Elitist Jerks Combat Ratings at level 85 (Cataclysm)

04/29/09, 1:47 PM   #451
Mavanas
Great Tiger

Night Elf Rogue

 Originally Posted by Slafsinator A peculiar effect of stacking Armor Penetration to the point of negative armor is that the closer the base/debuffed armor of the target is to the so-called Armor Penetration cap (see Ghostcrawler's post), the more effective Armor Penetration is. Using Shattered Throw to reduce the target's debuffed armor further than with Sunder/Faerie Fire can decrease the DPS of those having above 100% Armor Penetration. To simplify it -- in one case you debuff someone to 3000 armor and you get 200% Armor Penetration. You end up at -3000 armor on the target -- great. In the other case you debuff the target to 8000 armor and you still have that juicy, tasty 200% Armor Penetration. Result: -8000 armor on the mob.
There is an interesting implication of this observation that is within reasonably attainable armor penetration values. Suppose a rogue has 100 armor penetration from gear, and mjolnir runestone and grim toll proc at the same time, for a total 1324 armor pen rating or 108% armor penetration. Then there is a breakeven value of reduced boss armor at which the rogue will start doing LESS damage with shattering throw. This value is 7779 (assuming shattering throw stacks multiplicatively with other debuffs). Any value below that will make the rogue do less damage when ST is used. For reference, the reduced armor with 5 sunders and faerie fire is 8088.68 and if the rogue has serrated blades, the reduced armor is 7450, so in that case shattering throw starts hurting the rogue's dps. I know it's still not a very common situation, but with further gear changes, armor penetration can go up and the threshold of reduced armor will go up as well.

