War AIs

[Thalassicus]

I'll give some backstory.

In the spring Firaxis blocked AIs from attacking citystates in the early game. Firaxis also increased the likelihood of AIs to perform a rush in the early game. This forces militaristic AIs to rush neighboring major civs. This rush almost universally fails, so militaristic AIs never perform well.

Since that time I've been searching for anything we can use to disable these early rushes. Early contact is the first and only thing I've discovered so far that appears to work. In the games I've tried this it's been very successful, notably in the Game of the Month maps. It improves the performance of militaristic AIs because they delay war until they found at least a few cities. I prefer randomness to be based on the map terrain and AI personalities, instead of glitchy AI behavior.

 

[Txurce]

I'll give some backstory.

Thank you.

In the spring Firaxis blocked AIs from attacking citystates in the early game. Firaxis also increased the likelihood of AIs to perform a rush in the early game. This forces militaristic AIs to rush neighboring major civs. This rush almost universally fails, so militaristic AIs never perform well.

In my games these rushes work a minority of the time. Specifically, does Greece always get wiped out in your games?

Since that time I've been searching for anything we can use to disable these early rushes. Early contact is the first and only thing I've discovered so far that appears to work. In the games I've tried this it's been very successful, notably in the Game of the Month maps. It improves the performance of militaristic AIs because they delay war until they found at least a few cities. I prefer randomness to be based on the map terrain and AI personalities, instead of glitchy AI behavior.

I wouldn't call chronic denouncing "glitchy," but of course see why it can be problematic. Early aggressors wind up in trouble even if they don't denounce, because they often receive mass denouncements anyway. Your solution would address that as well

My concern at this early stage is that, because it's a universal change, it may well lead to a consequent deterioration of the performance of more peaceful AI's, and to a predictable and therefore exploitable period of early peace for the human player. If I knew the odds of early war are low, my starts would be very different than they are now, and not very different from game to game. That doesn't sound like a good thing.

 

[EsoEs]

Unless one civ in particular is always getting squashed as a result, I don't see why it's a problem

Im pretty sure this is the case. Unless a war AI picks up a VERY isolated start, their early rushes against other AIs will always fail, leaving them friendless, behind on tech and cities, and with a slew of outdated warriors/archers. Their rushes against human players can be effective, probably why Firaxis added it in the first place, but they fail horribly against other AIs

 

[Txurce]

This doesn't happen almost all the time in my games. It's more like a majority, which means a minority of the time, I have a successful early aggressor (a good thing). That's why I'm looking for specifics.

For example, the Mongols tend to do particularly poorly, but because they're often not overly aggressive early, they survive in my games until they hit a weirdly consistent mid-game wall. On the other hand, Rome, Greece and the Aztecs (and there are others) tend to go on a rampage early. Greece and the Aztecs also seem to denounce more than any other civ. So you would think that these two civs would be at the bottom of the standings in most games. But they're not in my games - their performances average out to second tier, more or less.

 

[bwoww78]

is there any progress on a UI update that indicates turns remaining on DoFs and deouncements (both player-AI and AI-AI). While it might seem gamey to be able to plan around this countdown, I think the benefits to the player definitely outweigh that cost.

(btw my apologies if this has already been updated in 10.8 betas; I'm still working with the 10.8 stable release)
Thanks

 

[black213]

I'll give some backstory.

In the spring Firaxis blocked AIs from attacking citystates in the early game. Firaxis also increased the likelihood of AIs to perform a rush in the early game. This forces militaristic AIs to rush neighboring major civs. This rush almost universally fails, so militaristic AIs never perform well.

Since that time I've been searching for anything we can use to disable these early rushes. Early contact is the first and only thing I've discovered so far that appears to work. In the games I've tried this it's been very successful, notably in the Game of the Month maps. It improves the performance of militaristic AIs because they delay war until they found at least a few cities. I prefer randomness to be based on the map terrain and AI personalities, instead of glitchy AI behavior.

To be honest, I haven't seen a civ attack a city-state (except if the said city-state is peaceblocked) in VEM for a long time. In NIGHTS and vanilla I see civs attack them more frequently.
Such as, one time in NIGHTS Mongolia captured all four city-states on his small continent shared with Gandhi (this would've been a lot more rewarding for Mongolia if it was VEM ). I hope to see something like this in your mod too!

