Elitist Jerks Druid Changes Patch 3.1

03/07/09, 8:38 AM   #466
a civilian
Piston Honda

Worgen Druid

Sargeras
 Originally Posted by nightcrowler - Rip+rake (40%): assuming no clipping, if you miss a rip or a rake you are missing 1 GCD of their dps, so -1% hit: 0.01*0.4 = 0.4% less dps
That's not quite an accurate estimate, for two reasons. The first is more obvious: if you miss rake, you lose 1/9 of its damage; if you miss rip, you lose 1/26 of its damage (assuming 9 second and 26 second durations, respectively). So with -1% hit, you're actually losing

0.01*(1/9*.18+1/26*.22) = 0.02% total damage, [edit: correct result of calculation is 0.03%, of course, but this does not significantly alter the final result]

given that rake is 18% and rip 22% of total damage.

The other reason is that you don't simply lose the damage of 1 second of rake or rip. Rather, you exchange it for an equivalent quantity of non-dot attacks, i.e. shreds and ferocious bites. You still lose some damage, of course, since shred and ferocious bite are less efficient than rake or rip. (You also lose energy, but that is already covered in the subsequent paragraph.) It's hardly worth making a detailed estimate of the damage loss here, so let's just assume shred and ferocious bite are half as efficient as rake and rip. This means we gain back half of the lost damage, so overall we are only losing 0.01% of total damage, for having -1% hit.

With this revision, we now obtain 1 hit = 21.5 points by the same calculation.

To be fair, we really should consider omen of clarity, as it is affected double by hit: first in the proc rate, and second because if you miss a clearcasted attack, you lose the entire cost of the attack, not just one fifth. My rough estimates have put omen of clarity at about 11% of overall damage in 3.0x; I will assume this is still the case in 3.1. The value of hit in this case is now simple: this 11% damage has two chances to miss, so -1% hit yields -0.22% damage (taking first-order approximation).

We also have to revise the lost-energy estimate to avoid counting missed clearcasted special attacks twice. However, I now notice a mistake in this estimate as well, for you assume that one special is used every second, when actually we should be using one special every 38/16 seconds by your estimate. An easier way to estimate this is to simply take the total yellow damage, excluding that resulting from omen of clarity, and consider that we lose one fifth of it on a miss. 70% of damage is yellow damage, so about 60% is yellow and non-clearcasted. -1% hit is then worth -1%(60%)/5 = -.12% damage.

I believe you also underestimated the effect of miss on white damage. For white attacks, -1% hit is worth (.5*2.266+.25*.7+.24)/(.5*2.266+.25*.7+.25)-1 = -.64% of white damage assuming crits are 2.266 damage and glances are 0.7, so if white is 30% of damage, then -1% hit is worth -0.19% damage.

Overall, then, -1% hit is worth
-0.19% damage due to missed white attacks
-0.22% damage due to omen of clarity (both lower proc rate and missed clearcasted attacks)
-0.12% damage due to missed yellow attacks (which weren't clearcasted)
-0.01% damage due to lost rip/rake uptime
which sums to -.54% total damage.

Using the same calculation, 1 hit rating is now worth 22.5 points.

Last edited by a civilian : 03/07/09 at 5:01 PM.

 03/07/09, 8:48 AM #467 spartakos Piston Honda   spartakos Tauren Druid   Genjuros (EU) Nature's grace buff now lasts 15 sec. There are are encounters where the new Nature's Grace will lose most of his potential.
03/07/09, 9:26 AM   #468
Rangifer
Glass Joe

Tauren Druid

Sen'jin
 Originally Posted by nightcrowler - All yellow attacks (70%): if you miss you are wasting energy. If we assume an average energy consumption of 38 energy and an energy regain x secods of about 17 energy. We have 0.2*38= 7.6 energy wasted. 7.6/16 = 0.47 = 48% dps loss due to energy. So: -1% hit = 0.01*0.7*0.48 = 0.34%.
I don't understand where that 16 came from. I don't really know where the 17 energy comes from either. As I see it, you have a 1% chance of losing 20% energy for no damage. Therefore, you lose 0.2% of your yellow damage, or 0.14% of total damage.

