AIs and the Art of War

[Thunderbrd]

Rams shouldn't be allowed for city defense AI roles but it's possible that you targeted an attack stack he was building and had brought into the city as a ditch effort to save it. Then again, it's possible that they were still for some reason selected for city defense which would be rather odd and certainly worth looking into at some point. I haven't seen this taking place myself so far.

 

[Noriad2]

I haven't played C2C in weeks due to other games (currently EU4) but I remember that if you don't have horse or obsidian or UU's in the 2nd half of the prehistoric era, for a very long time the only alternatives for the strength-5 battering ram are strength-3 units, until you get to bronze.

Also due to city defense % rising very high very quickly in late prehistoric, any attack stack must include tons of battering rams. However some additional other units would be nice in a stack because battering rams alone can't take a city if I remember correctly.

 

[Thunderbrd]

I believe that attack stacks are forming much better now than they were. There's still a LOT of AI work to be done there but it's sufficient for the next release to be a huge improvement at least.

Some of the newer archer units fill a good defensive gap at this point I feel as well.

 

[JosEPh_II]

I propose that the Default Stack Limit for the Mod should be set at 30. This will help mitigate some of the stack problems being reported. Of course the Player will still be able to change it to suit their play but if they use the higher setting they must realize they are creating their own "oddities".

JosEPh

 

[Thunderbrd]

Any stack limit will eventually cause the AI problems as it'll find itself incapable of ever mobilizing the stack it's trying to build because it can't reach the numbers it expects itself to be able to.

 

[JosEPh_II]

Any stack limit will eventually cause the AI problems as it'll find itself incapable of ever mobilizing the stack it's trying to build because it can't reach the numbers it expects itself to be able to.

And didn't koshling warn several years ago that letting the Stack Limit be Unlimited was a bad idea?

JosEPh

 

[Dancing Hoskuld]

And didn't koshling warn several years ago that letting the Stack Limit be Unlimited was a bad idea?

JosEPh

AFAIK - No. Stack limit is normally unlimited. The code that limits the the stack that causes CTD and Waiting for Civs endless loop problems. Unlimited just gets slower turns is all.

 

[JosEPh_II]

Hmmm..... guess I'm not remembering right then. I have my current game set at 50. I do remember Koshling saying don't go smaller than 10.

JosEPh

 

[Arkenor]

Just a thought based on my experiences in my current game.

The AI has a tendency to build its cities very close together. When conquering an enemy region, I frequently have to thin their cities out. Is it considering the 3 square radius that cities eventually achieve? As it stands, it ends up with an awful lot of cities, but they do not reach the potential that they could with a little more room.

If this is the case, then making them build a little further apart might have useful consequences for them defending, with defenders spread less thinly. Less cities, but better ones, might also have some performance benefits.

In the picture below, the Tupi have done themselves a mischief with this behaviour. I wouldn't build that close even with the 2 square radius.

 

[alberts8]

The MIN_CITY_RANGE is set to 2 in C2C by default but i play with at least 3 to avoid that. You can change this in the GlobalDefines.xml.

 

[Thunderbrd]

The MIN_CITY_RANGE is set to 2 in C2C by default but i play with at least 3 to avoid that. You can change this in the GlobalDefines.xml.

It'd be nice to set this up a a Leader personality difference. I've seen some players prefer to stack their cities in close and others prefer to keep them well separated. There's strategic benefits and penalties to both ways to go so if this was a leader variable it would be quite interesting.

 

[JosEPh_II]

Just a thought based on my experiences in my current game.

The AI has a tendency to build its cities very close together. When conquering an enemy region, I frequently have to thin their cities out. Is it considering the 3 square radius that cities eventually achieve? As it stands, it ends up with an awful lot of cities, but they do not reach the potential that they could with a little more room.

If this is the case, then making them build a little further apart might have useful consequences for them defending, with defenders spread less thinly. Less cities, but better ones, might also have some performance benefits.

