Argh's Theory of Game Balance (very long). - Page 2

Argh's Theory of Game Balance (very long).

Discuss game development here, from a distinct game project to an accessible third-party mutator, down to the interaction and design of individual units if you like.

Moderator: Moderators

User avatar
zwzsg
Kernel Panic Co-Developer
Posts: 7052
Joined: 16 Nov 2004, 13:08

Post by zwzsg »

They just don't. They like to attach all sorts of worth to stuff that's not even worth mentioning, like DPS, when in fact, there is no such thing, outside of isolation cases, where you are strictly comparing two units, without terrain as a factor.
Talking about DPS is already a great step forward compared to those still talking only about weapon damages without taking into account the reload time.
Sure, DPS alone isn't enough to judge a unit, as there are many case where a high DPS is counterbalanced by the unit not being able to deal its DPS, either because of short range, short life, innacurate aiming, etc... However, in the simplified test case of units standing still next to each other, DPS / hitpoint is about the only thing that matters.

Therefore, game designers should ask themselves, frequently, "what is the purpose of this unit? Why is it important? Why should it be in the game?" If you don't have a good answer, you should take it out, until you do, or perform a crazy experiment, like doubling / halving an important variable.
The approach where the game designer has decided beforehand the role of each unit, and refuse to have them used in any other way, doesn't lead to interesting game. When you're at the designing stage, it's too hard to think about all cases, you're often blinded by the role you want the unit to have, then when you release and it's played, good players might find way of using the unit the designer never tought of. By following your principle, the designer should then change the unit to prevent the uninted use to remain possible in the next version.

I disagree with that. The player base will be more happy, the game will be more interesting, if you allow unthought-of uses.

Especially for a game like TA that offer so many possibilities, the role of the designer should be to make sure his units are very varied and offer ample possibilities, not carve in the stone the role of each unit beforehand.

And by variety, I don't mean have 100 tanks, with just the cost and the gun caliber changing. I mean using every arcane properties Spring offer to create differentiation: slope tolerance, shape of the shot trajectory, area of effect, secondary gun aiming speed, position of the gun muzzle, acceleration, time to deploy, amphibiousness, etc... anything the player can intuitively feel the impact of, whithout being able to quantitize how much it change the unit worth.

As Cavedog's Total Annihilation proved, you can skip balancing, as long as you have a rich enough gameplay, the game will find its balance itself, despite no stat being ever changed, and units being used in role in total opposition with what the designer intented. Like I said in that other "balancing" thread, to strive for a perfect balance seem pointless when you realise how awesome, fun, hard core, and skilled has remained TA ten years after his release, despite Cavedog never changing the unit properties.

Sure, you should still avoid using senseless values that would make only one unit out of 200 worth building, and you should keep an eye open for game ruining exploits. However, when you find a unit used in role you hadn't thought of, or when you find the game mechanic abused in an almost cheaty way, you don't necessarly have to rush to fix it. For instance, TA never had line bombing nor off-screen bombing "fixed". These are clearly exploits, and give an "unfair" large advantage to whoever use them. Yet today you can still win a highly skilled TA game without using bombers. These tricks just added new possibilities, but without killing every other possibilities.

Bottom line is: What matters is NOT ultra-precise balancing. What matter is setting a rich enough environment. Then you can just let the gameplay evolve by itself. As long as the interactions were set complex enough at the designing stage, the game will remain playable.
BaNa
Posts: 1562
Joined: 09 Sep 2007, 21:05

Post by BaNa »

Argh, I read your OP up until the part about zero sum games.

I refer you to the definition of Zero sum games in wikipedia:

http://en.wikipedia.org/wiki/Zero_sum_game

From there:
In game theory, zero-sum describes a situation in which a participant's gain or loss is exactly balanced by the losses or gains of the other participant(s). It is so named because when the total gains of the participants are added up, and the total losses are subtracted, they will sum to zero.

