Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[NW_Kohaku]

Recently,

How, exactly, does the game determine which animals are related? Does it use the currently-existing CREATURE_CLASS system (or a variant thereof), or does it analyze the various creatures' tags/anatomy and compare how similar each one is to the others? Or is it that I'm completely off here and it works in a completely-different manner? How does PET_EXOTIC affect things compared to PET?

If you're referring to this part of the devlog:

Attempts to tame/train creatures will add to your site's training knowledge of that creature, which augments future training of all creatures of that type.

I read "all creatures of that type" as "all creatures of that species".  Setting up relationships between species manually would be a big boring task, and doing it algorithmically would be difficult. 

I disagree.  While setting up manual relationships would be difficult, catagorical relationships could be easier, and there are reasonable ways to add a few simple tokens that would make a reasonable algorithm possible.

I should also say at the outset that almost everything in DF that is suggested falls somewhere on a spectrum between being a quick and simple fix that would make a reasonable improvement for little effort and the long-term ideal goal.  The purpose of this suggestion is to be a long-term ideal, not to be a quick fix, and the quick fix (not to really distinguish at all) is basically in place already. 

There are a few ways in which animals can be organized to make specific exotics more or less easily trained:

Categorical

Spoiler (click to show/hide)

Perhaps the most obvious way to show relations between animals (besides the brute-force method of manually punching them in) would be to do what is done in real life:  Taxonomically categorize animals from broad categorical kingdoms down to narrow and specific genus and species.

While it would probably be overkill to go for the full Kingdom, Phyllum, Class, etc, however, having four or five layers of categories added onto the raws of every creature would not be unreasonable.  A very broad class distinguishing "natural animals" from something more exotic like "living magical rock creature" is a more meaningful distinction than the distinction between different types of lizard.  Comparably, a combined family/genus type of category would be useful in putting brown bears and black bears and polar bears together while only having to add the number of tokens as you have layers of taxa per creature.

The side-positives of this sort of system would be that it could work well with future changes to the creature system, such as if you wanted to make part of the sphere/god/magic system that relates to specific creatures work by having relationships between specific types of animals.  It could also help ease in notions of "breeds", such as having the capability to cross-breed horses and donkeys (or maybe wolf and dog or even dingo).  It would also work well with regards to something like the alchemical properties of the materials of an animal to have taxa for their creature type for algorithmic association of chemical properties to specific taxa. 

Finally, it could be useful in the creation of a full-scale procedural generation of creatures to simply have a "family tree" of "real" animals, while the game just populated the holes with new creatures.

The negative side of this, however, is that taxonomy obviously does not denote behavior.  When trying to train behavior, in some ways, a spider has more in common with a crocodile than it does with an ant, even if ants are closer neighbors as fellow invertebrates.

Algorithmic

Spoiler (click to show/hide)

In Romance of the Three Kingdoms, all of the characters are given a hidden stat that gives them more or less of a likelihood to join up with one kingdom or another, based upon how close or far away they are in this hidden stat from the leader of that kingdom.  It's an integer from 0 to 100, and represents how they see China as ideally ending the war for control.  Characters on Liu Bei's side of the spectrum basically want the re-establishment of the Han nobility, while characters on Cao Cao's side of the spectrum want a new system of meritocratic advancement is basically what the system evolved into.

Probably the simplest way to categorize animals for training that has the closest reflection of how difficult they are to train would be to simply create a few "spectrum" stats for every creature that dictates a few key behaviors, like how they hunt, how socially complex they are naturally, and how developed their brains are.

When training creatures, it's far, far easier to work with creatures with specific types of traits, completely regardless of what their actual taxonomic relationships might be.  Pack hunters are easy to train because they are inherently used to working with other creatures, while solitary ambush predators are almost impossible to train.  This is why dogs and cats are very similar, but it is basically impossible to train a cat - a cat is a solitary predator.  Combine this with a less evolved brain, like an alligator, and you have a creature that is basically impossible to "train", you can only breed and domesticate them.

Likewise, a herd animal like a cow is very difficult to train, since it is not socially complex, even though it would be easier to train a cow than it would be to train an alligator.  Herd animals have social groups, but they are critically social groups where individuals make no difference to the herd, only the presence of the herd itself matters.  This is why sheep are most easily herded by simply training dogs to help herd them, since dogs are much more easily trained.

Birds, especially corvids, would be much easier to train, as they have more developed minds (they are even smarter than cats and dogs, or even pigs) and many of them can develop social habits to the point of building communal housing, although many are strictly social only with their pair-mates. 

Abstracting all of this by throwing out a numbers in a few different axes of rating the sociability of different animals along how much they rely upon learned behavior from their parents versus instinctual behavior (making insects be the low, and creatures like dolphins the high ends of the spectrum), having a sociability spectrum (with alligators the low, and creatures like bonobos the high), and having a third spectrum rating a more general adaptability of a creature evolutionarily (such as a specialist that has to adapt to specific conditions) versus generalists, marking more easily domesticatable creatures.  For example, bees being more easily domesticatable than longer-lived, generalist raccoons.

