Combat

[big_canuk]

Mario:

Well great. At least I know I am going about it correctly. I now have more incentive to check more carefully.

Hey, thats when I learned my 123, back with version 1.0A on floppy. You could do amazing things with self-modifying code, if you could keep track of what you were doing. Problem is, for the last 10 years, I have just used computers periferally, and I still find myself doing /fs in excel. My sheets are simpler now than they were back then, cause I haven't sat down to figure out how to do things. Your "combat odds" sheet is a wonderful example, and therefore teaching aid for me. You have just brought me forward many generations in spreadsheeting... groan.

No really, wow, thanks a mint.

edit: Well I got the 55.93 as the top odds. Must have made a mistake in the bottom 1/2. Thanks again!
[This message has been edited by big_canuk (edited December 28, 2000).]

[Sikander]

Well, I'll just pop into the off-topic spreadsheet conversation to admit that I too really loved spreadsheets. My first experience was in 1988 with Quattro pro at work. I had taken a couple of Basic programming courses in College, but this thing blew me away. It was almost too easy to do things that would have taken hours of coding in Basic. And it was powerful, with it's own scripting language very much like Basic, and the ability to easily chain to other programs. I cranked out quite a few successful programs with that software, and not only for work.

Anyway, back to the combat topic. It seems as though we need to get some data samples from the game itself to see if we can get it to tell us how it works. Unfortunately, 'random' numbers being what they are this may not be all that easy. I know there are a few programs which will slow your processor down, which may allow enough time to actually see everything displayed upon the screen clearly. Does anyone know if it is possible to take screenshots of SMAC/X with free software? If so, it might be easier to compile the data from sequential screenshots at leisure after the fact.

[MariOne]

well, today was a kinda dullday, sitting there waiting for calls from customers who weren't in their offices...
so I kicked a test for combat on my slow work PC.
I had 20 runs each for the 1:1 basic scout patrols combats, then one with a hardened vs disciplined one, then a laser vs patrol, then the famous laser vs. fusion patrol.

I succeeded to note down almost every round of combat. I wrot'em down one by one on a piece of paper, because I'm a moron luddite.
And when I finally got really bored, I went home cuz it was about time... and I forgot the stupid "data support" on the office desk!!! all that collected data unreachable till Tue 2 Jan!!! groan!
I can recall tho that the 2:1 won all the 20 battles, the 2:1*2 won 18 out of 20, and the hardened1:discip1 (game odds 9 to 8) won 11 out of 20.

Quattro Pro! I remember the feeling of coolness when it came out!
beware, Lotus 1.0A users, you're giving out your age!
Brrr... still now when I see someone using Word for tables (!) I shudder and feel an itch to reach for an hammer...

---

BTW, anyone with Prima Guide can tell us whether there's anything more detailed there? Googlie?

At this point I would consider asking Firaxis if they could disclose a bit more of their combat resolution algorythm.
After all, it wouldn't be the key of their success, like Coke's secret formula...

[Helium Pond]

Wow, look at all this. Nice work, guys. But let me get this multiplicative-versus-additive thing straight. If the odds are multiplicative, doesn't that mean that it's better to have two +50% bonuses than having one +100% bonus? So, for instance, an ECM unit, on a forest square, with a sensor nearby, defending against a rover, would have a strength value of its defense multiplied by 25%x25%x50%, or, worked out a little, would get its defense value times 2.34375, whereas the same unit, *without* ECM, defending inside a base, would only get its defense value times 2. Is that correct?

[MariOne]

HP, correct, but you got a couple of values wrong.
(I still play my tests with SMAC, but I "assumed" that no combat modifier has been changed in SMAX, please warn me if I overlooked it)

Let me say that the multiplicative vs. additive nature of the modifiers is a NON-issue.
In the sense that anyone just has to look at the combat screen in a case when he has two modifiers, and it would be self evident, no need even to think to doubt about their multiplicative nature. Probably if you never turn "display odds before combat" on and you never bother to look at the details of the combat display, then you might overlook it.