So your understanding of said games is flawed, or you didn't express yourself well in this passage:
Zero-sum means that a possible move results in a loss for all parties.
Zero sum games are by definition games where ANY possible move results in gain for one side and an equal loss for another (this gain/loss can be 0), or a similar version with multiple people. This kinda discouraged me from reading the rest.


Edit: Heh, I read through and I see other people got there faster than me. laziness ftw.
User avatar
Argh
Posts: 10920
Joined: 21 Feb 2005, 03:38

Post by Argh »

Edited out all erroneous use of phrase "zero-sum". Corrected and added to statements, added sections talking about positive and negative-sum gameplay and their impact on game balance overall. Thanks to everybody who's pointed out my abuse of terms, if you find other such mistakes please feel free to point them out.

Also edited opening arguments, in light of multiple people insisting that single-player titles are subject to game balancing. I didn't concede much, but I did want to be a little more sensitive about my choice of words there.

@zwzsg:

While I agree with many of your points, especially including the "unforeseen uses" section, I still disagree, and refuse to budge on the following:
in the simplified test case of units standing still next to each other, DPS / hitpoint is about the only thing that matters.
What I've said, at length, is that that "test" is almost completely irrelevant to the actual game balance. Put a Punisher from classic OTA (as opposed to the relatively-wimpy ones in most alphabet-soup mods) next to 5 Flashes. The Flashes win.

Put the Punisher at max range, with something spotting for it. The Flashes lose. Put a hill into the mix. Now you have an indeterminate problem, answerable only if you know the location, size and slope of the hill.

DPS isn't worth jack shit. It's not a meaningful number. You can build a unit with a DPS of 1000000000 and make it never able to hit anything. What's its real effectiveness? Zero. You can make it hit stuff 1 in 100 times. Now, what is its real effectiveness? A lot higher than 1%, as my practical example in the essay shows ;)

Again, I don't care if players want to use crap, useless numbers like that, to talk about what they think is balanced. But game designers should ignore the numbers, and concentrate on what the players are really saying about the relative powers of things.
User avatar
Zpock
Posts: 1218
Joined: 16 Sep 2004, 23:20

Post by Zpock »

Looking at DPS and HP per cost isn't meaningless, you just have to keep those other factors in mind as well. Especially comparing DPS and HPs of similar units, is a really good idea.

You look at some units, consider what dps/hp per cost is appropriate for the type of unit, and then test things. For example a unit with longer range and all things else equal should probably be relatively weaker... simple

Those other factors may be hard to keep perfect track of, but raw damage output and health are clearly available and it is easy to see how they affect unit balance so there's no need to make things harder by ignoring them.
SpikedHelmet
MC: Legacy & Spring 1944 Developer
Posts: 1948
Joined: 21 Sep 2004, 08:25

Post by SpikedHelmet »

I think that's a rather pompous position to take, that DPS is only considered by know-nothings. DPS constitutes the balance of a unit at prime conditions of engagement, which, although unlikely, can and do happen. Of course one should take into account other variables; weapon accuracy, accuracy while moving/while the target is moving, target size, so on and so forth, which will tweak t he effectiveness of said weapon, but overall, the effectiveness of a combat unit at absolute prime value is an important base with which to start.

Also, nobody likes when people go around using themselves as a reference for being correct. There is a very thick line between saying "I did it this way and I found it very amicable" and "I did it correctly, so you should do what I've done." And while it's understandable that some people don't give a rat's ass what others think, you might as well save yourself the trouble and time of trying to address the community as a whole with some essay subject unless you are going to placate people's personalities and act a little more "friendly". Obviously you care enough to keep trying to talk to people, so you should realize that it would be more effective if you did it in such a way as to make people more responsive (in a positive way) to what you have to say.
User avatar
KDR_11k
Game Developer
Posts: 8293
Joined: 25 Jun 2006, 08:44

Post by KDR_11k »

