Moonkin Raiding - Page 112

Whispah · 06/26/09, 12:13 PM

barely lost any dps at all when i started gearing for crit pre ulduar patch, since i got my eclipse faster and more often and also critted more SF's during and after the eclpise duration.

even when gearing for crit id guess that you'll recieve enough haste to get the soft cap without even trying :P

• Hamlet · 06/26/09, 12:23 PM

I'm still highly skeptical of those marginal stat values. We should probably have a deeper discussion of the SimCraft vs. WrathCalcs models to figure out what's going on here at some point. But I'm note even sure that the low value for haste above passes basic sanity checks--haste would be worth more than that based only on the improvement it gives to Starfire spam.

dedmonwakeen · 06/26/09, 2:18 PM

Originally Posted by Starfox

Now that you point it out >:C

But 3.2.0 is giving me headaches with BiS gear list for SimulationCraft, as you don't only have to get near 263, but with the change to eclipse you get a new value to watch out where your one scale factors suddenly loses its value: 400 haste.
That's when you reach 1.0-gcd with NG, CF and 5% ToW.

Starfox, you might be getting bit by huge default deltas used when calculating scale factors. I know it turns your stomach, but consider using smooth_scale_factors=1 so that you can use the smaller scale deltas. When "caps" are involved, we need to be very careful how we perform our scale factor generation.

conghaile · 06/26/09, 3:10 PM

Originally Posted by Arawethion

I'm still highly skeptical of those marginal stat values. We should probably have a deeper discussion of the SimCraft vs. WrathCalcs models to figure out what's going on here at some point. But I'm note even sure that the low value for haste above passes basic sanity checks--haste would be worth more than that based only on the improvement it gives to Starfire spam.

My initial gut instinct made me think I should gear to 400 haste, and then take crit instead to maximize damage in patch 3.2. Doing a quick test on WrathCalcs v3.1.3 gave me self-confirmation, although my math is far from absolute. I looked at the effect 10 haste rating and 10 crit rating has on overall dps.

I assumed 100% eclipse uptime so the increase in DPS on wrath is weighted equivalent to the increase in dps on Starfire. I assumed also 100% NG uptime since it's close to 100% for wrath at any point and near 100% on lunar eclipse. My final assumption was that no idol is equipped, as to save me a couple of test cases at this stage.

My baseline inputs are as follows (I can't seem to attach a file):
Talents: Full points in IIS, IFF, Moonfury, Wrath of Cenarius, EaM, Vengeance, Master Shapeshifter, ImpMoonkin.
Spellpower: 2300
Spell Crit: 364
Spell Hit: 257
Spell Haste: 400
Spirit: 350
Int: 900
Reaction Time: 0.2
Latency: 0.2
T8 2-piece bonus: TRUE
T8 4-piece bonus: TRUE
Chaotic Diamond: TRUE
Blessing of Kings: 10%
MP5: 109 MP5
Spirit: Fel Intelligence
Intellect: Arcane Intellect
Mark of the Wild: 52
Heroic Presence: None
Food: 46 Spellpower
Battle Elixir: Frost Wyrm(F)
Major: Insect Swarm
Major: Moonfire
Major: Starfire
Spell Haste: Wrath of Air
Spell Crit Chance: Moonkin Aura
Spell Crit Debuff: Winter's Chill
Spell Damage Debuff: Earth and Moon
Spell Hit Debuff: Improved FF
All Haste: Improved Moonkin
All Crit Debuff: Totem of Wrath
Replenishment Uptime: 100.00%
Judgement of Wisdom Uptime: 100.00%

Increasing crit rating by 10 ups the eclipse/IS damage of Wrath (Sheet "Wrath Calcs," Grid H-13) from 7143.6 to 7255.8 with a casttime of 1.000 seconds (assumed) and starfire's eclipse/MF average damage (Sheet "Starfire Calcs, Grid C22) from 12083.2 to 12098.2 with a casttime (Sheet "Starfire Calcs" Grid F6) of 2.200 seconds. This equates to a 12.2 dps increase over the solar eclipse and a 6.8 dps increase during lunar eclipse.

Increasing haste rating by 10 changed the casttime of Starfire from 2.200 to 2.195 seconds. This does not affect solar eclipse in any way, but increases starfire dps from 5492.3 dps to 5504.8 dps during lunar eclipse.

So averaging the two phases out, increasing by 10 crit rating increases overall dps by 9.5 dps, while increasing by 10 haste rating increases overall dps by 6.2 dps. That doesn't look like crit outvaluing haste by a ~5:1 ratio as SimCraft is spitting out, but it is still outvaluing it. The biggest caveat to my results I see is the effect of NG uptime on Wrath, given that it varies wildly in game, and is affected more by latency. These effects start to dampen however as crit rating goes up. I'd be also interested in gauging the effects of the SF and Wrath idols, but I need to get back to work.

GKRBER · 06/26/09, 4:25 PM

I need some help. Below is my armory link, please let me know what I am doing wrong. I cannot seem to get past 3400 dps in 25m Ulduar Raids. I use the standard rotation (IS,MF,IFF,W,EC, SF, Re-Dot, SF) etc. I am beginning to think it's my gear or rather incorrectly gemmed. I would also like to know, after a balance druid is hit capped, what would you consider more important - SP, Crit or Haste.

