Posted - 2005.12.08 18:33:00 - [121]

 

Edited by: HUGO DRAX on 08/12/2005 18:33:47
I will simplify the formula.

Mods < 3 = good
=


Mods > 3 = bad

Posted - 2005.12.08 20:49:00 - [123]

 

Originally by: Naughty Boy

---

Considering how effective they already are... It's maybe not so good.
Anyway, i was just posting the figures i got, but it really needs to be tested. I really wonder if that's what happening on SISI currently.

---

As I said, I've not tested it out, but the numbers you ran there shows exactly what I saw as a possible problem. :/

Posted - 2005.12.08 22:46:00 - [125]

 

Edited by: Jim Hsu on 08/12/2005 22:53:25
Edited by: Jim Hsu on 08/12/2005 22:50:26
Edited by: Jim Hsu on 08/12/2005 22:49:11

Originally by: Aenigma

---

May i point out that this formula is likely based on the Gauss/ Normal distribution given the shape of the graph?

Generally one has P(x)=b/(w*sqrt(2*pi))*exp(-1/2*((x-m)/w)^2), in which b and w and m (mean value) are constants and x is the variable.

In this case it's most likely P(x)=exp(a*(x-1)^2), in which a is approximately 0,1403.

In this case constant a needs better approximation for better results. Keep machine errors in mind however since maths on a computer just ain't that precise.

---

Ok. {b -> 4.7324517818660508016, m -> 1., w -> 1.8879751057425276869}

b/(w*sqrt(2*pi)) = 1 (!!!). One piece of the puzzle solved.

Now, exp(-1/2*((x-m)/w)^2) ... um...

Of course, W^(-2)*(1/2) gives that 0.1402740... thing.

And W is another weird forum constant. Try searching "1.887975" in google.

Ok. So... w = Sqrt(2)*x/(2*sqrt(Pi)). So, that's where Sqrt(pi) comes from.

Posted - 2005.12.08 23:28:00 - [127]

 

Edited by: Naughty Boy on 08/12/2005 23:29:17

Originally by: Jim Hsu

---

Indeed, it is that essentially. 2 mods offers a singificant improvement over current tranqiulity, 3 mods offers a little bit of improvement. Once you get to 4 mods, you should start to look for other things to fit in those slots.

---

Actually, while i was computing the last table for my previous post in this thread, i noticed that some mods would give a better boost than now. Not by an impressive, or even significant margin, but still...

Sensor boosters t1 for instance (it's worse for sensor boosters t2):
* ((1.5^4)^((1/4)^0.25)=3.148192599 or ~314.8% or 214.8% boost (current penalty)
* (weird constants stuff)= 315.7% or 215.7% boost (rmr penalty)

H4x?

Sincerly Yours, The Naughty Boy.

Posted - 2005.12.09 00:16:00 - [129]

 

Maybe was that only for damage mods?

Else, it's pretty weird, i agree. It shouldn't be that hard to get the breakeven point, it's somewhere between 1.46 (46%) and 1.47 (47%).

Sincerly Yours, The Naughty Boy.

Posted - 2005.12.10 03:33:00 - [131]

 

Originally by: TornSoul

---

Originally by: Lorth

---

I actually enjoyed reading this thread. It was fun to watch you work out the problem.
Now I'm wondering what that damn 2.222 constant it. I'de be willing to bet that its something dumb like the 42th root of 1337. Just seems that it has to be, but thats just me.

Thanks for a good read.

---

Who says the number has to be mathematically derived.
It could be something much more fundamental - Well, to TomB anyhow.

Look at the top number : Yeast

We know TomB likes his beer
Yeast is used in beer production

Coincidence? - Hardly

(For the record - I have zero clue what that page is about )

---

errr ... omg.

Posted - 2005.12.14 17:04:00 - [133]

 

Edited by: James Lyrus on 14/12/2005 17:07:05
That falloff looks somewhat like a cosh curve.
Maybe 1/cosh(x)
Not quite a perfect fit, but it does have the right properties. Namely a 'curve' downwards that gets small > 4.

Trying a few values in excel, of things like cosh(N/2 - 1/2) gives a fairly good approximation of the curve.

