Post Wildmana AI Wonder/Ritual weirdness persists

[Izmir Stinger]

I abandoned the base game as unplayable after patch h, but have recently come back to it. The pathetically low attack courage barbarians, inability of the AI to build ships, automatic refusal of all trades because "WFYABTA" and some other wackiness seems to have been fixed. There still seems to be something going on with Wonders and Rituals, though.

Emperor Difficulty, Standard Map size (normal number of civs), nothing weird about the map script, Normal speed. I think I was the first to Philosophy, but I delayed building the Bone Palace until well after turn 175. In the same game I finished Pact of the Nilhorn after turn 250. I had to trade for Cartography, which several other civs had had forever. Now it is nearly turn 300 and the Great Lighthouse still hasn't been built, despite the Hannah being in the game, doing well, and having beelined sailing long before I even met her.

From what I can see, something is still dramatically reducing the AI's tendency to go for wonders. As it stands, I could basically have my pick, and it shouldn't be that way on high difficulty levels. Are other people experiencing this, too, or was this just an odd game?

I also found some of those large "Barbuster" stacks filled with archers and obsolete units, though it didn't seem as bad - and the two civs lucky enough not to have barb problems due to geography (Luan) or the barb trait (Clan) didn't have them. The Calabim had a big one filled with Maori that was hovering near a barb city to the south while the clan invaded successfully from the north. So there is still some of that going on.

 

[Emptiness]

I've found the L-M AI much more willing to build wonders and rituals than the H-K AI. Military is still the AIs' top priority, but if an AI is able to get an adequate military it will turn its attention to wonders or rituals. In my games, if I focus on a specific one then I will beat the AI to it, but if I do not then the AI will often build it first. Previously I was finding that the AI never built rituals and only built certain wonders, so for me this is a big improvement.

 

[kumquatelvis]

I watched a few hours of my friend's patch M game, and he had his choice of the wonders/rituals. Pact of the Nilhorn in turn 200 or so, no competition for the Bone Palace, etc...

 

[DaveGold]

It seems to me that the rituals and wonders are getting built, sometimes in a timely fashion but often slowly and late.

 

[Skyre Noktis]

I watched a few hours of my friend's patch M game, and he had his choice of the wonders/rituals. Pact of the Nilhorn in turn 200 or so, no competition for the Bone Palace, etc...

It would be interesting if you could post the difficulty level here. I'm starting to get a sneaking suspicion that the Wildmana AI only really works properly on difficulty levels where it gets massive production bonuses. On Noble, Prince, etc. it spends most of its production on defence and doesn't have much left for anything else.

That said, the OP is playing on Emperor so I assume this shouldn't be an issue for him.

 

[Angrenost]

It's the same on Deity, observing the game I'm playing atm.

 

[Morlark]

In my current game (patch m) on Noble, someone managed to construct the Heron Throne very early on. I was impressed with how quickly they managed it, since I hadn't even finished researching all of the key improvement techs yet. So the AI does work at least sometimes on the lower difficulty levels.

 

[Izmir Stinger]

In my current game (patch m) on Noble, someone managed to construct the Heron Throne very early on. I was impressed with how quickly they managed it, since I hadn't even finished researching all of the key improvement techs yet. So the AI does work at least sometimes on the lower difficulty levels.

The heron throne was the only wonder that went as early as I expected it to go in the game I describe above. I was a little miffed when I saw it go because I had just settled a double seafood city with lots of coast that would have been great for it. I also thought, "Oh good, they seemed to have fixed the AI not building wonders thing." Turns out it is only partly fixed.

 

[esvath]

Confirmed the OP. I played as Thessa in Lake map, Monarch, with other 6 civs. Alexis managed to build the Heron Throne. But then, I swept the Great Library, Catacomb Libralus, Bone Palace, Form of the Titan, Song of Autumn (no wonder for this one), Syliven Perfect Lyre. Now only turn 285.

The problem is, many AI civs should build those wonders way ahead before me. Since I beeline to WotF, the Clans managed to research Bronze Working and Warfare way beyond me. But Jonas never built the FoT. Falamar also never built the Great Lighthouse. I even gave Drama to Keelyn, hoping she would built Syliven's. Nope. Nil.

I don't think this was caused by AiWeight in BuildingInfo.xml since both Heron Throne and Catacomb Libralus have AiWeight of zero.

 

[kumquatelvis]

The game I observed was on whatever difficulty is right below deity (immortal?).

 

[Turinturambar]

I took a quick look at the gameutils.py file and here is a list of wonders that the AI will (almost) never build due to the AI_chooseProduction function.

BONE_PALACE: Completely missing, so the AI does not consider it, instead the forbidden palace is set as a generic worldwonder, which does not make a lot of sense.
GUILD_OF_THE_NINE set to 0.
BUILDINGCLASS_CITY_OF_A_THOUSAND_SLUMS set to 0.
BUILDINGCLASS_MOKKAS_CAULDRON' set to 0
BUILDINGCLASS_SHRINE_OF_SIRONA set to 0
BUILDINGCLASS_CATACOMB_LIBRALUS set to 0
Form of the Titan should be fine. Probably the level requirement prevents the CLan AI from building it.

Only the Svarts and the Ljos have a positive value for the pact of the nilhorn, but they should build it asap.

But the kicker is that once a civ enters conquestmode (basically reaching a milestone tech for the civ, can be as early as bronze_working for the Sheaim and generally is at iron working) the AI will only build very few select wonders (Guild of Hammers/Form of the Titan...) unless it has reached a very high unit threshold.

 

[esvath]

(I swear I already reply this but it didn't show. If there is a double post, please forgive me )

So would that means values in FavoriteWonder tag (in CvLeaderHeadInfos) never be taken account when deciding which wonder will be built? That is too bad. How can I modmod this? Editing AI_chooseProduction?

Thanks.