Warning: incredible geekiness ahead. Read at your own risk.
However, I'm not really going to introduce all of my thoughts at once (I'll probably add more as I get the time).
Given Southie's comments, I think I should provide some more background. Here it is.
The way I started thinking about designing a skill system is by starting with a simple example: strength. I am not very strong, and I have trouble carrying a 50-lb monitor. Let's be very generous and say my maximum lifting capability is 100 lbs. There are people out there who can carry significantly more, on the order of 500 lbs (the world record for snatch weight-lifting is ~469 lbs.). What's my point? This difference is linear, not logarithmic. The same holds true for most physical tasks I can think of, like running speed and jumping distance. The question is whether this relationship also holds true for other kinds of tasks, such as mental tasks. Is there a logarithmic difference in problem-solving speed between human beings? I seriously doubt it. For instance, I'm not the fastest Set visualizer in the world, but I'm not logarithmically slower either. How about something harder, like a difficult algebra problem? I'm not fast at algebra, but I doubt there is somehow out there who could solve a problem so much faster as to require a logarithmic scale (100 times? Sure, let's be generous. 1000 times? That's stretching it. 10,000 times? Provided that I can solve it at all, I doubt there is any human who can do it 10,000 times faster).
Thus, I argue that to measure real abilities of real people, we're talking about a linear scaling. How much should those abilities cost, in experience? I contend that there is an exponential drop-off in how common exceptional abilities are in the population. For instance, I know that IQ follows a bell-curve. I strongly suspect other attributes like strength also do. While I have no data to support this (and would desperately like to find some), I strongly suspect that if you look at the degree of knowledge of a specific field (say, physics), you'll find an exponential drop-off at each increasing level: people who know algebra-based physics, calculus-based physics, have a physics B.S., have a masters, have a Ph.D., are at the top of their fields in the world, have a Nobel prize, etc. Also notice that, despite how presumably people specialize in what they're good at as they become adults, it doesn't really take less time for people to learn the material. I think this reflects that achieving increasing degrees of specialization requires exponential amounts of experience. It is usually not too hard to become passable at a task, but it requires long devotion to become really good at it.
This is one of the fundamental mappings of my system: linear ability scale, exponential experience scale.
Now, what about tasks? Most tasks are not simple "do-or-do-not" tests like lifting. There are gradations of success, which I plan to use a normal distribution to represent (see the next section). Also, most tasks have a threshold, a minimum skill you need to succeed in the task. For instance, I could not run a 4-mile marathon, i.e., maintain a running gait the whole time. I'd collapse before I finished. Similarly, an illiterate is not going to succeed at reading anything without instruction or lots of time (equivalent to spending experience to learn it). A blind man will fail any task requiring sight, automatically. No amount of learning can help that.
This is all well and good, but it only really deals with tasks whose difficulty is equal to the skill in question. The simplistic solution is just to use another normal distribution. But I'm not sure I believe this. Thus, the next section.
My original question:
What I'm specifically looking for is the resolution of a dilemma: how I should scale the probability of succeeding at a task with the difference between the skill in question and the difficulty.
To be more algebraic, I have four variables here: difficulty (d), skill (s), degree of success (x), and of course probability (p). Holding d and s fixed, it seems to me that x should be normally distributed around a mean which depends on the difference between d and s. The variance should be set by s. (I have reasons for this, but I don't really have the time to explain them now). Now, ignoring x, I am interested in how s and d each change when the other is held fixed. The problem: I can think of no real reason to pick one distribution over another. A normal distribution implies assumptions a high degree of nonlinearity in how much improvement in success chances you get for a small increase in skill around the critical point (the difficulty); it's a sigmoidal curve. I've played a game with a system like this (Asheron's Call), and I'm not really confident this skill system is realistic. The real question is, what sort of distribution would be realistic? How much improvement in success chances should you expect? That's the issue I'm wrestling with right now.