In our case, say that our defending unit has no armor, for simplicity.

Combat Display Readout
1 - 1+ - 1
No Armor: 1
Disciplined: +0%
Terrain (Forest) +50%
Sensor +25%
Jammer +50%
Strenght: 2.81

Had they been additive, that strength would have been of 2.25.
Instead it's [ 1 ] * 1.5 * 1.5 * 1.25 = 2.8125
I find sometimes easier to keep the fractional notation to do those maths
e.g. +50% is +1/2, that is *(1 + 1/2) or *3/2.
The game rounds the digits but keeps actually track of the fractional values, and you can verify it looking at the (screwed) odds it shows you.

So in our case the modifiers would be expressed like:
3/2 * 3/2 * 5/4 = 45/16 = (2 13/16)
An attacking recon rover has a strength of 2 = 32/16, and indeed the game proposes me 32 to 45 odds for that attack!

About "the same unit without ECM defending inside a base"... well:

The modifier for base defense is 25%, NOT 100%!!!
Did you tinker with your alpha.txt?

So, let's rearrange your example and consider a recon rover attacking a scout patrol (NOT ECM).

In a sensored base, the patrol would defend with
Base +25%
Sensor +25%
Stregth: 1.56

or rather 1 * 5/4 * 5/4 = 25/16 = 1 9/16

On a forest (or rocky or fungus) tile w/out sensor
Terrain +50%
Strength 1.5 (3/2, or 24/16)


BTW, while you have lots of customizable combat parameters in your alpha.txt, those for terrain defensive bonuses are NOT in there! So you can't alter them!

EDIT:
< shame >
Why didn't you explicitly mention that your base has a Perimeter? That it's NOT a default!!!
Sorry for overlooking it, your example was comnistent then.
So, 150% * 150% * 125% vs. 200% * 125%

Yes, in the PERIMETERED base you get a 2 * 5/4 = 5/2 modifier (2.5 - NOT just 2 as you said, confusing me), or 40/16.
With the ECM/forest/sensor the math is above, 45/16.
[This message has been edited by MariOne (edited December 30, 2000).]

[Helium Pond]

MO: no shame for you, shame for me! I was sloppy and inaccurate. You, as always, are exhaustively clear, which I mean in the best possible way.

[big_canuk]

quote: Originally posted by MariOne on 12-29-2000 05:04 PM I can recall tho that the 2:1 won all the 20 battles, the 2:1*2 won 18 out of 20, and the hardened1:discip1 (game odds 9 to 8) won 11 out of 20.

So... the 20/20 is maybe consistant with our postulated combat theory to date, but the 18 of 20 for the 2:1*2 is not consistant with 56% as our theory says. The 18 of 20 is more consistant with what I would have guessed.

A two stage combat resolution scheme, may increase the odds even more, as they diverge from 1:1 as we seem to be seeing, and maybe explain the 3 0, 3 1, etc. results from resolution rounds. I like VoodooChilds idea of "collateral damage". This would explain the high level of loses, even at overwhelming odds, when you attack with only 1 hitpoint remaining. Perhaps these 1 1 rounds are thrown in at random.

Perhaps, battles are fought 1 hit point at a time, and we only see occasional updates. But maybe these mini- battles are resolved, and according to their outcome, the 3 0, (say for 3to1 or better), 3 1, (say for 3 to 2) is assigned to the updates.

In addition, or alternatively, "other" modifiers may be creeping in. I suspect a randomness increasing modifier for native life battles. I suspect a additional benefit to elite units (maybe just on defense, it always seems harder than it should to kill them). Other possibilities would be modifiers based on absolute power (versus just relative), and perhaps extra defense for AAA vs aircraft etc.


[This message has been edited by big_canuk (edited January 03, 2001).]
[This message has been edited by big_canuk (edited January 03, 2001).]