[ QUOTE ]
[ QUOTE ]
[ QUOTE ]
For the defense case, I have to be hit 10 (or more) out of 20 times. The chance of that happening is around 1 in 100, which is an astounding 10,000 times more likely than the resistance case.
So, if you the question you are asking is "what is the average damage taken?" 25% defense is the same as 50% resistance. However, if the question you ask is "how likely am I to die?" then 25% defense is no where near as good as 50% resistance.
[/ QUOTE ]
You're going to drag us all through binomial theory aren't you.
truly thou art evil!
[/ QUOTE ]
Average lifetime analysis is a Markov Chain analysis, not so much binomial theory. Last time I said that, someone actually did it. I'm pseudo-lazy: I didn't want to calculate it, but I was willing to write a simulator to model it.
[/ QUOTE ]
I have been using 1D random-walk theory, but in the world where there are only two possible outcomes (hit or miss), Markov chains, binomials, and random-walks all work out to be the same thing so it ends up not mattering which notation you use.
The trick I am using to vastly simplify the caculations is to approximate the damage distribution as a gaussian. That way all I need to keep track of is an average damage and a standard deviation. The method may be a tad mathematical for these forums, but if you can handle a square root, you can use the results.
On a side note:
Yes the range of possibilities for a defense build is much wider than for a resistance build. While you will sometimes die in a situation where a resistance build never would, if you get lucky you can do things that are impossible for a resistance build. One particular event triggered my investigation down this road:
I went into a mission to fight Luimary. This is the Elite Boss version not the Hero version (after the patch). The first time I faced her, she downed me in seconds. I came back from the hospital to try again and I took her down with barely a scratch. The difference in the two attempts was so striking that is really piqued my interest in the topic.
Of couse, the second reason I have been doing this investigation is to answer the question many people have asked: Are the EA passives worth taking?
[ QUOTE ]
[ QUOTE ]
[ QUOTE ]
For the defense case, I have to be hit 10 (or more) out of 20 times. The chance of that happening is around 1 in 100, which is an astounding 10,000 times more likely than the resistance case.
So, if you the question you are asking is "what is the average damage taken?" 25% defense is the same as 50% resistance. However, if the question you ask is "how likely am I to die?" then 25% defense is no where near as good as 50% resistance.
[/ QUOTE ]
You're going to drag us all through binomial theory aren't you.
truly thou art evil!
[/ QUOTE ]
Average lifetime analysis is a Markov Chain analysis, not so much binomial theory. Last time I said that, someone actually did it. I'm pseudo-lazy: I didn't want to calculate it, but I was willing to write a simulator to model it.
[/ QUOTE ]
Over on the Tanker forum, da5id has been working on a Markov Chain Time to Live model for the game - when not preparing for his Physics dissertation defense of course. 
Liberty
Mister Mass - 50 Inv/SS/NRG Mut Tank [1236]
Doc Willpower - 50 Grav/FF/Psi Mag Controller
Baron Wonder - 50 SS/Elec/Mu Mag Brute
Sound Bight - 50 Son/Son/Mu Tech Corrupter
[ QUOTE ]
[ QUOTE ]
[ QUOTE ]
[ QUOTE ]
For the defense case, I have to be hit 10 (or more) out of 20 times. The chance of that happening is around 1 in 100, which is an astounding 10,000 times more likely than the resistance case.
So, if you the question you are asking is "what is the average damage taken?" 25% defense is the same as 50% resistance. However, if the question you ask is "how likely am I to die?" then 25% defense is no where near as good as 50% resistance.
[/ QUOTE ]
You're going to drag us all through binomial theory aren't you.
truly thou art evil!
[/ QUOTE ]
Average lifetime analysis is a Markov Chain analysis, not so much binomial theory. Last time I said that, someone actually did it. I'm pseudo-lazy: I didn't want to calculate it, but I was willing to write a simulator to model it.
[/ QUOTE ]
Over on the Tanker forum, da5id has been working on a Markov Chain Time to Live model for the game - when not preparing for his Physics dissertation defense of course.