To make the obvious response, it depends on how you're measuring "skill" and "difficulty". You can say I'm "level 12" in video games, with "300 experience" until level 13, but that doesn't mean anything without knowing whether the experience scale is linear, quadratic, exponential... ditto with the level. I could require ever-increasing experience for linear gains in ability, or exponential experience for exponential ability (signifying large gaps in improvement, but similarly large jumps in skill level). For instance, I can juggle four balls (not very well). Because of this, most things difficulty 3 balls is pretty trivial (except for some reason my mind doesn't seem to work around "the crane"). Things difficulty four are pretty hard. Difficulty five and up is impossible. Now, it'd take me, say, a month to get to five balls (working on and off). After that it'd probably take me three months to get to six balls. This would probably be a normal distribution with exponentially increasing effort to increase skill.
For another thought, I'm ranked roughly in the 1500s for chess. There's a nice little table that says I've got a 75% chance of beating somebody rated (or 'difficulty') 1200. This is actually also probably a sigmoidal curve.
I'd like to work in a video game related thought in here, so let's take a look at SuperSmashBrothers. "levels" are pretty hard to assign, but we did do a brief thought of ranking at one time. I played a few games against SteveHaas one on one, and he won a few, while I won more (much to his anger at the random number gods). If we kept playing, odds are his win percentage would be about the same. Take people a bit lower than Steve, though, and I'll spike them into a pit without losing a life. Effectively, there are many people that I will not lose to, regardless of the game dropping bombs on me and giving them stars and hearts. Logically, there are people out there that I would lose to easily, although I would hope that I'd make a pretty good showing of myself.
Basically, you can make any definition on anything, as long as it works, is fun, and reasonable. These are some of my thoughts on "real skill/difficulty comparisions", though. I suspect the sigmoidal curve is a decent approximation. -- EvilSouthie
EvilSouthie's competition examples are very interesting. I think the juggling is an example of exponentially increasing effort (experience) to get a linear improvement in skill. As for the rest . . . they are examples of competitions, which may be a different kind of task entirely than ordinary tasks. Or . . . maybe not. Maybe a normal distribution is a reasonable simulation (it would certainly make my task easier).
Southie, could you provide a link to such a table? Also, what does the formula look like (if you know), or could you point to where to find out (my first-blush google search didn't pan out)?
Assuming you're talking about the USCF table - http://www.uschess.org/ratings/info/lifetime.html has a table about a third of the way down. No clue what formula they're using to generate said table, but with that many data points it shouldn't be hard to find a fit.
Since USCF ratings only show past performance and not current skill, the formula to correct a rating over time is shown at http://www.ohiochess.org/ratingch.htm .
To make this page even more long and rambling than it already is, I'm going to add some more.
Now, one of the fundamental principles which I established above is that a linear change in skill is linked to an exponential increase in the cost of the skill. In fact, I'm going to state it like this:
Note that this doesn't really mean anything yet, because I haven't established how the stats relate to the probabilities of success at tasks. That will come later. Now, I'm concerned with another issue: how to deal with the nature vs. nurture problem. Specifically, how should inherent abilities be handled? This is actually a very thorny problem, which has bedeviled far more serious undertakings than designing a role-playing game. However, the real problem is that it's not a simple case of what attributes should be assigned to inherent and which to learned, but rather a complicated interaction between genes and environment. I'm going to effectively dodge the issue by taking the same functional approach to genes that role-playing systems take to skills: rather than concerning myself with how real genes work, I'm going to classify them according to aptitudes.
Basically, under this system an aptitude is something inherent (to pick a real world example, a tendency to develop more slow-twitch rather than fast-twitch muscle). What it does in game terms is to reduce the cost of buying certain skills, which are learned abilities. Note that I use 'learned' here very loosely; for instance, your physical strength is learned because you can change it by lifting weights and performing sit-ups. How exactly the aptitudes work is something which I will work out later, but there will be one aptitude for every skill. Aptitudes will not (usually) change after character creation, except in negative ways (having your skull bashed open might decrease your intelligence, which would increase the cost of buying most mental skills).
The skills, of course, are everything else. As I said, I plan to include some very odd attributes under skills, things which you wouldn't normally expect to find. However, skills are the meat of what should normally change during the games. Moreover, skills and only skills determine the chance of success at a task. An aptitude should NEVER be used to make a check of that kind. This leads to:
Now, as I've argued before, the quality of success at a task should be normally distributed around some mean determined by the difficulty of the task and the skill of character attempting to perform it. In fact, when actually performing a task, the only quantity which matters is the difference between the skill and the difficulty.
