zwzsg, you still don't understand the argument presented. Try again.
Zpock, I'm glad that you two can start a productive discussion, but you still need to tell me what the purpose of accounting is, or figure out where I'm going with it. =P
Also, synergies are accounted for in the simple relative balance equations I used, which I've posted an example of in another thread, and a spreadsheet for on UF. The formula I used is a starter formula that only covers the basics and leaves a lot of wiggle room, primarily because the more variables I add in, the more I have to edit, and it's very time consuming for one person picking up a large product in the middle.
The really weird thing is that it's so simple to use that a child could open it up and have the whole thing cracked inside of 5 minutes, and none of the brain trusts who've tried so far've had any luck. =.=
Sometimes I forget that you're all kids, and most of you've never actually done anything with your lives yet.
For about 5 minutes. Then one of you cheerily reminds me. =)
TA Derivative Balancing Theory, Part Deux
Moderator: Moderators
-
[Krogoth86]
- Posts: 1176
- Joined: 23 Aug 2007, 19:46
The point I'm trying to make is that you just can't do a list of things like health, speed, size, dps, aoe, range and costs and get a wonderful balance you have nothing to do about anymore - it just gives you a very good point to start from...Zpock wrote:Krog I think you can still use the equations system, assigning a points system to each attribute that seems important like, area of effect.
That's why I made that Flash/Sumo/Gollie example. When you have an amount of flashes that kills a Sumo you just can't do a definite status list for your Gollie-like unit just from those stats and some maths. That's because lots of units and few to just one unit are two different cases which both are valid. Your shiny equations now can't bring you far if you want to measure your new units strength from the previous balancings. The only way to get rid of the antithesis of nearly beating a sumo and totally owning a flash horde would be to shrink down the AoE until there's no more of it and finally your new units are equal in power to both the flash horde and the sumo. But that's not what you intended to as you wanted it to have AoE - so you have to interfere here and "decide" which case to choose or even making em lose in both situations but nowhere with a big difference...
Well you started about setting up relationships between the stats to calculate a balance. In this case all 3 units have the same role: Skirmishers and with that said they are there for the same purpose. That's why you can do a direct comparison...Tired wrote:Krog, what is the point of light cavalry? You've played Medieval:TW or something. What purpose does light cavalry serve that heavy cavalry doesn't?
Flashes have little health, range and dps in comparison but are cheap and so come in masses while a Sumo has lots of health, range and dps but as he costs a lot he's just one unit. If we now don't mention that in TA those units are on a different tech level you here can really achieve what you want: A relationship based on equations concerning their stats...
Since you do such a poor job at refuting other people arguments, I'll have to do it myself.Tired wrote:zwzsg, you still don't understand the argument presented. Try again.
My reasoning present a fatal flaw. I assumed that because Rock beat Scissor, then ValueOf("Rock")>ValueOf("Scissor"). This is not true. Beating another unit doesn't make it better, because maybe it's the only thing that unit is good with, while the beaten unit has many other uses. And a multipurpose unit is often better than its specialised anti-unit.
So, if we take into account the "worth" of units overall, accounting for every situation, we actually have ValueOf("Rock")=ValueOf("Scissor")=ValueOf("Paper") and so RPS is balanced and I haven't proven anything.
My point that the "worth" of a unit is a point in a multidimensionnal space, and cannot be summed up into a one-dimensional value, still stands, I just have yet to prove it.
Copy'paste the formula here!The really weird thing is that it's so simple to use that a child could open it up and have the whole thing cracked inside of 5 minutes, and none of the brain trusts who've tried so far've had any luck. =.=
- If it's so simple a child can find it in 5min, you sure can to?
- Many a modder have been looking for such a formula for ages, it's like the holy grail of modding, give it to us and thou shall be whorshipped by generations after generations of modders!
Imo everyone who has seriously attempted to balance more complex mods for spring knows, you cannot design simple formulas and run them to predict outcome of battles between certain units.
Strictly mathematical balancing is only usefull for comparing very similar units and allowing you to make better judgements.
Such simple thing as turret turning speed on flash made it much more usefull raider and there are thousands of such subtleties in spring as Arrgh and others pointed out.
You cannot possibly account for all of them unless you simulate whole spring, and only way to do it atm is to play the actual game :) CA updater was created to speed up this process because newly added units like blastwings or egg need some real game testing.
