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

SOD - Ruins games

Discussion in 'Civ4 - General Discussions' started by ianfuture, Jul 23, 2009.

  1. bestsss

    bestsss Emperor

    Joined:
    Jul 12, 2009
    Messages:
    1,955
    Random is no problem since it has well estimated distribution. Basically one can work w/ the mean or median for simplification. I don't see any algorithmic issue to include random in the calculations. (or make random function distributes in any predictable way, use more than a single seed, etc)

    Exclude the AI bonuses, that create fake difficulty and presently one can hardly think of challenging AI. Imagine in chess starting w/o a rook (or one of the player starting few turns ahead) that's basically what we get at higher difficulties.

    On a flip note: sometimes I dream to have close to unlimited time and no other obligations and try my hand at programing a real AI for a strategic game. No bonuses, no cheating of any kind (no cheap stuff like 'you traded w/ our worst enemy'). Imagine playing vs human except w/ way, way better micro... As it is now (given the inherent bonuses) beating the AI is simply finding its lacks and blatantly exploiting them. For instance winning a culture game w/o enough military to defend, in a normal case (say vs another human) it would end exactly to what happens to the poor AI, a few cities mercilessly razed. (No random involved either for such a game, winning or losing and blaming it on a pure luck is lame in my book).

    On the topic: I dislike SOD since that's virtually the single possible way to wage war, later in the game it's possible to have more than one SOD, though :D
     
  2. PieceOfMind

    PieceOfMind Drill IV Defender Retired Moderator

    Joined:
    Jan 15, 2006
    Messages:
    9,319
    Location:
    Australia
    Variability of results implies branching of decision making though. What I mean is, the AI then needs to know what to do if it loses all its spears, or loses all its catapults in a city assault, for example. Working with the mean result is pretty ambiguous and not likely to be very effective unless you're talking about something like 100 unit vs. 100 unit battles. When it's only 2 units vs. 2 units the variability is greater.

    Firstly, the comparison to chess with the rook is IMO almost silly, again because of the simple fact that chess is a very different game to civ.
    That aside, the AI bonuses are pretty much a fact of life. You're never in your lifetime going to have an AI that doesn't need some bonuses to beat an experienced human player in a game like Civ. And besides, just because it has starting bonuses doesn't mean that is the complete explanation for how it can win. Again, it's not as if it's an accident that the AI can win the game by culture - specifically it was taught how to do it for Better AI in warlords and then ported into BtS.

    It's easy to criticise an AI for its failings but much much harder to make a better one. ;)

    To be fair, you're talking about a strategic element of the game where most already agree that luck or RNG doesn't play a big role. I think you'll find that the failure of the AI to stop a human cultural victory has been considered before, but suggestions to change it have been met with opposition from players who are more after the role playing game. In other words, it's "gamey" for an AI to declare war on you just to prevent you from a cutlure win. Heck, there have been players who have complained about the AI comment made near the end of the game when they refuse to trade techs "We'd rather win the game thank you very much."

    Teaching the AI to prevent a human culture win (or at least try) would actually be on the easy side of things for the AI programmers. As I said above, they just haven't done so because people don't want that. So I don't agree with your argument, or really that your example is a fair criticism of the AI.
     
  3. bestsss

    bestsss Emperor

    Joined:
    Jul 12, 2009
    Messages:
    1,955
    >>You're never in your lifetime going to have an AI that doesn't need some bonuses to beat an experienced human player in a game like Civ.
    That's too bold. I am quite sure it's doable w/ the existing resources. Even though I am not so young any more I preserve my right to disagree {of course a truck can always smash me). I am pretty positive (and I have 2 decades of programming experiences albeit not AI). Developing similar AI would require way more investment though and most likely will not pay off.

    However, you got my points mixed up. (or mostly i expressed 'em so)
    I do not criticize the current AI on its own [not the better AI either]. That's what we got, it can be tweaked but it's fundamentally flawed (I am to live w/)

    The entire luck argument is not towards the game of civilization, it's built in and designed in mind w/... I'd rather like to see a strategy game w/o that. That's all. Most competitive games include little to no luck.

    Comparison to chess is there only to illustrate the unfair advantage the AI gets, playing versus human and having the advantage would certainly result in win but w/o any pride in. It won't be any different in a game of chess w/ the advantage I mentioned. I do not compare the games at all. Civ4 may not run algorithms like chess to think ahead since in chess everything is well visible known.

    Ok, what I'd like to see as AI (just to put in the shape of the civ). Imagine you play vs real human, you don't see diplo screen w/ attitude and so. You don't read the AI mind (we don't like you enough). I did mention about the AI bonuses. No roleplay either (roleplay is good only in sex, imo).
     