There should be a function, which I haven't determined yet, that maps the difference in a character's skill and the difficulty of a task to the mean of the normal distribution. I'm not sure how to do that yet, but it will Axiom 4 when I do.
What about the variance of the normal distribution? This should be dependent on the skill, and only the skill. One of the effects of increasing skill is increasing reliability of performance, i.e. a decrease in variance. "Beginner's luck" is a superstition which I'm convinced stems from this phenomenon. I've seen it myself in my performance in video games and classes, where my homework scores tend to even out after the first few weeks of the semester. Thus, Axiom 5 connects variance in an as-yet-to-be-determined way with skill.
This is the skeleton of my system. I intend it to work on pen-and-paper, so I have to figure out a way of condensing this analysis into something that can be done with dice. Luckily, normal distributions can be easily simulated, provided you have a sufficient number (10 should be fine) of dice. Once again, I'm going to be adding more pieces as I think about them. I may have to wait awhile and see if I can think of some good justifications for picking the functions in Axioms 4 and 5.
After some more analysis, I think I have Axion 4: the difference between skill and difficulty maps linearly into the mean of the Gaussian distribution for Axiom 3.
Why? This satisfies the linear increase in capability I discussed above in my simple examples of lifting and some basic cognitive tasks. However, it also deals nicely with Southie's competitive examples. Assume, for a moment, that points in this system correspond to standard deviations away from the mean (this system is almost certainly too coarse to be used, but I will probably use something like 1/2 or 1/3 standard deviations for the real system). Then a competitive difference of 1 point will mean the stronger player will win more than about 2/3 of the time (presuming ties go to the higher skill); 2 points, more than 95%; 3 points, more than 99.9%. In other words, the chance of losing to a lesser player with much of a difference in skill drops very rapidly. In other words, the solution to this problem was there in front of me all the time, I just didn't see it.
I have a tentative dice rolling system to work with these probabilities. However, to make it work, I would want to renormalize all the dice so they have a mean of 0. Thus, a d4 has -2, -1, 1, and 2 on its faces, and so on. Why? This way I can easily sum a bunch of different die sizes together without changing the average of the collection. That these types of dice aren't standard is problematic, but it can probably be worked around in some way. The reason I want to use different die sizes with the same average is so that I can control the variance by changing die sizes.
Actually computing the combination of die sizes using a number for variance is probably going to be quite difficult. I believe there is a technique from linear algebra that I can use; however, for playing the game and for people who are less mathematically inclined, I think I will probably have to resort to a table. While not necessarily the best method, I think it is more justifiable in this case than when the d20 system uses tables to replace 1 variable linear equations with rounding.
Now I'm actually looking into the details of the mathematics of these dice. First, I found the correct formula for the variance of a die (use the formula for the sum of squares up to n, 1^2+2^2+...+n^2 = 1/6n?(n+1)(2n+1) and standard probability calculations), which is 1/12(n^2-1). The correct formula for a die with a mean of zero as above is 1/6(n+1)(2n+1); this latter formula is larger, so my dice will have larger variances than standard dice. Using the Platonic sizes, here are the actual numbers:
Die: Var. Std. Dev. d4: 1.25 1.12 d6: 2.92 1.71 d8: 5.25 2.29 d12: 11.92 3.45 d20: 33.25 5.77
Die: Var. Std. Dev. d4: 2.5 1.58 d6: 4.67 2.16 d8: 7.5 2.74 d12: 15.16 3.89 d20: 38.5 6.20
The standard deviation figures are not completely accurate, since these are not really normal distributions, but are relatively close approximations. The formula for the sum of the variances of independent random variables is simply the sum of the variances over the number of variables, so it is possible to create a distribution which has any variance below the largest variance here, at least approximately.
