3 ways to improve the AI in the next patch

[Txurce]

The AI is much improved in my opinion, but there are three simple ways the AI could be improved in the next patch:

• spend most of its gold above an era-based threshold (flavor-dependent, not all on units; AI bonuses would need rebalancing)
• pillage when not attacking (barbs, too)
• escort settlers without exception

What do you think?

 

[hobbsyoyo]

I agree. I'm not sure about #1 though for all difficulties though....imagine a deity AI who spammed infinite, world-spanning COD and built every building in every city in every turn. They could do it too...damn cheating AI.

 

[Txurce]

Increased gold spending would definitely require rebalancing the AI bonuses. I'll clarify that in the OP, since that answers the posts before and after this one

 

[Glassmage]

Watch MadDjinn Carthage lets play and you see Theodora has 1000000 gold or something like that and didn't spend gold on anything. Imagine now if she does. Good luck winning anything.

 

[BobDole]

The AI needs more than simple improvements. These might help a little, but aren't enough. The only challenge, in my opinion so far (Immortal difficulty atm) is the early game, where the AI has a ridiculous advantage. After that, assuming you play well, they're a joke. They might pile up a lot of gold, but they tend to fall behind in everything else. And I'm not sure that they could even spend that gold effectively since there's only so many buildings and units you need. Oh, and naval combat AI is still underwhelming (even on an Archipelago setting they don't build a large enough navy).

 

[antoca]

When will the next patch be released?

 

[gunnergoz]

I concur but do feel that the AI players having some gold does allow for them to buy my luxuries for cash, when they have none of their own for trade. If the AI has neither spare gold or luxuries to trade, then my excess luxuries are pretty meaningless.

I see much less pillaging than I would expect from barbarians. Barbs also don't do anything that would prolong their miserable little lives, hanging out within city defense range for no good reason. At least if they died while pillaging, their raids would be somewhat meaningful.

The escorting of settlers is really important too. The AI does it seldom but when it does it usually manages to found cities. As it is now, if I see an AI settler solo in territory that I plan to use, I usually just declare war and take it, rather than have to deal with a city to reduce later. But an escorted AI settler is a different matter and takes more of your own resources to take out, if that is your intention.

 

[Txurce]

I concur but do feel that the AI players having some gold does allow for them to buy my luxuries for cash, when they have none of their own for trade. If the AI has neither spare gold or luxuries to trade, then my excess luxuries are pretty meaningless.

That's how I see it as well, and clarified the OP to better reflect it.

 

[PhilBowles]

Increased gold spending would definitely require rebalancing the AI bonuses. I'll clarify that in the OP, since that answers the posts before and after this one

Wouldn't it be simpler simply to reduce the gold bonuses it receives, since gold the AI isn't spending is gold it isn't using to help it win the game? As such they aren't really bonuses to anyone except the player luring the AI into bad trades.

I've very rarely seen unescorted settlers on Emperor at least; the exception appears to be if the settlers haven't yet 'set out' - I've captured settlers standing around at the borders of Attila's territory that weren't escorted. Although if anything they seemed to escort settlers actively more in vanilla than I've seen now.

They also need to add something that tells Attila's AI how the battering ram works, and also tells the other AIs that it's a priority target unlike the spearman (the AI will routinely shoot horse archers, convinced the ram is just a unique spearman that can't do much if its ranged support is gone). They can code AI behaviour to consider other specific unit types; coding it to recognize 'siege melee unit' as a different category should be straightforward enough, it's just that only Attila would need to make use of it offensively.

 

[Peng Qi]

The AI is much improved in my opinion, but there are three simple ways the AI could be improved in the next patch:

• spend most of its gold above an era-based threshold (flavor-dependent, not all on units; AI bonuses would need rebalancing)
• pillage when not attacking (barbs, too)
• escort settlers without exception

What do you think?

It already attempts to do 3. The only time you'll see a settler with no protection is if that settler has embarked; since an embarked settler and an embarked unit take up the same "part" of the tile, the AI embarks the settler first and the escort on the next turn. This is why the AI couldn't colonize overseas on release; it would embark the settler, then have the settler "wait" until its escort was on the same tile it was again, which was impossible, so it waited forever and the escort would just sit there on the coast staring at the settler until something came along and sank the transports.

EDIT: Also in G&K I rarely see the AI racking up money into the thousands. It usually spends every penny it gets above 15K in the mid-game and spends like mad on city states once the UN is built until it's broke. Could probably be tweaked to spend more of that 15K though; it probably shouldn't sit on more than 1K.

 

[Txurce]

Wouldn't it be simpler simply to reduce the gold bonuses it receives, since gold the AI isn't spending is gold it isn't using to help it win the game? As such they aren't really bonuses to anyone except the player luring the AI into bad trades.

You would actually do both, since the benefit of spending its gold is big enough that the normal AI bonuses would need to be nerfed. The AI will always buy luxuries, but is more selective about strategics. Most of the extra gold it would spend should go toward improvements like buildings, which would help its overall performance.

