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 Difficulty Scaling: Is it time to lower A, increase B?

Discussion in 'General Balance' started by Stalker0, May 24, 2021.

  1. Stalker0

    Stalker0 Baller Magnus

    Joined:
    Dec 31, 2005
    Messages:
    8,205
    Though the AI has gotten a lot better at the late game, it still tends to struggle on optimally completing its VCs compared to a human. So while the AI shoots ahead quickly in the early game, and its a long time to become equal to them....there still is the feeling that once you catch up with the AI the game is effectively won.

    So I was wondering if people think its time to shift the difficulty a bit. Lower the A value (aka early yields) in favor of a much stronger B value. This would make the AIs start a little weaker, but it would start to get a bigger boost in the mid game.

    My first thought was lower A in favor of C....but its probably too long for the AI to get its needed bonuses to stay competitive, so a B increase is probably a better idea.
     
    DeAnno and InkAxis like this.
  2. 2506

    2506 Warlord

    Joined:
    Feb 23, 2020
    Messages:
    176
    Couldn't the problem be addressed more directly? I don't think the AI has any major issues here regarding Domination or a DV, agreed? So let's move on to an SV. The AI is way ahead in tech, but diddles around and loses the space race. Why not simply make the building of SS parts more of a priority?
     
    Tiberiu likes this.
  3. Stalker0

    Stalker0 Baller Magnus

    Joined:
    Dec 31, 2005
    Messages:
    8,205
    I personally think the AI is best at CV, even with DV they don't always prioritize their tech towards the key techs, get the right WC policies proposed, I don't think they try to deal with others for votes, etc. In comparison, I think the AIs SV is pretty good, its just they are bad at optimizing for the SS parts.

    But at the end of the day that is the saying "hey modders, make the AI better" which can be very tricky. So yeah I'm proposing a more brute force approach. Which also has the nice side effect that the game feels less different for the AI vs the humans...the AIs don't get such an insane start and then start to flounder....they hit the gas a bit less hard at the start but keep their foot on the gas for longer.
     
  4. CrazyG

    CrazyG Deity

    Joined:
    Oct 14, 2016
    Messages:
    5,740
    Location:
    Beijing
    The AI is just not great at actually winning (like the act of winning itself), aside from domination.

    It's pretty good at fighting and generating culture or science, it just can't formalize the victory. I don't think upping the C bonus would do much about that. Often I'm not the #1 civ but I just formalize the win first. I'm not sure what to do, I have noticed the AI not building spaceship parts or the cultural wonder thing when they are available, so that would be the only suggestion I could give.

    As a side note I'd guess that like 80% of the times an AI has won formally before me it was exactly Austria, she gets to enough votes pretty easily.
     
    Last edited: May 25, 2021
    stii likes this.
  5. azum4roll

    azum4roll Emperor

    Joined:
    Jul 17, 2018
    Messages:
    1,593
    Gender:
    Male
    I almost lost my last game when Assyria+Portugal (master and vassal) passed United Nations and World Ideology at the same time, and Portugal revolted a few turns before the Hegemony vote to have a chance to win. I'd say they know what they're doing.
     
    2506 likes this.
  6. Gazebo

    Gazebo Lord of the Community Patch Supporter

    Joined:
    Sep 26, 2010
    Messages:
    18,278
    Gender:
    Male
    Location:
    Little Rock
    The only victory the AI struggles with, consistently, is SV, mainly because - for whatever reason - it always seems to fall behind the tech curve a bit towards the end of the game. It is hard to diagnose specific illnesses of the AI at that point because there are so many moving parts.

    As far as actually prioritizing and building the SS parts, the AI doesn't dawdle as bad as it used to. It's getting there that's often the problem.


    G
     
    CrazyG likes this.
  7. CrazyG

    CrazyG Deity

    Joined:
    Oct 14, 2016
    Messages:
    5,740
    Location:
    Beijing
    I see you changed the thread title, I know that Milae made a difficulty mod that has been well recieved. We could take ideas from that in terms of addressing the difficulty curve.