When I was first thinking about this system, I considered the possibility in order to simulate a normal distribution, it might be necessary to generate "tails" using repeated die-rolling to simulate the exponential drop-off. What are the actual probabilities of a worst-case situation which needs tails, i.e., where all the dice turn up either at the very high or low end of the range? With 10 d4's (d20's have higher variance, but lower probability of hitting all at the bottom or top), the probability is 2*2^-20 = 1.91e-6. This is pretty negligible; if you role-play weekly and have 1000 die rolls per session, which is vastly in excess of the longest, most combat intensive session I have ever seen in White Wolf or DungeonsAndDragons, you will encounter approximately one incident every ten years (52 weeks in a year). Thus, I don't think I need to worry about the tails, unless I need to reduce the number of dice rolled for convenience reasons.
Continuing this ramble, while I'm still in doubt about Axioms 4 and 5, I can still think ahead some to what kinds of aptitudes and skills I should have. I am going to adopt the incredibly ambitious goal of attempting to create a system which will work regardless of setting, ala GURPS. This won't work, but I think I will at least be able to take a good first stab at it, and it might provide some useful insights for a setting-specific set of aptitudes and skills.
More to come . . .
Since the details of the die-rolling can easily be taken care of later, I suggest that you try to build the system first using a computerized RNG with a normal distribution to play-test early on. This will achieve two useful goals: You will be able to work out the appropriate variances and means using a much better approximation, and you will be able to balance the system without having to worry about quirks of the die-rolling system. Then you will be able to replace to the actual Gaussian with a die-rolling system that you can design based on actual experience as to how the the normal distribution affects gameplay. I agree, by the way, that making the game possible to play without a computer or calculator is a noble goal.
So, here goes. I need to create a list of aptitudes and skills which encompass all human achievement. A simple task, really. Not! Here's what I've come up with so far, and I'm definitely interested in hearing comments.
These are the easy ones. I'm combining smell and taste into one category, since for most purposes, the taste of something is primarily determined by its smell, and frankly taste is not important enough to deserve its own category. I'm not sure what to do with touch, and pain tolerance/sensitivty probably needs to be dealt with. Some of these aptitudes may influence other skills, but each also has its own skill which is used directly. Touch may not merit its own skill (how often would a touch skill check come up in a game?) but it probably would have influence on other skills.
I can see touch coming up in some situations, though most of them are activities likely to be performed by thieves -- picking pockets (or realizing that your pocket has been picked) or doing various stuff in the dark or out of sight. Whether you'd want to wrap this into manual dexterity is something else again.
Aptitudes: Sight, Hearing, Smell/Taste?, Touch?, Pain?
Skills: Sight, Hearing, Smell/Taste?, Touch?, Pain?
While physical qualities seem like they should be straightforward, when I actually got to thinking about them, they're not as clear as you'd think. Strength is fairly obviously an aptitude (build, gender) and a skill (working out makes you stronger). I will use endurance to represent a very specific thing: the efficiency of one's body at handling strenuous phsyical exercise over long periods of time, i.e. marathons and the like. I have good reason to believe this is strongly connected to the oxygen transport capability of the blood, and a lot of modern track training centers around boosting red-cell counts and the like. Next up: fine motor control. Some people are better at it than others, but I'm relatively certain that it too can be influenced by training (video games, anyone?). Size is a physical aptitude and, to a lesser extent, a skill (it's a skill which doesn't change much, but certainly can be changed by things like malnutrition during childhood). As an aptitude, it should certainly influence skills like strength (bigger people are stronger) and speed (they're slower, also). Pure speed, i.e. how fast can you accelerate, is another physical quality which is based on characteristics like fast-twitch vs. slow-twitch muscle tissue, and should likewise influence qualities like endurance (faster people have less of it). Reflexes is a quality which measures reaction time, more than muscular acceleration. Obesity is a quality which I feel ought to be included, but I'm not sure how I'm going to handle negative aptitudes. Similarly, I feel that I should include some aptitude (and maybe a skill) for body type, i.e. short and squat or tall and thin, big-boned or not, etc. Immune system is a quality which doesn't come up much in role-playing games (players don't like random diseases to fell their characters or even make the characters feel crappy for a while) but plays a big role in real life. Also, I might need a quality to represent the general resistance of a character to physical damage, i.e. being hit by a car or other such trauma; I'm calling that toughness for now.
