Community Feedback Needed: Biggest AI issues

[jacktannery]

The problem/great thing about FFH is that each civ has a distinctive style to play and win. Kael, you have a huge number of devoted fans who are excellent at programming, as can be seen in the numerous scenarios and mod-mods.

Why not ascribe each of your civs to one particular individual who is good at programming, and have them improve that civ-specific AI, allowing you to concentrate on the barbs? Have another competition, as the last ones were quite successful.

 

[xienwolf]

Actually if we had enough interested programmers I would be thrilled to see an AI battle Contest. Each programmer is assigned a specific Civ to code an AI for, and then they are placed in an auto-play scenario to see who can win.


Problem is that the AI is written to be very general, not Civ dependant. And with the diversity which FfH is attempting to provide, the AI needs to remain very general, even though the Civs are each very flavored. Otherwise you are writting a special AI for each Leader/Civ/Religion combination, so roughly 80 different AI programs, each of which would need completely re-written if the stats of even a single building shift too much (since that would possibly throw off the overall strategy goals, and you are no longer general enough to be basing it off the stats alone).

 

[jacktannery]

Good point. Oh, well.

 

[Chip56]

Up to Four. Why would you stop at getting the second level of a spell sphere if you could get the third?

There shouldn't be any limit at all.
Affinity for the win!

 

[Turinturambar]

The two biggest AI problems that make this game too easy on every difficulty level:

Bad early game strategy: The AI will build a couple warriors and then start a very early expensive settler. This delays growth for a long time, reduces the tech rate and the second city won´t be useful due to lack of buildable improvements. Most importantly this behaviour leaves any civ open to a warrior rush even with nonagressive/nonraider civs (up to immortal level at least, deity too probably).

Fix: reduce the cost of settlers to something more sensible?

poor research priorities in the early game: Early on the two most important techs by far are bronze working and education. Bronze working is a big boost to the economy and military and education helps you get out of the dark ages with little commerce faster. The AIs tend to delay these absolutely vital techs for a long time.

Fix: enable all workers to chop forests from the start; make cottages buildable earlier (new starting tech?)


Smaller stuff:

I´d really like to see a scary Hyborem sometime. Can his immortality be changed to the one of losha valas and additionally he gets some kind of always attack script? He might actually conquer an empire then. Or just give him a rather impressive army when he gets summoned.

 

[DemonMaster]

I´d really like to see a scary Hyborem sometime. Can his immortality be changed to the one of losha valas and additionally he gets some kind of always attack script? He might actually conquer an empire then. Or just give him a rather impressive army when he gets summoned.

I have played this mod (various versions) since ever and for the first time (not an exaggerating) Hyborem was on the move. He has so far conquered about 30% of the world and converted all (including me) except a sturdy dwarf (and Basium of course) to AoV. The game was a living hell lol to play. The reason can be that he started on a pretty good peninsula and two smaller civs as neighbors.

PS
And the damn bastard stole (his world spell) my greatest city. I have build (and prepared) for a really good city with a thousand slums. The city was also the home for the CoE and the wonder that allows you to heal one unit per turn

 

[Avahz Darkwood]

Limit the building of basic religous units to 4(?) or some other limit (unless granted via bonus, spell, etc) and allow more religious spread via trade.

(I know the problem with this is that it would limit more advance religious priest once they are available, but the number cap can be removed then)

 

[Verily]

Limit the building of basic religous units to 4 (unless granted via bonus, spell, etc) and allow more religious spread via trade.

(I know the problem with this is that it would limit more advance religious priest one they are available, but the number cap can be removed then)

Easily solved: just make the lowest-level priests national units, but not the others, and lower the priority of the higher-level priests (as otherwise the AI might try spamming those instead).

 

[MrPopov]

I believe some of the bad city placement issues are coming from the fact that Everyone is thinking like the Kuriotates' Sprawling trait. Also I believe the starting locations are being normalized that way too. I've noticed my precious food bonuses being too far for my capital's BFC leaving me with a mediocre city. That is less of an issue though because of the starting settler bonus (thank you so much for that, I wish vanilla civ had this!)

 

[Nimbus]

Limit the building of basic religous units to 4(?) or some other limit (unless granted via bonus, spell, etc) and allow more religious spread via trade.

(I know the problem with this is that it would limit more advance religious priest once they are available, but the number cap can be removed then)

even more easily solved, make the low level disciple a national unit (qty:4) but instead of sacrificing the unit to spread the religion they cast a spell in a city to spread its religion(but no gold cost like CoE)

 

[Kael]

Blocking Wane makes sense. What makes no sense is that I went into Worldbuilder and gave them Shades, in their only city, and so many they would lose money on support, so had to disband, settle, or live through a strike. They settled NONE of them (one went off to protect a worker... not sure about the rest).

Sure, but you must admit thats its not really worth the effort to program the AI to use the shades if it isn't possible for the AI to have them outside of your worldbuidler experiement.

