Forged Alliance Forever

Re: Calculating a balance "score" for units

2020-02-02T22:16:17+02:00

Post this somehow anyway maybe in format bit.ly / somereadablename ?
I thought current balance is something like this:
hp = 2(for t2) or (3 for t3) * masscost, range & firerate adds expontetial mass cost (but usually range & high damage is arty = low speed, no hp, low ROF), speed&turning kinda balanced by units with it having hp halved (average speed in games is ~2.5) & dps halved as well.
Shields have a very good shield-hp/mass ratio.
But range on shields makes it exponentially more costly (doubling range & hp on t3 roughly means x8 price from t2). Stealth/jamming have some weird costs, not really streamlined.

Statistics: Posted by techmind_ — 02 Feb 2020, 22:16

Re: Calculating a balance "score" for units

2019-12-28T22:11:12+02:00

Hi,
I'm new here but this topic pinned my interest because I'm involved in couple of similar project some already finished (one that I can freely share is one form Battle for Wesnoth - another open source project, I managed to have very good reasult there). I like doing things like that and I treat them as challenges, I have never done table for rts before so I expect this to be fun.

bugblatterbeast wrote:
Obviously there are limitations compared to actually playing the units (and especially playing them well).

There shouldn't be many of them. Just some things might be harder or easier.

Ithilis_Quo wrote:
but then there are speed row, range and stuff like that. So its not posible use only raw numbers

Hmm I don't really like that you just multiply two values because units without attack can exist and be great damage sponges(example: ram form AoE2 vs archers, it has attack but its very low and almost unused except vs buildings).
Speed is very doable, range should be too, micro might be the most problematic but should be also doable.

moses_the_red wrote:
This should probably be done via replay analysis.

There could be situations where useful units are built only in emergency situations, like firebeetles to push away a rushing com, but at the very least it should be easy to get an objective baseline for further work this way.

I dont really like replay analysis many people can think different things after watching the same replay, also player level there should be high but comparable and that is often not the case. Telling for experience it usually leads to arguing.

There is special category of units that is called "counter" usually counter unit don't need to be strong they just need to do one specific job well.

I already made first attempt on making this but because I'm new user I cant share link here. If anyone is interested please let me know so I can show you what I already did.

Statistics: Posted by Hejnewar — 28 Dec 2019, 22:11

Re: Calculating a balance "score" for units

2019-12-01T01:46:43+02:00

This should probably be done via replay analysis.

You should be able to determine whether building a unit implies a won game, and that should probably get you what you want. Any unit that consistently results in a lost game is probably a bad unit.

There could be situations where useful units are built only in emergency situations, like firebeetles to push away a rushing com, but at the very least it should be easy to get an objective baseline for further work this way.

Statistics: Posted by moses_the_red — 01 Dec 2019, 01:46

Re: Calculating a balance "score" for units

2019-11-25T02:43:37+02:00

bugblatterbeast wrote:
Just wondering if anyone has tried in the past to calculate a balance "score" for units, based on their characteristics (DPS, range, HP etc)? Given that I'm not the first nerd with Excel it seems likely. Just wondering what has been done before & what peoples thoughts are. Obviously there are limitations compared to actually playing the units (and especially playing them well).

I spend some time under excel, and far more in empirical tests to got some algorithm that works.

the base value is 200m = 1500hp 60dps 23r 3speed +-2s reaload time (yes its cca pillar)
for every +80m = +500hp + 30dps
for new tech add +5r or 500hp 30dps

if add +5r then -500 hp or 30dps
if reduce -5r range +500hp or 30dps

if add +1,5speed reduce 250hp or 15dps
if change rof to 4sec reduce 250hp or 15dps

By this, you will get all time nearly well-balanced units that you need only poolisch on end.

Then matematicly you can got some number to compare

hp/mass cost x dps/mass cost = raw balance indicator (but there ist nothing as cost hp and dps, so its very raw number, and for example on some actual units ->)
1500/200 = 7,5 x 0,28 = 2,1 (actual pillar)
1900/300 = 6,3 x 0,29 = 1,85 (rhino)
2750/360 = 7,53 x 0,33 = 2,51 (obsidian) -> but real price of obsidian is + 37,5m while energy drain put into mass of pgen so => 6,9x0,3=2

but then there are speed row, range and stuff like that. So its not posible use only raw numbers

Statistics: Posted by Ithilis_Quo — 25 Nov 2019, 02:43

Re: Calculating a balance "score" for units

2019-11-21T12:07:49+02:00

There is already a balance value inside every blueprint called "Threat".

GPG introduced this for the AI to decide if an unit is a threat to different layers.

As example from a unit blueprint UAS0302 (Aeon Battleship):

Code:: Defense = { AirThreatLevel = 16, EconomyThreatLevel = 0, SubThreatLevel = 0, SurfaceThreatLevel = 225, },

That means, this battleship is a low threat to air unit and a medium threat for units on the surface layer (Land and Water)

This is far away from balanced but maybe a good start.
Also if those numbers where updated it would be awesome for AI play.

Statistics: Posted by Uveso — 21 Nov 2019, 12:07

Calculating a balance "score" for units

2019-11-21T11:31:44+02:00

Just wondering if anyone has tried in the past to calculate a balance "score" for units, based on their characteristics (DPS, range, HP etc)? Given that I'm not the first nerd with Excel it seems likely. Just wondering what has been done before & what peoples thoughts are. Obviously there are limitations compared to actually playing the units (and especially playing them well).

Statistics: Posted by bugblatterbeast — 21 Nov 2019, 11:31