Realism in Roguelikes

[Gr3yling]

It's not very realistic to have attributes as numbers.

Although, to be fair, both the attributes that I mentioned in the part of my post that you quoted for this reply can very easily be represented numerically.  Perception (visual acuity) is routinely described by numbers (20/20, 10/20, etc).  I realize you could argue that some aspects of visual perception are harder to quantify, but I think you still see what I mean. 

Whole body strength seems like it would also be fairly easy to quantify using a weight lifting exercise that involved a number of widely distributed muscle groups, maybe something like deadlifting?  I'm not very athletic so I couldn't say for sure. 

I also like the idea of attribute values that allow the player to compare the PC to a normal human in an understandable way.  In ADOM, I think an average human being is described as having 10’s for all their attribute values.  So, you can easily imagine exactly what having a strength of 20 means: that the PC has the strength of two people.

You could think about perception the same way, I think.   For instance, if the PC had a perception score of 15, they would probably have they had 30/20 perception.  Again, I realize I’m oversimplifying the idea perception in a lot of ways, but you still see what I’m saying.

It could be possible to model a realistic enough game world, or at least give that kind of impression to the player. It wouldn't be impossible, but you should also think about how powerful monsters are, because with typical roguelike monsters the player (if human) would not survive.

There’s an argument that what really makes a hero is luck, not the more mundane attributes that they possess.  In most fantasy stories, the characters are stronger than the average person, but not strong enough to beat an orc, or even a goblin in single combat (I figure that goblins are at least as strong as chimpanzees).  They are more agile, but not agile enough to successfully dodge their enemies’ attacks over, and over, and over again.

So, to me, the answer is that a “normal person” could survive just fine in a fantasy world, if they had the gift of luck that is given by the gods or whatever other supernatural powers influence such a world.

[Vanguard]

Take the idea of attributes, for example.  When I think of increasing realism, increasing the number of attributes used to describe the PC is the first thing I think of.

What about heading in the opposite direction?  You could remove attributes completely.

Perk systems work really well in games like Infra Arcana and Mage Guild.  They carry most of attributes' mechanical benefits without any of their problems.  It's easier to form a mental image of a knight with the "very strong" perk than one with 52 strength.

[akeley]

My point is that my idea of realism is not tied to how detailed the game is.

For me, realism is something more like:
* The game's features are consistent with the setting.
* The game's mechanics, text, AI, etc does what you'd expect, based on reality. This reality may be modified to allow magic, elves, zombies, etc, it can still be realistic within that premise, and must still behave like you'd expect a world with elves to behave (I cringe every time I hear the argument "the game has elves and magic, don't complain that it's unrealistic that you can eat an apple to instantly heal all your wounds")
* Features should primarily be added to enhance immersion in the game world, not to add abstract gamey little mechanics. If you add dynamite, it should be because it would make sense that the game world has dynamite, not because you want to add an area of effect attack.

I'm mostly concerned with immersion and atmosphere.

This is definitely closest to my take on realism. Of course in some other genres it`s much easier to define - for example when somebody says "It`s a realistic flight sim" you instantly know what`s up, but upon hearing "realistic roguelike" I wouldn`t know what to expect to be honest. But, the above encapsulates the whole thing rather well - "consistency" being the main word here.

And definitely wouldn`t say that number of attributes affects realism - it can of course, but it can also detract from it, if done badly. Realms of Arkania spring to mind, an old school cRPG so loaded with stats and attributes that character creation is kind of a separate game. But the problem is, majority of these stats are just for show and either affect nothing at all or the effects are minuscule and invisible to the player (fact it`s an implementation of a tabletop system probably being the cause, it`d make much more sense in live DM session)

Linley`s Crawl has only 3 main stats and yet it`s superbly "realistic" for me - I simply feel during playin that that`s how things would turn out during real dungeon exploration. It`s full of beautiful little touches that should be mandatory - monsters fall into traps, there`s "friendly fire", a monster can pick up a weapon or potion and use it and so on (one thing it`s lacking and which is my absolute favourite is monsters duking it out with each other - think Powder has something like that. I love games that include this mechanic).

So instead of investing in countless attributes - these suit in-depth strategy games much better - I`d prefer to see better monster AI...though by better i don`t  mean more ruthless, in fact I`d love to see monstas make mistakes or even avoid me sometime and do some other stuff - why is every goblin hellbent on taking on heavily armed Draconian Skirmisher? Logically, some of them would think twice about that. And spellcasters could fail their creations and cause harm to themselves too. Or conjure a demon that would be a threat to them as well as a player.