 05/01/09, 2:24 AM #452 a civilian Piston Honda   Stenhaldi Worgen Druid   Sargeras A few comments on properties of this armor penetration system: Let A be the armor value before armor penetration, but after debuffs. This would be the armor value the target sees on their character pane. A' is the effective armor value after armor penetration. I will denote by x the fraction of damage taken from an armor mitigated attack (1-damage reduction). C is the armor constant, 15232.5 for level 80 attacker. k is the nominal armor penetration fraction. We know $1-x = \frac{A'}{A'+C}$ and $A' = A-k(A+C)/3$ . Thus, $x= \frac{C}{(A+C)(1-k/3)}$ . Some implications of this:This expression asymptotes at k = 3, so that if you could obtain 300% armor penetration, you'd do infinite damage (assuming there's no cap). Not realistic? You can actually get surprisingly close -- [Grim Toll], [Mjolnir Runestone], [The Dusk Blade], and executioner yield 151% armor penetration; another 25% can be obtained from mace specialization and battle stance, so that you "only" need 124% passive armor penetration from gear. This is little over 1500 armor penetration rating -- still not realistic. It's actually not possible to obtain enough armor penetration that an armor debuff would cause you to do less damage. $\frac{\partial x}{\partial A}= \frac{-C}{(A+C)^{2}(1-k/3)}$ , which is always negative (for k < 3). That is, an increase in armor (before armor penetration) always yields a reduction in damage. The equation $\frac{\partial x}{\partial k}= \frac{C/3}{(A+C)(1-k/3)^{2}}$ for the absolute effect of a change in armor penetration, quantifies the increasing returns of armor penetration. In particular, armor penetration is $\left(\frac{1-k/3}{1-k'/3}\right)^{2}$ as valuable at armor penetration k' as at armor penetration k (if other stats are held constant). Edit: indeed, these are only valid for A > C/2. I had forgotten there was a minimum function involved. Last edited by a civilian : 05/01/09 at 12:31 PM.
 05/01/09, 11:07 AM #453 Mavanas Great Tiger   Mavanas Night Elf Rogue   Lightning's Blade You cannot ignore the minimum function in the cap formula. The exact formula for effective armor is $A' = A-k*min(A, (A+C)/3)$ For high values of armor debuffs, A becomes smaller than (A+C)/3, and the formula simplifies to $A' = A(1-k)$ and armor mitigation formula becomes $x= \frac{C}{A(1-k)+C}$ The breakeven value of A is C/2, which can occur any time you have sunders and shattering throw up, or if you have faerie fire, sunders and serrated blades. This changes your analysis somewhat, in particular for A < C/2, you will do infinite damage if k=C/A+1, which given the restriction on A is at least 3. So the magic value k=3 remains unchanged. However the partial with respect to A becomes: $\frac{\partial x}{\partial A}= \frac{-C(1-k)}{(A(1-k)+C)^{2}}$ which is positive for k>1 (easily obtainable, see for example my previous post). This means that a decrease in armor decreases damage. The partial with respect to k changes to: $\frac{\partial x}{\partial k}= \frac{C/A}{(A(1-k)+C)^{2}}$ which is always positive.
 05/03/09, 5:40 PM #454 Mavanas Great Tiger   Mavanas Night Elf Rogue   Lightning's Blade Spell crit reduction I have done additional testing of spell crit reduction against boss-level target dummy. There have been some extensive tests of physical crit reduction against bosses, but very little testing of spell crit reduction. Dontmindme has shared with me the results of his previous tests. He found 1 crit in 1060 successful wound poison applications at 2.2% spell crit, and 2 crits in 3667 applications at 2.16% crit. He was under the impression that both tests seem to indicate a 2.1% spell crit reduction. My tests: Obs Rating Perc Crit Crits Observed Reduction 3722 96 2,0912% 0 0,0000% 2,0912% 3330 97 2,1130% 0 0,0000% 2,1130% 739 97 2,1130% 0 0,0000% 2,1130% 3724 98 2,1348% 1 0,0269% 2,1079% 1912 98 2,1348% 1 0,0523% 2,0825% Here are some conclusions: - Spell crit reduction against boss-level targets is no more than 2.1348%. - I have not been able to disprove a hypothesis that spell crit reduction is at least 2.1130%. - There is a 17% chance that spell reduction is 2.08% and there were no crits in tests 1-3 by random chance - There is a 37% chance that spell reduction is 2.09% and there were no crits in tests 1-3 by random chance - There is a 59% chance that spell reduction is 2.10% and there were no crits in tests 2 and 3 by random chance So while 2.1% appears to be a likely round number for spell crit reduction, more tests with higher number of observations are needed to pinpoint the exact number. Here is backup for key tests: 1 crit out 3724 observations with 2.1348% crit and no crits out of 3330 observations with 2.1130% crit In same tests, I have observed white damage crit reduction: Obs Theory Observed Reduction CI low CI high 12210 21,45% 17,23% 4,22% 3,55% 4,89% 13420 20,70% 16,14% 4,56% 3,94% 5,18% 4396 19,41% 14,83% 4,58% 3,53% 5,63% 2024 17,39% 13,04% 4,34% 2,88% 5,81% 8824 18,60% 14,36% 4,24% 3,51% 4,97% 8382 19,81% 14,83% 4,98% 4,20% 5,74% 12210 6,51% 1,83% 4,68% 4,43% 4,91% 11242 14,07% 9,21% 4,87% 4,31% 5,39% Conversion 4,56% 4,43% 4,89% Thus while this does not disprove the hypothesis of 4.8% crit reduction for white damage crits for rogues, 4.5% seems like a more likely number.
05/03/09, 5:53 PM   #455
• Vulajin
Vula'jin the Void, blessed by the loa

Mal'Ganis
 Originally Posted by Mavanas Thus while this does not disprove the hypothesis of 4.8% crit reduction for white damage crits for rogues, 4.5% seems like a more likely number.
Unless the mechanics changed at some point, the 12037 attacks I performed when we originally tested this with 4.79% tooltip crit without experiencing a single crit disprove any possibility of a 4.5% reduction for rogues.

Also from the PTR: Retesting hit table assumptions

 Originally Posted by Enervate Yep, still a fucking idiot.

05/07/09, 12:55 PM   #456
Roywyn
Bald Bull

Roywyn
Gnome Mage

No WoW Account (EU)
Spell Crit Reduction and Mages

 Originally Posted by Mavanas There have been some extensive tests of physical crit reduction against bosses, but very little testing of spell crit reduction.
Those results of spell crit for rogues also match the results of spell crit for mages, if I'm not mistaken.

