Denmark

[infidel88]

"AI is very good at pillaging". This is 10th turn of war. Oh, Harald has 12 policies, rest of the world has 17+

 

[BiteInTheMark]

"AI is very good at pillaging". This is 10th turn of war. Oh, Harald has 12 policies, rest of the world has 17+

AI excells at using civs with a UA or UB which activates passively. Cause of this, civs like China, Egypt, Ethiopia, Arabia, Korea, ... are in most games superior to civs which need effort to activate their UA. The more you go away from standard play (Venice, Aztecs, Denmark are the strongest examples), the more random their success is in the most games.

Unless you are able to completely rewrite AI code up to human level, you can't balance such civs for the AI AND human. I think, the only solution for this would be to create different UAs, one for the human and one, more passiv for the AI.

 

[Gazebo]

"AI is very good at pillaging". This is 10th turn of war. Oh, Harald has 12 policies, rest of the world has 17+

What?! The Danes don’t crush everyone, every single game?! Clearly the AI is brain dead!

You’re better than this nonsense, dude.

G

 

[Gazebo]

AI excells at using civs with a UA or UB which activates passively. Cause of this, civs like China, Egypt, Ethiopia, Arabia, Korea, ... are in most games superior to civs which need effort to activate their UA. The more you go away from standard play (Venice, Aztecs, Denmark are the strongest examples), the more random their success is in the most games.

Unless you are able to completely rewrite AI code up to human level, you can't balance such civs for the AI AND human. I think, the only solution for this would be to create different UAs, one for the human and one, more passiv for the AI.

That’s absurd and insulting. All the AIs understand their UAs. But understanding a UA doesn’t mean they’ll win every game. Someone has to lose.

G

 

[infidel88]

Can you please explain to me how AI is good at pillaging and for 13 turn war all AI did was swimming with their ships around and only managed to pillage 1 fishing boat. And left that fishing boat untouched for whole war (which is free ) and didn't move to other side of island with their gazillion ships just to pillage? Note - when war started I had 2 warriors, 2 papal guards (landsknechts) and 1 heavy skirmisher. Yet AI Harald didn't think about dropping a single land unit and pillage my tiles. Oh wait, he actually had tercio in my land when he declared war that lasted 2 turns. And you know what he did? Pillage? Nah, lets just go towards that Castel Gandolfo city. This is Denmark AI problem. When I was playing them I just had 3 highly mobile scouts getting me 12 pillages per turn. AI doesn't make strictly pillaging wars. Never. They pillage as a side job, just the way every other civ does.

 

[Gazebo]

Can you please explain to me how AI is good at pillaging and for 13 turn war all AI did was swimming with their ships around and only managed to pillage 1 fishing boat. And left that fishing boat untouched for whole war (which is free ) and didn't move to other side of island with their gazillion ships just to pillage? Note - when war started I had 2 warriors, 2 papal guards (landsknechts) and 1 heavy skirmisher. Yet AI Harald didn't think about dropping a single land unit and pillage my tiles. Oh wait, he actually had tercio in my land when he declared war that lasted 2 turns. And you know what he did? Pillage? Nah, lets just go towards that Castel Gandolfo city. This is Denmark AI problem. When I was playing them I just had 3 highly mobile scouts getting me 12 pillages per turn. AI doesn't make strictly pillaging wars. Never. They pillage as a side job, just the way every other civ does.

The AI knows how to pillage. Priorities and danger can limit where and when they do it, but to claim an understanding of the AI through observation alone (and not through analyzing the code itself) is prone to misunderstanding.

G

 

[infidel88]

Sorry to say this, but you sound a bit like "it's not a bug, it's a feature". I just fail to see how a civ, who benefits from pillaging the most, swims with their ultramobile navy next to a fishing boats for 13 turns and fails to pillage them. I know AI will never do stuff human can do. I don't expect them to use scouts the way I mentioned above, but I expect pillaging civs to actually pillage. And crippling players who use this UA (yes, yields were too hight, but it was fun) is not a solution.

 

[Gazebo]

Sorry to say this, but you sound a bit like "it's not a bug, it's a feature". I just fail to see how a civ, who benefits from pillaging the most, swims with their ultramobile navy next to a fishing boats for 13 turns and fails to pillage them. I know AI will never do stuff human can do. I don't expect them to use scouts the way I mentioned above, but I expect pillaging civs to actually pillage. And crippling players who use this UA (yes, yields were too hight, but it was fun) is not a solution.

The AI doesn’t know what you know or see what you see. The AI is an algorithm. It views things like tiles as a probability model. Likely it viewed the plot as too risky without proper visiblity.