In the picture below, the Tupi have done themselves a mischief with this behaviour. I wouldn't build that close even with the 2 square radius.

Resources and map configuration also plays a major part in this. I play more like the Tupi when resources for early city development is more critical than future spacing.

I would balk at being forced to having my cities mandated and dictated to be spaced further apart. Flexibility is key.

JosEPh

 

[Arkenor]

Resources and map configuration also plays a major part in this. I play more like the Tupi when resources for early city development is more critical than future spacing.

I would balk at being forced to having my cities mandated and dictated to be spaced further apart. Flexibility is key.

JosEPh

I wasn't talking about anything that would limit where you can build. My concern was whether the AI is making intelligent city location decisions. The Tupi in that picture had screwed themselves over horribly, and the situation is common across my entire map. Those are old cities, with feeble populations because they denied themselves the space to grow.

 

[JosEPh_II]

Again it has to do with When the cities were founded and early resources and not Down the Road decision making. The AI is not going to be looking eras ahead. They placed the cities based on the "then" surroundings. Not that the 3rd ring would be more important 4 eras later.

And trying to program/code the AI to do so might not be practicable, wise, or even prudent. But giving the AI the flexibility is probably the best that can be done. But I will defer to T-brd and alberts2 on that part of the Mod.

EDIT: And if it's not what you like just do what alberts2 said:

The MIN_CITY_RANGE is set to 2 in C2C by default but i play with at least 3 to avoid that. You can change this in the GlobalDefines.xml.

That should help your future games.

JosEPh

 

[Thunderbrd]

I wasn't talking about anything that would limit where you can build. My concern was whether the AI is making intelligent city location decisions. The Tupi in that picture had screwed themselves over horribly, and the situation is common across my entire map. Those are old cities, with feeble populations because they denied themselves the space to grow.

It's debatable (and we've debated endlessly on the subject) whether or not the tupi actually did himself a service or a disservice the way he planted his cities in that example.

There's pros and cons to close and far city placement.

Pros:

• Quicker for one city to be able to support the other with defensive needs or for response stacks to get from one of your defensible points to the other if the opponent changes targets suddenly.
  
• More cities means more buildings adding base commerces and so on means that while individually the cities are weaker, the nation has attained much more productivity per plot in its nation.
  
• Splitting plots more between cities limits growth while still giving more of the building based income sources so with limited growth comes less need for property control measures to be unit based so as to handle the excesses beyond what the city buildings can handle. Less pop, less problems.
  
• More commonly speaking, all the plots in the nation get used rather than having leftover plots that terribly weak filler cities MIGHT be able to put to use.
  
• Less distance to capital upkeep per city.
  
• More cities gain the benefits of overlapping resources that are counted in the city radius.
  
• Packing strong culture sources close together makes it harder for invaders to hang on to the territory they capture if they can't finish the whole nation off during the war.
  
• Putting cities closer to the resources they are intended to grab for the nation means they'll not often need to wait so long to get access to that further out resource.

Cons:

• More cities means more Num City Maintenance.
  
• It takes more cities to claim the same amount of land thus land claim processes can be slower.
  
• When you squeeze in cities and make two or more of your cities compete for the same plots, you limit their overall individual effectiveness. This is particularly a problem for specialized cities like your military building centers, which are given ultimately a lower production limit and therefore won't be able to pump out as many of the specialized high quality units as quickly.
  
• For those cities that gain wonder level local modifiers, they aren't deriving as much plot based base values to modify onto.
  
• Less food means less specialists so your cities won't be as fast to produce GPs either.
  
• More cities means having to maintain more defensive military.

For me, I prefer mega city planning at first and then allow the AI's more commonly packed in or not approach to further support the empire later.

 

[Yudishtira]

Those are old cities, with feeble populations because they denied themselves the space to grow.

This isn't vanilla. Population simply doesn't rely on working plots. I've seen an AI size 90 city on ice, so I don't think it is paucity of plots that is stopping those cities growing.