Btw, both demonstrate how useless is pure math - blastwing is a flying kamikadze bomb unit which does most damage due to its shrapnells and not actual explosion, egg is unit which can be turned off to disable weapons but get extra HP and run past enemy defenses.
I think that Tired is not so stupid and is well aware of the fact that mathematical balancing leads nowhere (as seen by his numerous changes between various revisions of "perfectly balanced" TiA).
I think he is just trolling around, he just likes trolling CA dev team, or generally anyone who doesnt agree with him. I must admit that it's mindly entertaining though :)
Strictly mathematical balancing is only usefull for comparing very similar units and allowing you to make better judgements.
Such simple thing as turret turning speed on flash made it much more usefull raider and there are thousands of such subtleties in spring as Arrgh and others pointed out.
You cannot possibly account for all of them unless you simulate whole spring, and only way to do it atm is to play the actual game :) CA updater was created to speed up this process because newly added units like blastwings or egg need some real game testing.
Btw, both demonstrate how useless is pure math - blastwing is a flying kamikadze bomb unit which does most damage due to its shrapnells and not actual explosion, egg is unit which can be turned off to disable weapons but get extra HP and run past enemy defenses.
I think that Tired is not so stupid and is well aware of the fact that mathematical balancing leads nowhere (as seen by his numerous changes between various revisions of "perfectly balanced" TiA).
I think he is just trolling around, he just likes trolling CA dev team, or generally anyone who doesnt agree with him. I must admit that it's mindly entertaining though :)
But the "tired balancing model" dosn't care about wich units beat wich? It deducts the value of units from their attributes, very much inline with TA design philosophy...zwzsg wrote:Tired wrote:My reasoning present a fatal flaw. I assumed that because Rock beat Scissor, then ValueOf("Rock")>ValueOf("Scissor"). This is not true. Beating another unit doesn't make it better, because maybe it's the only thing that unit is good with, while the beaten unit has many other uses. And a multipurpose unit is often better than its specialised anti-unit.
So, if we take into account the "worth" of units overall, accounting for every situation, we actually have ValueOf("Rock")=ValueOf("Scissor")=ValueOf("Paper") and so RPS is balanced and I haven't proven anything.
My point that the "worth" of a unit is a point in a multidimensionnal space, and cannot be summed up into a one-dimensional value, still stands, I just have yet to prove it.
A balancing method that instead works with how effective units are vs eachother would be an entire different thing, that I think would also certainly be interesting and not at all impossible to get use out of.
What is the point of assigning costs other than to translate unit effectivness into economy hit :)
And definition of unit "effectivness" is equally difficult problem.
Yes as Tired said, you can say that each HP is worth 1 cost, each range 1e etc to have unit costs mathematically derived by its other characteristics but it wont be in any respect better than human balancing which takes into account more variables and actual game experiences :)
Btw 1-2 wekks ago I run a program which calculated costs of all units using equations from other characteristics of CA units. Some units were hand classified differently to be less or more spammy or to prioritize their construction (mexes) but otherwise equations were used.
It produced really funny and weird mod very different from what I expected :) I had to revert it within 2 hours and since then Im banned from doing similar crazy attempts..
And definition of unit "effectivness" is equally difficult problem.
Yes as Tired said, you can say that each HP is worth 1 cost, each range 1e etc to have unit costs mathematically derived by its other characteristics but it wont be in any respect better than human balancing which takes into account more variables and actual game experiences :)
Btw 1-2 wekks ago I run a program which calculated costs of all units using equations from other characteristics of CA units. Some units were hand classified differently to be less or more spammy or to prioritize their construction (mexes) but otherwise equations were used.
It produced really funny and weird mod very different from what I expected :) I had to revert it within 2 hours and since then Im banned from doing similar crazy attempts..
Yeah, I supposed that if a unit beat another, it's automatically better, so whatever formula Tired use, it must outputs it's better too, because it is.
But right after posting I realised I was wrong, unit A beating unit B (at equal cost) does not mean unit A is better that unit B, as B may have others, and more useful, uses than fighting A.
Gah I should stop using mathematical terminology before you beat me at it again!
But right after posting I realised I was wrong, unit A beating unit B (at equal cost) does not mean unit A is better that unit B, as B may have others, and more useful, uses than fighting A.