 Originally Posted by a civilian The other reason is that you don't simply lose the damage of 1 second of rake or rip. Rather, you exchange it for an equivalent quantity of non-dot attacks, i.e. shreds and ferocious bites.
The assumption is that you try to Rake/Rip, fail, then try again, so you're losing 1 s of uptime, a GCD, and energy. You're not exchanging it for anything.

Here's an alternate OoC calculation:
5.833% chance per white of a proc and 0.893352 s/swing gives 4 procs/min (rounded up). Assuming the maximal value for a proc of 42 energy, that's 168 energy/min. Normal energy generation is 10/s + 60/30s for TF. That's a total of 888 energy/min. Losing 1% of OoC procs is worth 1.68 energy/min, or 0.19% of total energy, which has a value of roughly the same as the missed yellow attacks for about 0.14% total damage.

Summing up, 0.08% (white) + 0.14% (OoC) + 0.14% (yellow) + 0.02% (DoTs) = 0.38%. Following the rest of the calculations, that 25.3 dps/1% hit or 0.79 dps/hit rating. Giving a final value of ~16 points.

03/07/09, 9:57 AM   #469
dukes
Bald Bull

Dukes
Tauren Druid

No WoW Account (EU)
 Originally Posted by Starfox Your value for 1843 spellpower is way off here: (642-215)/1843 = 0.231687 perhaps a trinket procced and you didn't notice, because with a coeff of 0.2, it would need 2135 sp to get 642ish ticks. rest matches nicely (511-215)/1480 = 0.2
Sorry, forgot to say I had 2t7 on for the 1843 spellpower test.

03/07/09, 10:27 AM   #470
Starfox
King Hippo

Tauren Druid

Destromath (EU)
 Originally Posted by dukes Sorry, forgot to say I had 2t7 on for the 1843 spellpower test.
Ok, then it matches exactly the 0.2
 Originally Posted by spartakos Nature's grace buff now lasts 15 sec. There are are encounters where the new Nature's Grace will lose most of his potential.
Eh? You mean 3sec?
Yea, on things like heigan dance you kind of lose the gains of the NG, but i think the general buff this new NG is outweights this by far.

Last edited by Starfox : 03/07/09 at 10:39 AM.

Hello.
Light the fuse.
For all my homies.
Do not run, we are your friends.
SimulationCraft Druid Guy

03/07/09, 11:27 AM   #471
Taudark
Von Kaiser

Tauren Druid

Kor'gall (EU)
 Originally Posted by Wulfmanjack On PTR it seems we get more out of expertise. I have 25 on live and 29 on ptr (from 6,25% to 7,25%) On Char Screen its says on both 125 expertiserating.

It's most likely a bug. I very much doubt they'd buff expertise by 25% as well. They most likely buffed expertise when they buffed arp somehow.

Last edited by Taudark : 03/07/09 at 11:37 AM.

 03/07/09, 12:55 PM #472 alpiino Glass Joe   Muusikauha Night Elf Druid   Deathwing (EU) Had a chance for 1 test but If I saw right the GCD gets reduced below 1s with enough haste. Just tested my wrath spam with powerinfusion, natures grace and haste pot on ptr. Edit: There was quite some lag in ptr so someone should test this a bit more and could confirm it gcd can go below 1s. Last edited by alpiino : 03/07/09 at 1:21 PM.
 03/07/09, 12:57 PM #473 Lord BEEF Soda Popinski     Werebeef Tauren Druid   Mal'Ganis I'm hearing some reports of moonkin topping the damage meters on PTR. If it turns out that they need to nerf us, I hope they do it by reducing the amount of haste nature's grace provides. To something like 15%, or maybe 10% haste and 3% damage or whatever. The idea being that lower haste allows us to continue to enjoy haste as a stat
03/07/09, 1:16 PM   #474
Cryect
Bald Bull