Also simple physics could help a lot the realistic cause - it`d be nice to open the door and bash someone with it or maybe block it with a table (loved that in DDA). Or say you hit that suicidal goblin strong enough, he staggers and squashes the poor rat behind him. Things like that.

[Kevin Granade]

I agree with the others that number or granularity of attributes is mostly orthogonal to the realism of the game, whatever that means

I think the answer to "how many attributes" is "as many as you need for the playstyles you're going to support".
If you're trying to make a completely open-world game, where there are a huge number of ways to play the game, you might need more, if you have a tightly focused game where there are a limited number of well-defined actions, you might not need many at all.

For example, suppose your game is 100% combat oriented, no crafting, lockpicking, writing, or any other kind of activity requiring fine motor control.  Then depicting fine motor control is completely irrelevant.  Likewise it's not too much of a stretch to make a game where an "int" score is irrelevant, or "perception", or strength, con, or any other stat.  Maybe you have a space exploration roguelike where the dominant stat is mathematical ability so you can better plot courses with calculus.  In short, what attributes are significant is entirely based on what you're going to be doing with them, which leads to the question of...

What you want to accomplish with attributes?  Do you want attributes to provide a way to differentiate between character builds chosen at game start?  Do you want attributes to provide an aspect of advancement for the player?  Are they a mechanic to enforce character specialization? Do they provide a thematic grounding to relate character abilities in understandable ways?
Similarly, what do attributes themselves accomplish within the framework of the game?  Do they provide prerequisites for other aspects of character creation or progression?  Do they provide a bias to a RNG for various actions?  Do they act as prerequisites for performing actions directly?

Another question is "what's an attribute?".  Clearly you mean something like "numeric scores representing intrinsic ability to perform various actions", but the function that kind of thing performs can be replaced by other systems (e.g. traits as Vanguard points out) or as often the case in action games, removed entirely.  DDA has less than a handful of attributes (4), but a large number of skills(28), and a huge number of traits(195).

I'd be very interested in building a system that has *only* traits, possibly arranged in a hierarchy of prerequisites.  I don't see anything particularly unrealistic about that approach.  Some might be "attribute-like", like "strong", "very strong", etc, but it's not absolutely necessary, for example if you have multiple "trees" of trait dependencies, the lower-level traits can provide all the functions of attributes in other games.  Perhaps you have a tree of strength-based traits, and your effective strength is the sum of traits in that tree, so each time you chose a complimentary trait the related ones become more powerful.

The answer to "why numbers" is simply "so you can plug them into formulae", and the answer to "why show the player the numbers" is "so they can predict the outcomes of the formulae".  All other purposes can be filled some other way, such as boolean trait checks instead of threshold or RNG based checks, or descriptive labels.

[Gr3yling]

What about heading in the opposite direction?  You could remove attributes completely.

You could, and I'm open to the idea.  There are a couple of problems, though.  First, you are essentially asking the player to "just trust you" when it comes to game mechanics.  If I don't have at least some insight numbers the game is using to calculate the outcome of the PC's actions, how do I know that the mechanics are rational at all?

Second, you need numbers to make accurate comparisons.  Who will do more damage, the PC who is "very strong" and has a "dull" sword, or the PC who is "weak" and has a "very sharp" sword?

Honestly, though, I do really like the idea you are bringing up here, and I have thought a lot about it in the past.  I think that a definite plus of that sort of system is that you can have incredibly detailed statistics used to simulate the outcomes of player actions, because you don't have to their limit complexity to a level that the player can easily understand.   

It's easier to form a mental image of a knight with the "very strong" perk than one with 52 strength.

On the contrary, it's easier to form a mental image of a knight who is twice as strong as the average person (20 strength) than one who is "very strong".

[Gr3yling]

I hear this all the time but I don't get it.  Why is it good for video games to pretend that they're not video games?  Why is that necessary to achieve immersion?

Nobody is “pretending” video games are anything they’re not, because video games can be whatever you want them to be.  There’s not some sort of role boundary that is being inappropriately transgressed here, any more than there would be if any other medium incorporated realistic elements in an attempt to mimic reality more closely.

Why is realism, or detail necessary to achieve immersion?  I think you answer your own question later in your post.  Even though you are not interested primarily in using attributes to vividly describe the PC, you enjoy magic systems that are detailed and realistic.  You have the same desire for realism that I do, we just look for it in different aspects of games.   