How long do you have to play a save before you can work the third ring? It took me well over 6 months on the one and only save I've ever gotten to that stage on. The defensive benefits of huddling cities together far outweighs any such considerations from my perspective.

 

[Faustmouse]

I have the same opinions about city spacing that TB has. You can reduce the cons if you first space your cities a bit further apart; like 5 plots in between or so. This will claim you lots of land and resources and allow them to grow faster to a size 6 city. Later your civics allow you to have as many cities as you want (not because of limitation, but because of rediciously low maintenance modifiers) so if you found a new city, it gets a net + extremely fast.

Your military city and / or GP farm can either be placed a bit further away from other cities, or you can claim a plot for a city while in a city screen and double clicking it. This makes the cities around your military city a bit weaker, but the main city stronger. However, have a look on what percentage of production hammers: and ) actually comes from plots, and what comes from buildings and specialists.

 

[LDK]

Just a thought based on my experiences in my current game.

The AI has a tendency to build its cities very close together. When conquering an enemy region, I frequently have to thin their cities out. Is it considering the 3 square radius that cities eventually achieve? As it stands, it ends up with an awful lot of cities, but they do not reach the potential that they could with a little more room.

If this is the case, then making them build a little further apart might have useful consequences for them defending, with defenders spread less thinly. Less cities, but better ones, might also have some performance benefits.

In the picture below, the Tupi have done themselves a mischief with this behavior. I wouldn't build that close even with the 2 square radius.

the problem lies elsewhere, majority of yields mid/late game come from trade and specialist. Towns built on river or coast are valid locations, hell i would build on every plot if i could if it had river and/or coast.

What AI has trouble with is constant anarchy by changing civics randomly or to match friends civics or on diplo request and wrong build order for its cities, then it fails to understand how to grow cities by picking proper civics (feudalism squatter camps -25% growth threshold) also AI gets wrecked by pollution as it builds every building there is and it stacks very very fast +chops forest to build farm to compensate that further increase pollution spread/growth.

 

[Soulmate]

Spoiler :

Okay seriously this session expired stuff makes me want to kill people. I jsut made wall of text with screenshots and it just cockblocked me. FU forum.

Okay, second try ...

I don't see any problem for AI settling too close, neither it is a problem for a player, you must use every bit of space. And you can make prosperous city on any terrain in this mod.

What does realy bother me is that Ai settles too FAR away(contrary to previous accusions) he often ignores free lands near his capital dotted with resources for far reach frontiers, which ignites rebellions wars, etc etc. Here is promininent example :

Spoiler :

Why bother settling on the continent with lots of resources when u can have this juicy arctic island, right?

Now, about the siege ram stacks of doom :

Spoiler :

He would still conquer the city with his 5 elephant raiders and 2 horsemans he brought, IF only he attacked with them AFTER the precious siege rams. Does AI consider them his top priority units or what ?
Here's some more fun with the rams:

Spoiler :

Poor chinese have a mad king with unnatural wooden trains obseesion.
Seriously though, they are aalone on the whole massive island and even barbarians are too far away, but instead of settling they chose to build rams, without even option to actualy put them to use. I could understand archers at least....

 

[Thunderbrd]

What does realy bother me is that Ai settles too FAR away(contrary to previous accusions) he often ignores free lands near his capital dotted with resources for far reach frontiers, which ignites rebellions wars, etc etc.

Why bother settling on the continent with lots of resources when u can have this juicy arctic island, right?

The AI is inspired to try to reach out to squeeze in opponent civ growth then build in behind it. It's not something all players would do but it IS effective at frustrating players.

He would still conquer the city with his 5 elephant raiders and 2 horsemans he brought, IF only he attacked with them AFTER the precious siege rams. Does AI consider them his top priority units or what ?

The problem is that siege needs to have its own AI type. This area of invasion AI is something that does need desperate attention but I'm holding off til next version cycle to address it in full along with a complete unit review of land military units. Sometimes they do well here and other times not so much. I could tweak a few numbers to see if I can get them to build a little better balance in their stacks in the meantime.