  4. Junuxx

    Junuxx Emperor

    Joined:
    Sep 6, 2005
    Messages:
    1,153
    Location:
    the Netherlands
    Programming such an AI system is not impossible, but you don't see them in commercial games because:

    1) It will slow the game down, big time. When they have to wait, people complain.
    2) People (the customers, eh) like to win and to exploit the AI, not the other way around. If you actually make a system that gets the whole picture and carries out a solid plan, if you actually enable it to steal your workers, Quecha rush and betray you for a change, your precious tech lead will come to depend mostly on lucky starts and goody huts. Is that really desirable?
    3) Lack of time and resources. A top notch AI system specially tailored to Civ could easily take several man-years to develop. Sure, it would please hardcore fans, but will it pay off as an investment? If people want humanlike opponents, they can play multiplayer, right?

    What a game like Civ really needs is not optimal AI, but fun, adaptive opponents that give the right level of challenge. Maybe they could even imitate your tactics if you keep stealing their workers, but they shouldn't do anything you couldnt possibly have figured out without being able to do ten billion calculations per second.

    Now I don't believe such a 'fun and smart' opponent would be impossible to program either, but it is a hard and more importantly completely different task from making a perfect civ player. And for such a system, comparisons with chess really lose their meaning.

    Hmmm.. Maybe so, if you're over 70?

    I wish someone in the 50's/60's/70's had kept a list of all the things people said along the lines of "You're never in your lifetime going to have/see/get [something deemed impossible]"

    I'm sure reading that list now would have been a good laugh ;)

    Now that I can agree with :)

    Teaching the AI to axe rush would definitely fall in the same category, btw. It's a really simple sequence of decisions. But most people wouldn't want that either.
     
  5. PieceOfMind

    PieceOfMind Drill IV Defender Retired Moderator

    Joined:
    Jan 15, 2006
    Messages:
    9,319
    Location:
    Australia
    Not impossible, just improbable. ;) Or at least, infeasible, as highlighted by your explanation. It's just a commercial videogame in the end, and I can't imagine it getting anywhere near the sort of AI investment that a well known game like chess has gotten. And chess computers have only recently been able to reasonably outdo the best human opponents if I understand correctly, and chess is much much simpler than civ. Plus there's the speed issue you mentioned, which I too hard forgotten.

    Yeah I could be wrong but it would be bitter sweet! :D I just prefer not to have unrealistic hopes.
     
  6. attackfighter

    attackfighter Emperor

    Joined:
    Dec 25, 2008
    Messages:
    1,010
    Location:
    Intellectual Elite HQ
    i've never even seen an ai come close to a culture victory, and i play up to emperor
     
  7. PieceOfMind

    PieceOfMind Drill IV Defender Retired Moderator

    Joined:
    Jan 15, 2006
    Messages:
    9,319
    Location:
    Australia
    Play a game against Ramessess where you do not cripple him early, and you will have a good chance of seeing an AI go for cultural victory.

    If you were implying that you've never seen an AI get near a cultural victory because you've always won the game by then, then congratulations - you've probably mastered your difficulty level and will probably need to move up a level before you see an AI get close to a cultural victory. :D

    I find at Emperor that in games with about 8 or more players, at least one of the AIs will go for culture, and they usually need to be taken care of with nukes or some such.
     
  8. TheMeInTeam

    TheMeInTeam Top Logic

    Joined:
    Jan 26, 2008
    Messages:
    25,697
    We must be playing different games :sad:. I've seen the AI pull axe rushes against me an other AIs with >10 units by 1000 BC. Somewhat sloppily executed compared to humans that sell their souls to field axes in time, but if that AI gets a city or two the fun may never stop. It'll take peace, redeclare later (before of after feudalism) and either wipe the target out or vassal it when it has 3 cities left or so.

    Humans can block such an axe rush, but if this nonsense happens on other continents the AI that gets big might be troublesome.
     
  9. attackfighter

    attackfighter Emperor

    Joined:
    Dec 25, 2008
    Messages:
    1,010
    Location:
    Intellectual Elite HQ
    I play emperor, and I usually only cripple the civs starting right next to me, so the rest have plenty of chances to go cultural. Sometimes they'll have one or two of their cities with super high culture, but they've never had 3.

    Also, I was responding to the comment that 'the majority of us have probably lost 2 ai going cultural', and my point is that most people play prince and monarch, so the majority of us probably haven't seen ai even go for it...
     
  10. PieceOfMind

    PieceOfMind Drill IV Defender Retired Moderator

    Joined:
    Jan 15, 2006
    Messages:
    9,319
    Location:
    Australia
    In that case I just plain disagree then. I think they go for cultural victory often enough that most civ4 players in the forums here would have at least seen such an attempt. :) I guess it'd be impossible to know without directly polling them though.
     