    One thing I noticed about difficulty over time on my current game (May 19th patch, Deity) is that the AI has an explosive start but does start to stagnate a little bit later on. Inca had 5policies on turn 43! But he didn't get the 6th until turn 80 and I had almost caught up, I eventually got a slight lead around policies 7 or 8. It went similarly with techs, I was 10 behind by turn 50, but only 3 behind on turn 100.

    So if you just don't get intimidated and stick it out, and also avoid war, things do get better.
     
  8. Stalker0

    Stalker0 Baller Magnus

    Joined:
    Dec 31, 2005
    Messages:
    8,205
    Huh? That's been the title since I made the thread.
     
  9. CrazyG

    CrazyG Deity

    Joined:
    Oct 14, 2016
    Messages:
    5,740
    Location:
    Beijing
    Then I just have a bad memory.

    Regardless I don't think that B adjustments would do the trick.
     
  10. InkAxis

    InkAxis Prince

    Joined:
    Feb 24, 2020
    Messages:
    500
    Gender:
    Male
    Why not? Isn't the issue that the AI doesn't have enough yields mid game? So we should increase mid game yield bonuses. The thing about formalizing a VC is a different discussion.
     
  11. CrazyG

    CrazyG Deity

    Joined:
    Oct 14, 2016
    Messages:
    5,740
    Location:
    Beijing
    I don't know if the AI lacks yields in the mid-game, but I was responding the concerns about late game, which I think ought to be the C value.
     
  12. InkAxis

    InkAxis Prince

    Joined:
    Feb 24, 2020
    Messages:
    500
    Gender:
    Male
    I'm confused. So overall you do think C bonuses should be increased?
     
  13. CrazyG

    CrazyG Deity

    Joined:
    Oct 14, 2016
    Messages:
    5,740
    Location:
    Beijing
    In the first post I refer to policies 6,7, and 8, so by 'later on' I meant like classical or medieval era. The C bonus isn't too relevant here.

    I have noticed more than once that they weren't building the wonder for cultural victory in their capital when they were eligible too. My point is it doesn't matter what C is, if they don't put that in their queue they won't win. But surely it makes more sense to change C than B, since B is from the mid-game, not the late game.

    I think C bonus is fine.
     
  14. InkAxis

    InkAxis Prince

    Joined:
    Feb 24, 2020
    Messages:
    500
    Gender:
    Male
    Yeah that's what I meant, shouldn't we have an increase in B bonuses to fix that?
     
  15. CrazyG

    CrazyG Deity

    Joined:
    Oct 14, 2016
    Messages:
    5,740
    Location:
    Beijing
    Possibly, but I find the difficulty curve to be alright. The AI almost always reaches policies 4 and 5 extremely fast, so that slow down is needed to have a chance at all. If the AI kept their ancient era policy pace for the entire game, they'd have ideologies by turn 180.
     
  16. tothePAIN

    tothePAIN King

    Joined:
    Dec 31, 2018
    Messages:
    753
    Gender:
    Male
    For Deity, I think the suggestions in this thread re good. The AI bonuses should be slightly reduced earlier and increased later. Right now, Deity meta is basically survive the Deity early game lead. But that is very vanilla esque .
     
  17. Milae

    Milae Warlord

    Joined:
    Dec 17, 2019
    Messages:
    277
    Gender:
    Male
    I guess I would say this but I honestly would recommend using the changes I made in my mod. I think it addresses this issue pretty well and it's also pretty simple to make changes to if people want harder early/lategame.

    Basically it's just a percentage cost reduction for policies, tech, growth and golden ages from the start, as well as increase in cost reduction for unit/buildings/upgrades. Then the AIPerEraModifier is also applied to policies and tech costs to make sure lategame stays competitive.

    A/B/C is used to give gold and slight science and culture every 10 turns to make early game more consistent.

    I think with this and if they were somewhat hard-coded to construct Citizen Earth Protocol and Spaceships parts when possible this would be sorted.
     
    LifeOfBrian likes this.
  18. Milae

    Milae Warlord

    Joined:
    Dec 17, 2019
    Messages:
    277
    Gender:
    Male
    Also btw increasing C also does more for midgame than increasing B. It's more or less that the bonus is A + B + C in Ancient/Classical and then roughly C * era^2 for the rest of the game.
     

Share This Page