OT: Making a Game System (part 2)

[Omnirizon]

cleveland,

i havn't done any work on the magnitudes of these variables yet, that's the point I'm at now, but was just wanting to see what people thought about a system like this. the amount that these variables are allowed to get very large or very small can be tweaked, that's probably the most work if I decide to use this system.

my goal was to find a system were things are more interactive and less flat, that's what I'm hoping this will be.

coding the math won't be hard. I'm going to make a little application that will simulate a fight between two characters using the system. I can then use that to play with everything and work on getting a good balance.

I will post that app here as soon as I get it done; and others could experiment with it. I'm on fall break right now, so maybe I can get it done over the next week; however I have some papers to write also...

[Tyrant]

Very interesting, I'll look forward to your progress.

One question i have right off the bat...are you attempting "realism" or is this meant to be cinematic/video game style with heroic action heroes?

With regard to your question about grenades etc., if the basic system is clean, rough edges can be smoothed out by having damage types and limiting the percentage of damage that can be absorbed by vitality. For instance lethal explosive damage could only be 50% absorbed by vitality unless the target can dive for cover, or whatever.

[JimMorrison]

Just wanted to post some encouragement! 8 )

I've been studying game systems for a long time (since that wee age when I discovered that AD&D wasn't the only way), and I have to agree with the basic concepts of where you're going with this. For example, my own system (that I haven't worked on lately, for shame), actually has 12 statistics, with the intent that there are very subtle and complex interactions among them.

I think there are a few very good points to developing a system like this:

- The extreme complexity which would kill the game in PnP, becomes a great tool, when you decide to commit the system to only be used with a computer doing all the complex calculus.

- Decreasing the number of available stats causes two problems, first it waters down the personality of your characters, I have seen numerous systems that go the STR/DEX/INT route, and the degree of variance among successful characters drops precipitously. Likewise, there is an inverse relationship between number of stats, and ease of min-maxing at the expense of character building. Using 9 statistics, even if they can be said to overlap in certain ways, adds a lot of nuance to a character, allowing you to give much more depth and personality as they develop - as well as making more progression paths viable.

- Cause and effect can gain a much greater depth as well. Avoiding falling into a pit, for example, should not simply rely on your "Dexterity". Having a statistic that represents sensual "Perception" allows an interaction where you assume there is a point of observation of the pit, as well as a point of physical reaction - and someone who is either perceptive but clumsy, or nimble but dull may fall straight into a pit that a person who was more average in both respects may have avoided. This goes beyond mere realism, directly to immersion.


All that said, I would have to agree, the "Memory" stat is going to be the hardest to work with, without it becoming either a terrible annoyance, or a gaping hole in the system that either requires boosting at every opportunity to keep from being screwed, or to drop as much as possible to gain points. Of course, insuring that it has very deep interactions with other stats insures that it can't be ignored totally (make it the primary stat modifying learning rate?), but it will work better in the context of the game if it has an overt function, rather than being totally passive.


I'll probably discuss this more with you later, I'd like to help you as much as I can, without corrupting you with specific design points from my own system.

[Lokean]

Omnirizon, since you seem to be only vaguely familiar with the Pythagorean means, I'll give examples with the data set you gave earlier.

The Pythagorean means of the data set [1, 1, 1, 2, 10] are as follows:

The arithmetic mean is the sum of the members of the set divided by the population of the set.
μ = (1 + 1 + 1 + 2 + 10)/5 = 3

The Geometric mean is the product of the members of the set raised to the power of the reciprocal of the population of the set.
G = (1 * 1 * 1 * 2 * 10)^(1/5) = 1.82

The Harmonic mean is the reciprocal of the arithmetic mean of the reciprocals of the members of the set.

H = 1/((1/1 + 1/1 + 1/1 + 1/2 + 1/10)/5) = 1.39

It is vitally important to note that these functions only exist for positive, real numbers.

I think it essential that you realise how strongly the harmonic mean skews the results towards the low values and, quite importantly, the limit of the harmonic mean.
Consider a situation where a character's vitality cost for acting is the harmonic mean of 100 values, 99 of which are 10, one of which is 1. The harmonic mean is 9.17.
Consider the same thing, but now 99 of those values are one googolplex. The harmonic mean is 100.

This illustrates an important relationship with regards to the the population of a set, its smallest member and its harmonic mean. The maximum value of the harmonic mean of a data set is that of the smallest member multiplied by the population.

[Omnirizon]

Originally Posted by Lokean

I think it essential that you realise how strongly the harmonic mean skews the results towards the low values and, quite importantly, the limit of the harmonic mean.
Consider a situation where a character's vitality cost for acting is the harmonic mean of 100 values, 99 of which are 10, one of which is 1. The harmonic mean is 9.17.
Consider the same thing, but now 99 of those values are one googolplex. The harmonic mean is 100.

This illustrates an important relationship with regards to the the population of a set, its smallest member and its harmonic mean. The maximum value of the harmonic mean of a data set is that of the smallest member multiplied by the population.

thanks!

