I am really glad that I pointed out that I wasn't trying to establish fact. It's a really great thing that I explicitly said this was a question of matter of fact vs. matter of opinion. Even drrck agreed with me on this. You seem to be up a creek without a paddle, as they say. :shrug:


Oh, gee, now Great and Mighty Syn has decided that I am worse than moronic. Why do you even talk to me?

I'm not going to waste any more time on you… :sad:

Firstly, yes, it is true. Secondly, I explicitly stated "given only the race and class" (i.e., when only taking into consideration those two factors). Finally, two paragraphs after the text you quoted, I went on to point out what happens when you add on other factors: "Obviously in a real system you would add on other factors (equipment, skills, etc) - but the point is that everyone can add on those factors. While you might argue that the halfling has a greater chance win if he has magic armour and a high weapon skill, there's no reason why the orc wouldn't also have magic armour and a high weapon skill, thus once again leveling the odds."

I understand why you feel the way you do, because I had much the same opinion when I started out. But I've now been developing and balancing competitive PK muds for 13 years, and I've got a pretty good idea of what works and what doesn't. What you're describing can work for a non-competitive game, or one where group play is expected or required, but for a competitive mud where solo play is supposed to be viable your approach is simply going to give the impression of poor game design, and will detract from the overall playing experience. You don't have to believe me, but eventually you'll find out for yourself one way or another.

This discussion about a halfling vs an orc is pointless. The winner is always the halfling.

Good point Darwin, good point… :lol:

You care about learning vs. conditioning because something can't be skilled unless it's able to learn. Therefor, things that don't have the capability to learn (read: computers) can't be skilled. Subsequently, whether or not an RPS-balanced system "works" for PvE is irrelevant, because mobs have no skill. This procession of the conversation was all presented in very clear ways over the past few posts; I really don't see why you're getting worked up over it.



I never said anything of the like. I think you're misremembering things.



If you want to get technical instead of just apologizing, I didn't use any terms incorrectly; you merely made an incorrect assumption based on wording that wasn't as clear as it could have been.









I'm detecting a theme with all the misremembering.

Here you go again, talking about skill vs. learning vs. conditioning. You still haven't said why we care about this difference. You also keep dismissing the fact that an AI agent can over time get better at a task. It's getting old to keep coming back to this.


This is what you said:

"Neural nets only work properly when there are goals to work towards, or "bad things" to avoid. Supplying these conditions is a low-level form of conditioning. Without them, a computer doesn't know what it should or should not do and can't analyze past outcomes because it doesn't know what is favorable and what is not. It would just be the most complicated random number generator ever."

That you said "only work properly when […]" suggests that for some reason we do not have those elements in this case.

But ok, fine, I accept that you are now saying that a neural net can indeed get better over time. If that is what you are really saying, it makes me wonder why you keep implying that the AI agent cannot get better at a task…


Are we, uh, talking about the same thing? I was referring to matter of fact vs. matter of opinion, which is something you agreed with. Go look at the text I was replying to. I don't know why all that stuff you cite is supposed to be relevant.




All of the above said, this discussion is getting kind of boring (to me, at least) and has gotten far off track. I'd like to nominate a repeat posting of Darwin's picture, and with that just let this die.

I quite clearly stated why you care. A few posts ago, I believe it was quixadhal who posited that because an RPS-style balance "works" for PvE, it must also work for PvP. I disagreed, and what followed (all these points about learning vs. conditioning, AI, etc.) were my explanations. If you, personally, don't care, then simply stop responding.



If it's getting old, maybe you should go back and re-read my first 27 responses to the same question, so I don't have to keep giving it over and over. They won't "get better" than their current designer, so it's irrelevant that they have the ability to adapt.



Did I say "in this case, they can't have goals"? No. I merely made the general statement that neural nets do not work properly without goals or anti-goals in order to set up the next statements in my post. In fact, if you read the last of such statements, my use of the word "would" quite obviously indicates the hypothetical nature.



Again, I never said that an AI agent cannot get better at a task. Misremembering FTL…



This is what you replied to:



The phrase "whether or not they ever will is a matter of opinion" is not some metaphysical complex statement claiming that there exists some concept between true and false in some conspiracy theory to overthrow common logic. It means that whether you believe they will or whether you believe they won't is your opinion. I'm fairly sure you knew this and were just being an ass.