However spell crit reduction for mages is completely different from your results. Which makes it likely that it's also different for other pure caster classes.
Melee mechanics are also different for mages, so there's a precedent of split mechanics (a weapon skill capped mages will almost only get glancing blows against raid and boss mobs, not just the usual 24%, unless they recently changed it).

Mages below 2.1% or even below 2.0% spell crit (that's as low as you can get due to your base intellect) will easily get a crit or two when testing for short amounts of time. Some more thorough tests suggested that the spell crit reduction is a lot lower than 2% without gear, but slowly increases up to around 3% when you get crit and intellect from gear and buff.

There has been no real conclusion that is universally agreed upon for mages since the things are pretty obscure.

Chaotic Meta Gems in Cataclysm: http://elitistjerks.com/f75/t106009-...2/#post1794256

DPS spec and class comparison in Naxxramas gear: http://code.google.com/p/simulationc...i/SampleOutput
The Blue Bar and you - the complete Fire Mage 2.4 mana compendium: http://elitistjerks.com/658230-post3191.html

And [Timbal's Focusing Crystal] doesn't proc on AM.
Neither does [The Egg of Mortal Essence] since 3.1.

05/09/09, 1:00 AM   #457
Mavanas
Great Tiger

Night Elf Rogue

 Originally Posted by Roywyn Those results of spell crit for rogues also match the results of spell crit for mages, if I'm not mistaken. However spell crit reduction for mages is completely different from your results. Which makes it likely that it's also different for other pure caster classes. Melee mechanics are also different for mages, so there's a precedent of split mechanics (a weapon skill capped mages will almost only get glancing blows against raid and boss mobs, not just the usual 24%, unless they recently changed it). Mages below 2.1% or even below 2.0% spell crit (that's as low as you can get due to your base intellect) will easily get a crit or two when testing for short amounts of time. Some more thorough tests suggested that the spell crit reduction is a lot lower than 2% without gear, but slowly increases up to around 3% when you get crit and intellect from gear and buff. There has been no real conclusion that is universally agreed upon for mages since the things are pretty obscure.
I noticed that phenomenon with poisons too, in two cases, when I put on more crit gear and calculated the crit reduction, I ended up with 2.4-2.5% crit reduction, however the confidence intervals still included 2.1%. It may well be that the reduction formula is more complicated, it could be in relative terms.

 05/12/09, 2:08 PM #458 landsoul Myrmidon Champion     Landsoul Worgen Warrior   Alterac Mountains With sunders and fearie fire only, it seems that the intersection is at 1530 penetration rating. Anything more than 1530 rating will decrease your effective dps if shattering throw is applied to the target. Here's some graphs to show. x is armor penetration rating, y is percent damage done /100 sunders, faerie fire sunders, faerie fire, shattering throw intersection point As said before, if you have too much ArP, a shattering throw will be bad for you specifically, probably not for the rest of the raid at that given time though. Further evidence that shattering throw and the new armor penetration system is terrible design. Shattering throw is especially terrible for the warrior casting it, especially if they have a high amount of ArP (which is what best possible gearing leans towards) >--Coaching Site--< Private coaching / Warrior Resource >--Stream--< Tues, Wed, 7 Eastern.
 05/12/09, 4:19 PM #459 Slafsinator Von Kaiser   Slafsinator Night Elf Warrior   Stormscale (EU) I honestly have no idea how you came to that conclusion, Landsoul. According to my tinkering, the point where using Shattering Throw potentially reduces your DPS does not have any relation to the player's ArPen at all (barring really extreme, ludicrous amounts of it). Instead, only the target's debuffed armor matters, and the optimal amount of armor for a target to have for your 100%+ ArPen character is: 7616,25 (=15232,5/2). Anything more will initiate the ArPen "cap", and anything less will yield a lesser effect on ArPen, crossing over into negative values. What you can do though, is conjure a graph on how much armor reduction you need from debuffs to make the target stick at 7616,25 with different base armors, but it's really not that interesting thanks to the amount of useful armor values for raiding and DPS. Against a boss you want 28,38% armor reduction from debuffs, which is just more than FF + sunders, and quite a bit less than FF + sunders + shattering throw. I'm coming to these figures through my spreadsheet, feel free to poke around and see if I made any mistakes.
05/13/09, 4:57 AM   #460
suicuique
King Hippo