[/ QUOTE ]
He would be the one.
[Guide to Defense] [Scrapper Secondaries Comparison] [Archetype Popularity Analysis]
In one little corner of the universe, there's nothing more irritating than a misfile...
(Please support the best webcomic about a cosmic universal realignment by impaired angelic interference resulting in identity crisis angst. Or I release the pigmy water thieves.)
[ QUOTE ]
[ QUOTE ]
[ QUOTE ]
[ QUOTE ]
For the defense case, I have to be hit 10 (or more) out of 20 times. The chance of that happening is around 1 in 100, which is an astounding 10,000 times more likely than the resistance case.
So, if you the question you are asking is "what is the average damage taken?" 25% defense is the same as 50% resistance. However, if the question you ask is "how likely am I to die?" then 25% defense is no where near as good as 50% resistance.
[/ QUOTE ]
You're going to drag us all through binomial theory aren't you.
truly thou art evil!
[/ QUOTE ]
Average lifetime analysis is a Markov Chain analysis, not so much binomial theory. Last time I said that, someone actually did it. I'm pseudo-lazy: I didn't want to calculate it, but I was willing to write a simulator to model it.
[/ QUOTE ]
I have been using 1D random-walk theory, but in the world where there are only two possible outcomes (hit or miss), Markov chains, binomials, and random-walks all work out to be the same thing so it ends up not mattering which notation you use.
The trick I am using to vastly simplify the caculations is to approximate the damage distribution as a gaussian. That way all I need to keep track of is an average damage and a standard deviation. The method may be a tad mathematical for these forums, but if you can handle a square root, you can use the results.
On a side note:
Yes the range of possibilities for a defense build is much wider than for a resistance build. While you will sometimes die in a situation where a resistance build never would, if you get lucky you can do things that are impossible for a resistance build. One particular event triggered my investigation down this road:
I went into a mission to fight Luimary. This is the Elite Boss version not the Hero version (after the patch). The first time I faced her, she downed me in seconds. I came back from the hospital to try again and I took her down with barely a scratch. The difference in the two attempts was so striking that is really piqued my interest in the topic.
Of couse, the second reason I have been doing this investigation is to answer the question many people have asked: Are the EA passives worth taking?
[/ QUOTE ]
And the answer is?
Inquiring minds want to know.
[ QUOTE ]
I have been using 1D random-walk theory, but in the world where there are only two possible outcomes (hit or miss), Markov chains, binomials, and random-walks all work out to be the same thing so it ends up not mattering which notation you use.
[/ QUOTE ]
In our case, random walks and markov chains are functionally the same thing, except markov chains give you a direct method to resolve the random walk, something looking at random walks statistically won't give you quite the same handle on. Binomials make sense if you are attempting to determine one discrete event, but the conclusions you draw might fall apart unless you resort to some possibly model-breaking simplifications.
[ QUOTE ]
The trick I am using to vastly simplify the caculations is to approximate the damage distribution as a gaussian. That way all I need to keep track of is an average damage and a standard deviation. The method may be a tad mathematical for these forums, but if you can handle a square root, you can use the results.
[/ QUOTE ]
Like those. You'll lose the effects of discrete damage on lifetime, which are the opposite extreme to compounded probability distributions. Its not a large counterbalance, but in effect you'll be removing the only term in the model that has a chance of counterbalancing the effect you're trying to measure, which is not so good.
If you aren't quite getting what I'm trying to say, I'll give you a quick pointer to the issue: lets look at the point near the end of life of a defense-protected player and a resistance protected player. At some point, health for the defense player will drop to the point where one more shot will kill her. The probability of a kill at that point cannot be averaged: the probability of a kill is constant regardless of whether the remaining health is 90% of the damage of one shot, or 5% of the damage of one shot, or anything in between. This skews the lifetime percentages by a small but not insignificant factor.
