FfH2 AI Issues Thread

[Lone Wolf]

The AI will sometimes use the useful mobility promotions on its 1-move melee and archery units, if you give it the AI weight of 50.

See attachment screenshots for proof.

 

[Lone Wolf]

It would help AI Hyborem if the AI had a large weight on the Add Manes to City spell.

 

[Lone Wolf]

In my (limited) experience, the AI is rather reluctant to summon Hyborem with the removal of forced Infernal Pact beeline. Maybe that tech needs a higher weight.

 

[Emptiness]

I agree. I don't think I've seen the AI summon Hyborem at all since the change. AV is rarely researched by the AI, and even when it is and an AI ends up following it they just never seem to get around to researching Infernal Pact.

 

[Nor'easter]

I've actually seen the AI summon Hyborem in my two most recent games, including the one I'm in right now. Both games are with patch d. Maybe it's not that frequent, but it did happen.

 

[Lone Wolf]

I've seen AV being researched OK (esp. by Os-Gabella), although me telling the Sheaim leaders to focus on religion more and limiting the spread of RoK might have something to do with it. But yeah, poor Hybo is probably very upset that everyone is ignoring him. What it will do to his self image?

 

[Selrahc]

I've seen a fair amount of AI's who just have no desire to expand. 2 or 3 hundred turns into a game and they sit in a capital surrounded by mines, building warriors.

Generally it's the dwarfs that do this the most, but in a pinch the Malakim or even the Sheaim have a fondness to spend the entire game in fairly well defended 2 population mining colony/desert outpost.

 

[Lone Wolf]

I've seen it happening sometimes because their settlers are struck in a loop near barbarian fortresses (a settler-escort pair moves to a goblin fortress, is afraid, retreats. Rince and repeat).

 

[Emptiness]

Perhaps the code that is used to determine where the AI decides to found its capitol should be reviewed:

Spoiler Ljosalfar start area :

Settling 1E or 1SE provides: River, 1 flood plain, cows, horses, clam.

Spoiler Clan start area :

Settling in position provides: Fresh water, wine, pigs, reagents. Settling 2NW provides: Fresh water, pigs, reagents, horses.

Spoiler Khazad start area :

Settling in position provides: River, cows, bananas, reagents. Settling 1NW2W provides: River, 2 flood plains, irrigated rice, cows, cotton.

Spoiler where they settled :

The Ljosalfar have: Fresh water, cows (that they can't develop), mithril (which won't help early game development).
The Clan have: Horses, life mana, with an option on irrigated corn (they must outculture the Khazad to get it).
The Khazad have: Iron (which won't help early game development), with an option on irrigated corn (they must outculture the Clan to get it) - and 4 jungle tiles in their BFC.

In addition to opting for worse starting resources, by relocating these civs have now clustered their capitols. In the cases of the Clan and the Khazad this has limited their options for (close) expansion, and has also put all of them into immediate border contact with other civs (ie early diplomatic penalties that may escalate into early conflict, plus the risk of losing the ability to work tiles of the capitol if outcultured). They all would have been better off just settling in place than moving to where they did (even the Ljosalfar, who would have destroyed the flood plain by doing so). If they were going to move then they all had a much better option that was closer to their starting position than where they settled.

 

[ADHansa]

I have a little suggestion on how to hinder the Ai from settling in each others knee. When making the map the game makes an x and y limit wherein the starting settler may create the first city. To explain each civ start with x and y limit of starting settler plot then try to increase those limits 1 step in a time in both directions until they interfere with another players starting limits. When this is done as much as possible in all directions withdraw the limits 2 points to create a little spece between the limits.
This creates a little box where the settler may create the city at least 4 tiles away from any other starting city. Of course this is a little cheating and a human will not have the same knowledge but i think it's fair, we do have other clear advantages in this decision.

thoughts for how the structure of what fakecode creating starting settler lime could look like

Spoiler :

get the players and index them pplayer(1) to pplayer(numberofplayers"NoP")
get the locations of each player starting settler
for i= 1 to NoP
	pplayer(i).startx=startingsettler.getX
	pplayer(i).starty=startingsettler.getY
	pplayer(i).firstcitynorthlimit=pplayer(i).startx
	pplayer(i).firstcitysouthlimit=pplayer(i).startx
	pplayer(i).firstcitywestlimit=pplayer(i).starty
	pplayer(i).firstcityeastlimit=pplayer(i).starty
for k= 1 to 10
	for i= 1 to NoP
		bexpandnorth=1
		bexpandsouth=1
		bexpandeast=1
		bexpandwest=1
		if pplayer.firstcitynorthlimit=maplimitnorth:
			bexpandnorth=0
		if pplayer.firstcitysouthlimit=maplimitsouth:
			bexpandsouth=0
		if pplayer.firstcityeastlimit=maplimiteast:
			bexpandeast=0
		if pplayer.firstcitynorthwest=maplimitwest:
			bexpandwest=0
			for j= 1 to NoP
				if not i=j:
					if pplayer(i).firstcitynorthlimit>=pplayer(i).firstcitysouthlimit and pplayer(j).firstcityeastlimit>=pplayer(j).firstcitywestlimit:
						if pplayer(i).firstcitynorthlimit>=pplayer(j).firstcitysouthlimit:
							bexpandnorth=0
						if pplayer(i).firstcityeastlimit>=pplayer(j).firstcitywestlimit:
							bexpandwest=0
					and the same for the 3 other directions
		if bexpandnorth:
			pplayer(i).firstcitynorthlimit+=1
		and the same for the 3 other directions
for i= 1 to NoP
	pplayer(i).firstcitynorthlimit+=-2
	if pplayer(i).firstcitynorthlimit<pplayer(i).startx:
		pplayer(i).firstcitynorthlimit=pplayer(i).startx:
	and the same for the 3 other directions

 

[ArkhanTheBlack]

Didn't have much time to play lately, but now I tried a patch m game on monarch with a friend and the AI seemed to be a lot improved.
The hippus made a most brilliant first strike on one of my cities (1 turn declare war, 1 turn invade) and kept me on my toes with very dangerous attacks of mixed troops.
They both destroyed tiles and attacked my cities. I could fend them off and get some time for teching but I wouldn't have survived an attack of a second faction.

Didn't play enough to give a complete review, but the first impression was quite good.
Very good balancing of defense, expansion and teching.
I think factions of opposing alignment could be a bit more aggressive, but I'd have to play a few more games to be sure.

Very nice work!

 

[ArkhanTheBlack]

Played another quick tech game with a friend on emperor. Unfortunately, this time the AI's played pretty bad. They never attacked and just waited like cattle to be slaughtered. Although there were no wandering stacks, the AI's didn't seem to be able to switch to military production and war after they reached a decent economic power.
I think Patch G and before was still better if it comes to AI aggression.

Another problem was the Hyborem rush of my friend. The power of the infernals is just insane. About round 110 his 27 size capital had a production of 200 per round. And his other cities weren't much weaker. Add the power of Hyborem + body guards + Cultists with fire ring + Infernal Grimoire => Beasts of Agares, Eidolons, Mardero & Co. shortly after that, and you can do the math.

Actually, we decided to ban (player controlled) Hyborem from our games since he's simply too overpowered.

There was also sync error one time at the switch to hyborem (we had to reload anyway because of the switch) and one time 20 - 30 rounds later. Not sure why.