AI Too smart by half.....

[lonewulf911]

Been playing Civ a long time. I get frustrated....quit...and return months later. But the same pattern persists: a very good AI that takes the fun out of the game. In many games, the AI will generate units or items (i.e. enemy city walls) at rates that exceed what you can do. As one example, in a recent game, a three city nation all within 6 of each other and visible to me as I attacked generated TWELVE combat units in a span of 10 turns or so. In the early game, I (or anyone else) normally has six or seven at most. I was going for an early domination capital...the AI RNG stopped being random and simply wouldn't let it happen. Anything you can do...I (the AI) will do better.

Is this just me or have others experienced the same thing? I can see making Deity ridiculously hard but at the King level...come on. All people end up doing is moving on to another game as I often do.

[Tyyne]

Lower difficulty if you can't manage it.

[plaguepenguin]

Even Prince is not even, since the AI have a slight yield bonus. That bonus is bigger on King. The bonuses are greater for gold and production than they are for the other yields -- research, culture, and faith. At King the AI get +40% to their gold and production, and +16% to faith, science, and culture. They also get 2 tech or civic boosts at the start of the game.

The +16% and the boosts mean that you have to be more efficient and strategic than the AI at developing your economy just to stay even, much less to get ahead in tech so you can throw more advanced units at them. Once you do start a war, that +40% lets them churn out units, walls, and encampments 40% faster than you can. The gold advantage especially hurts just after you declare war, because the AI tend to keep a whole lot more gold in their savings accounts than you and I do, so they go on a unit spending spree as soon as you attack.

It's easier to use strategy in the wargame aspect of the game, because it's more intuitive than the economic development game. But, unless you luck into a really good starting location, the economic development game requires you to use strategy to keep up and get ahead, so you have to master its less intuitive mechanics.

You can pick better city sites than the AI, and that's a key advantage you have, but to exploit that advantage you almost always have to do some peaceful expansion to get good cities that can support your war effort before you start your career of conquest. Because you can do strategy and the AI has to rely on a program, you're also better than the AI at sequencing the order in which you go after techs and civics to support the particular strategy you are pursuing. You also do a better job at selecting which tiles to improve, and where to put your districts, and that of course reflects back on selecting sites for new cities.

Early conquest is a very powerful overall strategy playing as any civ/leader, and on just about any map the game hands you. But to make it work you have to build an economy that supports the tools you need to make it work. If you find yourself out-tooled by the AI as you describe, you probably have to regroup and retool your strategy around better early development.

[Oaks]

I don't think the AI civs can just spawn units from thin air; they are bound by the same rules as you. They do receive bonuses to various outputs on higher difficulties, they can't just miraculously spawn troops though.

[SlasH]

Originally posted by Oaks:

I don't think the AI civs can just spawn units from thin air; they are bound by the same rules as you. They do receive bonuses to various outputs on higher difficulties, they can't just miraculously spawn troops though.

Are you sure or do you just think so? And if sure, can you point to some detailed info about it (e.g. website etc.)?

[Oaks]

Originally posted by SlasH:

Originally posted by Oaks:

I don't think the AI civs can just spawn units from thin air; they are bound by the same rules as you. They do receive bonuses to various outputs on higher difficulties, they can't just miraculously spawn troops though.

Are you sure or do you just think so? And if sure, can you point to some detailed info about it (e.g. website etc.)?

I always go to civ fandom for information. Here's the part on difficulty levels;

https://civilization.fandom.com/wiki/Difficulty_level_(Civ6)

[Ghaleon]

IIRC Prince difficulty is the hardest setting where the AI doesn't have any advantages or 'cheats'.

Honestly the thread is pretty strange to me since this is one of very few games that has such a large selection of difficulty levels to play with.

There is one thing I find weird even on prince difficulty though (understandable if it's on a harder setting). I'll take another civ's city. They'll bust out city walls like 2 turns after I start attacking them, and they made units beforehand, and their science/culture is... fine.

But then I capture the city and it's like... zero districts in it. Half the time it doesn't even have a monument or granery. Like how do they get resources with these cities that build jack all?

[Liquidlazor]

Originally posted by Ghaleon:

IIRC Prince difficulty is the hardest setting where the AI doesn't have any advantages or 'cheats'.

Honestly the thread is pretty strange to me since this is one of very few games that has such a large selection of difficulty levels to play with.

There is one thing I find weird even on prince difficulty though (understandable if it's on a harder setting). I'll take another civ's city. They'll bust out city walls like 2 turns after I start attacking them, and they made units beforehand, and their science/culture is... fine.

But then I capture the city and it's like... zero districts in it. Half the time it doesn't even have a monument or granery. Like how do they get resources with these cities that build jack all?