There are extreme cases where DPS is not a factor at all but in most cases it is at least an important factor. DPS for a longer ranged unit dictates the amount of damage it can do during an x second approach. DPS also dictates effectiveness when attacking targets that can take a lot of damage, especially when other factors are neglectible. When I want to down a Gundam comm I'll use machineguns, not beamrifles or rocket launchers because MGs have the highest DPS and a com can take a whole lot of shooting. You should at least calculate the DPS for your units since you don't want to notice too late that your super-powerful doom cannon takes so long to build and recharge that it's cheaper to just spam crappy lasers.
User avatar
Argh
Posts: 10920
Joined: 21 Feb 2005, 03:38

Post by Argh »

Before I get into further discussion of these points, I'm going to put something else on the table.

Here are my rules that I have been using for PURE. They're the only things that aren't flexible- anything else can be done.

1. Every unit must be "distinctly different" from every other unit. Distinctly different is defined as: more than one major balance stat is very different than other units of a given cost tier, or that they have a very different weapon system than other units of that cost tier.

2. No units that have no niche uses will be included. A "niche use" is a nonstandard case, where a unit can perform tasks that might be novel or unexpected. For example, I have a unit that is most efficient when the battlefield is covered in Features, and when dead converts to a Feature that is an effective obstacle to most other units of the same combat power.

Most niche uses must be discovered and then improved upon, rather than by design. Call it, "improving on happy accidents", if you want, but I always do stuff like this.

3. No units that do not specifically counter at least one other unit will be included.

That's it. If a unit fits all 3 definitions, it might be a good choice- after that, it's all down to what I want, what I think will be fun, what the artists come up with. Simple, really. It doesn't cover "need to have" units- things that are in place to provide strategic-level gameplay- but it works for everything else.




Now, onto DPS.

Basically, you're all saying that, despite giving you a concrete and easily-tested method to demonstrate that DPS isn't a reliable figure, you want me to say, "hey, but it's useful, if you have a bunch of units that are remarkably similar in function".

If it will make you happier- fine, point granted.

Got a lot of tanks that are basically just tanks? Fine, use DPS.

Except, what about armor values, hitpoints, top velocity, acceleration, brakerate, turret rotation speeds, weapon velocity, range? Terrain? Gravity?

DPS is just ROF * Damage (which damage? do we use multiple Armor values?) * FlankingBonus... it doesn't even take into account area-effect, relative velocity of projectiles, guidance, or even relative percentile chances of a hit :|

Even compared to rough calculations done in a typical tabletop wargame, DPS is not likely to lead to statistically-accurate results. While I may not be being as polite as you might like, and I'll go ahead and grant "pompous" if it makes you feel better, I'm being accurate.

It'd be one thing, if Spring was a 2D, flat-world RTS, like StarCraft. Then we're just talking about relative DPS in a very small context- I don't think that, except for a few exceptions like the Yamato Beam, that Starcraft's weapons even took character's angles into consideration before firing calculations.

It's another thing in Spring. DPS isn't very useful as a yardstick- the game engine is too complex, if you're using all of the bells and whistles. It's just that simple.



Instead of just tearing down your arguments, though, which leaves nothing positive in its wake, let's talk about practical alternatives.

The first, and probably best alternative is for somebody to make a map designed just for testing units out with, under different conditions, but in a scientific way.

So, maybe we should just open up a discussion about making a map to run the "random enemies" script on, to do balance testing with, including areas like chokepoints of several sizes, positive and negative slopes, broken ground with heavy cover, forest, and a water area with a shoreline that is passable by almost anything that can go in the water. It should be easy to design something like that, with a player start on the bottom and a single start for the AI up at the top, where the AI is forced to go through lots of terrain types to get to the player... heck, if the resources were balanced well, it might even be fun for MP- one player gets resources, but has serious terrain disadvantages, whilst the other player gets fewer resources, but is in a much better defensive position...

That'd be a pretty good starting-place, and would provide a better experimental setting to evaluate things with, and test assumptions.