Healing/regeneration complicate things even further, but the basic principle is there: averaging that effect out disadvantages defense.
Put it in ultra-extreme terms: suppose something existed that did 200,000 points of damage with 75% accuracy, say. Damage averaging would nullify the inherent advantage defense would have verses resistance in such an extreme situation. At more reasonable damage levels, that advantage gets increasingly smaller, but doesn't disappear. Averaging presumes the contribution is exactly zero.
[Guide to Defense] [Scrapper Secondaries Comparison] [Archetype Popularity Analysis]
In one little corner of the universe, there's nothing more irritating than a misfile...
(Please support the best webcomic about a cosmic universal realignment by impaired angelic interference resulting in identity crisis angst. Or I release the pigmy water thieves.)
This problem strikes me as something solved more easily by iteration than analytically.
Model a set of defenses, resists, regeneration.
Set a series of damage types and magnitudes occurring at fixed intervals, with fixed to-hits. For argument's sake, even 10 actors would be reasonably easy to arrange.
run it through a loop 1000 times
Output a text file of the lifespans.
Yes this is brute force, but it could be relatively flexible.
For starters, just one damage type, one resistance type and one fixed regen rate could all be done in an excel macro.
If I get really possessed, I'll take a stab at it this weekend,
If you have any suggestions or pointers please, please throw them out here before then.
The cake is a lie! The cake is a lie!
[ QUOTE ]
[ QUOTE ]
[ QUOTE ]
[ QUOTE ]
[ QUOTE ]
For the defense case, I have to be hit 10 (or more) out of 20 times. The chance of that happening is around 1 in 100, which is an astounding 10,000 times more likely than the resistance case.
So, if you the question you are asking is "what is the average damage taken?" 25% defense is the same as 50% resistance. However, if the question you ask is "how likely am I to die?" then 25% defense is no where near as good as 50% resistance.
[/ QUOTE ]
You're going to drag us all through binomial theory aren't you.
truly thou art evil!
[/ QUOTE ]
Average lifetime analysis is a Markov Chain analysis, not so much binomial theory. Last time I said that, someone actually did it. I'm pseudo-lazy: I didn't want to calculate it, but I was willing to write a simulator to model it.
[/ QUOTE ]
I have been using 1D random-walk theory, but in the world where there are only two possible outcomes (hit or miss), Markov chains, binomials, and random-walks all work out to be the same thing so it ends up not mattering which notation you use.
The trick I am using to vastly simplify the caculations is to approximate the damage distribution as a gaussian. That way all I need to keep track of is an average damage and a standard deviation. The method may be a tad mathematical for these forums, but if you can handle a square root, you can use the results.
On a side note:
Yes the range of possibilities for a defense build is much wider than for a resistance build. While you will sometimes die in a situation where a resistance build never would, if you get lucky you can do things that are impossible for a resistance build. One particular event triggered my investigation down this road:
I went into a mission to fight Luimary. This is the Elite Boss version not the Hero version (after the patch). The first time I faced her, she downed me in seconds. I came back from the hospital to try again and I took her down with barely a scratch. The difference in the two attempts was so striking that is really piqued my interest in the topic.
Of couse, the second reason I have been doing this investigation is to answer the question many people have asked: Are the EA passives worth taking?
[/ QUOTE ]
And the answer is?
Inquiring minds want to know.
[/ QUOTE ]
Energy protection is not. Dampening field is a good power IMO, it's not crazy good like energy drain or a must-have like the defense toggles but it's a solid choice. Some people disagree and think it's crap, but I'm reasonably happy with it.
[ QUOTE ]
[ QUOTE ]
EC doesn't stack with other stealth powers, so removing it from the stalker set was presumably less of a balance issue and more about not wanting a completely worthless power in the stalker set.
[/ QUOTE ]
i think it was more of a "removing repulse from the brute set was less of a balance issue and more about not wanting a completely useless power in the brute set".
Stalker EA predates Brute EA. and Havind Hide and another stealth power in a stalker secondary would be absolutely huge for stalkers. no more dipping into concealment for stealth to stack for pvp. EA would be FOTM.