I figure that it is only polite to let you know that I won't reply to this beyond saying the present. I see no point in rehashing the same stuff yet another time: I do not believe that it serves any purpose other than wasting everybody's time and diminishing the worth of this thread. I'm neither declaring victory nor defeat; I'm just stating that I think this has run far beyond its useful course. :smile:

Agreed.

I will just address one small point regarding "conditioning." Drrck has correctly stated that NPC opponents must be pre-conditioned to perform well against players. Since NPC opponents are algorithms, which may or may not use any combination of heuristics, neural nets, or expert systems, this is an obvious fact.

What may not be so obvious is that the players also require pre-conditioning. When a player beings playing a MUD, they have to learn what the victory conditions are, what the failure conditions are, and what actions tend to guide them towards which state. For some of us, this pre-conditioning started before we ever saw a MUD. Playing tabletop games introduces you to the concepts of "hit points", and the idea that reducing someone to zero will bring victory for you and defeat for them. Earlier still, the idea that being hit is bad comes from being spanked (if you're old enough that your parents wouldn't be arrested and hauled off to prison for doing that), or perhaps having your peers assault you.

The only point I'm trying to make here is that humans don't come with a built-in ability to do well in these kinds of games, just because they're human. They will generally learn much(!) faster than any AI can adapt. They have a broader scope of knowledge and memory to draw upon, and can make connections between seemingly unrelated events to yield an intuitive grasp of a problem. But they don't come to the game unprepared.

In my example, I program the AI to know that a shield bash is bad, and I teach it how to counter such a move. If you're a new player who has never played anything like this kind of game before, you also have to learn that a shield bash is a bad thing (when used against you), and you also have to learn what move you can make to counter it.

As a human, you have the ability to rewrite your programming as you live out your life. You learn from each encounter and become better at judging what actions lead you towards goals you desire. For an NPC to do the same, it will require a programmer continue to update its code to allow for new things players try against it. This is the fundamental difference.

I don't believe conditioning, in any form, has any bearing on the ability to learn. Furthermore, if you have a dedicated programmer who has time, energy, and sufficiently full logging to analyze every encounter in your game, there's no practical reason the AI can't be taught and adapted as players use new tactics against it.

Will it ever surpass the ability of a human? No. Will it seem to anticipate human actions and perform well enough to beat many of them? Yes. That's what's important. It doesn't have to think, it just has to be an entertaining challenge.

And, now that we've long since scared off poor Avaeryn (the original starter of this dead-horse flogging thread), I'll also let it rest in peace.

That, unfortunately, is a distinction which doesn't matter very much.

Let's say you play a halfling, and another player plays an orc. Let's say your player-skill level is similar to, or perhaps very slightly higher than, his. But the fact that he's an orc and you're a halfling gives him enough of an edge to kill you in 95% of the encounters between you, assuming equal levels of player skill, equipment, etc. Finally let's say that this other player starts a vendetta against you, and kills you at every conceivable opportunity. Are you comfortable with the fact that he will virtually always beat you despite not being more skillful than you are, and that you will never, ever be able to consistently beat him purely because you're a halfling and he's an orc?

Saying "oh, but the imbalance is not 100% overwhelming in every case" simply isn't good enough. If the imbalance is smaller then, okay, it's a smaller problem; but it doesn't become a negligible problem unless the imbalance itself becomes negligibly small.

This was, I think, the original intention of the "real life racial prejudice" analogy. We would agree that a manager being prejudiced twards one race and against another is wrong; if there are other factors which can sometimes outweigh that prejudice it doesn't stop being wrong. You are, in essence, suggesting that our problem of balance stops being a problem simply it isn't a problem 100% of the time. It doesn't. It remains a problem, even if it's only a problem 1% of the time.

The other thing to say is that there's an important divide in this thread between those who are focusing on PC vs NPC, and those who are more interested in PvP. In the players-vs-mobs case I think just about everyone would agree that there isn't a problem with a rock-paper-scissor-like model; so long as each race or class has a similar proportion of "easy" and "difficult" opponents, and there is balance in terms of the rewards to be gained by defeating them (i.e. one race's "easy" opponents give rewards on a similar level to another race's "easy" opponents) the game is balanced. But it becomes much more of a problem in a PvP game, because there you can't always choose your opponents. By and large, mobs don't actively hunt you down everywhere you go in the game. Other players may well do that.