 

[Thalassicus]

To make it clear, I don't have a problem with the capability of AI rushes against the human player. It encourages us to prepare for anything in the early game. The problems are:

• AI blocked from targeting citystates. This is the key problem that appears to be hardcoded in the game core, and would make everything else go away.
• AI-AI rushes. Since both AIs have similar bonuses and de-facto identical strategies and tactics, it results in a stalemate in almost all games, and weakens both players. I've only seen early AI-AI rushes succeed two or three times in the entirety of my time playing Civ 5.

Early contact between AIs reduces the second problem.

My concern at this early stage is that, because it's a universal change, it may well lead to a consequent deterioration of the performance of more peaceful AI's, and to a predictable and therefore exploitable period of early peace for the human player. If I knew the odds of early war are low, my starts would be very different than they are now, and not very different from game to game. That doesn't sound like a good thing.

It has the opposite effect because the AIs befriend one another, yet not necessarily the human, so the human is a more likely war target.


@bwoww78
I researched the DoF timer. Since they are not done on the trade deals screen (why? doesn't make sense they aren't there), we cannot use the normal trade deal tools to determine a DoF or denunciation expiration date. It's therefore necessary to detect when a player instigates the deal, and store a timer in the savegame file that counts down once per turn. It's on my todo list but somewhat of a hassle to impliment.


@black213
They attack citystates in the late game but not the early game. I tried changing every variable I could think of that's available to modders, and the AI never even moved units to a citystate, even when it's at war with the citystate. If possible I'd make militaristic AIs favor citystates two to three times as much as major civs... they are much easier to capture, and killing one has fewer downsides than warring with a major civ (loss of trade deal possibilities).

 

[Txurce]

It helps if the human is a more likely war target, but my first concern was that by helping aggressive states survive into the mid-game, they would then have an advantage over peaceful ones, making the overall "fix" a wash. But this is just speculation - let's see what happens.

 

[Txurce]

Is "Adjusted the city creation & development A.I." the toning down of the early denouncers, or something else?

 

[Thalassicus]

I don't mind if aggressive states have an advantage over peaceful ones, since it encourages a "superpower AI" in the endgame, which people seemed to indicate is a popular concept. Also, aggressive AIs currently get a number of small bonuses to prop them up (like extra workers) which I could remove if they make better overall strategic decisions. Leaders have a single "denounce priority" number which controls their likelihood of doing so for the whole game. If it's lowered two or three points, the game turns into a peace rally. I did drop it one point. It's not possible to change only early denouncements with our current modding tools.

What that line in the release notes refers to are a bunch of complex changes to how the AI founds and develops cities.

For example, it did not favor military training in hill and forest cities. I also discovered that in cities with a barracks, all AIs had +6 flavor for "offense" and "defense" units. This is why even peaceful AIs were building tons of swordsmen and catapults, much to my frustration. These units waste of resources if the AI doesn't intend to go to war to use them! It also had no modifiers for recon, naval, or mounted units, which meant AIs didn't build a good mix of units.

The city founding values are things like how much it weighs the 1st, 2nd, and 3rd rings of a potential city site. I reduced the value of the outer rings since border expansion is slower now. I've also been searching for ways to fix the glitchy fixed-distance city placement behavior. If the AI does not have tiles precisely 3 tiles away from existing cities in the early game, it gets stuck and does nothing. I haven't figured out a way to solve this yet.

 

[albie_123]

Wait, so an AI will always found its first few cities three tiles away from the capital, regardless of how good the site is?

 

[Thalassicus]

Right. Firaxis introduced this about half a year ago to force the AIs to infinite-city-spam so AIs were competitive with human infinite-city-spam. These "band-aids" to underlying problems are a pet peeve of mine. The better solution is to fix the underlying problem! ICS has since been solved, but the band-aid remains.

I've been trying to get rid of this ever since, but haven't found a way to do so... either it's in the game core, or it requires a complex interaction of the AI variables in a way I can't discern. There was a long discussion about this between several modders in the main C&C forum in the Spring. Unfortunately, we couldn't figure out a way to really control the AI city founding search range, and my experiments since then have been fruitless.