By "immersion" you mean becoming completely absorbed in a task or a piece of media until it has your complete attention and the rest of the world just sort of falls away, right?  Heavily abstracted games can do that just as well as simulationist games with a thousand character stats.

For some people they can, sure.  But not for me.  And, honestly, some realism is required to be able to relate to the PC, to role play.  I can’t identify with a tetrad the way I can with my PC in ADOM.

And we are focusing mainly on role playing games in this discussion, right?

Like, one of the most immersive things about ADOM is how the world is really mysterious in a way that directly affects you, the player.  You want to give the game your full attention and figure out what's going on.

Can you please explain that more, Vanguard?

I agree with the others that number or granularity of attributes is mostly orthogonal to the realism of the game, whatever that means 

But, I don’t understand.  You’re trying to implement an extremely detailed system for modeling injury to the PC.  It’s at least as detailed as adding a few more attributes.  Why do you prefer one over the other?

I think the answer to "how many attributes" is "as many as you need for the playstyles you're going to support".

Right, and I’m mainly thinking about a fantasy themed game world similar to ADOM.  So, that may help narrow it down somewhat.

What you want to accomplish with attributes?  Do you want attributes to provide a way to differentiate between character builds chosen at game start?  Do you want attributes to provide an aspect of advancement for the player?  Are they a mechanic to enforce character specialization? Do they provide a thematic grounding to relate character abilities in understandable ways?

Similarly, what do attributes themselves accomplish within the framework of the game?  Do they provide prerequisites for other aspects of character creation or progression?  Do they provide a bias to a RNG for various actions?  Do they act as prerequisites for performing actions directly?

Yes to all of those.  Were you thinking some of them were mutually exclusive? 

I have to think more about the rest of your questions.

[Vanguard]

You could, and I'm open to the idea.  There are a couple of problems, though.  First, you are essentially asking the player to "just trust you" when it comes to game mechanics.  If I don't have at least some insight numbers the game is using to calculate the outcome of the PC's actions, how do I know that the mechanics are rational at all?

Second, you need numbers to make accurate comparisons.  Who will do more damage, the PC who is "very strong" and has a "dull" sword, or the PC who is "weak" and has a "very sharp" sword?

None of this is unique to perk systems.  Every game expects the player to trust its mechanics.  Attributes can be unclear too.  Perks don't prevent the developers from revealing their games' internal formulas.

On the contrary, it's easier to form a mental image of a knight who is twice as strong as the average person (20 strength) than one who is "very strong".

It isn't, though.  What we call strength is really a combination of a huge number of different traits.  To say that a given person is numerically and exactly twice as strong as another is nonsense.

"Very strong" is subjective but there's nothing wrong with that.  Maybe in your mind a very strong character is a towering mass of muscle or maybe they're really lean and fit.  Either is fine.  The important thing is that it gives your imagination a starting point.

If you still feel that it's important to know exactly how strong your guy is, you could always include flavor text on the perk screen that says "You are twice as strong as the average person."

[AgingMinotaur]

The tendency to include a bunch of stats is probably a legacy from tabletop RPGs, where the stats also serve as aids to roleplaying. Eg. a good player will reflect the character's Charisma score in dialogue. (Nonetheless, back in the days when I played a lot of tabletop RPGs, the most immersive sessions came after we ditched stats and dice altogether.)

In a CRPG or RL, I think you should only have as many stats as are meaningful. If Intelligence is only used for a few rare skills, better to drop it or merge it with another stat. To take an example from my own game Squirm, I had a single stat for skillfulness/finesse: basically intelligence, dexterity and charisma baked into one.

I did at some point toy with the idea (again lifted from RPGs such as Nephilim and Everquest) of a game using the four classical elements as stats: Fire, Water, Earth and Air. Some skills/feats might test a single stat (eg. Fire to determine melee damage), others could rely on a combination (eg. Fire+Air for archery, Earth+Water to resist magic effects). That might not be the most realistic system, to try to get back on topic, but it could make for interesting and immersive gameplay, methinks.

As always,
Minotauros

[Gr3yling]

None of this is unique to perk systems.  Every game expects the player to trust its mechanics.

That’s kind of a vague statement.  Every game does necessarily expect a player to trust it’s mechanics to some extent, just like every game has to reveal those mechanics to the player to some extent in order to be playable. 

I can give you plenty of examples of successful games where the player has total access to the mechanics, like D&D, for instance.  The mechanics of crawl and DDA also seem to be mostly, if not completely, transparent to players.  I don’t think they are any worse off for that.