Night Elf Druid

Stormreaver
 Originally Posted by Valerian This looks fine at a glance to me. The real question is then why does both the new Rawr (2.2.0b3) AND Tossk's 3.1 simulator have Hit and Expertise as significantly lower than Agi/Str?
Well in the case of Rawr it used to undervalue it because it just assumes the only loss for misses is Energy and never takes into account the GCD and assumes magical perfect up times etc. The way the Rawr model works is very questionable value for accurate results besides its fast to calculate versus something like a Markov Model. In the latest 2.2 they decided to undervalue hit/expertise even more for some reason (haven't examined the code closely to see what they decided to change). Pretty sure Toskk's model is very similar to Rawrs so will have similar issues.

My limited experiments have shown that nightcrowler's results are much closer to being correct than the numbers I see from Rawr's or Toskk's models (don't have anything very exact since hard to replicate but tests show regemming for +hit/expertise if below the cap ends up being higher DPS than Str gems when I ran tests a while back).

I need to do something useless.

03/07/09, 1:45 PM   #475
Allev
King Hippo

Tauren Druid

Mal'Ganis
 Originally Posted by nightcrowler About high value of hit/exp in Night's Sim First of all, I'm 99.99% sure that the simulations are ok. I've tested cycles on live and they work fine. Second. If you don't use FB, hit=exp=38, still high.
So, the last time you showed your simulator, you claimed that losing the GCDs was pushing back your finishers enough to cause hit value, and now you say that FB doesn't make a difference in the value of hit, when you clearly won't have any trouble maintaining CPs for a simple 4SR/5Rip cycle.

Justification of the simulator in that "I've tested the cycle on Live" isn't much of a justification. Even if the final DPS is in the same neighborhood of correct, it may just mean 2 things are wrong, and they're compensating for each other. Also, the fact that you're sure doesn't make you right, given that your math doesn't agree with everyone else's (in these posts and with the existing theorycraft). I'm not saying it's wrong, I'm just saying its credibility is questionable, especially when it isn't reviewable as a whole.

Finally, if you have an average 29 second cycle, you have downtime on rip, and if you're only doing a 3/4 point SR, then SR is dropping too. I can't believe that you aren't going to want to conditionally exclude FB-- it's not worth letting those two drop. You're averaging at least 3 white attacks that aren't getting SR, and because it's an average, you're going to be missing SR up on rakes/shreds on the bad cycles as well.

03/07/09, 1:55 PM   #476
Lilija
Piston Honda

Worgen Druid

Bronzebeard (EU)
 Originally Posted by Lord BEEF I'm hearing some reports of moonkin topping the damage meters on PTR.
I did some tests on dummies however no replenishment so I gues 2min tests aren't that much representative. Anyway, I haven't noticed any significant difference in dps. Actually the live one tend to be higher but I would say that's lucky crits.

I also noticed one more thing. Dps based on solar eclipse is very stable but lower from what can be reached by lunar eclipse. However, lunar eclipse still suffers from not proccing when you want it to.

 03/07/09, 2:17 PM #477 Tristanian Dreamwalker   Tristanian Night Elf Druid   Kor'gall (EU) Solar eclipse always produced more reliable DPS, but wasn't optimal due to wrath clipping. Judging from my own tests, I'm more inclined to agree with Adorielle's assessment that there is indeed a "haste breakpoint" (essentially the amount of haste required to drop the NG'd wrath below 1 second), after which, lunar eclipse will still come on top, despite the RNG factor. It would be interesting to see some numbers on the next few days, as more testing is in order. Then there is the matter of what happens while under the effects of heroism/bloodlust, though I suspect that Starfire spam will still be favored, in the vast majority of occasions.
03/07/09, 2:39 PM   #478
a civilian
Piston Honda

Worgen Druid

Sargeras
 Originally Posted by Rangifer The assumption is that you try to Rake/Rip, fail, then try again, so you're losing 1 s of uptime, a GCD, and energy. You're not exchanging it for anything.
You lost some rip/rake uptime, but you did not lose the energy/combo point cost of that uptime. (You do lose some energy, but that was already accounted for in a separate estimate, so we do not want to account for it here.)