Good reply. However lots of information is lost in the norm. I'll argue that by normalising a unit, you get a value that sure can be compared with other units norms, but has lost the info meaningful for balancing. While we'll all agree that the unit itself is an element of multi-dimensional space, the question is, is the "unit's worth" space, and not the "unit" space itself, an isomorphism of |R?Zpock wrote:zwzsg: ever heard of norms?
Gah I should stop using mathematical terminology before you beat me at it again!
A good example would be Stars! In this 4x you as the player design your own race. This is done using a complex points system that put different prices on the many attributes you can choose, including interdependencies. This is very similar to the "balance units by equation" idea that tired is exploring if you think about it. So I wouldn't be so quick to dismiss it as useless... although I certainly wouldn't consider it the perfect balancing model or anything.
The point of assigning mathematical values to an abstract concept such as efficiency is to allow somebody to model the problems of game design using a fixed set of rules, in this case, the rules of mathematics and deterministic logic.
This should, ideally, reduce all ambiguity and subjectivity out of the discussion, given that we all adhere to the same system. However, there are two weaknesses to this argumentative decision.
First, the rules can be interpreted differently. Euclidean geometry, is not the only variant, etcetera.
Second, the act of reduction to mathematics entails some loss of resolution in the elements, which means that you can object to the mathematical interpretation of the abstract or qualitative on the basis of imprecision in simplification/translation.
However, the model can be useful without being perfect.
This should, ideally, reduce all ambiguity and subjectivity out of the discussion, given that we all adhere to the same system. However, there are two weaknesses to this argumentative decision.
First, the rules can be interpreted differently. Euclidean geometry, is not the only variant, etcetera.
Second, the act of reduction to mathematics entails some loss of resolution in the elements, which means that you can object to the mathematical interpretation of the abstract or qualitative on the basis of imprecision in simplification/translation.
However, the model can be useful without being perfect.
It's even hard to find the correct numbers to put into formulas. For example, the damage values you type in the weapon TDFs can be far from the effective damage the unit deals in-game, even assuming that a weapon always hits. Here are some measurements I took in July:
http://caspring.org/discussion/2/24
In my tests, the effective damage went from half to twice the expected value.
http://caspring.org/discussion/2/24
In my tests, the effective damage went from half to twice the expected value.
Has Stars! been pounded by hard core players? If so, have they found "the ultimate template" that is better than any other point repartition? But I guess it's a single player, so there's no harsh competition.
How complex is Stars! ? Sure, when the mechanics are simple, the variable few, and the interactions straighforward, you can write out formulas, end up with a set of equation, solve it, and have perfect balance. However, I'll argue that TA & Spring is far more complex than the average x4. Each projectile has its own trajectory simulated, units don't jump from square to square but have smooth acceleration, varied hitsphere, terrain slow down movement and block shots, and well, all this with smooth variations, and tiny characteristic that you'd think are eye-candy but actually have a small impact on the game, such as the length of the barrel. 4x usually have much cruder engine, not taking that many things into account, or if they do, with white/black variation and not the smoothness Spring has.
In short, it's not because you've solved perfect balance for one game that you can for every game.
How complex is Stars! ? Sure, when the mechanics are simple, the variable few, and the interactions straighforward, you can write out formulas, end up with a set of equation, solve it, and have perfect balance. However, I'll argue that TA & Spring is far more complex than the average x4. Each projectile has its own trajectory simulated, units don't jump from square to square but have smooth acceleration, varied hitsphere, terrain slow down movement and block shots, and well, all this with smooth variations, and tiny characteristic that you'd think are eye-candy but actually have a small impact on the game, such as the length of the barrel. 4x usually have much cruder engine, not taking that many things into account, or if they do, with white/black variation and not the smoothness Spring has.
In short, it's not because you've solved perfect balance for one game that you can for every game.
Stars! are infinitely less complex than Spring and much easier to calculate using few simple equations ..
Zwsg I dont think that term isomorphism is usefull in this context :) Since there is isomorphism between RxR and R but you cannot even get reliable unified single dimensional "cost value" from current Energy, Metal, Time variables.
To calculate unified cost from EMT you would have to take into account stuf like E producing structures EMT costs, nano producing structures EMT costs, distances of metal spots, time for worker to open nano spray, changing effectivness of E->M conversion through the game with mexes, moho mexes, metal makers etc.
Obviously conversion rate would differ from map to map and from vehicle to kbot etc..
