I'm part of a team in the process of designing a skill structure for a MUSH. Some fairly unique problems have cropped up; I'm hoping you guys are able to offer some advice.

Let me preface this with some background information regarding the game, the goals we want the skill system to achieve, and some of the ideas we've tinkered with.

The MUSH is being designed to be as flexible as possible. There are no restrictions on what a player can try to accomplish; if, for instance, someone wants to play a lawyer, the game is going to be set up to allow that. At the same time, if someone wants to play a martial arts master, this too is acceptable. Some of these things require game mechanics to handle (for instance, the martial artist needs a combat system), others do not (a politician does not require a "politics" skill; such things are role played).

Though any and all character ideas are welcome, it is expected that military conflict will be a focal point for the MUSH. With this in mind, we have hit an interesting snag; allowing character advancement while at the same time preventing enormous power discrepancies. It would be unacceptable, for instance, to have a linear advancement model (such as the d20 system). Can you imagine a level 20 character (in d20) engaging in combat with level 1 character? Such a system would make it impossible for newbies to survive any kind of combat situation with more experienced characters.

On the other hand, we do not want to see a system where advancement is meaningless. In Shadowrun, for instance, new characters have a decent chance in combat with more karma-laden opponents. But the benefits from advancement are too limited in that system; higher target numbers are nigh-impossible for everyone, regardless of skill. Not to mention the advancement pace; it takes far too long to see character growth in that system.

Thus, the crux of the problem: what kind of system can we use that rewards advancement properly, but also prevents experienced players from dominating the game?

One system that has been proposed (though it is incomplete, at the moment) is as follows:

Skill levels accumulate a number of behind-the-scenes "points" which are then used to determine outcomes of skill tests. The points given per skill level decrease as the level goes up. For instance, while going from skill level 1 to skill level 2 might yield 10 points, going from skill level 150 to skill level 151 would yield only one point. (For the purposes of this example, assume that the number of points yielded per skill level decreases by 1 every 15 skill levels)

Success for a given skill test would be determined in a manner such as the following:

Joe has a rank of 50 in "Acrobatics". Fred has a rank of 25 in the same skill. In an acrobatics test where Joe has a 50% chance of success, Fred would have about a 28.75% chance of success. This is calculated as follows: Because 50 ranks of Acrobatics gives Joe 400 points and 25 ranks gives Fred 230, we can boil it down to the following equations:

--------------------------------------------
400*x = .5
230*x = ?

x = .5 / 400

230*x = (230 * .5) / 400 = 115/400 = .2875
--------------------------------------------

Arbitrary success rate caps can be implemented. For example, chance of success can never be higher than 95% or lower than 5%. This gives approximately the kind of diminished returns we are looking for, though playtesting would be required to fine tune the numbers.

The problem with the system I outlined above comes when you tackle opposed tests; who, for example, would win an opposed armwrestling test? Without a specific difficulty to test success against, the system breaks down. It becomes an especially tricky problem when you want to maintain a remote chance for a newbie to succeed, even against a hardened professional.

It should be noted that the system that has been described deals solely with reaction/active skills; that is to say, it does not encompass knowledge based ones. Someone who has never lifted a gun before has a remote yet non-zero chance of shooting a target that is a hundred yards distant; someone who has no concept of zero doesn't have any chance of solving a complex calculus equation. Knowledge skills are different than physical skills, and are to be treated as such.

Also important to note: As the format we're dealing with (code, versus pen and paper) is not restricted to things such as dice, we have decided to embrace that freedom when designing a system. This is obvious, I think, when looking at the example I provided. If a dice-based solution can be found, however, please don't hesitate to bring it to my attention.

Now that I've outlined the problem, some of our attempted solutions, and the goals we have, do any of you have some suggestions for tackling the issues mentioned? Any and all responses are vastly appreciated.

I've helped run a number of MUSHes in the past, including Ambermush, Hellmush, and a few others.

There's one thing I really hold strongly as my impression regarding creating advancement situations where a grand disparity between characters is not eventually created if the MUSH lasts long enough, and that's this:

If the amount of sheet-affecting advancement per time unit is greater than zero, it can't be done.

At the least, it's fantastically difficult. And with the restriction that it can't be too slow either, you're more or less killing your already slim chances.

Given my druthers and a freedom to ignore your restrictions, I'd split advancement up much the way we've done it in Fate; slow advancement on the sheets, but plenty of rewards in the form of "fate points", which can be spent (and then are gone) on individual die rolls, nudging results in the spender's favor. This way they're getting some kind of frequent reward, which players in an advancing game seem to want, but their actual capabilities do not grow at too fast a clip.

