Community Feedback Needed: Biggest AI issues

As it has been since the very first Civ, the game is all about the first hundred turns. Get those right and many of the other problems will take care of themselves, with a couple exceptions.

As far as I can tell, the AI is building workers/settlers too early (given the higher costs of these in FFH and the longer time it takes to grow the subsequent population due to the worker tech delay). This causes them to also often miss the chance to get a couple scouts out early enough to survive/pop huts and to build an adequate warrior force to fend off the first barb/human rush. City defense bonus for warriors was a great idea, but why not have an earlier archery tech so AI can get them early enough to defend their cities with them?

BTW, letting archers get the bronze boost might also be a good idea. Adding 1 commerce for river forests could too, since chopping is so late. I disagree that the AI should beeline Bronze - economy/happiness (?) is where they are hurting.

In the next phase of the game, in general City States is too strong and/or the financial penalty for expansion is too high. If this won't be changing, the AI needs to have some threshold that makes them get Education ASAP and switch to CS, as their economy noticeably starts slowing when upkeep becomes an issue.

Prioritizing Trade would help, and this also means they should be able to build a few horsemen on that tech path. Maybe let these be built without stables or make stables cheaper. The Great Library is also on this path, and the AI should put a much higher priority on going for this - even if they miss, they get a nice cash infusion.

When new unit techs become available, they won't build many/any, since their economy is already hurting and they can't afford the upkeep. Later on, they can't figure out how to upgrade adepts because doing so requires a mage guild. I don't think cash is the issue as they usually have 100-200 around to trade.

City size is another issue I've noticed. Late game they'll often be stuck at 13-15. Sanitation/Public Baths should have a higher priority (for non-FoL civs, the +1 for farms is huge, letting you switch to Foreign Trade), and I'm guessing Gambling Houses would net 5 happies.

Finally, on any of the unique features that make this game so fun for humans, the AI has no clue. I.e. I've seen Loki wander aimlessly through my territory for 200 turns without once spending the night in a city. Given this disadvantage, perhaps it wouldn't be so bad to let the AI build Archmages/High Priests from scratch or upgrade without buildings...
 
Team AI doesn't know how to handle Barnaxus. I was on a team with the Elohim when big B got killed by bears (bears!) and before I could grab him, an Elohim horseman picked up his pieces and ran off into enemy territory! Poor guy was never heard from again...
 
BTW, letting archers get the bronze boost might also be a good idea. Adding 1 commerce for river forests could too, since chopping is so late. I disagree that the AI should beeline Bronze - economy/happiness (?) is where they are hurting.

Maybe set up an AI vs AI game both of the same Leader but with one leader set to prioritise bronze working and another set to prioritise an economic line. Rinse and repeat for e.g. 30 games and if one comes out vastly superior.
 
Bannor should be taught how to switch into their Crusade civic during war, since they are the only one who have a civ specific civic, and a powerful one that is. It is sad to see the Bannor does not bother to use it properly when their tiles are full of mature town.
 
1. AI should not use Rangers and other units with a city attack penalty to attack cities. In my latest game at Monarch level the Bannor and Svartalfar both over-favoured Rangers and sent them in useless waves against my cities.

2. The AI is poor at city breaking. Even when the mix of units is good, the AI rarely uses fireballs or catapults to reduce city defence to zero before attacking.

My process for taking a strongly defended city would be something like:
a) reduce the city defences to 0% or thereabouts with fireballs/catapults.
b) if your strongest attacking unit does not have odds of over 90ish% against the defender then use a couple of summons or throw-away units to soften them up. Alternatively, just use fireballs or catapults to soften up the defenders.
c) with each attack make sure that the odds are high. I rarely attack with odds less than 65% and almost never at odds below 50%. If you're not reasonably likely to win then wait, otherwise you are just wasting units.
d) make sure you have sufficient units to hold the city against counter-attack once it is taken.

I know that's fairly complicated to turn into code, but too often I've seen the AI throw away wave after wave of units. This just gives my defenders stackloads of xp and makes their next poorly executed assault even less likely to succeed. The occasions when the AI does take cities are usually down to sheer weight of numbers or my failing to anticipate it.

I think the kamikaze nature of the AI is often what leads to their cities being poorly defended, one Warrior in the capitol, etc.
 
I'm actually really surprised this hasn't been mentioned yet. I generally play on Monarch or higher difficulty level on large or huge maps, and later in the game have massive amounts of workers. By this point in the game, I have better things to do than stand around and tell my workers what to do, so I automate them. 20 turns later, I'll have cities that have nothing but forts on their land, and I'll have workers building workshops over my food resources, etc. I think I recall reading something saying workers would now check to make sure they weren't too close to another fort before building one, but it doesn't seem to work on my automated workers. Maybe there should be some way to customize the automation, at least a little bit so they're not going around starving my cities and ruining their economies.
 