[/ QUOTE ]
I'm only lvl 20 and play mainly weekend nights...but I like EA quite a bit so far.
[ QUOTE ]
This problem strikes me as something solved more easily by iteration than analytically.
Model a set of defenses, resists, regeneration.
Set a series of damage types and magnitudes occurring at fixed intervals, with fixed to-hits. For argument's sake, even 10 actors would be reasonably easy to arrange.
run it through a loop 1000 times
Output a text file of the lifespans.
Yes this is brute force, but it could be relatively flexible.
For starters, just one damage type, one resistance type and one fixed regen rate could all be done in an excel macro.
If I get really possessed, I'll take a stab at it this weekend,
If you have any suggestions or pointers please, please throw them out here before then.
[/ QUOTE ]
It takes closer to about 5000 interations before it converges to the average normally.
[Guide to Defense] [Scrapper Secondaries Comparison] [Archetype Popularity Analysis]
In one little corner of the universe, there's nothing more irritating than a misfile...
(Please support the best webcomic about a cosmic universal realignment by impaired angelic interference resulting in identity crisis angst. Or I release the pigmy water thieves.)
shouldn't be a problem with any computer available since the mid 90's
The cake is a lie! The cake is a lie!
[ QUOTE ]
shouldn't be a problem with any computer available since the mid 90's
[/ QUOTE ]
Would be nice to have something to cross-check mine on.
[Guide to Defense] [Scrapper Secondaries Comparison] [Archetype Popularity Analysis]
In one little corner of the universe, there's nothing more irritating than a misfile...
(Please support the best webcomic about a cosmic universal realignment by impaired angelic interference resulting in identity crisis angst. Or I release the pigmy water thieves.)
Gee, it's almost like you guys are speaking English or something. 
SOOOOOOOOOOOOOO
have we come across any hard numbers yet? was it 12.75 per shield? or 16%ish.
20% 3 slotted DEF
is energy cloak 5%?
to recap castle's numbers
<font class="small">Code:[/color]<hr /><pre>S/L : 12.75 %
F/C : 15.00 %
__E : 15.00 %
__N : 10.50 %</pre><hr />
I gave TopDoc my test for Kin shield against lethal damage. I will say that even without taking the streakbreaker into account, it looks to aggree with Castle's 12.75% number.
i did some testing against an even level Luddite lambent of light last night. Used only power shield for energy defense and let him use his nrg/smash range attack.
463 attacks
169 hits
270 misses
miss streaks
1:36
2:24
3:18
4:10
5:4
6:7
the seven hits that came after the 6miss streaks were streakbreaker hits, so taking them out we have this
162/463 = 34.989% hit rate, or 15.01%defense
that's the number Castle gave too
i don't think Castle took into account the energy defense of Kin shield, or if he did -he made a mistake and subtracted it from the Negative def% instead of adding it to the NRG def%.
i'm going test again tonight using both power and kinetic shield against the white luddite.
Level 50 is a journey, not a destination.
▲Scrapper Issues List - Going Rogue Edition▲
[ QUOTE ]
to recap castle's numbers
<font class="small">Code:[/color]<hr /><pre>S/L : 12.75 %
F/C : 15.00 %
__E : 15.00 %
__N : 10.50 %</pre><hr />
i don't think Castle took into account the energy defense of Kin shield, or if he did -he made a mistake and subtracted it from the Negative def% instead of adding it to the NRG def%.
[/ QUOTE ]
Another possibility is that _Castle_ mistakenly added the extra Energy DEF of Kinetic Shield to Fire & Cold, instead of just to Energy. In this case base numbers would be 12.75% DEF to S/L/F/C, with Energy receiving the extra 2.25% bump from Kinetic Shield to get to 15%, while Negative gets 2.25% less & so stands at 10.5%. But certainly something about _Castle_'s numbers is wrong.