For another thing, you can simply assign units a generic "effectiveness" rating. That's what I've done in PURE. It's not a "real number". It's just a concept. But, combined with the rules I started with, it's been a fairly effective strategy, thus far.

I've done it very simply, via cost. Units that cost 4X more than another unit should be 4X more "effective". What does "effective" mean, though?

Well, that's a toughie, in a game that doesn't include a lot of similar units, which is one of the unit design goals I think that most games should have. The way that I do it is that I decide what each unit's job is- what it should be good at- and then I do some playtesting, and see how well it works, doing its specific job. Whatever that is. If it's a scout, I scout with it. If it's an artillery piece, I use it as artillery. Then I keep tweaking values until it feels like it's worth using, in that specific role. I'm not gonna apologize for the "squishy" sounding words, either- "feels like" is what I want to achieve, because, as I've stated before, game balance is perceptual, not necessarily actual. I want to know that if I was a player, using the unit correctly, that it does the job. I assume that players aren't stupid, basically.

Sometimes, that's all that I need to do. Sometimes, it's more difficult than that, because the specific role was either too-narrowly defined, and makes the unit useless within a larger mix of things, or was too-broadly defined, and I've come dangerously close to making something that's omni-purpose, and can be used for anything. When that happens, I redefine its role in some way.



For example, let's go look at the OTA Panther (a unit that has a close family resemblance with a unit in PURE, btw).

The OTA Panther was a unit that exemplifies the pitfalls and dangers inherent in looking at DPS values. Against ground units, the Panther is a fairly fast tank, that can deal two types of damage, at varying time intervals. One of the hidden factors in a unit with more than one weapon system that can fire at any target, btw, is how many fewer shots get wasted, in actual combat- the lightning weapon may have already killed that pesky AK, no need to waste a rocket. So, it's pretty dangerous to just look at DPS because of that factor alone.

However, it gets worse. The rocket is somewhat effective (in OTA, mind ye) against aircraft, putting this unit into a somewhat unique category, along with the Jethro and a few other dual-purpose units in OTA. Therefore, when you consider its real costs, you're basically getting a fairly fast tank that's moderately effective against most units, and can kill aircraft besides. It's not as cost-effective as units that just do one thing, but it can do many things.

Now, was it OP? No, I don't think it was. I think it was UP. Why? Range problems, a big corpse and footprint, and not quite fast enough to be really dangerous as a flanker. Many things were more effective for their costs, even in the dual-use roles. I'll take a huge stack of Jethros (IIRC, it was something like 5-6:1, on costs) for a Panther any day of the week. But, enough about that.



I have a unit like that, in PURE. In fact, I have three- but two of them were relatively easy balancing issues, because they aren't really multi-function- they do one job far better than they do the other, so it's obvious what they're for.

One of them, however, has proven to be one of what I think of as "fulcrum" points in the game design- if I don't balance it appropriately, then I might break the entire game. So, every time I've made major changes that effect game balance overall, like introducing all of the lovely FlankingBonus code into PURE... I have to rebal that unit pretty aggressively, along with the other fulcrum points... then adjust the other points until things are right again. I don't just adjust one thing at a time- every element in a game like PURE depends on every other element. But some are more important than others.

Identifying "fulcrums" in your game design is hard but necessary work. That's something that's individual to a game design, and they aren't always obvious. Don't assume, for example, that since Unit A is supposed to be a bread-and-butter unit, that it's a fulcrum... because in fact, it may be that the real balance depends on how well Unit X, which is one of the counters to Unit A, is working in that context. When I design a unit, I very specifically set out what role I'd like to see it in, mess with it until it seems to fit that role, then observe what happens when I take it out of that role. If a unit is supposed to counter another one, I look at that, then I look at what other things it is countering, and how well, and try to see it within the totality. It's not a perfect system, but it's fairly effective.

There are other methods, of course. One is to limit complexity, so that you can keep strict control over what balance means. I reject that approach, because, as people have pointed out in this thread, it means that you're providing players with a straightjacket, forcing them to wear it, and then asking, "hey, are you comfortable"? Not a good plan, imo. However, it's always an option, and in small doses, it's an effective tool.