...so I automate them. 20 turns later, I'll have cities that have nothing but forts on their land, and I'll have workers building workshops over my food resources

There's a setting in Options to have your workers not destroy existing imrovements. I've had this on since vanilla. Also, the option to not have them chop forests is nice if you're the elves, although I don't know if the workers are smart enough to build improvements on top of forests with that on.
 
I'm actually really surprised this hasn't been mentioned yet. I generally play on Monarch or higher difficulty level on large or huge maps, and later in the game have massive amounts of workers. By this point in the game, I have better things to do than stand around and tell my workers what to do, so I automate them. 20 turns later, I'll have cities that have nothing but forts on their land, and I'll have workers building workshops over my food resources, etc. I think I recall reading something saying workers would now check to make sure they weren't too close to another fort before building one, but it doesn't seem to work on my automated workers. Maybe there should be some way to customize the automation, at least a little bit so they're not going around starving my cities and ruining their economies.

Ths is exactly the reason why I don't like to automate workers (although in past civ versions, I've used automate fallout cleanup and automate routes to convert roads to railraods). I know it involves more micromanagement to not automate, but I'm used to doing it by now.
 
Well now I feel like a huge noob, thanks for the tips guys. However, there should be a way to make it so workers build the optimal improvements to keep cities functioning at their maximum capacities while of course keeping you supplied with resources as well, and not spamming forts.
 
Just finished a game last night, the AI really needs to know that cities and ice don't mix. I was assembling my stack 'o doom in a conspicuously visible spot, and Varn not only didn't do anything to counter, but he built two lightly defended cities with nothing but ice in the BFC (it's not like we were friends either; -10 and furious in the diplo). If he would have focused his resources, he could have held out a few more turns.

Also, there was a beautiful castle on a hill at an impenetrable choke-point lying completely empty. Dunno if the AI can understand choke-points, but with just a handful of longbowmen and some fireballs/sand lions and I would have actually had a fight on my hands.
 
I want to mention something, which was not posted yet, i think.
Since the Religion FoL ist usually founded very early in the game, it is most likley that it will spread to non-elven civs, who will convert to this religion. When this happens, there are often ~60-70 percent of their empire covered by woods. FoL of the leaves will convert them to ancient forests.
The bad point is, that all of these non-elven civs wont chop the forests anymore. That means, that more than half of their plots will have poorly 3/1/0 or 2/2/0 yields. This is definetely too little to run an empire in the middle- or lategame. So i think is very important to tell the ai, that these forests must get chopped to build improvements.
 
I suspect geography plays into this a little (particularly the founding-cities-on-ice thing mentioned earlier), but based on my last few games Decius appears to be just terrible, even for an AI, at city placement and expansion generally when he's running the Malakim.

In the most recent, he did neatly execute a hunter/catapult attack on me despite that, but if he'd had more than one city that was any real use it would probably have gone better. Generally he seems to have some issues or weird priorities about expanding.
 
Regarding Sidar and Wane, the AI should consider units like the immortal and marksmen so they don't wane them after upgrading them. Or before for that matter. If they're reasonably close to the tech for them they should hold on to four of the units that can upgrade to them.
 
After playing some high-difficulty games some things become obvious which are real major issues.

Let's start with the general things that cover all civs.


first and foremost:

The AI doesn't aim to win. At all (even on Deity it has plenty of resources but no aim to wrap the game up whatsoever.).
And it really should. Since there are quite different options which should cover most situations it should also be possible to find something for every civ and most situations.
Thats the first thing you should focus your attention on. Its the one thing that should most seriously enhance the AIs ability to compete.
If it will try that in some way other than on accident players really will need to tone down the difficulty at least a bit.

To fix this lets have a look at the conditions available:


Altar: All Civs which use alot of Priests or Desciples and are good or neutral should definately try for the Altar of the Lunator. And hard. It yields really powerful benefits in Economy as well as the possibility to win (And even if AI doesn't suceed in building all even getting to step 3 or 5 will yield some serious economic + Military benefits. So nothing wrong for the AI if it tries in Vain.).