 

[Txurce]

I don't mind if aggressive states have an advantage over peaceful ones, since it encourages a "superpower AI" in the endgame, which people seemed to indicate is a popular concept.

I hope it works this way. My concern was that we would have superpowers like the Greeks and Aztecs replace superpowers like the Iroquois and Siam.

What that line in the release notes refers to are a bunch of complex changes to how the AI founds and develops cities.

For example, it did not favor military training in hill and forest cities. I also discovered that in cities with a barracks, all AIs had +6 flavor for "offense" and "defense" units. This is why even peaceful AIs were building tons of swordsmen and catapults, much to my frustration. These units waste of resources if the AI doesn't intend to go to war to use them! It also had no modifiers for recon, naval, or mounted units, which meant AIs didn't build a good mix of units.

I've also been searching for ways to fix the glitchy fixed-distance city placement behavior. If the AI does not have tiles precisely 3 tiles away from existing cities in the early game, it gets stuck and does nothing.

This is really interesting. If the AI starts building more appropriately, it will be huge.

Off-topic: do you ever send the devs suggestions, like for example dropping the overly rigid AI spamming mechanic?

 

[Seek]

I don't mind if aggressive states have an advantage over peaceful ones, since it encourages a "superpower AI" in the endgame, which people seemed to indicate is a popular concept. Also, aggressive AIs currently get a number of small bonuses to prop them up (like extra workers) which I could remove if they make better overall strategic decisions. Leaders have a single "denounce priority" number which controls their likelihood of doing so for the whole game. If it's lowered two or three points, the game turns into a peace rally. I did drop it one point. It's not possible to change only early denouncements with our current modding tools.

Please consider reversing this! VEM games are generally so peaceful that they risk getting boring. But of course I will test the new version and report.

 

[EsoEs]

Please consider reversing this! VEM games are generally so peaceful that they risk getting boring. But of course I will test the new version and report.

In the middle of a game where the AIs are: Rome, Japan, Russia, Songhai, Aztecs, India, Siam. Not a peaceful game

Also if you think you games are too peaceful, try a higher difficulty. Emperor games I find a bit too peaceful for my liking, go Immortal if you want a real challenge

 

[Seek]

In the middle of a game where the AIs are: Rome, Japan, Russia, Songhai, Aztecs, India, Siam. Not a peaceful game

Also if you think you games are too peaceful, try a higher difficulty. Emperor games I find a bit too peaceful for my liking, go Immortal if you want a real challenge

Yes, I've been planning on moving up to Immortal recently. My next game will be a domination game, I think; I haven't done one of those in quite some time!

 

[Atlas627]

It seems to me that whether or not the AI declares war has little to do with its flavor and everything to do with what difficulty you are playing on.

When I play on King, the game is super peaceful. When I move up to Emperor, there are some early wars. When I move up to Immortal, it is painful. When I try Deity, I KNOW there will be a war (if not every single possible war) against me.

Instead of giving the AIs bonuses at start, can we give them gradual bonuses? Like workers produce 20% faster instead of the free workers, similar for settlers, +20% production when constructing units, etc.

On harder difficulties it isn't a problem of if I can keep up, its a problem of can I survive early game? The only changes to late game are the setbacks from early, which means if I survive well early I still own the AI late game and have a really boring game as on a lower difficulty.

 

[EsoEs]

On harder difficulties it isn't a problem of if I can keep up, its a problem of can I survive early game? The only changes to late game are the setbacks from early, which means if I survive well early I still own the AI late game and have a really boring game as on a lower difficulty.

Immortal is very challening, I've yet to experience the "boring late game" except on Emporer and lower. Regardless of how well you do early, the AI still provides quite a challenge, they have a lot of units and aren't afraid to declare you.

 

[Txurce]

Instead of giving the AIs bonuses at start, can we give them gradual bonuses? Like workers produce 20% faster instead of the free workers, similar for settlers, +20% production when constructing units, etc.

On harder difficulties it isn't a problem of if I can keep up, its a problem of can I survive early game? The only changes to late game are the setbacks from early, which means if I survive well early I still own the AI late game and have a really boring game as on a lower difficulty.

This is also being discussed right now in "Another thread about capturing cities."

I don't think it's a question of "instead." The AI already has both % bonuses and extra units at the start.