Another is to just do whatever you think looks cool, or fits into a storyline or a given universe, and let balance sort itself out purely by what playtesters have to say about it. That's OK, except that you end up with tons of things that are probably going to be "obsoleted" by each other, to steal Storm's phrase (DRB and I used to laugh, both with him, and at him, depending on what we read- I'll never forget his hilarious put-downs, though, they were massively entertaining, even if his understanding of balance was sometimes dubious).

To solve that problem, you can either just ignore it and accept that it's going to happen, which is probably bad game design, but might make your fans happiest, or you can aggressively target the worst cases and remove them from the game in the interest of gameplay, then just don't talk about it in your hype and refuse to answer questions like, "hi, one of the novels about the Space Ninjas described how the Boolean Monsters had like, this huge army that was able to clone itself, and if they're not in the game, you obviously suck"

Please, resist the urge to make Boolean Monsters ;)
User avatar
smoth
Posts: 22309
Joined: 13 Jan 2005, 00:46

Post by smoth »

holy smokes... wall of text. Well since this thread has gone beyond your theory into a collection of thoughts you are accumulating. It may be appropriate to merely call it a design thread.
User avatar
Argh
Posts: 10920
Joined: 21 Feb 2005, 03:38

Post by Argh »

Or a blog. Which would be lame.

I dunno, maybe I should request that it gets killed, and try to write a concise version, expressing it as haiku or something:

Game balance is
Solving the unsolvable
Whilst seeking fun choices

I personally think it's at least somewhat useful to talk to people who have reasonable objections, though, if nothing else it helps show me where my ideas have serious holes, either in perception or in the way that they're being argued. The whole bit about my blatant mis-understanding of zero-sum was valuable in and of itself, frankly, and helped me restructure my thoughts about how gameplay works, and why.

And the DPS thing is useful because it's an example of how we all get into patterns of perception and belief that are resistant to factual argument- people found DPS was useful in the past, and they expect it to remain valuable in the future, even if it's not.
User avatar
Fanger
Expand & Exterminate Developer
Posts: 1509
Joined: 22 Nov 2005, 22:58

Post by Fanger »

Argh, DPS by no means should be used as an absolute when balancing a mod, however to completely dismiss it is a bit silly. Yes your right when you do DPS you dont take into account: accuracy, range, AOE.. etc.. However DPS does serve as a good benchmark to see what the relative approximate damage per second of a unit is vs another unit. This can tell you whether the damage of a unit is perhaps lower or higher than another, and allow you to evaluate whether you want this. Basically you can use DPS like a rating system, its not absolute but it gives you a rough idea of what the damage of that unit is in a vaccume which can be compared to another unit and then decide whether you want that unit to do more damage or less.. it may not take all factors into account, but that doesnt mean you can use it to gauge effectiveness. It just shouldnt be used to nerf or boost something solely based on its DPS in a vaccume without considerations...
User avatar
knorke
Posts: 7971
Joined: 22 Feb 2006, 01:02

Post by knorke »

Of course DPS is not always reliable a very vague number.
But you do not always need precise numbers. Escpecially not in a game where you never have the same situation twice.
Look at this picture:
http://www.latrobe.edu.au/maths/smith/i ... 12poly.gif
The blue line is the actual function, the red line was calculated to be close to the blue line. It does not really fit 100%, but its close enough.