  11. Negator_UK

    Negator_UK King

    Joined:
    Nov 14, 2007
    Messages:
    758
    Good advice if you have them, but not much use in defence, as if I have that much seige I'm usually on offence myself :D
     
  12. bestsss

    bestsss Emperor

    Joined:
    Jul 12, 2009
    Messages:
    1,955
    1997 deep blue crashed G. Kassparov. First win was in 1996 vs reigning world champion. Yet deep blue was far from a standard computer, by that time (according top500.org) was an extremely powerful supercomputer (ranked 259).

    Since then the evaluation algorithms have been greatly improved and now desktop computers can consistently outperform any human.

    As for axe rushes and any other rush by the AI. W/ the current bonuses in place it is way too unfair if attempted to execute in human alike form vs. the one w/o the bonuses (ai vs ai shall be fair :D ).

    Again, my gripes for the artificial intelligence is the lack of the intelligence in virtually any game. Relying on unfair advantages to balance the scales makes me displeased from scientific point of view. Usually I still enjoy the game, though. But, that's just me.
     
  13. TheMeInTeam

    TheMeInTeam Top Logic

    Joined:
    Jan 26, 2008
    Messages:
    25,697
    If you're in a position where you can be declared on there's little choice but to hole up on a hill city with a TON of units, use siege, or just nail/stymie the problem civ ASAP.

    I've seen experienced players pull worker steal/harass/choke on deity. This keeps jerks like GK, Shaka, etc. in check by substantially slowing their expansion, ability to hook up strategic resources, and get into a position where they can make a good stack in the first place.

    If they don't border you, a declaration usually means you played your cards wrongly.
     
  14. Junuxx

    Junuxx Emperor

    Joined:
    Sep 6, 2005
    Messages:
    1,153
    Location:
    the Netherlands
    Mm, if seen Monty's axe stacks coming for me a few times when I got carried away wonder trippin'.

    But if it's sloppily executed it isn't really a rush, is it?
     
  15. TheMeInTeam

    TheMeInTeam Top Logic

    Joined:
    Jan 26, 2008
    Messages:
    25,697
    That's like saying the AI doesn't play at all if you continue down that logic path.
     
  16. Tatran

    Tatran Deity

    Joined:
    Aug 23, 2002
    Messages:
    4,184
    I've lost a game to Mali.
    First check, the victory screen showed his cities were at 50k (marathon speed)
    Nothing to worry about, enough time to build a space ship.
    Second check, while building the last space parts, the 3 Malinese cities were at 140k. :eek:
    Those cities created ~1000 :culture:/turn
     
  17. Junuxx

    Junuxx Emperor

    Joined:
    Sep 6, 2005
    Messages:
    1,153
    Location:
    the Netherlands
    I don't think so, at least I didn't mean to go on such a path. I meant to say that the AI doesn't really focus on a rush, and doesn't actively attempt to attack as soon as possible. Those seem pretty essential to rushing, the word rush itself seems to imply haste. The bonuses the AI receives just happens to allow them to produce a decent stack pretty quickly if they feel like it, but it will be just an ordinary sort-of-early attack. Of course difficulty matters, I'm only your average Monarch player, not a Deity vet.

    The AI does play, but obviously with other objectives than the human player.

    Really, most people? I think most people by far play in the Warlord-Noble-Prince range... Remember that experts tend to have a higher profile on the forums, and that CivFanatics in general probably attracts more dedicated players on average.
    I don't have any facts on this but I really doubt your assumption.
     
  18. KaytieKat

    KaytieKat King

    Joined:
    Jul 29, 2007
    Messages:
    999
    Hi

    They AI goes for culture wins on noble alll the time. You can tell when its latish in the game and all of a sudden their tech rate just DROPS like a tech that should take em like 20 turns or something goes into like 250 turns and then stays that way. Its cuz they just ramped their slider to 100% culture and is going for the culture win.

    Kaytie
     
  19. Junuxx

    Junuxx Emperor

    Joined:
    Sep 6, 2005
    Messages:
    1,153
    Location:
    the Netherlands
    I agree with Kaytie here. All you have to do to see it happen is, err, nothing. Be inactive enough and it will happen before 2050.
     
  20. TheMeInTeam

    TheMeInTeam Top Logic

    Joined:
    Jan 26, 2008
    Messages:
    25,697
    The easiest way to detect culture attempts is in the victory screen. It will tell you the top 3 culture cities of your opponent...if you're on normal speed and the AI is over 10k with all 3, it's a red flag. Once you get used to the patterns you can see the tendency even sooner.
     

Share This Page