Gameplay ideas and AI questions.

[hiphopin]

Well first I want to ask 1.) Can the AI effectively upgrade their adepts to mages and mages to archmages yet, and if so do they know how to promote them. 2.) If not is there an option in this mod to do that. 3.) If not is someone willing to make a tiny module to allow this, their is one that does this but its outdated, I LOVED that module, it worked perfectly.
Next, I want the AI to be more challenging but I don't want to increase the difficulty past noble (I HATE having any setting other than noble as its basicly cheating, faster research for AI past noble, faster research for human prior to it.) Any ideas?

And finally how do you play this mod, what settings do you use, and just what do you like to do on the game period?

 

[Charybdis]

If you use 'no AI level restrictions' in custom game, I believe for the units that require X level to be promoted it will just build them instead. This ups the difficulty a bit, also turn off raging barbs. AI doesn't react well to it, and it cripples/kills them more while giving your troops feeder exp.

 

[Sarisin]

I have seen AI Archmages in the game, but now that you mention it, I am having a hard time remembering when I saw Mages. This doesn't make sense, of course, unless the AI is somehow making the jump from Adept to Archmage.

I think the problem there is that the AI often uses Adepts as combat units and they just are not around long enough to level up to Mages/Archmages.

You will see plenty of High Priests though. I played one game when I was at war with the elves and they were somehow spamming High Priests - which requires regular priests to be Level 6, I think before they can be upgraded. Usually the AI will use Priests as combat units too and they don't often stay around long enough to level up - that's why this spamming of High Priests surprised me.

 

[MagisterCultuum]

I'm still wondering if higher AI weightings for adepts than mages or archmages are part of the problem.

With "No AI Level requirements on," the AI can build mages and archmages directly. I'm assuming that Archmages have higher weightings than mages, so the AI prefers them when building things directly.

You know, a downside of this is by not going though the process of upgrading them they are potentially missing out on lots of free promotions.

 

[xienwolf]

FF has the weights set up so that it increases in value from Adept->Mage->Archmage. So that part isn't an issue at the least.

AI are allowed to build archmages and mages directly if you set the No Level Requirements option (as mentioned), which achieves what the old module did in a clunky fashion.


Yes, a directly built Archmage is not as strong as a naturally promoted one because it will just start with the basic free upgrades due to multiple mana nodes and the 1 free promotion for being an Archmage, but it starts with channeling 1/2/3 so it will gain XP very quickly to make up for it. End result is the loss of the 2 free promotions from being an adept/mage earlier. But overall it is still a large asset for the AI to be able to use Archmages at all, if they ever bother to research down the sorcery line (few AI do that until nearly the end of the game).

 

[hiphopin]

Thats the main problem, magic in FFH is extremely powerful and EVERYONE can get value from it, but since they research it so late (btw I have been playing with AI can directly build units the whole time anyway) you can just faceroll over all the enemy AI by turn 200-250ish, it makes it so unfair later on as it is NO compitition, in fact I usually quit the game before finishing it many times just because sending in a stack of 2 heroes 2 archmages a few catapults and some champions basicly will destroy any city, hell 2 archmages alone can destroy most.

I want challenge but I dont want them to cheat (IE harder difficulty settings) I just don't know what you guys could do to achieve that.

Also the fact that AI cities are poorlu defended in most of my games doesnt help. 15 warriors doesn't cut it.

 

[Sarisin]

Thats the main problem, magic in FFH is extremely powerful and EVERYONE can get value from it, but since they research it so late (btw I have been playing with AI can directly build units the whole time anyway) you can just faceroll over all the enemy AI by turn 200-250ish, it makes it so unfair later on as it is NO compitition, in fact I usually quit the game before finishing it many times just because sending in a stack of 2 heroes 2 archmages a few catapults and some champions basicly will destroy any city, hell 2 archmages alone can destroy most.

I want challenge but I dont want them to cheat (IE harder difficulty settings) I just don't know what you guys could do to achieve that.

Also the fact that AI cities are poorlu defended in most of my games doesnt help. 15 warriors doesn't cut it.

You made some excellent points, hiphopin.

I have tried going just from Monarch to Emperor (in FFH, not FF) and the handicap the AI gets made for very unpleasant games for me. I just cannot understand how the guys play at those super difficulty levels.

I think if you play with the building requirement option when you start a custom game, you will find the AI DOES field stronger units and defenders - no more Warriors defending cities.

However, there are some problems even then as they lack the required techs to get the stronger units. Some of this is caused by AI civs beelining techs that just take a very long time to research. They are spending 100+ turns to research Taxation (epic speed) so they can get the civic Republic. Meanwhile all their cities have Warrior defenders as you mention.

I would like to see the whole beelining dynamic removed from the game/mod/modmod so that there is more of a level playing field. I know some would say the poor AI needs every break it can get, but I think by letting them beeline it hurts more than it helps them in the long run.

 

[xienwolf]

Kael did something with the beeline issue in 0.34, hopefully got rid of it rather than just placed a check for if the tech is allowed or not for the game. Depending on what he has done we'll see if further tweaks are needed on that line.


I'll be playing with the way that cities decide what to work on when I get into the code again (possibly this weekend). I intend to make the AI work on buildings which enable units for them as a priority, then work on defenders (and increase the desired number of defenders in general), then it will probably stick to the old priority system while I see how that works out.


All in all, the basic rundown of any AI function is a simple series of checks done in a specific order. Whichever one is set first in the order will be done the most. I believe that what we are seeing is that in BtS few units actually REQUIRE a building, so the first check done is to see if there is a need for any units, and if there is to fill them with the best it can currently make. In FfH it is suicide to keep working on outdated units, tossing them to their deaths, and building replacements. You need to get out your vital buildings, THEN ensure sufficient units.


After some work on prioritizing proper build orders in cities, I'll start looking for all uses of AIWeight and removing them. They are nice when you are trying to inform the AI that there is a nifty Python function that is good (like Planar Gates), but it is HORRIFIC when the AI is trying to make longterm plans. If I am seeking to get the AI to build good defenders, I will have it prioritize defensive strength, and free promotions that enhance defensive capabilities. If I am seeking to get the AI to build good offense, I will have it prioritize Collateral Damage and Offensive strength. But in both cases, AIWeight is JUST AIWeight. I cannot tell the computer if that means "Defend Good!" or "Generate Money!" or anything else. Overall, it is something which should have only been used as a band-aid, but has become a crutch.

About the same can be said of Flavors I suspect. They are meant to be small, just to provide some "random adjustments" between each leader. But instead they are currently fairly large, and thus they dominate over everything else. Thus instead of considering 2 technologies as "This one enables a Wonder, but the other one enables better tile Yields. My economy is floundering, so I cannot afford to work on a Wonder, so I'll take option 2" the AI is forced to say "This one is Culture, which I LIKE, and this other one is Science, which I don't like. So I'll take the first one!" In the appropriate case, it accounts primarily for the current situation, and uses the flavor of the Civ/leader as a balancing factor for close tie situations. In the inept case, it just seeks out arbitrary flavor which may not even be properly set up.



Anyway, all of that is just so those of you who don't think "computer-esque" can understand that it can be quite tricky deciding how to go about making the AI "smart" about things.