I’m less concerned with the expectations of other game developers than I am with making rational design choices.  And I think having transparent mechanics is such a rational choice.

Any game is going to have some mechanics that seem counterintuitive.  You can’t just give the player a game and say: “play this in a way that makes sense” because the mechanics that make sense to the developer may not make sense to the player.

Honestly, I think the main reason a developer would want to hide mechanics from players isn’t to generate a mystique about the game world, but because those mechanics aren't rational. 

And there are at least some other people who think game mechanics should be transparent.  Games like final fantasy tactics have lengthy documents exclusively devoted to explaining game mechanics (check out Aerostar's Battle Mechanics Guide on gamefaqs for a good example of what I am talking about). 

Attributes can be unclear too.  Perks don't prevent the developers from revealing their games' internal formulas.

It sounds like you are saying here "bad systems of character development are bad."  Sure, any type of character development system can be done well or badly.  However, in the specific example we are talking about, the numbers give you a lot more information about the PC than just saying they are “very strong”, for all the reasons that I already pointed out.

Am I misunderstanding what you were saying?  Would your "very strong" perk explicitly say "+3 to strength" in its description?  If it would, well, that would require using attributes too, wouldn't it?

It isn't, though.  What we call strength is really a combination of a huge number of different traits. 

What I call “strength” is a single trait, and unless you want to get incredibly complicated, that’s probably fine.  I of strength think of it as being a measure of the contractile force of a muscle.  That, or something similar to that, is what just about anyone else who hears the term “strength” is going to think. 

We don’t need to delve into muscle physiology. I am aware there are *many* complicated ways of measuring muscle performance.  Even I’m not crazy enough to stumble down that rabbit hole, and it’s hard to believe that anyone else would feel anything was being lost by the omission of that type of esoteric information.

To say that a given person is numerically and exactly twice as strong as another is nonsense.

You know I’m going to ask you why you say that, Vanguard. 

Are you wanting to model each muscle group individually, now?  I think I was wrong to have suggested that originally.  The characters we are modeling aren’t just going to the gym and just doing upper body exercises, they are engaging in activities that use almost all their major muscle groups when they do things like fighting in heavy armor.  So, they wouldn't rationally have really strong biceps but really weak quadriceps, for example.
 
As I already pointed out, deadlifting is probably a decent way of measuring whole body strength.  It uses a number of different muscle groups (quadriceps, biceps, pectorals, a number of back muscles, etc).  I’m not sure why you couldn’t reasonably say that if person A could dead-lift twice as much as person B, A was twice as strong as B.

"Very strong" is subjective but there's nothing wrong with that. 

I already mentioned what I thought was wrong with it.  I’ll be happy to debate those claims with you, but right now you aren’t specifically addressing any of them.

*Edited because I can't spell "sense", apparently.

**Edited again because there were a lot of errors that I missed.

[Gr3yling]

In a CRPG or RL, I think you should only have as many stats as are meaningful.

And I realize you are right.  Attributes should probably just be a way of choosing the PC's specialization, rather than a source of realism.

EDIT: The next question would be what attributes should be used.  I think the four element idea that you pointed out is very cool, AgingMinotaur, but my personal preference is to go with a more traditional system.  It would probably include things like strength, dexterity, intelligence, and so on.

I really like the idea of attributes or skills opening up new content to the player.  For instance, in fallout 2 (I think) if the PC has a high enough perception score, they can notice that the village guard is using a sharpened stick as a spear.  And this allows them to ask the guard to show them how to make such a spear.  In fallout 3, if your strength score is high enough, people will notice that you are strong and give you quests that they otherwise would not have.  That sort of thing. 

Any other ideas?

[Krice]

EDIT: The next question would be what attributes should be used.  I think the four element idea that you pointed out is very cool, AgingMinotaur, but my personal preference is to go with a more traditional system.  It would probably include things like strength, dexterity, intelligence, and so on.

I like the way how talk about realism degenerates back into traditional rpg system.

[Xecutor]

Increased number of attributes have one serious problem. It's much harder to balance them.
At least if you have attribute bonuses on items.
The more attributes you have, the harder it is to focus (equipmentwise) on attribute that class of your character depends on.

About developing strength thru training. When completely untrained person comes to the gym, he will develop
quickly for several weeks. Then his development suddenly stops. Any further advancement requires huge effort.
That is if you don't use drugs. You need an enormous willpower to continue to develop and stay natural.
JFYI.