You're right in that humans can be (and are) conditioned, but the conditioning you're speaking of is not what I was referring to.

If you break it down to the simplest of concepts, humans have a survival instinct from birth. They know that living = good and pain/dying = bad. They also have the innate ability to comprehend abstract ideas, and when you combine the two, humans already know what the theoretical goals and anti-goals for a game in which you can live and die are. Computers, on the other hand, do not have built-in anything. They are a blank slate and can't comprehend abstract ideas (or "comprehend" anything, for that matter). This is why they can adapt, but do not have the capability to learn, and thus will never be "as good as" a human in contexts such as these without hard-coded advantages in stats, et al.

The difference between learning and conditioning is a very slight one in terms of understanding, but a gigantic one in practice. It's really the fundamental core of next-generation AI, and why neural nets were developed to begin with.



It's not the actual "shield bash" that we're referring to; I agree that both would have to be conditioned to know that shield bash does damage to your character. However, it's basic concepts like damage = bad and dying = worse that are what I was talking about earlier. Humans know these innately and don't have to be conditioned in this respect, whereas computers do not, because they have no concept of "damage" or "dying" outside of what you program them to "know".

While discussing and debating the original question may help someone to come to an informed decision on how to proceed in implementing a resemblance of balance between races in their mud, it doesn't help at all when the discussions and debates fall into the abyss of nit-picking each others' words.

This is not an exact science. If it was, there would already be clear guidelines on how to achieve a set of balanced races.

It doesn't help when post after post contains "you said this" followed by "and this is how I read it." That is then followed by a post stating "you thought I said this… but what I actually said was this" which then goes into another rather lengthy explanation detailing the same points made in a previous post that may or may not have any relevance to the original topic. It doesn't further the discussion when post after post there is someone saying someone else's post was stupid, moronic, lame, idiotic, ect.. Those are personal opinions from one poster about another poster and do not add anything relevant to the topic at hand.

While I find it somewhat amusing that this pattern happens again and again in various threads and happens more frequently the longer the thread gets, after a while it becomes boring and I fail to find anything relevant to read regarding the original topic. That is really a shame. If I wanted to watch people bickering with each other I could watch Jerry Springer.

So, once again, here is a monkey flinging poo. (Which is what this thread has turned into.)

at least mine doesnt stink.
;)

-Syn

Really?




So what exactly are you saying? That when you say something is a fact, what you really mean is that it is opinion, and you have no idea what a statement of fact really is?

You explicitly said, it was a fact. You obviously have no idea what it means. Kudos. The great Syn did call you worse then a moron, and still agrees with the assessment.

-Syn

Matter of Fact and Fact are not the same, Syn.

A matter of fact does not claim something specifically to be true, it claims that something COULD be true, with enough evidence present. Fact states that there is evidence present and it has bee found valid, thus making it fact.

So.. Sorry, great Syn, but as David said, You're up the creek without a paddle here.

And this has what relevance to balancing races?

Could this thread at least try to get back on topic?

Thanks Kayle, I was starting to think that the binary goblins intercepted "matter of" somewhere between my keyboard and the MudBytes forums. :tongue:


Chances are slim, to be honest. :rolleyes: I think there would have to be another question asked; although it may not appear as such, the 9 pages of verbiage were actually (by a few degrees of indirection, yes) related to the topic. Also, I think that many of the participants have said everything they wanted to say without going into yet another round of repetition.

It occurs to me that the format of a straight-line sequence of posts is not the best suited to this kind of discussion. By its nature, this kind of discussion tends to spawn all kinds of tangents whose relevance seems slim 10 pages later, even though if you trace it back there is relevance, but in the meantime those tangents dilute the content of the thread. A threaded format, where you reply to points in particular, would probably make it easier to follow the core points and leave the tangents to the side, or perhaps more importantly block out the tangents entirely. (You see, another tangent… what would have been the best place to post this? A new thread, where it's coming entirely out of the blue? Here, where its relevance is clear, yet it is not relevant to the topic?)