 

[MagisterCultuum]

Would changing the <UnitAIs> fix this? currently they have <DefaultUnitAI>UNITAI_ATTACK</DefaultUnitAI> and

<UnitAIs>
				<UnitAI>
					<UnitAIType>UNITAI_ATTACK_CITY</UnitAIType>
					<bUnitAI>1</bUnitAI>
				</UnitAI>
				<UnitAI>
					<UnitAIType>UNITAI_RESERVE</UnitAIType>
					<bUnitAI>1</bUnitAI>
				</UnitAI>
				<UnitAI>
					<UnitAIType>UNITAI_COUNTER</UnitAIType>
					<bUnitAI>1</bUnitAI>
				</UnitAI>
				<UnitAI>
					<UnitAIType>UNITAI_CITY_COUNTER</UnitAIType>
					<bUnitAI>1</bUnitAI>
				</UnitAI>
			</UnitAIs>

Since Shades cannot attack, this seem wrong. None of these AIs really fit Shades. Perhaps changing the Shades' AIs to great person AIs would fix the problem?

 

[hbar]

The AI as Sheaim attack with their pyre zombies last, not first (for the collateral).

On a plus note, I just had an AI radiant guard cast shiny light on me, 1st spell I've ever been affected by (and man is that an irritating spell)

Also, the AI as Sheaim adopted Kilmorph, and then Empyrian. Not so much AI as flavor.

 

[xienwolf]

Sure, but you must admit thats its not really worth the effort to program the AI to use the shades if it isn't possible for the AI to have them outside of your worldbuidler experiement.

Valid point which I simply hadn't thought about

 

[vorshlumpf]

Agreement on the initial strategy. I think it's ridiculous to get a recommendation to build a worker/settler right after founding a city. If the AI is guided by these recommendations, then that would explain why it has so much trouble fending off barbarians in the early game.

Better stacks would be nice. Barbarians, especially - if they have individual units within a square or two of each other they should group up instead of hitting you one at a time.

Healing, as has been mentioned many times. Since the AI will often throw its units at you with horrible odds, it occasionally gets out of combat with a very nice XP reward. However, this is eliminated when they attack next round with their unit still injured.

I second the suggestion for more pillaging when assaulting a city or fortified unit is low-odds. Nothing frustrates me more than not being able to do something about enemy units running rampant in my flower beds.

Better unit selections. I've seen stacks of disciples used for combat, including Varn's Lightbringers. Having the AI build those required buildings and get the required techs would certainly help.

More sensible diplomacy. I have had Friendly allies refusing to gift me their maps when they would sell them to me for 5 gold. I'll gift an ally an expensive tech (at their request) and they'll refuse to gift me some gold right after. I feel the relationship is largely one-sided. (I'm generally unhappy with diplomacy in Civ IV, anyway, so this it not so much a FFH thing)

More conditional strategy would be nice, but I'm sure would be quite time-consuming to implement. For instance, elves terraforming to get forests under their improvements. Other civs Blooming when they adopt FoL, at the very least on any non-worked tile so they can get Ancient Forest and potential treant defenders.

 

[MagisterCultuum]

For instance, elves terraforming to get forests under their improvements.

Not really relevant, but I still think that the elven civs should be able to cast bloom on already improved tiles instead of needing to pillage, bloom, and rebuild. This isn't hard to implement at all if you use python instead of only XML/SDK.

 

[Jie]

What I always found annoying is AI building cities (not just the Kurios as mentioned earlier) too close to each other, and not being able to utilize the terrain properly. I find myself razing a lot of cities I capture.

I think it's been mentioned but there's the stack of doom issue. If I damage the stack with rings of fire, they don't move forward with the attack, even though they would prob still be able to overwhel with the sheer numbers.

In addition, sometimes they do crazy suicide attacks with the stack of doom, that there is no chance of them breaking through my city, so better to have them not attack than to attack.

 

[Tirailleur]

AI should not be able to found city without unit to protect it. In my game Lannun don't stopping to found a city on my continent and each time the city is empty. I declare war and raze the city without any combat then i asked for peace and this stupid Falamar accept and give me gold I ever have razed three time this city and get three times a good amount of gold for the peace...

Fix :

AI do not found city without defense

AI do not accept to pay gold for peace if there are no threaten on it

 

[Yong]

AI should protect their Hero units better.

In 0.32 they won't attack if win rate < 95%, this is good but not enough.
Probably one additional rule is need:
After the combat the Hero should be able to rest in a protected tile.
Currently it is not rare to see AI Hero winning a combat with few HP left, standing in a tile alone, then got killed in the next turn.

Furthermore, I'm not sure if AI knows that their Hero got 100 free exp points.
Maybe they should consider waiting a few turn to let Hero gain one more level before enter a combat.

 

[Bad Player]

Generally, but imo it makes for a more interesting game if more religions get founded, even if that means the ai is more likely to fight amongst itself.
Besides, diplo bonus is less important for ai, since they can't count on it mattering from the player's perspective (You might kill them even if they are your allies.)
So I'd like trying to get their own religion to be a priority, unless they had heavily invested in their first.

Although if there was a reason for you not to attack them it would be better IMO (back in the days of holy wars you are better off not attacking other ppl from your religion so that you could fight alongside them). Another idea might be giving a diplo penalty "You attacked one of our brethren" (for civs of the same religion as you) or even "We can't trust you because you attacked your own allies" (for all civs).