I suggest the following family of functions for chance of success, C(x), as a function of rank:

C(x)=1.8/(1+e^(-k*x))-0.85

These functions are all strictly increasing, equal to 5% at zero and tend toward 95% as x goes to infinity. The smaller one makes the constant k, the more gradually the function tends toward 95%. If you want someone at rank 50 to have a 50% chance of success, try k=0.022. Plot this function using your favorite plotting tool (e.g. a graphing calculator).

If you like these functions, the problem is reduced to handling interactions between characters with known success chances.

Thanks for the input! It's good to know that we aren't the first to encounter frustration in this kind of environment.

iago wrote:
If the amount of sheet-affecting advancement per time unit is greater than zero, it can't be done.

At the least, it's fantastically difficult. And with the restriction that it can't be too slow either, you're more or less killing your already slim chances.

Given my druthers and a freedom to ignore your restrictions, I'd split advancement up much the way we've done it in Fate; slow advancement on the sheets, but plenty of rewards in the form of "fate points", which can be spent (and then are gone) on individual die rolls, nudging results in the spender's favor. This way they're getting some kind of frequent reward, which players in an advancing game seem to want, but their actual capabilities do not grow at too fast a clip.

It seems that you're dead set on this "perceptible" advancement. But I agree with Fred that it's quite problematic. For one, there's a general law of perception which says that the more of something you have, the more marginal increase you need to notice a difference. For example, to have a noticeable difference in a noise, you have to nearly double the actual volume in terms of decibles.

So, other than the little number changing on the sheet, players won't notice a difference, reliably, until there's a large difference in actual effectiveness. So, either rely on that little number (which is valid, having a guage makes a large difference), or give up on the idea as of having both perceptible differences, and simultaneous balance.

Have you considered not having any advancement at all? I've been waiting for the MUD, MUSH, etc, that embraces this principle. If you can "do anything" in your game, then why do you have to have increasing abilities in order to make it interesting? Just a thought.

Mike

Mike Holmes wrote: Have you considered not having any advancement at all? I've been waiting for the MUD, MUSH, etc, that embraces this principle. If you can "do anything" in your game, then why do you have to have increasing abilities in order to make it interesting?

Mike Holmes wrote: It seems that you're dead set on this "perceptible" advancement. But I agree with Fred that it's quite problematic. For one, there's a general law of perception which says that the more of something you have, the more marginal increase you need to notice a difference. For example, to have a noticeable difference in a noise, you have to nearly double the actual volume in terms of decibles.

So, other than the little number changing on the sheet, players won't notice a difference, reliably, until there's a large difference in actual effectiveness. So, either rely on that little number (which is valid, having a guage makes a large difference), or give up on the idea as of having both perceptible differences, and simultaneous balance.

Mike Holmes wrote: Have you considered not having any advancement at all? I've been waiting for the MUD, MUSH, etc, that embraces this principle. If you can "do anything" in your game, then why do you have to have increasing abilities in order to make it interesting? Just a thought.

I'd be happy to discuss that particular issue with you via PM, if you'd like; it is an interesting one. Again, however, I'd really like to prevent the thread from getting off the topic I originally posted about;

Mike Holmes wrote:
Can I ask, what merits advancement in this game? What behaviors in general are you trying to reward with your system? That would help in determining what makes for a good system with which to provide said rewards.

What are these misgivings? It seems like such a system, applied as the only means of character improvement, would be a powerful tool, so long as the character improvement specifically allowed the characters to perform more interesting pieces of roleplay (the +noms allow you, eventually, to interact with your environment instead of just being in it, edit areas, create objects, etc, but don't create combat effectiveness or anything like that.)

It sounds to me a bit like your team is suffering from what I tend to call Final Fantasy Syndrome. I call it that because it struck me as an influence the video game was having on some D&D players when it first came out; you could call it Big Numberism if you prefer. The idea is that we need to have huge numbers to feel like we're getting somewhere. Final Fantasy gives you thousands and thousands of experience points, gold coins, and whatever other rewards they have, advancing your levels swiftly toward triple digits; yet you're not really any better, you just feel like it. After the release of the popular video game, we saw a lot more players who wanted weapons and magic that did incredible damage, protections that were nearly impervious, and hit points through the roof. The only way to keep the game in balance, of course, was to kick up everything else around them. As I told one referee, multiply all the numbers by one hundred, and let them play that way. If they've got a thousand hit points instead of ten, they'll feel powerful, even though goblins typically have one to seven hundred each.

In Multiverser, any skill you've got starts at a value equivalent to 11; nothing has a value greater than 40. However, there are three hundred nineteen steps to get from eleven to forty. We call them skill marks, and the better you get, the more of them you need to get better.

These things make it work:

• At the lowest levels, each earned skill mark buys the next improvement; so you're getting incrementally better faster at the bottom end.• As you advance, it requires more skill marks to get the next improvement. We have two break points in the game, one after which it costs four skill marks to improve and another after which it costs twenty-seven.• You can get more than one mark at a time if you do something more challenging, but you lose any extra marks beyond what you need for the next improvement (you can't go up more than one notch at a time).• Although each improvement goes to chance of success equally, at the break points skill improves in other ways representing speed and quality of success.


