the human that created it.
    2. It is IMPOSSIBLE to create a perfect strategy for a game that often requires the player to change his or her tactics on the fly to adjust to a new situation. (games as "limited" :rolleyes: as chess or as big as Civ) Artificial Intelligence is just that, artificial. It appears to be intelligent, but it is actually following an extremely complicated set of instructions.
     
  19. Dr Corbett

    Dr Corbett Chieftain

    Joined:
    Jan 19, 2004
    Messages:
    76
    Personally, I think you've misjudged Otto von Bismarck. He's a more interesting character than that, which is what made him historically dangerous.
     
  20. jeremiahrounds

    jeremiahrounds Prince

    Joined:
    Nov 3, 2003
    Messages:
    404
    "1. A program created by a human cannot be superior in its logic to the human that created it."

    I dispute that. The whole idea of adapated AIs is to finess optimality out of a situation that humans arnt able to discriminate in a expressed way. For example humans are horribly bad at writing logic that identifies letters. They can do it. But they cant write the logic that identifies letters. Those are two wildly different things.

    However a human can design an algo that mathematically searches for the logic to identify a letter in a background a noise. Note that the algo being written isnt HOW to identify the letter. Its how to search for the mathematics that identifies the letter. So that when the process finishes running you now have a matrix that takes bitmap inputs and tells you if a letter is in them. And in that process it exceeds the human. Since now in that adapative signal/matrix is something a human has never articulated--namely the logic to identify a letter in alot of noise.

    A rule based system without any sorts of weights to train does indeed reflect the sort of "never" be more then the sum of the knowledge of the person that wrote it. However once the weights or even the interaction of the weights are free then the entire system can become more then the original human was capable of understanding.

    In the end humans claim these innovations as the domain of the tool user. However that doesnt mean in all endeavors you are as capable as the algorythm you write out can become. And that is the ever present and expanding art of AI.

    As for the original topic. I think the problem with alot of these "suck" AI criticism is the alternative "rule" is just to flat to be dangerous. For example when the AI only attacks your unit in their turf predictably when it isnt better defending then you have a situation where they get pillaged to death. And yes i have pillaged empires to death if they refused to come out.

    Take for example one thread above where the person complained that the AI should stay home instead of dropping units off in his turf to take a cheap shot. Thats a horrible rule!

    The worst the worst THE WORST rules for the AI involve being predictable. [though predictable gives character we are talking about power here]. If the AI always stays home if it thinks it needs everything to defend then you will quickly realize that and exploit it. If the AI sometimes does the stupid move and test your defensives and sometimes does the smart move by staying home to defend then in the larger sense the AI is much stronger. In the future you will always no that the AI sometimes drops most of its forces on your shore when its weak and that will force you to deal with that "stupidity" in your defence resulting in you sending less troops off to the fight in the future.

    I guess when people are designing these "Smart rules" they should always maintain that "sane unpredictability" is stronger then "best predictability" every time when designing rules to play against a human.

    Ill give you an example. My radioshack chess game in 1994 didnt play randomly at all. Always the same move in every situation once it got past say move 7. So it was stronger then if it had some randomly made moves right? NO! Only if i didnt play it twice in a row! If i played it twice in a row i had an edge the computer could never had! I knew what it was going to do before it did!

    In the end I learned to exploit this by memorizing a sequence of moves that caused the chess board to [unavoidable] fall into a 2 prong attack loosing fatal material. And all fo that could have been avoided if the computer sometimes played one of many "sane" moves instead of "always the best"

    So! I say to the world before writing "rules" and calling rules stupid remember predictable is always loosing against a human.

    And i think that applies in this case. I think sometimes the AI has to run out and exchange blows on your troops in a position to pillage even if better is defending. It has to to keep you honest and from thinking "those troops are safe because the computer always does the smart thing"

    And alas beyond that i have serious reservations about how smart a defensive posture is. I rarely wait for the AI to pillage me I run out and take him if I think I can do it even if i loose hammers in the exchange. The reason is im buying a "nonpillaged improvement" in what ever the loss of hammers was.
     

Share This Page