So how do these interact with fatigue? I can see endurance helping with fatigue checks for physical activity. I personally wouldn't think that video games are the best way to train fine motor control, and caffeine reportedly has deleterious effects on it. Video games train reflexes (of a sort, anyhow). Does toughness affect rate of blood loss, or just how severe the fatigue check for an injury is?
Are you lumping arm strength in with leg strength? Does strength therefore affect jumping ability? Is "agility" just a combination of speed, reflexes and coordination (I'm imagining the skill used in gymnastics/acrobatics)? Sense of balance? --AndrewSchoonmaker
Endurance certainly affects fatigue; it's the skill that's appropriate to use to make fatigue checks for most physical exertions. I didn't discuss drugs in the fatigue section; that has now been corrected. I disagree about video games training fine motor control; I actually had a psychologist who suggested that I play video games to improve my motor control when I was in early elementary, because I was frustrated at my lack thereof (my reflexes still suck, but I've learned to cope better). Toughness might affect how severe an injury is, and possibly bleeding. That's one of those issues I haven't sorted it out yet. Yes, I am lumping arm strength in for leg strength; for jumping, I might have a separate skill, or alternatively, have it determined by a formula like Strength - Weight. Agility is as you describe, a combination speed, reflexes, and coordination.
Thus the list:
Aptitudes: Strength, Endurance, Fine-motor Control (for sake of wieldiness, maybe call it Coordination), Speed, Size, Obesity?, Reflexes, Toughness?, Body Type?, Immune System
Skills: Strength, Endurance, Fine-motor Control?, Speed, Size (will have to be handled differently than most skills), Obesity?, Reflexes?, Toughness?, Body Type?, Immune System (yes, it's a skill; why else did you get sick so much as a kid?)
Heaven help me. Mental qualities are incredibly difficult, much harder than physical attributes and about as nebulous as social ones. I'm still working on them, so I'll add more later. Of course, the 900-pound gorilla of this exercise is the IQ question: is there really some g factor, as some psychologists argue, which represents general intelligence and is strongly genetically linked? I'm not sure, but I'm inclined to think that even if such a thing exists, its influence is damped by other random genetic factors. Moreover, this is a role-playing system, not a treatise intending to solve enormous outstanding problems in psychology and the philosophy of the mind. Thus, for balance reasons and with some support from other theories of intelligence, I'm going to go with a more complex model.
Aptitudes: Memory, Language, Mathematics/Logic?, Spatial Reasoning, Musical Ability, Naturalist Ability, Introspection, Memory, Common Sense . . .
Skills: Not there yet; so many I don't even want to think about it yet
I am frankly a little stuck with skills. I'd like to hear other people's ideas on this, to see if discussion will help me think more clearly about the issues.
It looks like you have mostly "pure" skills here now. Are the more complicated things which are clearly skills going to simply require multiple checks, or do they get skills of their own? For example, I can think of at least four things you'd reasonably need to check for Archery -- Strength, Coordination, Sight, and Spatial Reasoning. Firing a handgun would be similar, minus the strength check (though you'd want to check something to see if the character is put off-balance by the recoil); does this mean that any good archer is also de facto suited to wielding guns? Other martial arts (not just of the Eastern variety)?
What about stealth?
It seems as though, no matter how comprehensive a list of skills you come up with, somebody will occasionally need something outside the system. Any ideas on how to handle that? (Or is that why you don't have any listed second-order skills?) --AndrewSchoonmaker
At the moment, the system I have in mind can be described as a set of aptitudes, with edges going to the set of skills. There are no limits on how many skills an aptitude can affect, though of course those that affect more skills will cost more. The original intention I had was to have, for instance, Archery, affected by aptitudes like Strength, Coordination, Sight, and Spatial Reasoning. This would be entirely separate from the skills Strength, Sight, etc., on the theory that while being a good archer might give you strong arms, this doesn't necessarily mean you're "Strong" in the game sense. However, examining the issue more closely, I'm worried that the non-linearities in the real world are going to completely destroy this model. I.e., it seems like Archery should be affected by the skill version of Sight, for instance, and that Archery should have a cross-effect on Strength. I'm not sure how to treat this mixing. I'm also not sure if I'm producing anything like a reasonable list for mental aptitudes, much less mental skills, and there are least two more frustratingly ambiguous categories to deal with: social aptitudes/skills, and personal characteristics like temper, ability to defer gratification, etc. These issues are why I haven't gotten past where I have in skill design.