You don't really need big numbers to do it; you just have to provide something that tells the players they're getting better. This isn't terribly new, really--most of that is reflected in OAD&D, including the use of experience points to mark improvement without realizing it in play, gradual increase in difficulty improving, opportunities to earn experience faster at high levels, and break points at which skills improve in other ways. It's just applied a bit differently in a skill-based system.

I hope this helps.

--M. J. Young

Though everyone's posts are appreciated, unfortunately the thread has gotten off track. The main issue that I has hoping to get some advice on is inventing a skill system that does not suffer from an immense power discrepancy between new players and old.

Once again, I'd *love* to discuss this stuff with everyone, I would just ask that it be done via PM to ensure that this thread remains focused on the original issue.

Seidaku wrote: The main issue that I has hoping to get some advice on is inventing a skill system that does not suffer from an immense power discrepancy between new players and old.

Make skill rating the denominator in determining chance of failure. Let's say an unskilled person (rating 0) attempting a task of minimal difficulty (rating 1) has a 75% chance of failure. This could suggest a formula like:

Chance of Failure = ( 2 + difficulty ) / ( 4 + skill rating ).

Thus, 3/4 = 75%.

Chance of failure on a minimal difficulty task would progress like this:

[code]
Skill Failure vs Difficulty of 1
0 75% (3/4)
1 60% (3/5)
2 50% (3/6)
3 42% (3/7)
4 37% (3/8)
5 33% (3/9)
10 21% (3/14)
15 15% (3/19)
20 12% (3/24)
50 5% (3/54)
100 3% (3/104)
[/code]

And so on.

Not perfect, but it's pretty easy to explain, and does allow you to infinitely scale skill and difficulty rating (which may sometimes be the other guy's skill, armor rating, whatever). The chance of failure may be too large for your taste, but I'm hearing a likelihood of ratings that scale up pretty rapidly, and as such you'll want a system that can scale wildly on the difficulty axis as well...

Good idea, but there's a problem Fred. Already we can see that the odds of 1 v 1 is not 50%. That's OK if it's an IGO UGO system, but the effect does decrease with higher levels. So in a fight between two guys with level 100, you have more than a 98% chance of failure. Talk about your whiff factor.

No, I think that intuitively, the odds shouls remain about 50% for failure for equal skill, and work off that.

If we want a diminishing return we should use some sort of curve that starts at zero difference from 50 for zero difference, but then reaches an asymptote of 100% at maximal difference. So, how about this.

C = % chance of higher score to succeed
L = lower of two ratings
H = higher of two ratings

Chance = 50+((H-L)/H*50)

For code for use with a RAND statement, it would simply be

C=.5+((H-L)/H)

I can "parabolize" this curve, too, if that's desirable. As it stands, basically you have to have twice the skill of your opponent to have a 75% chance of winning. I could push that out to any point you like.

Mike

Mike Holmes wrote: Good idea, but there's a problem Fred. Already we can see that the odds of 1 v 1 is not 50%. That's OK if it's an IGO UGO system, but the effect does decrease with higher levels. So in a fight between two guys with level 100, you have more than a 98% chance of failure. Talk about your whiff factor.

Hi Fred,

Go with Mike's suggestion if what you want is for the probability of success to depend on the ratio of the two skill levels, so a 2 vs. a 1 would have the same chance (75%) as a 200 vs. a 100, while a 101 vs. a 100 would be very close to even.

On the other hand, you might prefer the probability of success to depend on the difference between the two skill levels -- while still staying below 1.0 no matter how high the difference becomes. So, a 2 vs. a 1 would have the same chance as a 101 vs. a 100, while a 200 vs. a 100 would have the same (extremely high, but not quite guaranteed) chance as a 101 vs. a 1. In that case, you want (using Mike's notation):

C = 100% - (50% / (x ^ ((H-L)/y)))

This gives a 50% chance when H=L, and the chance of failure for H decreases by a factor of x for every y points by which H exceeds L. For instance,

C = 100% - (50% / (2.0 ^ ((H-L)/4.0)))

means that H has a 75% chance of success when H is 4 greater than L, an 87.5% chance of success (that is, one eighth chance to fail) when H is 8 greater than L, a 93.75% chance of success (that is, one sixteenth chance to fail) when H is 12 greater than L, and so forth.

See my old Symmetry thread for a discussion of the implications of this. You can ignore the parts about the actual dicing mechanism, and concentrate on the discussion of the underlying math and the ramifications in play.

- Walt

PS Mike's formula for code use should be C=.5+((H-L)/H*.5). Mine would be C = 1.0 - .5 / (x ^ ((H-L)/y)).