The balance formula thread

[Regret]

Post here your balance formula ideas.








I present you my balance formula:

45.81 = (hitpoints*acceleration*maxvelocity*turnrate*weapondamage*range*areaofeffect) / (reloadtime*buildcostenergy*buildcostmetal*buildtime)

Explanation:

All variables in dividend increase units 'strength' as they increase, and decrease it when they decrease.
All variables in divisor decrease units 'strength' as they increase, and increase it when they decrease.
Quotient (45.81) was calculated by using variables from an arguably balanced unit, the core instigator from BA.

[1v0ry_k1ng]

right.. so does it give you the same value.. or even near the same value.. if you put another BA unit in?

[TheFatController]

unit rating = (((how good players perceive the unit to be) * ((how good the unit actually is) / 10)) * (number of times I won a game with the unit)) / (metal cost)

[smoth]

doesn't account for a setup anim or limited arc.

[MidKnight]

Sorry in advance for the tongue-in-cheek reply but...

My formula:
Don't play BA

I have nothing against BA, but the entire idea of balancing units so that they are equal in strength defeats the point of an RTS. Why do I need 5 units that fill the same role equally well if I'm just as well off using only one?

[Pxtl]

MidKnight, I'm trying really hard to say this nicely, because my first instinct was vitriol and Spring forum has enough of that: you're completely missing the point. BA's units are the exact opposite of what you just said. Every unit has a very specific role that it was made for, and it is moderately useful to utterly useless at other things. Some units have a more limited role than others.

There are no redundant units. There are no useless units. There are units where there usefulness is too limited, and there are units where their usefulness is too expansive, and there are units that reward too much micromanagement, and there are units whose proper use depend on some very obscure knowledge. This is why I don't like BA very much.

But what you said was incorrect.

[imbaczek]

using a scoring formula will only get you so far. you need to judge every unit with every unit pair-wise, create a matrix of what's better than what and look at the RPS cycles that emerge. this will get you a much better assessment of unit balance, but it's a lot of work. also, scoring should probably be done in several ways (micro/no micro, etc) and quite possibly you need a human to do it, since the stupid machines aren't good at noticing small, important details like turret turn rates and such.

[luckywaldo7]

The use of a unit is not always determined by good stats, but often by the right combination and values. There is a huge difference between just outranging an LLT, and having the same range as an LLT. Try to think more in terms of unit relationships than pure numbers.

This beautifully concise and rather self-explanatory piece of balancing advise just about sums it up. (not sure which dev wrote it though)

[1v0ry_k1ng]

you guize just wait, my formula is going to be the knees bees

[MidKnight]

@Pxtl: I concede that, to some degree, I was wrong. I probably should have left BA out of it; when I last played it continuously, I was a noob (and this was 6 months ago XD)Actually, I've played BA actively since December 2007 (at least 10,000 minutes ingame) but I never learned anything until I started playing CA, and resultantly speccing more. .

However, please allow me to revise my argument just a little tiny bit:

People seem to think that "balancing" units Is about making them comparable in strength. In my opinion, balancing is about making factions comparable in strength, making the units produced by similar factories comparable in usefulness, and possible making units from one factory comparable in usefulness and strength to units of a different factory. As such, using a "balance formula" makes no sense, as all it does is make sure that whatever units it is applied on end up performing similarly. The only conceivable use of such a formula, as far as I can see, is making a standardized system of "ratings" for units, so that the average strength of average units may be calculated.


On another note (and this is a form of TLDR) , Special abilities, like out-ranging other units, special abilities like impulse and paralysis, even (especially) projectile speed and arc, play a huge role in differentiating units in any RTS, and unless a formula includes, prioritizes, and assigns values to all of these, it, in my newbish opinion, is not suitable for large-scale use.

Darnit, I made my TLDR longer than my actual post again! >_<

[Google_Frog]

I didn't see this thread, my post in SL should probably be here instead.

I've got the CA formula though, here it is:

If it's faster and has more range than AK it's OP.

[1v0ry_k1ng]

what are you talking about? the AK is first class op

[KingRaptor]

CarRepairer gets it right.

[1v0ry_k1ng]

unit cost = (pwn - fail) * baw

its science !

[KDR_11k]

45.81 = (hitpoints*acceleration*maxvelocity*turnrate*weapondamage*range*areaofeffect) / (reloadtime*buildcostenergy*buildcostmetal*buildtime)

So, defenses can't move so at least three of the factors are zero... LLT is UP, give it rapid-fire D-shotgun!

[Regret]

Instead of zero use 1 for stats that don't apply to certain units.

D:

[SpliFF]

Yeah, I tried the formula approach once before. It's total fail. The real attack value for unit A vs unit B formula includes:

for EACH weapon on A:
per shot damage
shells in burst
burst length
reload time
accuracy / tracking
aim speed
projectile speed (including instahit)
area of effect (may have multiple rings)
target armor type
cost
build time

chance to hit includes:
range
turret turn speed
turret max/min pitch
turret max/min arc of fire
unit max speed
unit acceration
unit turn speed
target range
target speed
target turn rate
target size
target layer (ie, air, sub)
direct vs. indirect fire
shot propagation (ie, dgun continues through target and obstacles)

I tried to come up with factors for each property, ie the property's value / average of property value for all weapons in game.

That was easy enough. The trouble started when I tried to figure out how important each factor is RELATIVE to the others. Is shot damage more important than accuracy? If so, by how much? If an AOE weapon hits water (and therefore doesn't explode) how much should that count against it? What about engine bugs that prevent units from aiming properly? What about weapons the AI doesn't use effectively, should they be worth less?

You also have to deal with the way some factors influence each other. For example a high damage could actually have a negative effect when combined with a low accuracy and low range (you'd be destroying your own units 30% of the time).

And this isn't even counting 'specialist' weapons like paralysers, capturing, jamming, terrain modifers, etc...

Also the real unit value you're trying to calculate must include the attack value against all other units accounting for how often those units are used (having a unit than pwns a krogoth would worthless if nobody built krogoths).

If you spent a couple of years on this you might get something approximating reality. It'd probably be about 200 lines of code, slow and difficult to validate. By the time you were done, engine and mod changes will have invalidated many of your values and assumptions.

My solution was to use a modified version of CarRepairers formula "unit cost = (pwn - fail) * baw - n00b". It might seem like a joke but if GOOD players consistently spend money on something then ipso-facto it MUST be a good thing. A clever mod would take units that are purchased "too much" and nerf them (or raise the cost) in small doses until it's used, on average (accounting for game time, map and unit cost) about the same as other units. This way the game would end up being perceived as balanced - and most likely that would be close enough to the truth. Constantly monitoring these usage values would also allow a mod to adapt to engine changes which could make a previously balanced unit unbalanced again.

Most mods do balance like this, but in a haphazard way (relying on humans in the loop). It would be very interesting to see if an automated system (watching 10's of thousands of games) could do better. I was going to do that (and still might) with a replay analyser but I have other priorities for the moment and don't see myself having time for a few years.

[Gota]

so you might do it but not in the next few years?lol.

[SpliFF]

Pretty much, I have a mod to write. This is more relevant to my experiments with AI.

[Gota]

I'm not trying to make fun but it just sounds hilarious.
"i might do it but not right now...maybe IN A COUPLE OF YEARS!!"