Civs that instantly spring to mind fastly in that regard are:
Malakim (what a surprise...), Elhoim (another really odd one, no?), Khazad (get some milage out of priests as wel), Kuoritates (due to being a powerful builder and being able to use the Benefits more tangiably due to fewer cities), Sheim (if they oddly turn neutral or good during the game. Especially if lead by Os-Gabella. If they later turn veil its still very good allowing to build Savants upgrading to Mages very fast among other things.), Embers under Jonas if they turn good or neutral, for Ljasalfar it might be ok as well, especially under Arendel. Mercurians also should have some serious incentive to go that route since they really have some leaning gamewise if not flavorwise.
Also its the best Condition to try for smaller empires where the other options are out of reach.

This is also a victory condition which is very viable / suitable to rather peaceful, trade-happy, tech-focused empires.
Sidar (because Shades are much easier to get in a city with the altar which is a real boon to them. Since Priests also Level on their own. Perhaps also fits them not so badly flavorwise) / other industrious civs could be made to try it as well due to being able to get the final one way cheaper.

Also the Cities with a higher level altar should be more prone to producing desciple Units than others (and at least the State Religion as well as perhaps others should really be spread there) since that makes the desciples built there potentially very powerful. (Like abovementioned Instantmages from Savants.)

Grigori even if not very "spiritually inclined" and without much benefits from the respective techs should seriously get their Lunator beeing able to construct the altar back in some way or the other (and they don't get much from those techs so some edge on that victory might seem fair) and go for that victory after SoW is researched. (Unlike the others they should aim for that one rather late though.).
And its the single most fitting civ going for that victory flavorwise. Its the ultimate in human empowerment after all. Even if going a bit farther than cassiel might have hoped... :p

Interesting Civics here are: Theocracy (and to a much lesser extent republic or God-King depending on size of Empire) / Pacifism / Arete, Slavery or Caste System / Agriculture or Guardian of Nature / Public Healers / Overcouncil


Religious: Every Civ that has founded a Religion early + built a shrine (which it should build more reliably) and converted some civs / allready has some considerable religious influence should seriously try for that victory. And fast. Time is a big factor here.
Thanks to inquisition (which the AI also needs to start to use if running a focused relgious game) this one should be quite achievable (especially if one big block forms). This is especially true for civs with a religious leaning. Looking above for the list of altar-lovers might give an insight.
Unlike going for altar this should rather be pursued by larger empires though. (Those should use Theocrathy and try to impose (by means of force if necessary) their religion as well as the civic on as many civs as possible and use Desciples quite liberaly. Also Purge the Unfaithful might be good for them.) (Followers of the founder should perhaps accept their founder going for that victory. Since flavorwise it is a Victory for them as well more or less. Perhaps big religious blocks vieing for religious dominance might be intersting)
Members of the Undercouncil might try this one perhaps since open borders can be enforced and CoE is spread rather easiely and hard to resist. Those Civs which are CoE and try for this one should also unlike others use the Spread liberally. Yields a lot of Gold from the shrine as well.)

Interesting Civics here are Theocracy (what a surprise...) / Nationhood , Social Order, Consuption or Crusade / Aprenticeship, Slavery, Caste System or Guilds / Conquest, Mercantilism or foreign Trade / fitting Compassion / Undercouncil.


ToM:
At least all Arcane Civs should try on this one if they at least own 4 Nodes. (A routine for building the towers and dispelling the mana might be hard to teach but it still sounds at least possible.) after Researching Arcane Lore / SoW.
So Sheaim, Amurites, Balspheraps, Svartalfar are obvious here. The towers are good in their own right and more
That could be done in conjunction with teaching the AI how to use magic properly. Others might fit as well.

Amurites should try this one from 3 Nodes ownership upwards since they start with Metamagic. Also they should start for it as soon as they get the magical-spheres/aspects-techs not just later since they get way most milage from the towers. And that should be their prefered victory since they have it easiest to do and also beeline that path anyways allready (at high difficulties they should be able to build all 4 aspect-towers before even reaching sorcery. Let alone SoW. That whould set some serious pressure on the player / other AI to stop them from winning by arcane might.).
Sidar could try it as well due to being industrious and having some advantage from different Mana (even though that one is far less interesting than the Altar victory for them.).
Grigori could also try that one if its a bit later in the game and much nodes are available.
All civs owning 6+ Nodes should perhaps try this one. Since its classic Builder, and the AI can rather easiely grasp how to build, it seems. Exept Khazad and perhaps some more simple Civs naturally...

