This is an important observation.
Indeed, more than any other factor, score is most closely related to turn count.
So logging scores only at the point of victory is less useful than it should be, since this the turn counts won't be the same.
Some statistics pertaining to relative civilization/victory type power in ~100 AI-only games
- Thread starter L. Vern
- Start date
Indeed, more than any other factor, score is most closely related to turn count.
So logging scores only at the point of victory is less useful than it should be, since this the turn counts won't be the same.
pineappledan
Deity
From a human player perspective my own opinion is that this is fine. Militarist AIs don’t have to win the game, they just have to kill you and force a restart. If players can win with them just as they can with other civs, and AIs can kill players (or at least immiserate humans enough that they quit), then that’s sufficient.
This is where I think this idea of changing DomV so that AI can win is missing the point. We aren’t making this mod so that AI can just play against each other, we are making it so humans can have fun and be challenged
This is where I think this idea of changing DomV so that AI can win is missing the point. We aren’t making this mod so that AI can just play against each other, we are making it so humans can have fun and be challenged
pineappledan
Deity
Yes and no. Even though Arabia wins almost 1 of every 2 games, he still ends games in 5th place out of 8, on average. So his scores aren’t just low in a vacuum, though they are also low.
Arabia’s just a win machine it seems. Other AI are playing the scoreboard and he’s playing for the win condition. Not exactly sure how the AI for this 1 civ manages to optimize like that, but he is just obviously better at playing lean.
Bazaar + traderoutes is a very forgiving win mechanic.
jdevo
Chieftain
Yeah, I wish we had a conquest victory for capturing capitals, and a domination victory for controlling maybe 65% of the land.
Rekk
Deity
- Joined
- Dec 9, 2017
- Messages
- 2,725
The warmonger AI needs to be able to knock out other players while they are reaching for their Victory Conditions. If they can do that, this should result in at least a couple wins by the Warmonger Civs.
That they as a group were only able to get a single victory (Japan's Cultural Victory) is very troubling; it suggests that warmongering against the leading players in the late game is not something the AI can profit from.
Changing DomV to allow AI to win "earlier" (thus actually makes it a viable win condition for AI) can push player into action, because it can be seen as having another contestant you have to deal with.
Currently if there's a militaristic civ popping up ahead, nobody (human player) would even bother to response unless you're their neighbor, because it's so easy to just focus on doing your own thing and win earlier (and even better if your competitor got slowed down). And even when you're neighbor of such civ, it's not hard to focus purely on defense because of the issue with taking cities I was talking about in new version 2.6 thread, you don't even need to be as strong, just stronger than their next target; not to mention it's possible for the whole world to gang up on them because of huge warmonger penalty - big diplomatic boost for free. There're just so many bonus having such civ in your game as long as you don't aim for DomV yourself.
Meanwhile for all other victories if there's a run away civ you have to response in one way or another regardless of location, and it's a big stretch if you can even do anything at all to them. There's a huge difference in gameplay decision based on this alone.
bad news: unfortunately this analysis is at least partially based on faulty data. there was a rather severe bug in the tactical AI which is now fixed - this should improve the performance of the warmongers. so take everything above with a grain of salt.
Great news that this is fixed. Would it be possible for you to release a hotfix DLL to address the issue in the current release, as it's quite a serious issue as the player is never in any real danger?
Maxxim69
Chieftain
Since @ilteroi hasn't replied in three days, I'll go ahead and venture a guess: the hotfix DLL may be the one he attached in a reply to my issue report on GitHub.
He did reply here, in the thread I made about AI not using their armies. Thanks for the heads up anyway. And it's not a complete fix, but an improvement.
L. Vern
Warlord
I played a game of Arabia after seeing the stats and what I can say is, their UA is stupidly good - making great people being a historic event that triggers more progress towards great people makes it scale with itself so hard that great people are popping out every 1-3 turns. I was culturally dominant over the whole world by turn ~210 from all the great work tourism+artistry
besset
Warlord
- Joined
- Dec 25, 2017
- Messages
- 206
I think the strongest civ's in the hands of the AI are the ones with the best bonuses for "passive" play, where you don't have to do much/strategize in order to get the bonus. Human players will always be better at civ's that require more strategy & micromanaging, e.g. Venice (or warmongering in general). That's why civ's like Arabia and China have always been strong for the AI, but that doesn't necessarily mean the civ is imbalanced for a human player.
Would probably be ok if it was just favorable peace deals and wonders.
As is it now bazaar mean you get it from trade routes, I think this is what make it go really bonkers.
Most games you can snag at least one of the trade route wonders (on emperor probably 2) especially if you have desert capital for petra.
If you dont want artistry go statecraft for +1tr.
Industry opener for +2tr.
Similar threads
