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

I will share with people when the AI will win the game

Discussion in 'Civ6 - General Discussions' started by tranduchau, Aug 9, 2018.

Tags:
  1. tranduchau

    tranduchau Chieftain

    Joined:
    Aug 8, 2018
    Messages:
    2
    Gender:
    Male
    I have read a content
    In autoplay games I have seen AI win a domination victory, more than once actually. Even efficiently so.
    (When I run autoplay games I use prince/standard settings. There is little point for higher DL on autoplay since they are all AIs and DL affects AI vs Player mostly).

    It is a little bit strange to see an AI, effectively win a domination victory one time and then in another game see an AI stop a domination attempt even though they where close. Sure, I can't see why the AI stopped, if they are in a dark age, have bad economy or because of happiness can't take more cities and in autoplay games, I see the entire map, the AIs don't. But still it is strange.

    So if you play a domination only game, you have to play several times in order to have some chance to see an AI even trying to match you and even then humans will always be better when it comes to tactics than any AI (in any game). But I must say if I would have been a player in one of those autoplay games where an AI won a domination victory I don't know if I could have stopped the AI. Perhaps if I started close to the warmongering AI I could have stopped them before they started to snowballing.
     
  2. Haig

    Haig Chieftain

    Joined:
    Aug 3, 2010
    Messages:
    1,977
    Location:
    Finland
    AI has won me in science, culture and score victory.
     
    RohirrimElf likes this.
  3. tranduchau

    tranduchau Chieftain

    Joined:
    Aug 8, 2018
    Messages:
    2
    Gender:
    Male
    Right! It has an indispensable foothold
     
  4. Leyrann

    Leyrann Warlord

    Joined:
    Jan 11, 2015
    Messages:
    3,539
    Gender:
    Male
    Location:
    Netherlands
    Curious.

    I haven't heard of AI winning domination victories yet, would be nice to see a game where it happens here.
     
  5. Trav'ling Canuck

    Trav'ling Canuck Chieftain

    Joined:
    Feb 7, 2018
    Messages:
    998
    Gender:
    Male
    I've completed 13 Deity test games, and no AI has ever held more than 1 other AI capital. Capitals once captured often flip back to their former owner due to Loyalty issues, and the AI tends to seek peace after capturing a city, making it unlikely to pursue a successful war the way a human player will.

    Admittedly, my tests are on Continents maps, and the AI could potentially do better militarily on Pangea maps, but no AI in my test games has ever controlled even their starting continent, so I don't think that would have an impact.

    So assuming the original post is talking about standard map sizes, that leaves three possibilities:

    1. That the AI does better against itself on Prince settings than it does on Deity. I'm not sure why this would be, as on Deity the AI gets bonuses against City States and Barbarians that it doesn't get on Prince, but it could be a possibility. Maybe the slower tech progress on Prince allows the AI to better use it's military units?
    2. My sample set is too small. Which it is to make any conclusive statements. But considering that with 7 AI per game x 13 games no AI has ever held more than 1 other AI capital, there's been no indication of variability in the success of the AI militarily to suggest that a domination victory would be possible at the Deity level even on an extreme outlier result.
    3. The AI behaves differently in autoplay games with no human player. I don't know why that would be, but possibly the AI handles autoplay differently.
     
  6. Elda14

    Elda14 Chieftain

    Joined:
    Aug 10, 2018
    Messages:
    3
    Hi. First time for me posting here.

    Until last week, in spite of a lot of games, the AI never won a game against me and my friend. (We always play with 2 humans, lots of AI, normal or huge map, immortal)... and finally Nubia crushed us in a wonderful way : science victory, the AI ended our game with a score of 1700, twice our score. And the AI also repaired A LOT its spaceport...

    Nubia was very threatening too, placing armies and armadas at our borders. We never saw that before... Quite interesting.
     
    RohirrimElf likes this.
  7. Leyrann

    Leyrann Warlord

    Joined:
    Jan 11, 2015
    Messages:
    3,539
    Gender:
    Male
    Location:
    Netherlands
    How many turns did it need?
     
  8. Elda14

    Elda14 Chieftain

    Joined:
    Aug 10, 2018
    Messages:
    3
    We didn’t save after we lost but it was before turn 290.
     
  9. Aristos

    Aristos Lightseeker

    Joined:
    Dec 11, 2001
    Messages:
    2,733
    Gender:
    Male
    Location:
    Deep inside...
    Knowing Firaxis, this smells like a lot of player:isHuman() IF-THEN tests within the code... which is poor coding.
     
  10. yung.carl.jung

    yung.carl.jung Morose & Lugubrious

    Joined:
    Apr 12, 2015
    Messages:
    1,870
    Location:
    franconia
    the answer to me is pretty obvious. with deity the AI has insane production bonuses, which in turn means that way more units will be flooding the map. the AI also has pathing issues, meaning wars take forever since the AI is also bad at abusing ranged units.

    with fewer units fight are won/lost much more clearly, it is easier to move and advance, the game snowballs more.
     
  11. Trav'ling Canuck

    Trav'ling Canuck Chieftain

    Joined:
    Feb 7, 2018
    Messages:
    998
    Gender:
    Male
    That's a really interesting hypothesis.

    I'm almost finished my deity tests (I'm getting less and less information from each game, and anticipate ending them after each AI leader has been included in 3 games). I'm tempted on this basis to run a series of tests at Prince level, to see if perhaps the human player has less time to win the game on Prince than they do on Deity?
     
    tzu and Leyrann like this.
  12. Leyrann

    Leyrann Warlord

    Joined:
    Jan 11, 2015
    Messages:
    3,539
    Gender:
    Male
    Location:
    Netherlands
    I think it's mostly just that, on Prince, an AI is more likely to win a domination victory. The other victory types I would expect to be similar or slower. I also really like the hypothesis, I should mention.
     
  13. Archon_Wing

    Archon_Wing Vote for me or die

    Joined:
    Apr 3, 2005
    Messages:
    3,324
    Gender:
    Male
    Well, I guess they have trouble with invading other continents, but here Montezuma was pretty vicious and destroyed nearly everything in his continent. The rest were probably lucky he didn't cross the sea. He completely took over Netherlands while demolishing England and Scythia.



    This is from an OCC on King Difficulty. Still nowhere near winning by turn 314 though but it did look scary.
     
  14. yung.carl.jung

    yung.carl.jung Morose & Lugubrious

    Joined:
    Apr 12, 2015
    Messages:
    1,870
    Location:
    franconia
    wars between AIs become both stalemates and, because during war more units are produced, intense ressource drains, which slow down AI development. as a side factor this also allows other AIs to catch up easier, especially if those are AIs with good economy bonuses that the AI actually uses. pathing issues were a very detrimental problem in Civ V and it's only very slightly better in Civ VI. It would be very interesting to just watch and analyse a war between two runaway AIs and see how certain pathing and targetting issues are handled.

    props to you for doing all the testing, I've been following your work :)
     
    Last edited: Aug 12, 2018

Share This Page

Ebates: Get Paid to Shop