Night Elf Warrior

Antonidas (EU)
 Originally Posted by Slafsinator I honestly have no idea how you came to that conclusion, Landsoul. According to my tinkering, the point where using Shattering Throw potentially reduces your DPS does not have any relation to the player's ArPen at all (barring really extreme, ludicrous amounts of it).
I'm speculating here.

Assuming shattering throw is similiar to sunder/FF mechanics (i.e. it is applied as a debuff on the mob and then ArP comes into play), we get following standard scenario:

1) without shattering throw
ca 8089 = boss armor post sunder/ff
ca 7774 = amount of armor which is base to the ArP calculations

2) with shattering throw
ca 8089*0.8 = 6471 debuffed boss armor = amount of armor on which ArP calculations are based (because it is < 7616)

The implications of this would be:
100% ArP is enough in 2) to get boss to 0 armor, while in case 1) you'd need ca 104%.
The point is, that in case 1) the armor value affected by ArP is higher while at the same time more armor has to be offset tto reach zero, while in case 2) ArP is less effective once you reach negative armor regions.

Rough calculations point me to 1527 being the lower bound for ArP when case 1) starts becoming more effective (read: reaches higher negative armor values) than case 2).

 05/19/09, 7:35 AM #461 Lollersk8er Piston Honda   Déathknight Human Death Knight   Azjol-Nerub (EU) Is the new 100% cap applied before or after the 8,1% effectiveness 'cap'?
 05/19/09, 10:48 AM #462 kalbear Bald Bull   Fellhoof Tauren Druid   Balnazzar They haven't actually announced that they're going to do a 100% cap yet. No one knows what the mechanics are.
 05/20/09, 4:43 AM #463 • Aldriana Mike Tyson   Aldriana Night Elf Rogue   Doomhammer Brief testing on target dummies tonight indicates that ArPen is indeed capped, and at 100% (not 0 armor).
 05/20/09, 10:01 AM #464 hellord Absolute Arms     Hellorco Orc Warrior   Hellscream (EU) I just did few tests and can confirm that battlestance is just reducing the cap by 10%. I have to guess mace stance does the same so reducing the gear cap to 75%. ArP Whore
 05/20/09, 11:45 PM #465 landsoul Myrmidon Champion     Landsoul Worgen Warrior   Alterac Mountains I just confirmed tonite with accuracy of 99.98 percent that you cannot remove more armor than the effectiveness cap. I did one test: Sunder x5, target dummy, 4037 AP with battlestance, 1216 penetration rating with grim toll proc. Bloodthirst multiplier: 0.5 (ap scalar) * 1.1 (unending fury) * 1.06 (2H spec) = .583 dmg per ap 4037 attack power strike yields 2353.57 expected damage, I did 2265 damage. 2265/2353.57 = 96.2367398% damage applied. effectiveness cap with 5 sunders = (10643*.8+15232.5)/3 = 7915.63333 1216/12.3216 + 10%= 108.7314403% armor removed (clearly over 100%) The boss debuffed, has 10643*0.8 = 8514.4 armor. Assuming the entire effectiveness cap of armor was removed, the boss is left with 8514.4-7915.633 = 598.7666 armor. Damage applied with 598.766 armor = 15232.5/(15232.5+598.766) = 96.2178221% which exactly matches the actual bloodthrist damage I dealt. Conclusion You can only remove armor up to the armor penetration effectiveness cap of (debuffed armor + 15232.5)/3 and any penetration after 100% tooltip is ineffective. >--Coaching Site--< Private coaching / Warrior Resource >--Stream--< Tues, Wed, 7 Eastern.

 Elitist Jerks Combat Ratings at level 85 (Cataclysm)