I had an interesting discussion of the fatigue system, well-written with good examples. This is what I remember of it.
Most role-playing systems assume that humans are always operating at their optimal capabilities. This is not realistic. For instance, at a given time on a given day, depending on how recently I've eaten, whether I got a full night's sleep the night before, whether I'm sick, and how much stress I'm under, my capabilities can vary significantly. The fatigue system is intended to represent these effects.
One of the goals of this system is to model real humans with the minimum complexity necessary to really capture how humans behave in everyday life. In the case of fatigue, I'm going to do this by lumping together all the possible sources of fatigue into a single system with regard to their effects. The sources of fatigue, in no particular order, tend to fall into these categories.
Whenever a character engages in one of these activities, that character must make a fatigue check to avoid accumulating more fatigue points. Fatigue checks have basically two parameters: frequency and difficulty. GM's should pay careful attention to the difference between the two. For instance, easy, frequent checks are for tasks like a light jog that wouldn't really strain a fit person, but that would prove a considerable challenge to a 500 pound couch potato. Hard, infrequent checks are for behaviors which no one is unaffected by, such as fasting; note that in some cases, like going without water for days, checks should be bypassed and fatigue points should accrue automatically. Easy, infrequent checks are rarely worth rollilng, while hard, frequent checks are for situations involving maximal stress and exertion, like combat. In some rare cases, it might be best to treat fatigue by assigning a few, permanent points to a character: for instance, a character with a debilitating, chronic, painful disease should be thought of as in a constant state of fatigue. Fatigue checks occur against an appropriate skill. Endurance, e.g., would be appropriate for most physical exertion fatigue checks.
The above paragraph describes how fatigue points are acquired. The next question is how they map to penalties to capabilities. This question is something of a dilemma because of a contradiction between two goals in my system design. The first goal is to minimize dice rolling, simply because it interferes with the flow of the game. The second goal is to eliminate "magic numbers." In order to get a semi-continuous (no magic numbers) mapping between fatigue points and fatigue penalties, dice-rolling is going to have to involved. I think that in this case, dice-rolling is the lesser of two evils. The effect of this decision is to require two rolls instead of one every time a fatigue check occurs: the first to determine how many fatigue points accrue, and the second to determine the current penalty. This penalty remains until the number of fatigue points a character has changes. The exact mapping function is yet to be determined, and that depends a lot on the details of how large the scales on the skills are and how dice rolling will work. However, it will be Gaussian, just like almost everything else in this system. One final note: fatigue checks are based on skill checks with the fatigue penalty added in. Thus, as a character becomes more exhausted, it becomes increasingly hard to pass fatigue checks, increasing the rate at which a character becomes more fatigued.
After a full night's sleep, provided the character has access to food and water, fatigue points should usually reset to zero. At the GM's discretion, resting, eating, or performing relaxing activities during the course of the day might restore lost fatigue points.
Is "just waking up" to be treated as fatigue? Some people do pop out of bed like toast out of a toaster, but I am very rarely one of them... --AndrewSchoonmaker
Yes, definitely. (I, incidentally, do pop out of bed like toast out of a toaster, but I should have thought about that, because I know a number of people who don't).
The traditional way of handling injury in role-playing games is hit points. This is a really, really bad method, particularly when there are no penalties to actions from losing hit points. In the d20 system, for instance, you can lose 90% of your hit points and still perform everything you were able to do before, at the exact same efficiency. I will trade some simplicity for much more realism in the handling of injuries.
The first stage in determining what happens when a character is injuried is to check for death. There are certain types of injuries that are more or less invariably fatal: serious damage to the heart, brain, or both lungs will usually cause death within minutes, if not sooner. Most injuries have this chance to "critical," killing the target instantly.
If an injury isn't instantly fatal, two more things need to happen. The first is determination of bleeding. Injuries which are not instantly fatal may still cause enough bleeding for the victim to lose blood pressure, go into shock, and then die. The other effect is the mechanical effects of the injury itself: broken legs will prevent a character from walking, while having both eyes poked causes blindness. Injuries also cause fatigue checks, depending on their severity.
