Mathmatical Balence Model?

[BackTrak]

QUOTE (Vegeta @ Jul 8 2007, 10:33 PM) Actually this iz a way wrong way to think of alleg BALANCE.... U should think of it the way I do when making a new faction..... U should balance out all strengthz with other weaknessez... AKA Giga'z MAD money and LUX shipz.. Balanced BY weak Hull etc....Dreg MADD speed..balanced BY large hitboxxez....etc..etc...BIOS superior tech..balanced BY huge research time... ITZ not math that will balance out and keep to the fun of the game.....ITZ logical and Imaginative thinking that will work in the end....To bad no haz even attempted to enlist my help on this core.. Az I think I really have a great grasp on how to balance and Maintain the underlying fun of alleg at the same time

Hi Vegeta,

I couldn't agree with you more (except about the part where you're saying that this is the wrong way to balance alleg /wink.gif" style="vertical-align:middle" emoid=";)" border="0" alt="wink.gif" /> ).

All I'm looking for here is a way to determine if a particular core is balanced. We would still need experianced faction designers to maintain the flavor of the factions by moving the appropriate tech parameters to get the core to balance. IE: you wouldn't give bios combat pods to increase thier rating vs giga. So, I'm envisioning a utility to help you out when you balance the core so that you could say: wow, something is making Phoenix way over powered compared to every one else, or these three factions are lower that those two, etc.

I'm certainly not claiming any skill in this area at all, but I'm stuck with an idea that much more complex systems (weather patterns, CAD designed airplanes and cars) all use math to determine balances. Allegiance should also be able to be modeled in a repeatable, reusuable fashion, but as you say, with the goal being to preserve the fun!

[Snack]

MadPeps, there are several stop conditions for GA. You could stop the GA after certain time has passed or iterations instead of waiting for certain "goodness" condition. True, GA's are friendly to parallel processing, but they can last a very long time anyway. The main problem with GA seems to be representing your problem with chromosomes. And this might be the situation here. Also if you have a huge number of parameters, over 64-bits of it, you have to fiddle around in code to represent a single chromosome etc...Anyway, the point is, it can get very messy...just as my GA code. /mrgreen.gif" style="vertical-align:middle" emoid=":D" border="0" alt="mrgreen.gif" />
Not only that, but your results can and will depend on parameters like, population size, mutation type and probability, number of iterations you will perform, etc. So you have this calculation which needs like a week and then you need to try and change the parameters a couple of times...you get the picture. The result can be quite spectacular though, even though the results may vary since it's a stohastic process.

[FlingPu]

BackTrak, I say build it and find out for yourself. Even if you can't make it where you want to go, you may gain a deeper understanding of core balance in the process.

[holtonster]

Everything being discussed here sounds rather interesting. I think Im going to head down to the local used bookstore and pick up some books on mathematical modeling/soft computer/fuzzy logic.

As far as long term goals, I think it would be interesting to try to create a core using the mathematical model(s) if they are ever completed. Itd be interesting to see how it turns out.

[Tkela]

Trying to come up with a formula that could be used to compare two different cores is, I think, impossible. There are too many numbers (with many discontinuous effects: increase miner top speed a little and production increases slightly, increase it a lot and miners are much more survivable) that any such formula would be more complex than the game itself. And while, in theory, you use something like genetic algorithms to generate the formula that would imply playing billions of games to evolve the right formula.

That said, when I was working on the original cores, I made a lot of use of formulas in excel to balance individual ships or weapons against one another. It still took a lot of guesswork, fudging and playtesting to get things close to right (and the number of changes you all made to the cores since then is a good indication that I didn't get it right /blush.gif" style="vertical-align:middle" emoid="" border="0" alt="blush.gif" />).

Answering the question of is this ship "unbalanced" against its peers (by comparing times to kill, frontal areas, fudge factors to account for missiles, agility, sensors, signature, speed) is possible. Telling whether or not it was actually balanced is harder but reducing 30+ numbers to a half dozen composite numbers did make it easier to get things in the right ballpark before throwing it out to the testers.

Having a tool that let you simulate a "real" combat between a small number of ships (even with a relatively simple AI piloting each ship) would be very useful in designing cores. You'd still need to take the results with a grain of salt (possibly a very large one, depending on how accurately the AI simulate real human dogfighting) but it would be better than what I was using.

[spideycw]

Hi Tkela!

[Snack]

Quail, you might want to get these:
H.J.Zimmermann, Fuzzy Set Theory and Its Applications, Second Edition, Kluwer Academic Publishers, 1991.
Z.Michalewicz, Genetic Algorithms + Data Structures = Evolution Programs, Springer Verlag, Berlin, 1992.
S.Haykin, Neural Networks, Comprehensive Foundation, Second Edotion, Prentice Hall, New Jarsey, 1999.

There are bound to be some subjects on MIT web as well about it...have fun, but beware, a lot of math in it. /wink.gif" style="vertical-align:middle" emoid=";)" border="0" alt="wink.gif" />

[holtonster]

QUOTE (Snack @ Jul 11 2007, 11:59 PM) Quail, you might want to get these:
H.J.Zimmermann, Fuzzy Set Theory and Its Applications, Second Edition, Kluwer Academic Publishers, 1991.
Z.Michalewicz, Genetic Algorithms + Data Structures = Evolution Programs, Springer Verlag, Berlin, 1992.
S.Haykin, Neural Networks, Comprehensive Foundation, Second Edotion, Prentice Hall, New Jarsey, 1999.

There are bound to be some subjects on MIT web as well about it...have fun, but beware, a lot of math in it. /wink.gif" style="vertical-align:middle" emoid=";)" border="0" alt="wink.gif" />

Thanks! Even if this thing never gets off the ground, should be an interesting read.

[MrChaos]

nueral networks
fuzzy logic
least squared regression

Snack /wub.gif" style="vertical-align:middle" emoid=":iluv:" border="0" alt="wub.gif" />

[Elephanthead]

I posted about measuring the average amount spent on each tech per victory, sorted by faction to get a rough idea about balance. It really is all about the money.