"Then I keep tweaking values until it feels like it's worth using, in that specific role."
Thats probally the best way. There are just too many factors and you can not pre-calculate how the game will play.
Its like the choaspendel:
http://youtube.com/watch?v=CqmKs-sCCtc
Move it just 1/10 mm and it will behave completly different - or maybe not.
User avatar
Zpock
Posts: 1218
Joined: 16 Sep 2004, 23:20

Post by Zpock »

Nah, most attributes you can change dosn't lead to a huge difference in the end result, if you only change it by a small amount, so it's not really chaotic. 5% dps buff, is likely a 5% increase in units overall worth.
User avatar
Evil4Zerggin
Posts: 557
Joined: 16 May 2007, 06:34

Post by Evil4Zerggin »

Zpock wrote:Nah, most attributes you can change dosn't lead to a huge difference in the end result, if you only change it by a small amount, so it's not really chaotic. 5% dps buff, is likely a 5% increase in units overall worth.
Actually closer to 2.5% (unless you're talking about a fire support unit, in which case it is closer to 5%). If you had 5% more resources, you could build 5% more units instead, which is 5% more firepower and 5% more hit points.
User avatar
Zpock
Posts: 1218
Joined: 16 Sep 2004, 23:20

Post by Zpock »

Hmm, this is besides the point, but if you make a unit have twice as much firepower and hps, it's actually 4times as powerful. It's easy to see if you consider that it can apply twice the firepower, for twice as long.

Interestingly, this does mean that an armys power is proportional to the number of units squared. A twice as large army will beat the smaller one with much less then 50% losses as we all know...
SpikedHelmet
MC: Legacy & Spring 1944 Developer
Posts: 1948
Joined: 21 Sep 2004, 08:25

Post by SpikedHelmet »

But then what happens when you factor in more complex variables? For instance, in S:1944, we've got a realistic version of armour penetration implimented; armour-piercing weapons become less powerful the further they travel, and other weapons can't harm certain units; for instance, a Panther won't even be scratched by 20mm fire (or even 75mm fire from a long distance). So it's not so much a matter of health or weapon damage, but damage over range, armour thickness compared to certain weapons, etc.
User avatar
KDR_11k
Game Developer
Posts: 8293
Joined: 25 Jun 2006, 08:44

Post by KDR_11k »

In Company of Heroes people list DPS stats for different ranges. That's also good to tell you which enemy you should engage at which range.
SpikedHelmet
MC: Legacy & Spring 1944 Developer
Posts: 1948
Joined: 21 Sep 2004, 08:25

Post by SpikedHelmet »

It's harder to do that, as the actual damage of a weapon varies not only by range but by target. A 37mm gun will do, say, 1000 damage against a truck at 1000m distance, 1000 damage against a medium tank at 500m distance, and 0 damage against a heavy tank at 100m distance...
User avatar
Zpock
Posts: 1218
Joined: 16 Sep 2004, 23:20

Post by Zpock »

I'd recommend grouping armor and damage types together in that situation. You could have (none?)/light/medium/heavy armor and divide damage types into something like SmallArms/HE/penetrating. Then it's easier to know how much damage units do, then if every unit has it's own independent values.
User avatar
Fanger
Expand & Exterminate Developer
Posts: 1509
Joined: 22 Nov 2005, 22:58

Post by Fanger »

Even if you make the damage setup more complicated than a smiple X damage at all ranges.. DPS is still useful as a gauge. Of course in such a situation you wouldnt compare the DPS of a unit that cant hurt something with something that can.. and of course given your range equation you could then apply that. So for example you wouldnt compare the DPS of a infantry rifleman with a Tiger Tank, but perhaps a Tiger Tank with a Panther.. the point of this is not for absolute balance, no instead your getting a benchmark. This is just a easy to use method to guess whether your damage values are in the right ballpark area..

DPS is a tool, just like a anything else for measuring, in certain situations you wouldnt use a yard stick to measure something, but that doesnt invalidate having a yard stick...
User avatar
Argh
Posts: 10920
Joined: 21 Feb 2005, 03:38

Post by Argh »

Ok, before I read another page of practically identical arguments, lemme try to move this forward:

1. DPS may measure things. It may even be useful. But it's frequently the wrong tool for the job.

2. If you really want to evaluate game balance, you need to use the right tools.

Maybe a discussion of what the right tools look like is in order. I actually have quite a few concepts I use, but I've never bothered writing them down.
Post Reply

Return to “Game Development”