Bleeding is handled by assigning a wound a certain rate of blood loss. The total rate of blood loss is simply the sum of the individual wounds' loss rates. Each individual wound has a chance to stabilize (clot, more or less), decreasing or stopping the loss from that wound. Medical attention can also stem blood loss, and is recommended to players who want their characters to live :). Characters, in general, can stand a certain amount of blood loss, but eventually they will go unconscious and die. Bleeding is the part of my system which is closest to hit points; however, note that while injuries in a hit point system affect hit points directly, injuries affect blood loss rate, not blood volume directly.
Mechanical effects have to be dealt with on a case-by-case basis. I'd like to come up with some guidelines and general categories of injuries, but that will have to come later.
The perverse section of my brain wants to know what happens to a character who is hit by a flamethrower, fireball, or what-have-you. I'd think the chances for shock would be pretty high (depending on severity of burns), despite a lack of blood loss.
Severe burns would fall under mechanical effects and bleeding (here, it is isn't so much "bleeding" specifically as "leaking all sorts of vital bodily fluids", but for game purposes, the effect is about the same). If the burns are bad enough, you're going to lose a lot of blood very quickly.
Also, how does internal bleeding factor in? Is it just lumped together (but harder to fix)?
It's lumped together. Perhaps an example will make this clear.
Player 1: "My character looks Bob over. Do I see any visible injuries?" GM: "No, Bob appears fine, if bruised a bit from the fall. Bob, you don't feel so well; you're bleeding at a rate of 1 point per turn." Player 2: "Aw, crap, internal bleeding. Does anyone here have the First Aid skill?"
I don't know if this is the kind of feedback you're looking for, but I have some fundamental questions about this system.
This system is partly an outgrowth of my frustration with being unable to reasonably simulate ordinary people in most of the systems I'm familiar with. The StoryTellerSystem?, for instance, doesn't allow you to create a character who has anything resembling the skill set of your average American: most people would reasonably have a dot or two in a wide variety of skills, but characters in the system tend to be either highly specialized or mediocre at doing a wide variety of things (I don't have my books with me, so I can't really craft a good example right now, but I tried in the past and generally failed). GURPS is interesting, but I have serious philosophical issues with the "fake stats" approach of strength, constitution, intelligence, etc. The d20 system is oriented primarily towards combat, but it doesn't simulate even the mundane melee skills in anything resembling a realistic fashion. I want to avoid crushing the story with realism, but I want to hew closer to the real than most systems do.
Ideally, one of my goals for the system is to have most of the relevant tables fit onto one sheet of paper. Rather than establishing a specific metric for each skill, the idea is to establish a general scale of how much proficiency a numerical value in a skill represents. I think the best way to think about this is to look at individuals who possess a certain degree of proficiency, and then establish a metric according to that. For instance, a very crude scale could go something like newbie-amateur-professional-"best in the world." My final scale should have a lot more gradations than that, but that's the idea. An important thing to remember about the metric is that everything is relative between the numerical difficulty of tasks (determined by the GM during the game) and the numerical ratings of skills.
As for the number of skills . . . this is a problem. It's a problem in any skill-based system which is trying to be remotely realistic. I'm not sure how I'm going to handle it.
This is another problem, and one which is very deeply rooted in role-playing systems. I suspect that one of the reasons role-playing is popular with nerds and other physically unfit individuals is that a character can be as physically capable as the player isn't, while the player's mental advantages (and disadvantages, but that's another issue) remain present during the game. Ideally, the players are mature enough and skilled enough at role-playing to separate their mental abilities from their character's abilities, but this is hard to do even when you are consciously trying.
To some extent, in my system, this would have to be moderated by the GM. If a player comes up with a brilliant plan to take over the world, the GM might have the player roll the character's "Mad Scientist Scheme" skill to determine whether the character would have been able to think of that. If a player overlooks something the GM thinks the character should think of, then a discreet roll of the appropriate skill might occasion a comment of, "Alice notices that letter is almost certainly a trap planned by Bob." It is going to be a bit different in style than I've seen many games run, but I don't think it's fundamentally impossible, provided players and the GM are paying attention to what's going through the character's heads. Since I also want to represent personality characteristics, I suspect a similar process will have to be used for them.