Think of it this way: After you try again and succeed in getting rip or rake up, you now have more time before the rip/rake has to be refreshed than you would have had if your attack had landed the first time. As a result, you don't have to spend as much energy/combo points on rake/rip, and this leaves more energy/combo points to use on shred and ferocious bite.

 Originally Posted by Rangifer Here's an alternate OoC calculation: 5.833% chance per white of a proc and 0.893352 s/swing gives 4 procs/min (rounded up). Assuming the maximal value for a proc of 42 energy, that's 168 energy/min. Normal energy generation is 10/s + 60/30s for TF. That's a total of 888 energy/min. Losing 1% of OoC procs is worth 1.68 energy/min, or 0.19% of total energy, which has a value of roughly the same as the missed yellow attacks for about 0.14% total damage. Summing up, 0.08% (white) + 0.14% (OoC) + 0.14% (yellow) + 0.02% (DoTs) = 0.38%. Following the rest of the calculations, that 25.3 dps/1% hit or 0.79 dps/hit rating. Giving a final value of ~16 points.
You neglect to consider that when you miss a clearcasted attack, you lose the clearcast, effectively losing the entire energy cost of the attack. This is why I said omen of clarity is affected twice by miss. 1% miss yields 1% fewer clearcast procs, and 1% chance for a given clearcast proc to be lost.

Last edited by a civilian : 03/07/09 at 5:09 PM. Reason: trying to make this a little clearer