Liberty
Mister Mass - 50 Inv/SS/NRG Mut Tank [1236]
Doc Willpower - 50 Grav/FF/Psi Mag Controller
Baron Wonder - 50 SS/Elec/Mu Mag Brute
Sound Bight - 50 Son/Son/Mu Tech Corrupter
I PM'd him and showed him his numbers along with what i think the numbers are. he's usually pretty good about replying. I'll post again when he does.
if he does say he made a typo about the numbers, and my idea is correct, then i'd be confident enough in those numbers to stop testing.
I'd move onto Nrg Cloak then.
Level 50 is a journey, not a destination.
▲Scrapper Issues List - Going Rogue Edition▲
Just to clarify, things guys.
The numbers I posted were the highest value of each armor. There is no stacking taken into consideration at all.
now i'm more confused. 
gonna PM Jonyu
Level 50 is a journey, not a destination.
▲Scrapper Issues List - Going Rogue Edition▲
[ QUOTE ]
now i'm more confused.
gonna PM Jonyu
[/ QUOTE ]
What it looks like is that when the shields were reworked, the energy defense from Kinetic was rolled into Power Shield and added to the fire and cold defense there, so Kinetic no longer provides any Energy defense. That all makes perfect sense to me. If that is the case, then really the only outstanding question is how much defense is provided by Energy Cloak.
it doesn't make sense to me why'd they do that without also updating the powers descriptions. or why they would suddenly make fire and cold def as strong as energy def, leaving negative out in the cold.
i don't like it at all. I'm gonna test Kin tonight to see if any energy defense is in there.
Level 50 is a journey, not a destination.
▲Scrapper Issues List - Going Rogue Edition▲
/em smashes his head agaisnt his desk... repeatedly... for 15 minutes
Ok, here's what I would like to have now:
A run down, FROM THE DEVS, on what each power in EA actually protects against and/or provides, including the specific long and short descriptions that the powers SHOULD be giving, AND the base values of every power in the set, broken down by type and/or position if present.
Please. This is getting ridiculous. Thank you in advance.
Be well, people of CoH.
[ QUOTE ]
it doesn't make sense to me why'd they do that without also updating the powers descriptions. or why they would suddenly make fire and cold def as strong as energy def, leaving negative out in the cold.
i don't like it at all. I'm gonna test Kin tonight to see if any energy defense is in there.
[/ QUOTE ]
Remember that Fire Tanks are weak to Cold, Ice Tanks are weak to Fire, and DA is weak to Energy. From a story point of view, it makes more sense that energy aura should be weak to negative energy than that it should be weak to toxics.
On a side note: remeber that when two different defense numbers apply, the better one is used. For attacks like Shadow Punch and Smite we get the 12.75% S/L defense rather than the 10.5% negative energy one. Pure negative energy attacks are acually quite rare.
i know how def works.
i can understand neg being weaker than the others. my problem is that why are the elemental numbers as strong as the energy numbers?
either way, i don't trust any of it right now and I'm gonna test every single combo. if it ends up agreeing with these goofy numbers then cool.
Level 50 is a journey, not a destination.
▲Scrapper Issues List - Going Rogue Edition▲
[ QUOTE ]
i know how def works.
i can understand neg being weaker than the others. my problem is that why are the elemental numbers as strong as the energy numbers?
[/ QUOTE ]
My guess is that the set is being conceptualized as "average smash/lethal defense, strong against energy but vulnerable to negative energy." Elemental defense gets to be high mainly from a balance perspective (so that energy isn't only strong against a single damage type) and is rationalized as energy defenses being equally good against heat-exchange attacks (fire/cold) as pure energy attacks.
[Guide to Defense] [Scrapper Secondaries Comparison] [Archetype Popularity Analysis]
In one little corner of the universe, there's nothing more irritating than a misfile...
(Please support the best webcomic about a cosmic universal realignment by impaired angelic interference resulting in identity crisis angst. Or I release the pigmy water thieves.)
Thank you for taking the time to do this, gang.