I've very rarely seen unescorted settlers on Emperor at least; the exception appears to be if the settlers haven't yet 'set out' - I've captured settlers standing around at the borders of Attila's territory that weren't escorted.

I play on Emperor and Immortal, and see less unescorted settlers than I before, but recently ran across two captured American settlers. They may well have been captured before leaving their home borders.

They also need to add something that tells Attila's AI how the battering ram works, and also tells the other AIs that it's a priority target unlike the spearman (the AI will routinely shoot horse archers, convinced the ram is just a unique spearman that can't do much if its ranged support is gone). They can code AI behaviour to consider other specific unit types; coding it to recognize 'siege melee unit' as a different category should be straightforward enough, it's just that only Attila would need to make use of it offensively.

Agreed. This is probably harder to code, which is why I focused on stuff I know can be done without too much trouble.

 

[smallfish]

Yes they do escort their settlers, with Tanks if necessary. The only problem seems to be on the water where their escorts will outrun the embarked settlers and ground units.

 

[Zaimejs]

One thing I notice is that AI has a tough time taking cities. I don't know what was going on in this one game I played. Arabia had a Roman city surrounded with advanced troops for 10 or more turns, and it never took the city. Maybe it's a lack of siege? Or something else?

They just need to realize when they are way stronger than an opponent and learn to wipe them out.

 

[smallfish]

One thing I notice is that AI has a tough time taking cities. I don't know what was going on in this one game I played. Arabia had a Roman city surrounded with advanced troops for 10 or more turns, and it never took the city. Maybe it's a lack of siege? Or something else?

Did they bring in enough siege units, and are those units getting hammered into oblivion? That's the crucial part.

In that, AI needs to learn how to maneuver its ranged units behind melee screen when attacking cities.

 

[Jaeger5]

I would like to see a sphere of influence. Now i play on Play the World G&K edition map. I get tired of telling no to say...Korea asking me to go to war against Egypt and I'm playing Sweden. How in the Hades do I even get over there with a 5 unit medieval army? It took my ship 20-30 turns through barbarian infested waters to get over there. And why are you POed at them with you squeezed in between China and Japan. Am I getting gigged for not helping you in this impossible task? Now, if you want help with Russia, Germany, France, England, I might think about it. But it appears I'm going to do all the fighting since you're on the east coast of Asia.

Sphere of Influence script: The AI won't ask me to go with war with somebody outside say...a 10-20 hex range from my borders.

 

[PhilBowles]

I would like to see a sphere of influence. Now i play on Play the World G&K edition map. I get tired of telling no to say...Korea asking me to go to war against Egypt and I'm playing Sweden. How in the Hades do I even get over there with a 5 unit medieval army? It took my ship 20-30 turns through barbarian infested waters to get over there. And why are you POed at them with you squeezed in between China and Japan. Am I getting gigged for not helping you in this impossible task? Now, if you want help with Russia, Germany, France, England, I might think about it. But it appears I'm going to do all the fighting since you're on the east coast of Asia.

Sphere of Influence script: The AI won't ask me to go with war with somebody outside say...a 10-20 hex range from my borders.

You can still declare war to make them happy - chances are Egypt can't reach you either. You probably won't get the "common enemy" bonus if you don't do any actual fighting, but you won't get a penalty for declaring war but not participating.

Still, that's a good idea - or more simply still, the AI won't ask you to declare war on civs that don't share your continent(s) until you've researched Astronomy.

 

[Piousflea]

I completely agree with #1 and feel like AIs need to be rebalanced.

AIs piling up 100s of gold early game and 10,000's of gold late game makes the game MUCH easier for players (because you can trade for the gold, and because they have less units and buildings than they should) and also makes the optimal singleplayer builds feel very artificial. Rushing for excess copies of a luxury or SR that you already have is only useful because you can trade for big time gold, and the ability to easily purchase settlers for 500g makes the "bonus settler" from Liberty feel much less powerful.

If AI's got less bonus gold but would always spend that gold on things it wanted (buying craploads of buildings and units), it would change singleplayer strategy a lot, and make it resemble multiplayer strategy a little more.

 

[antoca]

When will the next patch be released?

4 months and still no patch... what happened?

 

[Deaths Advocate]

It was released!

http://forums.civfanatics.com/showthread.php?t=479298

 

[bcaiko]

One thing I notice is that AI has a tough time taking cities. I don't know what was going on in this one game I played. Arabia had a Roman city surrounded with advanced troops for 10 or more turns, and it never took the city. Maybe it's a lack of siege? Or something else?

They just need to realize when they are way stronger than an opponent and learn to wipe them out.

Firstly, taking cities is much more difficult now. Barring that, the AI really does not like to lose units, and pre-patch will attempt to heal them while taking more damage...effectively losing a war of attrition. The patch is supposed to fix that. Though I still don't see the AI using suicide runs to take a city when necessary.