I once criticized a looter-shooter for changing enemy behaviour based on what you equip, basically rendering the underlying concept of looting for better gear rather meaningless to some extent (That extent being if you purposefully equip under-powered gear you get annihilated so hard you dont notice the difference) because you can get better and better gear but the enemies just keep scaling to your equipment to maintain the difficulty.

This was based on intuition of playing the game until another smartypants came around to the forum accusing me of making things up and I dont have proof of what I'm saying so therefore I'm in the wrong for making false accusations. So.... I looked into the GitHub code for armor modifications I actually found proof of what I'm saying in unaltered code for many armor modifications and I shared that. Unfortuantely the person that contested idiotically kept denying the straight facts that I was posting calling me stupid etc but alas I felt justified in my observations.

My point has many sub-points. I feel that same thing with the point you are making sometimes. I have mentioned that on these forums fairly recently. The only logical explanation I have is that the AI is smart enough to keep gold reserves instead of a standing army for that purposes, which seems a long shot given the requirements in accordance with the volume produced. So my intuitions on this remain. But the second point is that this would probably be auditable by doing the same thing, looking at the code.

Over and above that, there is a loss of innocence when realizing the parameters and the limitations of the construct of a game, like youre playing something unfair meant to cheat you into 'simulating an experience' rather than presenting a fair game. For that, the best advice I offer now is that it is a game, like a boardgame, and that is the rules of the game.

As such, I've seen a similar thing in a racing game, one of these games where you grind new cars the whole time. But the AI rubberbands like crazy because your car 'needs to be a certain rating to beat the challenge' so you can drive a world record lap but still be beaten by fake boosts because the game needs you to lose because the 'car rating' requires you to. There are many other things wrong with the game too. Nitro boost straight up not working, its just a shift of field-of-view to create the illusion of speed, or you drive a legit A-class hypercar but the boosts kick in slower than a D-rated car etc.

I dont put it beyond developers to take liberties in design in order to 'deliver a balanced experience'. Its very irksome to the player once you notice that.... What I do notice is that these AI bonuses do not carry over to multiplayer. I think this is something that is under-recognized in game design. Another obvious example is in football games, how your players start playing like amateurs to give the other side ascendancy, sometimes for predatory reasons like making you think your team is weak so you spend more real world money to buy more players. As with another game where you're constantly grinding better loot only to learn that loot drop rates are tied to what you have in inventory and storage already, ensuring you never get the best loot in the game until you buy them from the marketplace with in-game currency bought from real money.

So thats a lot said but is somewhat a tangent from what is going on with Civilization. Their business model is centered around unit volume and DLCs and therefore do not need to include corrupt mechanics on this basis of sales pushes. But there are elements of the mathematics have to make sense. Now I have been an advocate of AI design that strives for mathematical perfection as a cornerstone of difficulty but I think that is hard to deliver constantly with all the variable factors in every unique game.

I've left out some game names, but I'll mention 7 Days To Die, I mean its the same in many games like Minecraft or whatever, where the player has access to console commands and you can set your character to level 300 instantly etc, basically breaking the game. There is an essence of handing the player the power to customize the game to their liking, because once you have the power to beat the game in five seconds, one adopts a new world-view of gaming and the challenge the game programs by understanding that for the most part you set the level of challenge for yourself.

But in that many like cinematic games only have one difficulty and fair play to them, some games lend themselves to that. You could argue Civilization could be a better game if it had just one difficulty. But to deliver a product to masses you can understand people want that flexibility to have many playing options, including over-powered difficulties that increase replay value.

So part of me believes in that, that the AI can be better, and the purist in me believes AI should not bring uncalculatable practices injected on the side maliciously and deceptively. But just so I also realize a perfect AI will absolutely break any human player. So we take it at face value, AI to a large part is meant to simulate an experience. And its just a game, dare I say unfortuantely its only a game..., the loss of innocence lingers.

EDIT: Regarding the districts and conquest etc, I would totally see how districts would disappear for game balance, the AI already has a large layer of reacting to your military build-ups and prevention thereof through training times etc. I think we would live easier if we knew this is a game rule, that districts raze upon conquest because of their bonuses, even if it sucks that I have to comment on it in such a 4th-wall breaking way. The only other counter I can mention right now is also tied to the amount of attacks it takes to siege a city.

Now I cant imagine the game determining that the amount of turns/attacks it takes to siege a city being tied not only to military strength but also to potential yield bonuses to your empire, or you can sit their sieging one city for the entire history of the game, so the short fix could be to just raze districts. If I get accusations about what I say I might look into it more. I believe sometimes devs dont want underlying mechanics exposed too much, but on the other side I'm more of trying to preserve the mystery and make determinations through natural discovery, where my calculations of AI implementations stem from as well, as I dont need to look at the code for that, and when I dont have to, that is a compliment.