I think you're misinterpreting what's going on (which is understandable, since this page is more stream-of-consciousness than well-ordered essay). E.g., what I intend: Player A says to the GM, "I want my character to do X." GM thinks about it a moment, says, "OK. Task X has a difficulty of Y and uses skill Z. What's your character's rating in skill Z?" The skill check is rolled, degree of success or failure is determined, and then the GM uses that to determine what happens. The only wrinkle I have added is the possibility that a character is fatigued, which will result in a penalty applied to the skill rating.
The aptitudes modify experience and character creation costs for skills, not the skills directly. Character creation will be complicated, and might benefit from the use of a computer (a spreadsheet, more or less). In general, I want to cope with as much complexity during the design process as possible, so that it can be dealt with/simplified/abstracted/ignored. Then, most of the rest of the complexity should be pushed into character creation, so that the course of the story is interfered with as little as possible.
I agree. My goal is to minimize complexity during the game, at the cost of worrying about it during the design process and character creation. I also think that the most interesting stories come from improbable characters doing improbable things. Nothing I'm talking prevents the creation of improbable (read: interesting) characters. As for improbable things: well, I prefer if doing improbable things requires beating odds or persisting despite them.
You know, you've introduced so much complexity and DM discrection that you may as well forgo the whole "system" thing and just have the DM say what happens.
I think you're starting to run up against a fundamental limitation of RPGs, to wit, the two competing forces of realism and playability. Yes, realism is a good think, in general. But if one goes too far towards realism, things start to suck for two reasons:
1) The system becomes massively unwieldly. The ideal system interferes with the story/action/whatever as little as possible. This is the motivation for diceless RPGs... don't wreck the mood by stopping to role dice every time someone tries something, and just make up something appropriate. And its a decent idea, just hard to deal with for the GM, as coming up with realistic results for fanstastical characters on the fly is, to say the least, challenging.
2) The system becomes too real and ceases to be fun. Let me give you an example. When I was about 10 my brother brought home this new game system he had found (Runequest, if you care), and we decided to try it. It payed exception attention to detail... character creation took hours... characters had fully fleshed-out skills, abilities, and equipment consistant with the surrounding world. Whereupon we (the NPC and my character) walked out of town, ran across a monster, and attempted to slay it.
At that point I needed to create a new character, as my first one had gotten himself impaled on a spear and found himself rather dead.
So I did. This time optomizing for combat ability. And in all fairness, that character faired slightly better. In his first combat he managed to slay the orc (yes, singular) and survive... though it did cost him the use of one arm for the rest of his life, whereupon my brother and I concluded the system was dumb and went back to AD&D.
My point? Realistically, the average starting adventurer has an exceptionally short lifespan. Realistically, any combat that is dangerous enough to be interesting will wind up crippling or killing much of the party. Realistically, after the first time you get bitten by a dragon, hit with an axe, or shot with a shotgun, you should die or, at the bare minimum, be out of the fight. But that makes fights either exceptionally fatal (everone hits a lot and people drop like flies, as Runequest above) or extremely boring (no good examples of this, though Star Wars first edition comes close... to kill a typical stormtrooper you need to shoot him about 4 times, you only get one shot per round, and only hit about a third of the time. It takes an hour to roll out a fight between a party of three and five stormtroopers. And the party lives... though most probably need medical attention...) as hit probabilities are minute.
Thus, the ideal RolePlayingSystem, in my book, is fairly realistic but also ignores the more cumbersome details in favor of 1) what is easy/fast to adjucate and 2) what makes for a good story. Its fun to slay the dragon. No one wants to hear about how the party needed to sew themselves up afterwards.
I don't think such a system exists yet. There are a few that are pretty good, at least over some level range, but all systems break to various extents at some point and tend to fail in one of the above ways anyway. But I think it would be fun to try to design a system that didn't break down like this. But that's a bigger project than I really feel like messing with at the moment. -- SteveHaas