Any creature with similar numbers to a creature you have already trained will have a bleed-over bonus to training that creature, as it is "similar to train" that type of creature.  By doing this, you could set up a relationship between every conceivable creature by simply adding three number tokens to each creature.

Domestication

Spoiler (click to show/hide)

Many creatures need domestication to be truly trainable, and domestication of both animals and plants (such as making wild strawberries just plain normal strawberries) are a key part of civilization's development. 

Any creature you train routinely, generation by generation, will become domesticated, and a great many creatures are not really trainable without domestication.

The simplest and most reasonable way to work it would be to abstract this by making the "civilization skill" actually represent how much domestication they have done of the creature (making creatures of that type that come from that civilization have that domestication, while wild specimens do not). 

Domestication itself could also lead the way towards unnatural selection, which would, in turn, mean different breeds of animals, such as having different breeds of dogs, or having different creatures like Giant Cave Spiders that are domesticated and trained to spit their webs in specific ways to help your dwarves, or else even alter the way in which they spit their webs, depending on how advanced and nuanced the mechanics for this eventually become.

EDIT:
From the FotF thread:

Creature Classes could be such a powerful thing if applied more.

Say, you give Beak Dogs, Grimelings, etc, the [CREATURE_CLASS:EVIL_PET] or [CREATURE_CLASS:EVIL_WAR] and put something like [ALLOWED_PET_CLASS:EVIL_PET] on the goblin's entity so they'd use that in their sieges. This could allow more customization than the standard evil/good/underground classification. A wierd gnome civ could even have robots with the creature class [CREATURE_CLASS:ROBOT_GNOME] so that they would acquire them automatically, while other entities cannot access them.

Categorical classes can also have the benefit of making a more easily defined "this is a pet of this specific type of civ" command on the game. 

In a sense, if it happened naturally through gameplay, it would be better, but also just simply having a "all goblins have troll pets" command may be more to a modder's style.  It would also give players an option to say "this culture doesn't have horses" instead of all common domestics being... common.

[tsen]

It seems like doing it via algorithm would suit DF best as a fantasy world generator. That way you can easily generate numbers to fit those scales without the program having to know how to classify species.

[peskyninja]

+1

[NW_Kohaku]

When I got going, I thought about it, and there were probably ways to do it algorithmically without having to actually change all the raws if you wanted a really loose definition.  You could simply sort animals by their hunting/predatory/grazing tokens, the size of the groups they appear in, the presence of tokens like AMBUSH_PREDATOR, and body shapes like INSECT or general size. 

A spectrum type of raw editing (while raw-breaking on the update) would have the most controlled and predictable of results, however.

[King Mir]

I don't think that training one type of bird leading to an easier time training another type of bird would be a worthwhile benefit. Spending a lot of time on taming animals would take a dedicated effort. But further subdividing the associated entity knowledge probably wouldn't make people dedicate time to taming a different portion of the passing wildlife. People will likely just tame whatever they find cool, whatever is easiest, or tame everything they can.

There might be world gen effects to making taming of similar animals easier, like having a civilization of bird tamer dwarves, who over the years come to have all manner of exotic domesticated birds. But the player isn't able to influence that kind of thing much.

And then there's the fact that the effect may realistically be too small to be worth including. How much knowledge in training parrots will help in training Giant eagles?

[NW_Kohaku]

I don't think that training one type of bird leading to an easier time training another type of bird would be a worthwhile benefit. Spending a lot of time on taming animals would take a dedicated effort. But further subdividing the associated entity knowledge probably wouldn't make people dedicate time to taming a different portion of the passing wildlife. People will likely just tame whatever they find cool, whatever is easiest, or tame everything they can.

There might be world gen effects to making taming of similar animals easier, like having a civilization of bird tamer dwarves, who over the years come to have all manner of exotic domesticated birds. But the player isn't able to influence that kind of thing much.

And then there's the fact that the effect may realistically be too small to be worth including. How much knowledge in training parrots will help in training Giant eagles?

I would ask the question "How much different can training Peach-Faced Lovebirds be from training Masked Lovebirds?"

They're almost the same species. 

Likewise, I would say that there are some things you can learn from taming a peregrine falcon that would still be applicable in most other raptors, especially with regards to body language, and some of the more general habits and ways of dealing with the claws that catch. 

When it comes to ways in which parrots are similar to giant eagles, no, there wouldn't be all that many things that are the same, but there would at least be something to learn as far as ways to breed and read their body language, and they are certainly a less alien creature to train if all your experience was with parrots than to try training a giant spider. 

Having a civilization of bird-trainers makes more sense than a civilization of giant eagle trainers who have no idea how regular eagles work, but do know how to train giant cave spiders but not cave spiders, etc.

[King Mir]

Yeah, I agree that training very closely related species, such as an animal and it's giant variety should give a big boost to training each other. But that is hard to recognize algorithmicly.

[tsen]

It wouldn't be all that difficult to first sort by body plan, then apply what NW_Kohaku was talking about.