I do realize the degree to which the harmonic mean skews towards the smallest members of the set, and that is exactly my reason for using it; it creates a strong reverse salience. Reverse salience is where a single component of a system is impeding the performance of the entire system; the term originates from technology studies, where within a system some components are being developed very quickly, while others are allowed to lag behind, and despite all the advances in one technology, the overall system performance does not improved because of the reverse salience of other components in the system.

I liked the principle of harmonic means, and thought they fit a fighting system best. To my understanding, the idea of the harmean in a combat system like mine would be to say that "the percentage of 'attack' contributed by each variable is due to the degree of that variable's operation in the generation of 'attack'." This also means that variables can be weighted to change their degree of operation in attacking. I simply have to decide how much each variable contributes to it, and weight accordingly. Thus in melee, perception might be weighted less than strength and dexterity, for example.

[Omnirizon]

I would say its writing, not talking, that is important for computer programming. And in reality every computer programmer is going to be an above average writer, mathmetician, and... computer programmer. Take the following examples:


Hello my name is Donald Knuth; not only am I quite likely the most noted figure in modern computer science, I have written about 15 MASSIVE books about it. My father owned a printing press and this had a HUGE impact on my skills with computers later in life. I've also written the de facto standard software for doing academic publishing, LaTeX.

Hello my name is Eric Raymond, I'm a hacker extraordinaire who has wrote a category killing mail client for Linux, as well as hacked on Linux, Nethack, and tons of other computer software. Most notably though, I've written the most famous books on modern hacking, The Cathedral and the Bazaar, and the Art of Unix Programming. Some English student didn't write those, I did. Lastly, according to me and other books such as "Rebel Code", it was the highly social nature of computer scientists over networks such as usenet that allowed Open Source Software to become a reality. In other words, computer scientists are really good at communicating.

Hello, we are Linux. According to Raymond, we represent the top 5% of computer programming talent in the world. One of the most notable correlations we have is the fact that Linux is one of the most well documented pieces of software in existence; could this be because computer hacking and writing skills are related?

Hello, I am the liberal university model. Virtually every university in the world follows this model. Even MIT, known for computer science stuff, has English, philosophy, and literature courses on par with their technical courses. This is because knowledge in these areas improves performance in technical areas. Only community colleges and trade schools do not offer such well rounded education, and consequently their product is ill-suited for the new technological economy.

[JimMorrison]

Originally Posted by Omnirizon

I would say its writing, not talking, that is important for computer programming. And in reality every computer programmer is going to be an above average writer, mathmetician, and... computer programmer. Take the following examples:

Your assumption that you are trying to illustrate, seems to be that communicating with human beings and communicating with machines are inextricably interconnected. You may be correct to a certain degree, at a certain level of proficiency, however there is no clear absolute here. Most of the people whom I have known in my real life dealings, who were talented programmers, all happened to have trouble with people. But here is where your problem starts, you are trying to forge an immutable link between "writing", and "programming". I would argue that in this context, that there must then be an immutable link between "speaking", and "writing". Thus if the only way to have the best rating in programming, is to be an excellent writer, but the only way to excel in writing is to be a talented speaker, then speech becomes a necessity for programming skill.

The problem is that you are drawing a causal relationship from a circumstantial relationship. Perhaps what you are overlooking, is that while the most well known programmers happen to be good writers - it is because it is through that writing skill, that they become well known, more so than through their programming skill. There are droves of incredibly talented programmers out there, that you have no knowledge of (and who do not rise as high in power in a corporate world, usually), who are indeed very poor at dealing with people in any capacity.

[Lokean]

Omni, allow me to retort...

Hello, I'm John Warner Backus, arguably the most important figure in digital computer programming, ever. You have never heard of me, because I did my job and never made a fuss of myself.

Seriously though, from the method you are using to model calculations (largely in terms of reciprocals and products) the 'correct' mean to use would be the geometric. The harmonic mean is more properly used for data which represent ratios and rates of change of processes relative to some fixed, common unit.

'Reverse salience' is misleading. Reverse salient is a noun, referring to a particular weakness which holds back some process (specifically, holds back an armored column advance; the term is military in origin). If you wish to use a harmonic mean you are essentially telling players that their character is *weakened* by having even a single relevant attribute at a value less than any other.

[lch]

Originally Posted by Lokean

Omni, allow me to retort...

Presumably one of the two people involved in the development of the Backus-Naur-Form? I would expect that everybody involved in computer science in academics must have learned about the BNF, and not only if your area of expertise is theoretical computer science or compiler construction. It's the same like everybody should know about Von Neumann. There are obviously a lot of people that did important little things who you might never have heard about, though.

[Omnirizon]

Originally Posted by Lokean

Omni, allow me to retort...

Seriously though, from the method you are using to model calculations (largely in terms of reciprocals and products) the 'correct' mean to use would be the geometric. The harmonic mean is more properly used for data which represent ratios and rates of change of processes relative to some fixed, common unit.

'Reverse salience' is misleading. Reverse salient is a noun, referring to a particular weakness which holds back some process (specifically, holds back an armored column advance; the term is military in origin). If you wish to use a harmonic mean you are essentially telling players that their character is *weakened* by having even a single relevant attribute at a value less than any other.

Ok I see your point. I will consider the geometric mean then.