Its just not possible :)
Zwsg I dont think that term isomorphism is usefull in this context :) Since there is isomorphism between RxR and R but you cannot even get reliable unified single dimensional "cost value" from current Energy, Metal, Time variables.
To calculate unified cost from EMT you would have to take into account stuf like E producing structures EMT costs, nano producing structures EMT costs, distances of metal spots, time for worker to open nano spray, changing effectivness of E->M conversion through the game with mexes, moho mexes, metal makers etc.
Obviously conversion rate would differ from map to map and from vehicle to kbot etc..
Its just not possible :)
This is pretty much all that remains of Stars! sadly:
http://starsautohost.org/stars.htm
I would have to disagree that stars! is simplistic when it comes to race design. There's maybe 50-100 different choices in the race design, many of them gradual.
It's mostly multiplayer, limited SP against bots, just like spring, and yes huge amounts of hardcore players have hammered away at it during the years. I don't want to go into more detail... you'll have to try and read up if interested on the site above, altough there might be much lost information since the game is pretty much dead.
http://starsautohost.org/stars.htm
I would have to disagree that stars! is simplistic when it comes to race design. There's maybe 50-100 different choices in the race design, many of them gradual.
It's mostly multiplayer, limited SP against bots, just like spring, and yes huge amounts of hardcore players have hammered away at it during the years. I don't want to go into more detail... you'll have to try and read up if interested on the site above, altough there might be much lost information since the game is pretty much dead.
What we are currently doing in CA is
a) watching - watching the real game to judge subjective balance and pick candidates for further testing
b) checking basic stats (like ranges, DPS, AOE, HP/cost ratios)
c) testing units in more or less real situations with or without micro management. Usually outcome is determined before test by unit purpose set by designers, and test serves to check if actual game behavior fits designed and desired one.
For example we expect flash to beat janus in most situations because by design flash is a raider and janus is a skirmisher (in CA), on the other hand we expect group of stumpy + janus to kill flashes. So in tests we usually pick equal cost of januses vs equal costs of flashes and test it.
Of course results vary depending on micro of both units, depending on LOS, on extra room both units have, on the number of units etc.
So even such tests are very crude and very limited compared to actual game situations.
However I dont know better way of doing it. Especially if we want to keep unique units in game and increase difference between ARM and CORE playstyles.
a) watching - watching the real game to judge subjective balance and pick candidates for further testing
b) checking basic stats (like ranges, DPS, AOE, HP/cost ratios)
c) testing units in more or less real situations with or without micro management. Usually outcome is determined before test by unit purpose set by designers, and test serves to check if actual game behavior fits designed and desired one.
For example we expect flash to beat janus in most situations because by design flash is a raider and janus is a skirmisher (in CA), on the other hand we expect group of stumpy + janus to kill flashes. So in tests we usually pick equal cost of januses vs equal costs of flashes and test it.
Of course results vary depending on micro of both units, depending on LOS, on extra room both units have, on the number of units etc.
So even such tests are very crude and very limited compared to actual game situations.
However I dont know better way of doing it. Especially if we want to keep unique units in game and increase difference between ARM and CORE playstyles.
Yeah, that's how I would work, mainly looking directly at what players use and what not. It's the method of gathering and processing this data that is the hard part...
Seen this?
http://www.battle.net/war3/ladder/repor ... sage.shtml
I think blizzard uses some pretty heavy duty statistical methods in their work. Maybe some gadget could be made to send information about what players use to the developers? Probably not practical or worth the effort for spring ATM, no playerbase, easy to screw with it for individuals and so on.
Seen this?
http://www.battle.net/war3/ladder/repor ... sage.shtml
I think blizzard uses some pretty heavy duty statistical methods in their work. Maybe some gadget could be made to send information about what players use to the developers? Probably not practical or worth the effort for spring ATM, no playerbase, easy to screw with it for individuals and so on.
-
Forboding Angel
- Evolution RTS Developer
- Posts: 14673
- Joined: 17 Nov 2005, 02:43
all this talk is cool for AI vs AI gaming debating, but different people have different levels of perception, economical macro, micro, apm, playstyle, ETC. What values can you give to a player who has high micro level? What if 20% of that amount of micro is "useless" micro like spamming move orders to get a bigger apm count?
also how can you take points such as turret height and turn rate and give them default values? how much bigger of a value does a instigator have compared to a flash when it can shoot over flash and gator wrecks where a flash cant, and how much value does it have when it cant shoot over stumpy/janus wrecks? What values will stumpies then have AS A STANDARD because they are able to shoot over some wrecks. And what values will wrecks have as they can be used to block incoming flash damage?