Interesting Civics here are: God King (for smaller empires) or Republic / any fitting cultural / Slavery, Serfdom or Guilds/ any fitting econ/any fitting compassion/Undercouncil (for Slave Trade + free sience and immunity to tech-bullying.)/


Cultural:
All Civs leaning towards Culture / Bards / Recon could and perhaps should try this if resources are available and some Wonders or other Factors run nicely.
Balspheraph, Sidar, Svartalfar, Kuoritates should be quite suited for that one.
Favored Civics then should then be Religion or Liberty / Republic / Caste System / Foreign Trade or GoN or Agriculture / Public Healers or StW. Need not be to hard but are definately helpful.
Hall of Kings, Grand Menagerie (if possible in any way) and Theatre of Dreams sure help achieving that victory. As does having many religions and there temples in the cities. (Civs aiming for that victory should try to amass many religions. and focus their attention on some core cities.)
Also Civs in the Overcouncil might start to try on this one once Liberty is imposed councilwide.
Especially if done rather early.
Grigori might try as well in abscence of the possibility to win religious. (another one among the many reasons to give Honor / Deception and access to the councils back to the Grigori. Without overcouncil Cultural is so much harder.)
Due to the whole setup of the Civ and playwise this one should be one of the prominent Aims of the Kuoritates. Big Cities which are concentrated on surely helps achieving that victory.
Members of Overcouncil should perhaps also accept someone reaching cultural since it fits their flavor and is most likely to go that way.
Putting up the Culture-Slider and building Culture-Boosting Buildings might help to get there.



Domination: Civs that can easiely support huge empires and are powerful enough should try this.
This should include all followers of the Order as well as Veil and to lesser extent might be FoL, The Bannor and other Military Heavy Civs at least. Or anyone powerful enough to be reallistically able to round this one up. Other Civs might come to mind.

Civics which are interesting here are: City States (in times of peace) or Republic (otherwise) / Any fitting cultural which yields most happy/ any labor thats fitting the circumstance / agriculture or GoN (for pop) / Stw or public healers (depending on alignment) / most fitting Council but should lean to council and try to lead one.

Dominators should go for Wars rather late when blocks are formed (and the situation is clear) and Expansion is harder to do peacefully (they should expand "peacefully" very agressively / expansionist though.). But shouldnt restraint from taking something from a weak target if the opportunity araises.
Wars by dominators should aim for capitulation first and foremost and should aim at taking cities as oposed to razing them as well as be lead rationally and ended if its convenient. Partial Gains are fine if they offer a good deal.



Conquest: All the Civs that are more Militarily inclined (as represented by modifications to War-Weariness) could try this one if one of the other victories is out of the question. So clearly infernal+ Sheaim but also Calabim, Embers, Doviello, Bannor, Malakim, Khazad, Hippus and Mercurians might try for that especially one if nothing else is available. Should be prone to raze (depending on alignment) as oposed to domiation which should capture.

Obvious Civics are: Despotism or Theocracy, perhaps Republic / Nationhood, Social Order or Crusade / Aprenticeship or Military State / Conquest (could really feature reduced War-Weariness btw.) / StW, Fend for themselves or no care / Undercouncil or better no Membership to not be restrained by a council. (That should only be a rough guidline though.)
Oh and Warfare. Lots of Warfare... :p The more fighting and the bigger, the better. World war is all fine if you want to take the world by conquest.

Conquerers should use force as the main way to expand if any opportunity araises and should war early and decisively. Elimination of Rivials should be priority. (Capitulation should be accepted though if potential Vassal is beaten weak enough. Elimination might be an option as well naturally. Especially for harder to controll civs.)
Razing should be the preference and resetteling burnt territory the better way to expand (unless the prize is very interesting like a big city / wonders or a really good spot). But War weariness and revolts just slow down in question. The War machine should be kept going and support down / advance up.
Important thing to note: Concerers should really dig diplomacy. Simply fighting everyone at once doesn't make one the only civ on erebus that lightly. Even Hyborem knows that. :p So one at a time or at least managable wars. And Friendly / pleased ones should rather be last targets. Trading can help in warfare.


Oh and for the sake of the dominators and conquerers make War-Weariness less severe in some way (perhaps tied to the Armageddon-counter). And really go away faster. Its really a game stopper for all but the most extreme militarists. Only real peace-lovers should feel some sort of pain later on. FFH2 seems more combat focused in flavor and feature. This will help Players and Ai and perhaps makes waging war more fun. Im confident you may again find a way to even make an unfun feature a fun one via some tweak of mechanics. You have a real history for that. :)


Lastly Time: Civs that can't get one of the others or see other civs racing towards one of the other victories fast should aim for that one or at least try to prevent others from winning if its ok from the flavor side. (like everyone declaring on builder of ToM and evils on builders of the altar. Expanded on the other conditions if appropriate). Alliances should form under fitting circumstances against the whould be-winner.
Clear and major score leaders are perhaps the ones who should try for that kind delays actively especially if no other victory is near but the time-limit is. They should also try to dominate one of the councils and give resolutions which bind others civs attention like Councilwide-Proxy-Wars and the like. Also they should naturally try to inflate their score in any way possible. Especially if the end of history is near.