Anyways if you think something is a bug, report it and I’ll take a look, don’t moan on here. Not useful.
G

 

[BiteInTheMark]

That’s absurd and insulting. All the AIs understand their UAs. But understanding a UA doesn’t mean they’ll win every game. Someone has to lose.

G

Why is it absurd? The AI was written do to the normal stuff. But not such specialized things like "pillage as much as you can", "get a warscore of 25 to get a GA", "Increase your trade route cap max possible cause your a trade nation".....
And sorry to say this too, but how are you able to say the KI understand it? Do you see flying green digits and letters over your screen like in Matrix? Are you Neo?
I do not care if Denmark "understands" its UA (if you can say such a thing about such a limited AI), all that counts is the result I see in the end. And it doesnt make any difference if you have more vision into the code than me. If the AI "understands" its UA but didnt act like a human which understands it, isnt acting correct. The transition from understanding to acting is wrong, no matter what happens between.

And even with your insight into the game, you completly missed the bug, AI wasnt building walls, even its so obviosly by playing only one game till renaissance?

I never.... NEVER.... saw a Denmark coming to my land and plunder as much as he can. Even it would be impossible to me to stop him cause of low amounts of units. Never. All hes doing is the same stuff as other AIs do. Gather units in front of my city and siege the city. Tiles are only pillaged if it would heal a wounded unit or have a strategic ressource on it. (even before the pillage restriction was made). They are simply repeating the stuff they have learned from vanilla, but doing it now more efficient.

 

[chicorbeef]

Is the AI aware of the fact that pillaging a normal tile gives no bonus but pillaging a resource does? Is it aware of the difference between sea/land tiles?

 

[Gazebo]

Why is it absurd? The AI was written do to the normal stuff. But not such specialized things like "pillage as much as you can", "get a warscore of 25 to get a GA", "Increase your trade route cap max possible cause your a trade nation".....
And sorry to say this too, but how are you able to say the KI understand it? Do you see flying green digits and letters over your screen like in Matrix? Are you Neo?
I do not care if Denmark "understands" its UA (if you can say such a thing about such a limited AI), all that counts is the result I see in the end. And it doesnt make any difference if you have more vision into the code than me. If the AI "understands" its UA but didnt act like a human which understands it, isnt acting correct. The transition from understanding to acting is wrong, no matter what happens between.

And even with your insight into the game, you completly missed the bug, AI wasnt building walls, even its so obviosly by playing only one game till renaissance?

I never.... NEVER.... saw a Denmark coming to my land and plunder as much as he can. Even it would be impossible to me to stop him cause of low amounts of units. Never. All hes doing is the same stuff as other AIs do. Gather units in front of my city and siege the city. Tiles are only pillaged if it would heal a wounded unit or have a strategic ressource on it. (even before the pillage restriction was made). They are simply repeating the stuff they have learned from vanilla, but doing it now more efficient.

I hate to play this card, but you see there’s this thing called ‘learning a programming language.’ I know it. Others know it. You don’t. Your post highlights your ignorance of that fact. This is what I mean by ‘observation can lead to false reporting without looking at the code.’ You’re talking out of ignorance on a topic you don’t understand from a programmatic point of view.

Yes the AI knows how to use UAs. Yes it understands what its UA is and how to use it. But that doesn’t mean it willl do so without fault every time. The AU is not scripted. It is given thousands of inputs and variables and told to make specific choices. Sometimes those choices mesh wel- sometimes they don’t. But to say that failing to make the right choice every time means that they’re completely incapable is absurd.

G

 

[Gazebo]

Is the AI aware of the fact that pillaging a normal tile gives no bonus but pillaging a resource does? Is it aware of the difference between sea/land tiles?

I don’t want to answer without checking the code again, but last time I looked at the pillaging code, the answer was yes to all of these. But, as I said, I’m always keen to find bug, and I appreciate these kinds of specific, testable questions.

 

[amateurgamer88]

This is something I'm curious about. Let's say the Denmark AI is processing its turn and prepares to go on the offensive. Does it take into account that it gets a bonus while fighting in pillaged tiles and account that for its decision making? Or is this too complicated for the AI to do?

 

[Gazebo]

This is something I'm curious about. Let's say the Denmark AI is processing its turn and prepares to go on the offensive. Does it take into account that it gets a bonus while fighting in pillaged tiles and account that for its decision making? Or is this too complicated for the AI to do?