What value will a solar collector have by default, how much of a bigger value is it compared to a llt, since without a solar the llt is useless, but the solar is not useless without the LLT as it can be used as a tank in front of defences, or to be used as a nanoshield?
You cant give set values on things like these.
all this talk is only useful for ai creation, regarding player vs player balance discussion, 95% of things said here are bullshit
also how can you take points such as turret height and turn rate and give them default values? how much bigger of a value does a instigator have compared to a flash when it can shoot over flash and gator wrecks where a flash cant, and how much value does it have when it cant shoot over stumpy/janus wrecks? What values will stumpies then have AS A STANDARD because they are able to shoot over some wrecks. And what values will wrecks have as they can be used to block incoming flash damage?
What value will a solar collector have by default, how much of a bigger value is it compared to a llt, since without a solar the llt is useless, but the solar is not useless without the LLT as it can be used as a tank in front of defences, or to be used as a nanoshield?
You cant give set values on things like these.
all this talk is only useful for ai creation, regarding player vs player balance discussion, 95% of things said here are bullshit
You can give set values for these things. Duh.
40% slower turning rate is worth 20% overall unit cost. There - I just gave a set value. Bitch about whether it's a good value or not, but it's a value. This is balance.
Sleksa, my mod is balanced. Yours is not. This was defined by simple values. How can you all use the word "balance" as though it's some magical cureall that defines all elements of gameplay when it's not? The word simply doesn't mean what you think it does. Other people who speak English should've corrected the rest of you by now, but no one has, which is a fault of English words often having multiple meanings. In this case, it has but one.
I've explained this so many times now that I reach the conclusion that you aren't skipping what I've typed every single time. You simply don't understand it. I don't understand how that's possible, since you each have opposable thumbs, but you just can't be smart enough to track what I'm saying. Not a single one of you understands the purpose and means of balance, which I've already stated here - what it can do and what it can't. You think you do - you're wrong; you don't. You obviously don't.
The purpose of accounting is to create a zero sum balance between Assets + Liabilities and Equity. You select tax payment options as part of accounting, but that's not the purpose of it - just an ends to a means. You do not use accounting to "keep track of what your company is doing" in the kind of vague sense by which you define it.
All that accounting does is make sure that money doesn't disappear - that no one's withdrawing cash and not spending it on anything; fraud. All accounting does is make sure that you don't think that you have a fortune built up in your company that you can withdraw and spend when you have a huge number of unpaid debts. Accounting is not creative - it's just dull as dishwater book keeping. Accounting is nothing but balance.
40% slower turning rate is worth 20% overall unit cost. There - I just gave a set value. Bitch about whether it's a good value or not, but it's a value. This is balance.
Sleksa, my mod is balanced. Yours is not. This was defined by simple values. How can you all use the word "balance" as though it's some magical cureall that defines all elements of gameplay when it's not? The word simply doesn't mean what you think it does. Other people who speak English should've corrected the rest of you by now, but no one has, which is a fault of English words often having multiple meanings. In this case, it has but one.
I've explained this so many times now that I reach the conclusion that you aren't skipping what I've typed every single time. You simply don't understand it. I don't understand how that's possible, since you each have opposable thumbs, but you just can't be smart enough to track what I'm saying. Not a single one of you understands the purpose and means of balance, which I've already stated here - what it can do and what it can't. You think you do - you're wrong; you don't. You obviously don't.
This is the wrong answer, Zpock.OK, Purpose of accounting as far as I care:
A: to keep track of what your company is doing
B: to pay the proper taxes
The purpose of accounting is to create a zero sum balance between Assets + Liabilities and Equity. You select tax payment options as part of accounting, but that's not the purpose of it - just an ends to a means. You do not use accounting to "keep track of what your company is doing" in the kind of vague sense by which you define it.
All that accounting does is make sure that money doesn't disappear - that no one's withdrawing cash and not spending it on anything; fraud. All accounting does is make sure that you don't think that you have a fortune built up in your company that you can withdraw and spend when you have a huge number of unpaid debts. Accounting is not creative - it's just dull as dishwater book keeping. Accounting is nothing but balance.