If this is possible in any way AI should recognize if a condition is disabled and don't try that route if its not possible. No point in building the Tom and inciting world-anger on you without any gain after all. The Altar is the exeption since the last step is really grand. So going for that one still makes sense

More later. Other very serious issues are Use of the Seas / Naval Part. Good usage of promotions and hatching of high-level units and usage of the magic system + Promotions.
Later still more on my favorite civs which include the sidar and shading. To tired now to continue and elaborate on that points. But will.
 
As it has been since the very first Civ, the game is all about the first hundred turns. Get those right and many of the other problems will take care of themselves, with a couple exceptions.

As far as I can tell, the AI is building workers/settlers too early (given the higher costs of these in FFH and the longer time it takes to grow the subsequent population due to the worker tech delay). This causes them to also often miss the chance to get a couple scouts out early enough to survive/pop huts and to build an adequate warrior force to fend off the first barb/human rush. City defense bonus for warriors was a great idea, but why not have an earlier archery tech so AI can get them early enough to defend their cities with them?

Ive run maybe 20-30 AIAUotPlay games and I think you have hit the nail ont he head. The AI really isnt ready for the danger in the first stage of FfH and overexpands rather than growing his infrastructure to handle the challenges that will come.

Im working on getting the AI to expand more slowly and focus more on gettign a military backbone early on.
 
So, I've never had the AI get rolled by Barbarians. Is this more common on lower difficulties, or are we assuming raging barbarians? Only time it occurs is when the Pyre gets owned early on.

I'd love to see the Bannor use Crusade... but I think they'd more than likely own themselves. Doesn't it prevent you from going to peace? AI's don't know how to finish wars, so I doubt this would be a good thing.
 
It can happen if they start too close to barbatos, but that is a more extreme case of bieng rolled over by the barbarians.

Anyways, I agree with the war weariness, it's too harsh.
 
So, I've never had the AI get rolled by Barbarians. Is this more common on lower difficulties, or are we assuming raging barbarians? Only time it occurs is when the Pyre gets owned early on.

I'd love to see the Bannor use Crusade... but I think they'd more than likely own themselves. Doesn't it prevent you from going to peace? AI's don't know how to finish wars, so I doubt this would be a good thing.

Its not getting rolled by the Barbs as much as getting rolled by players. Spending the first 100 turns on settlers/workers/growth techs make them ripe targets when the player comes calling, well past their their capacity to defend. Meaning the players best strategy isnt in expanding himself but in taking the already beautifully developed land the AI has been kind enough to provide for him (and still defended by warriors/scouts).

As for crusade thats a specific function that may benifit from tweaking but right now I want the big global issues (trust me we have them). War Weariness may be a good thing to modify too, but thats a balance issue, not AI.
 
Well the AI seems to go into one or two modes early on. Either it overexpands with minimal defence (often forcing the human player into rushing it because it grabbed too much land) or it decides to warrior/scout rush. The trouble with these rushes is not so much that they kill you, as the AI refusing to accept peace so the war often drags on, retarding the development of both combatants, until some other AI dogpiles in and claims most of the spoils. I think the frequency of these rushes should be toned down because they generally benefit some other AI than the original rusher.
 
Its not getting rolled by the Barbs as much as getting rolled by players. Spending the first 100 turns on settlers/workers/growth techs make them ripe targets when the player comes calling, well past their their capacity to defend. Meaning the players best strategy isnt in expanding himself but in taking the already beautifully developed land the AI has been kind enough to provide for him (and still defended by warriors/scouts).

As for crusade thats a specific function that may benifit from tweaking but right now I want the big global issues (trust me we have them). War Weariness may be a good thing to modify too, but thats a balance issue, not AI.

Ah, okay, this I'll agree with. I think the major disconnect with the AI, comes from the fact that early aggressive units are much stronger at capturing cities than the AI's give them credit for, so think that a warrior or two can defend a city. Anyone whose played an aggressive civ knows they can easily wage war once they get to education, because combat 1 + 2 XP = shock which is a real city buster.

Edit: I disagree that war weariness is too harsh, by the way. FFH gives you access to Despotism and Nationhood from right out the door, and some civs even have a bonus to negating WW. The real issue is that people are afraid to switch civics during wartime. Maybe if Despotism didn't have the maintenance nerf, it'd be a more acceptable alternative?
 
Back
Top Bottom