Some attributes are better as numbers. Strength for example. It can be objectively measured.
On the other have IMO charisma is more subjective attribute.
Strictly speaking can only be estimated by other person. And might even depend of mood of that person.
Dexterity and agility are more like attributes that can define whether you can or can't perform some trick consistently.
IMO it is almost impossible to develop dexterity. You have it, or you don't have it. You can train some trick to perfection, if you can do it at all.
Agility feels more like production of reflexes and flexibility...

[akeley]

I like the way how talk about realism degenerates back into traditional rpg system.

Agree, I must admit I find the whole fixation on attributes a bit bizarre.

About developing strength thru training. When completely untrained person comes to the gym, he will develop
quickly for several weeks. Then his development suddenly stops. Any further advancement requires huge effort.
That is if you don't use drugs. You need an enormous willpower to continue to develop and stay natural.

This is quite a weird statement - I suppose you refer to bodybuilding and steroids as "drugs" - but even so, it`s definitely not as straighforward as that. I disagree that your development will "stop" - it will slow down, obviously, and become less observable compared to the first stage, but I wouldn`t call that stopping. Also, you don`t need "enormous" amounts of willpower - it all depends on the person`s mentality, and what his/hers goals are too.

[Vanguard]

It sounds like you are saying here "bad systems of character development are bad."  Sure, any type of character development system can be done well or badly.  However, in the specific example we are talking about, the numbers give you a lot more information about the PC than just saying they are “very strong”, for all the reasons that I already pointed out.

Am I misunderstanding what you were saying?  Would your "very strong" perk explicitly say "+3 to strength" in its description?  If it would, well, that would require using attributes too, wouldn't it?

It could be something like that.  But it doesn't have to.  The cool thing about perks is that it's easy to mix and match attribute-like perks and perks that are nothing like attributes.  "Very strong" could be +2 to hand to hand damage and +20 to maximum carrying weight, like a strength attribute.  Or it could just be one of those things.  Or it could unlock a better outcomes to events where you need to lift something heavy.

Perks are good for believability because you can be as vague and subjective as you want, while attributes are inherently ultraspecific and imply objective correctness.  If you want a perk to describe exactly how your dude looks or what they can do, you can put that down.  If you want to just give a general idea, that's fine too.  It gives you, as the developer, more direct control so you can easily address implausible elements.

I’m not sure why you couldn’t reasonably say that if person A could dead-lift twice as much as person B, A was twice as strong as B.

Because a person might be stronger in some muscles and weaker in others in ways that aren't easily reflected in deadlifting.  Because being able to lift something quickly takes more strength than lifting it slowly even though that isn't being measured at all.  Little things like that.

I dunno where I was planning to go with this when I first brought it up.  Maybe I was just being pedantic.

I already mentioned what I thought was wrong with it.  I’ll be happy to debate those claims with you, but right now you aren’t specifically addressing any of them.

Is this what you're referring to?

Second, you need numbers to make accurate comparisons.  Who will do more damage, the PC who is "very strong" and has a "dull" sword, or the PC who is "weak" and has a "very sharp" sword?

You can tell the player the exact numeric values or you can assume they will know that sharp swords are better than dull swords.  It's fine either way.

The point of perks isn't to pretend that games aren't really a series of numeric calculations or even to hide the numbers.  It's just a way of giving the developer and player more control and better variety while avoiding a lot of the baggage that comes with attribute systems.

[Gr3yling]

*sigh*  I have disappointed Krice:

I like the way how talk about realism degenerates back into traditional rpg system.

...And Akeley just thinks I'm weird:

Agree, I must admit I find the whole fixation on attributes a bit bizarre.

Regarding what Krice said: okay, okay, I take it back.  We don't have to limit ourselves to discussion traditional systems in this topic.

As far as Akeley's comment: I think the reason that attributes are interesting to me is that I have an interest in anatomy and physiology.  And I keep asking questions about them because there are a lot of questions that are unanswered and are apparently very hard to answer. 

For instance, a lot of people in this thread seem to prefer attribute systems that have 4-7 attributes.  But it's hard to get a concrete answer why.  Some might say any more than that would make the game too complex, but there are games with zillions of perks and skills that people find completely playable.

And there really is more to this topic than just discussing attributes.  I wanted to look at the role of other realistic elements, like the need for food/water, maintaining body temperature,  and in depth modeling of injuries.  The question was, essentially, which of these elements has a payoff that is worth the complexity that they add to a game.  I thought we could try to come up with some sort of objective criteria to evaluate each of these elements.

So that would be phase two of the topic.  Just a preview.