Discussion in 'Community Patch Project' started by Gazebo, Feb 17, 2019.
It does. But it only asks if it is worthwhile. Otherwise it just attacks.
That thought never crossed my mind, but it must be as you say. I always play continents, large map, and add an extra civilization (11). Perhaps the living space often isn't big enough for picking progress!? But carthage never seem to think so, since they forward settle everybody anyways.
Well, Carthage is basically a Progress civ, so it has a expansionist behavior. This behavior leads to the civ picking progress more often, but it is not hard coded. Sometimes, AI does not see it clearly and tosses a dice.
I think, immortal and deity AI picks always the best calculated option, below that, they randomly choose one of the 2 best options. (or was it emperor too?) And on the lower difficulties, one of the best 3. But having only 3 options for the first policy tree, it's really like throwing a dice .
I think the AI should do always the best option, it's anyway the handicaps which plays the major role in difficulty setting.
Its variety and replayability for the AI to be less predictable. On Deity, things like pantheons and religious belief selection can go very similarly every game. Certain wonders are safe just because the AI never seems to take those techs when pursuing best options (looking at Terracotta Army).
Also, I don't the AI decision making is so refined that the second best option is consistently worse than the first option.
The AI tactical decision making these days is very strong. On deity, my only really consistent strategy for conquering is naval. Otherwise, they can stop me just like I would stop me. They're even executing naval invasions!
Prior to this post, I would have agreed with BITM. This is a good example of the complexity involved on designing a longterm playable game.
Yeah, I thought Gazebo said some time back, that the AI had become almost "as good as it gets", sort of... Well... Apparently not! I'm really amazed by how good it has become; I couldn't see it coming in my wildest fantasies.
Define 'best.' Go ahead, I've got time.
Its an AI, in the end, its only solving equations, based on real numbers. And with it, the result is also a number. (I remember the issue with nearly all AI picking freedom ideology, cause it gave the healthcare tenet an astronormus high value, 10-100 times greater than any other tenet, leading to this result.)
For the AI, the best choice is the option with the highest score. (But the numbers for the equation are not coming out of the blue. Its using base values and modifiers, which are set by a human/Gazebo. So, the player is literally playing a game against a shadow version of a human/Gazebo)
He doesn’t need to G, you already did in your calculations.
He is just saying that because the policy tree has such a long term impact on a civ, thst regardless of difficulty, the AI should choose the best policy (based on its internal criteria), rather than randomly from 2-3 options.
I think it’s a reasonable position with benefits and drawbacks.
However, this randomness is just for lower difficulties, which reduces challenge (nice) and increases replayability (great). So I don't see an issue here.
Will you ever be willing to improve the AI for Civ VI when it's possible? Just wondering.
Thank you for your hard work and the amazing mod, Gazebo and other contributors!
I think no one claimed the civs will be always picking the same pantheon/policy tree/...
If it would work that way, the AI would pick even in emperor always the same 2 pantheons, and the replayability would be gone too. There are enough things which can happen. Every trader civ will probably prefer God of Commerce, but starting in mid of mountains or desert will lead to a stronger opinion about God of Nature or God of Desert. I dont think the replayability is in danger, if the AI is changed to pick always the best pantheon, cause the "best" is driven by the circumstances, and those are changing in every game.
And humans are just bags of water that exchange oxygen for co2. But we both know there's a lot more to it than that, right?
The AI is not just 'solving equations based on real numbers' – there are multitudinous factors that play into decisions beyond just 'a chunk of numbers.'
And it's not a shadow Gazebo, as I don't inject my judgment.
Having a neighbour picking a bad pantheon or a bad policy is an opportunity. As long as not every AI is making mistakes, it's fine.
I think you didnt want to talk with me about "the godly spark", the human soul or want to sell me one copy of the book of mormon. In the end, we are a result of genes and education, driven by an electrochemical engine. Sorry, Iam atheist .
I never denied the fact a lot of factors playing a role in the decision making. But those factors are all driven by numbers or "yes"/"no" questions ("1" or "0"). For an example, one of the factors which decides about the friendlyness or hostility towards a civ is the presence of the own UU. Having atleast one copy of the own UU increase the tendency for hostility by 10 times of the civ related hostility bias. The hostility bias is a number, written by a human. The factor 10 is a number also written by a human. Neither you nor the AI know if the multiplicator of 10 was chosen correctly (optimal). Cause it was driven by the opinion of a human. But the AI will simply use the given code and calculate numbers, and will decide based on that calculation.
Are Pickelhaubes the German equivalent of a fedora?
Religion is a tool to get yields and congress votes/
What's funny about this, to me, is that
a.) You have no idea what this means within the function
b.) You're operating off of now-obsolete numbers
c.) You have no idea how that function plays into the actual diplo AI decision-making.
You're essentially no better than a parrot in this situation. You repeat what you heard because you think it's worth saying or - gasp - it is somehow a solid argument, but you reveal your ignorance in the process.
Separate names with a comma.