Yes it does. There’s a mondo huge function that tells a unit what it’s maximum striking power from a tile will be (also can tell it what it’s striking power against another tile will be if we need it to). One of @ilteroi’s big tasks lately has been rewriting the model under that which makes the AI choose the best tiles for all units in an area to move into, based on attack power, defensibility, proper placement of unit types in layers, and potential danger for a plot next turn.

G

 

[amateurgamer88]

Yes it does. There’s a mondo huge function that tells a unit what it’s maximum striking power from a tile will be (also can tell it what it’s striking power against another tile will be if we need it to). One of @ilteroi’s big tasks lately has been rewriting the model under that which makes the AI choose the best tiles for all units in an area to move into, based on attack power, defensibility, proper placement of unit types in layers, and potential danger for a plot next turn.

G

That's pretty cool. Then I guess my next question is whether or not the AI is capable of deciding that, upon pillaging a tile with at least one more movement left afterwards, it would get more CS and thus can further maximize its striking power. It know it might conflict with AI units pillaging to heal up so I'm curious if that's possible. Then, we might see Denmark play a bit differently than other AIs where it uses pillages to make a more aggressive push on its enemies.

 

[Gazebo]

That's pretty cool. Then I guess my next question is whether or not the AI is capable of deciding that, upon pillaging a tile with at least one more movement left afterwards, it would get more CS and thus can further maximize its striking power. It know it might conflict with AI units pillaging to heal up so I'm curious if that's possible. Then, we might see Denmark play a bit differently than other AIs where it uses pillages to make a more aggressive push on its enemies.

I'm honestly not sure - the AI definitely prioritizes pillaging for injured units over allowing 'fresh' units to do it. I could probably add some forced code to tell the AI to pillage a tile regardless if it gets a yield or CS bonus, however it might result in some lower-health units getting wiped if they don't have access to a quick pick-me-up from a pillage.

G

 

[BiteInTheMark]

I hate to play this card, but you see there’s this thing called ‘learning a programming language.’ I know it. Others know it. You don’t. Your post highlights your ignorance of that fact. This is what I mean by ‘observation can lead to false reporting without looking at the code.’ You’re talking out of ignorance on a topic you don’t understand from a programmatic point of view.

Yes the AI knows how to use UAs. Yes it understands what its UA is and how to use it. But that doesn’t mean it willl do so without fault every time. The AU is not scripted. It is given thousands of inputs and variables and told to make specific choices. Sometimes those choices mesh wel- sometimes they don’t. But to say that failing to make the right choice every time means that they’re completely incapable is absurd.

G

Yeah, I have only minor understanding of the code. And you have a major insight to it.
But who of us observed first an AI which think its a good idea to have no walls in renaissance? Was it you with your great insight or was it me, who played the game and used logic and obersvation?

 

[amateurgamer88]

I'm honestly not sure - the AI definitely prioritizes pillaging for injured units over allowing 'fresh' units to do it. I could probably add some forced code to tell the AI to pillage a tile regardless if it gets a yield or CS bonus, however it might result in some lower-health units getting wiped if they don't have access to a quick pick-me-up from a pillage.

G

That's certainly not what we want. I like the idea that the AI tries to save their units. However, it does seem like the humans can use that to their advantage but that's a different topic altogether.

What's the order the AI makes its decisions for its units? Let's say we have a melee non-mounted unit pillage at the start of the turn and have 2 movements still due to Viking. Then, it withdraws to heal up and fight another day. Will another unit now notice the pillaged tile and account for a boost to its attack? I know it can be hard but it would be cool if a wounded unit can heal and withdraw while setting the stage for another unit to use that pillaged tile to continue the push. I haven't seen quite something like this in my games but maybe I just haven't played enough games.

 

[TheBigFundamental]

But who of us observed first an AI which think its a good idea to have no walls in renaissance? Was it you with your great insight or was it me, who played the game and used logic and obersvation?

Not cool, man.

 

[Gazebo]

Yeah, I have only minor understanding of the code. And you have a major insight to it.
But who of us observed first an AI which think its a good idea to have no walls in renaissance? Was it you with your great insight or was it me, who played the game and used logic and obersvation?

Here’s the difference between me and you: you see something, but you don’t know why it is. You observe, but you cannot interpret. You are like the man who witnesses advanced science and calls it divine: you see the event, but you lack the language to explain it. If you stopped at observation, that would be fine. But you persist in your blind explanations of the scientific, and in the process reveal your ignorance.

I am always thankful for observation and reporting. I love to answer questions and be challenged on topics or ideas. What irritates me is when people without access or insight into the code decide to use a bug or a quirk to explain away the thousands of man-hours spent making an AI that thinks and acts as close to a human player as we are capable of achieving on a single threaded, 32bit program.

G