 03/07/09, 3:34 PM #479 Garanthir Piston Honda   Garanthir Night Elf Druid   Eldre'Thalas This is probably stating the obvious, but it is worth noting. The Savage Defense "shield" gains value on trinket procs from [Mirror of Truth]. My shield went from 1312 absorbed to 1562 absorbed or roughly 5249AP and 6249AP, in addition Mongoose procs raise the damage absorbed as well as one would guess. It would seem that Mirror of Truth (and others like it depending on the selection available to you) can be a very good tanking trinket for heavy physical damage type fights. I think Greatness/Valor Medal would stay on top(for those that favor Dodge/Agi), but in the absence of a Greatness card Valor Medal/Mirror would be a very solid combo. The big weakness with SD is the RNG aspect, when testing against the Dark Rune Giants in HoS it is bothersome that I went on a string with no crits right when the mob gained a 20% damage buff. Last edited by Garanthir : 03/07/09 at 3:55 PM.
 03/07/09, 4:18 PM #480 nightcrowler Don Flamenco     nightcrowler Night Elf Druid   Runetotem (EU) Well I need to reply to A civilian and to Allev. I'll try to explain my self better and to do the best math I can (If I did a direct simulation there is a reason...). +++POST MODIFIED TO FIX AN ERROR with dots+++ (thanks to A civilian to point it out) First: less hit, less white dps With my stats I have 52% crit, 24% glance, 24% normal hits. An average white hits for (2*1.1*1.03*0.52+0.7*0.24+0.24) = 1.58632 times a normal hit. Let's add 1% miss -> 2*1.1*1.03*0.52+0.7*0.24+0.23+0*0.01) = 1.57632. Let's see the average dps loss on white hits = 1-(1.57632/1.58632) = 0.0063 = 0.63% In my sim white hits are 28.19% of total dps, so I'll lose 0.63% of 28.19% = 0.178% of total dps due to white misses. Second: less hit, less ooc procs If you don't white hit you don't proc ooc. We are not interested on dps here so 1% less hit (from the cap) means 1% less hit (for istance from 100 hits to 99 hits) so you are missing 1% of the ooc procs. ooc proc rate is based on weapon speed (normalized to 1). We have 5.83% chance to proc on each hit. With 0 haste we can find proc x minute. 0.0583*60 = 3.5 ppm. With windfury+moonkin aura+haste from gear, I have a 0.73s attack speed. So I'll have 3.5/0.73 = 4.79 ppm. Reducing hits by 1% actually reduce ppm by 1% or simply we will have 4.746 ppm. Because we will use ooc procs to do shreds 1 ooc proc = 42 energy. So 42*4.79 = 201.18. 1% less hit = 0.01*201.18 = 2.01 less energy every minute (on average). Third: less hit, less rip/rake uptime We have assumed that you want 100% rake uptime and 100% rip uptime. Under that assumption you: - don't clip dots - don't use other abilities instead of them - try to maximize the uptime - always micromanage energy to have enough energy to use those abilities. Under those assumption you are only GCDs limited. If due to misses you have less energy you will do less shreds, but you will always have enough energy to rip. But you also don't want to clip dots so if you miss you lose 1 second of uptime. Let's do an example: Rip ticks every 2 seconds for 26 seconds. A ticks happen at time 2, 4, 6, ... 26. At time 26 you will do another rip having a tick at times 28, 30, etc.. If that rip misses you will wait GCD and do the rip at time 27 and it will tick at times 29, 31.. etc.. -> If you miss with a dot you siply lose dot uptime. If you have 1% miss, you will lose on average 1 seconds of uptime every 100 times you refresh the dot. 0.01111*(22%/26+17.5%/9 )= 0.031% Forth: less hit = energy wasted on dots As already said we can only lose gcd on dots, but energy lost on dots will be less energy for shreds = less shreds. Sussose tha you have 35+42-10 = 67 energy. At time 0 you rake ->33 energy, alt time 1 you will have 42 energy and shred, after exactly 4.1 second you will have another 42 energy and shred again. You have done 1 rake and 2 shred in a 5.3 sec window. Now suppose that you miss the rake -> 42+0.8*35 =42+27 = 69 energy at time 1, you rake -> 44 energy at time 2 -> shred, next shred after 4.0 seconds -> 1 rake and 2 shred in 6.0 seconds. The main problem here is not time for shred but that you will likely lose some shred (if for istance you miss enough time that you need to rip again and so on. So basically we need to put the dot energy loss on shreds dps. With a missed dot I lose 7 energy on rake and 6 energy on rip. In 60 seconds assuming about 100% uptime i'll do about 6 rake and 2 rip. Those are 35*6+2*30=270 energy. With 1% miss we will lose 0.01*0.2*270 = 0.54 energy every minute Fifth: less shred hits = energy wasted If a shred miss we will waste 9 energy. But we need to know how many shred are ooc-shred and how many shreds are normal. I do 14.6 shred every minute and I have 4.79 ooc ppm. So I can say that 9.8 shred are normal, 4.8 shred are ooc shred. A shred cost 42 energy, so 0.01*0.2*42*9.8 = 0.82 energy every minute Six: less ooc hits = energy wasted 4.8 shred are ooc shred. An ooc shred miss is completly wasted. A shred cost 42 energy, so 0.01*42*4.8 = 2.02 energy every minute Point six is different and complementary to point two. Now we can sum up energy wasted = 2.01 + 0.54 + 0.82 + 2.02 = 5.4 energy x minute. That extra energy impact to shred damage. As already stated I do 14.6 shred per minute, so -> -1% hit = 5.4/(42*14.6) * 27.4% (shred dps %) = 0.242% Final value Now we can sum up all the dps loss: 0.178% white 0.031% dots (GCDs) 0.242% shreds,ooc, etc (energy) 0.45% total 0.45% of 6665 = 30 dps = 1% hit -> 1 hit rating = 0.94 dps = 20* 0.94 = 18.8 points very very far my sim data. P.S. To Allev: In my previous sim I simply said that i think that high hit could be explained by FB, as you can see in this post hit it's high also on a theorycraft base and w/o FB. I don't know what's the theory behind rawr and toskks, probably they don't take into account gcd? I dunno. As for SR uptime I have 99%, basically due to GCD (SR expires and you can't refresh it). 3 cps are the minimum but usually I use 4/5 cps, it's not fixed. Last edited by nightcrowler : 03/07/09 at 11:43 PM.

 Elitist Jerks Druid Changes Patch 3.1