Leads me to suspect we've got some e^n - e^-n factors.

Actually, using those values on the first post,

cosh((N-1)/1.7) is a pretty close fit, although not quite there. Trial and error doesn't get the curves overlapping at any point either unfortunately.

Not sure where that 1.7 (it's an ish, too, it's not a perfect fit) comes from mind.

EDIT: Actually, if you've a penchant for pi, then sqrt(pi) also works as the divisor, Still don't have _quite_ enough falloff at 4/5/6 though.

Hrm.

Posted - 2005.12.17 21:20:00 - [135]

 

Originally by: Matthew

---

Originally by: Jim Hsu

---

I can get it to arbitrary precision (if my inputs (existing data) are arbitrary precise). But .. what in the world is that number? Some "magical constant"?

---

It's entirely possible that there is some "magical constant". Thinking about how game design would go about making the curve, it's entirely possible that they'd start with with the type of curve they wanted, then use one factor in it to drag the curve around until it hit the key points they wanted it to. They would be far more concerned with getting the curve in the right place than having a pretty number in formula.

---

Or there is possibility that there is multiple magical constants for every stacked module added to make penality curve to look like they wanted. Or just maybe five or so, first four to these markable penalities and one for rest of modules to render them useless. (After a study there might be just bunch of magical constants that are not reflected anyway to any realworld curve. Just numbers those brilliant coders did to satisfy calculations... )

Maybe trying same might help - get curve that is close to current situation and adjust with constants and do testing for some other modules. Might work out - or not.

Posted - 2006.02.14 01:03:00 - [137]

 

Edited by: Naughty Boy on 14/02/2006 01:09:13

Originally by: Matrices Reborn

---

How do the skills factor in to various things affected by the stacking penalty?

---

Skills and stacking penalty are not directly related to each other. Only the stacking of modules is subject (sometimes) to the stacking penalty. There's no real stacking of skills, so no stacking penalty either. However, as modifiers are sometimes affected by skills, there's a slight effect of some skills in the stacking penalty of some modules (the skill modifier is applied to the modifier of the module just before the stacking penalty is applied).

Originally by: Matrices Reborn

---

Would you just add them on or assume the stacking affect them as well?

---

It depends on what you want to compute. For damage, tracking related skills, for instance, there's no impact of the skill effect into the stacking. For skills affecting directly the modifiers, there's an impact (ex: armor/shield compensation skills, tracking disruptors/sensor dampeners spec skills). Here's a particular case, with the shield/armor compensation skills.

Originally by: Matrices Reborn

---

BTW it does look like sensor damps. and tracking disruptors are going to be rendered useless by eccm.

---

...by the stacking penalty. Not useless, but there's an issue - changed already on SISI.

Originally by: Matrices Reborn

---

Is this stacking formula in effect already or will it happen in Kali.

---

It's on TQ, but changed on SISI.

Sincerly Yours, The Naughty Boy.

Posted - 2006.12.14 03:23:00 - [139]

 

my head is going to explode , the time put forth in the calculations here could have been made a compilation of every set up of mods on every ship to see different resistances..... my ears are bleeding, i think im going to go die somewhere...

Posted - 2007.09.24 12:39:00 - [141]

 

umm, I read all 5 pages - good stuff - is there any newer thread on this topic or should i post here if i have a comment?

Posted - 2007.11.20 01:14:00 - [143]

 

Where is the newer thread listed at?
IM new and dont have access to many mods and cant test resaonably, but seems to me a decent starting formula would be:
For shield resistances:
Sr=[1-[(1-{R1*sp1})*(1-{R2*sp2})*(1-{R3*sp3})...(1-{Rn*spn))]]

where:
Sr=shield resistance
R1=resistance of module 1 sp1=stacking penalty of module 1(1.00)
R2=resistance of module 2 sp2=stacking penalty of module 2(0.86911)
Rn=resistance of (n)th module

Basically I think Vishnej was correct, but didnt include stacking penalities.
In either respect it seems hardeners harden the remaining shield, after the first resistances have applied (but in no particular order(commutative)).