Thank you!

The World of Warcraft Armory

• Hamlet · 06/26/09, 4:31 PM

Originally Posted by GKRBER

I need some help.

http://elitistjerks.com/f73/t59457-m...dated_3_1_3_a/

e: also, http://elitistjerks.com/f73/a45-here...p_me_not_suck/

dedmonwakeen · 06/26/09, 4:47 PM

Scale factor generation via simulation is a painful endeavor. Sure.... It is easy to do (run twice and dDPS/dGear), but it is difficult to do WELL.

In this particular case, the large haste values were pushing Wrath below the GCD. Since SimulationCraft uses different lag values for different scenarios (spell queue, channel, gcd) we were crossing over from "queued lag" to "gcd lag". While this may still be the correct model, it plays havoc with scaling analysis.

When performing scale factor generation we apply a number of methods to reduce unnecessary variation: Allow degradation in accuracy of "absolute" dps values if we can improve accuracy of "relative" dps values. These methods are not present when we vomit out the reams of data. They are only present when calculating the dDPS.

Long story short: We now only use gcd lag for instant casts when performing scale factor generation.

EDIT: Deleted the raw data..... I'm still not happy with the result.

First line is "smooth" scale factors. Second line is "standard" scale factors. Sigh. I need to re-run with large iterations and see if "standard" Haste values are just simulation variance raising its ugly head or if the "smooth" processing has bugs I need to chase down....

Druid_T8_58_00_13_PTR      Int=0.46  Spi=0.50  SP=1.52  Hit=3.15  Crit=0.94  Haste=1.15   (smooth, 1k iterations)
Druid_T8_58_00_13_PTR      Int=0.58  Spi=0.49  SP=1.52  Hit=2.96  Crit=1.11  Haste=0.62   (standard, 10k iterations)
Druid_T8_58_00_13_PTR      Int=0.59  Spi=0.50  SP=1.53  Hit=2.96  Crit=1.12  Haste=0.63   (standard, 100k iterations)

EDIT2: Looks like I have to investigate the "smooth_rng" scaling issues. Bummer.

EDIT3: Well... I'm standing by the "smooth" values. When smooth_scale_factors=0, we use a whopping 250 delta haste rating in order to minimize variance. I believe this pushes us into the haste soft-cap, making the dDPS/dGear formula erroneous. When smooth_scale_factors=1 we reduce variance in alternate ways, enabling a delta of only 100 haste rating which is less prone to soft-cap issues.

Starfox · 06/26/09, 10:26 PM

Originally Posted by dedmonwakeen

Scale factor generation via simulation is a painful endeavor. Sure.... It is easy to do (run twice and dDPS/dGear), but it is difficult to do WELL.

In this particular case, the large haste values were pushing Wrath below the GCD. Since SimulationCraft uses different lag values for different scenarios (spell queue, channel, gcd) we were crossing over from "queued lag" to "gcd lag". While this may still be the correct model, it plays havoc with scaling analysis.

When performing scale factor generation we apply a number of methods to reduce unnecessary variation: Allow degradation in accuracy of "absolute" dps values if we can improve accuracy of "relative" dps values. These methods are not present when we vomit out the reams of data. They are only present when calculating the dDPS.

Long story short: We now only use gcd lag for instant casts when performing scale factor generation.

EDIT: Deleted the raw data..... I'm still not happy with the result.

First line is "smooth" scale factors. Second line is "standard" scale factors. Sigh. I need to re-run with large iterations and see if "standard" Haste values are just simulation variance raising its ugly head or if the "smooth" processing has bugs I need to chase down....

Druid_T8_58_00_13_PTR      Int=0.46  Spi=0.50  SP=1.52  Hit=3.15  Crit=0.94  Haste=1.15   (smooth, 1k iterations)
Druid_T8_58_00_13_PTR      Int=0.58  Spi=0.49  SP=1.52  Hit=2.96  Crit=1.11  Haste=0.62   (standard, 10k iterations)
Druid_T8_58_00_13_PTR      Int=0.59  Spi=0.50  SP=1.53  Hit=2.96  Crit=1.12  Haste=0.63   (standard, 100k iterations)

EDIT2: Looks like I have to investigate the "smooth_rng" scaling issues. Bummer.

EDIT3: Well... I'm standing by the "smooth" values. When smooth_scale_factors=0, we use a whopping 250 delta haste rating in order to minimize variance. I believe this pushes us into the haste soft-cap, making the dDPS/dGear formula erroneous. When smooth_scale_factors=1 we reduce variance in alternate ways, enabling a delta of only 100 haste rating which is less prone to soft-cap issues.

Apparently SimulationCraft had quite some issues when a caster hit the GCD cap with haste and scale factors, this affected balance druids the most, because with NG+CF it is an easy to reach cap for us.
Thanks to everyone who said those values can't be true.

Now I'm getting

  Druid_T8_BiS               Int=0.59  Spi=0.49  SP=1.49  Hit=2.91  Crit=1.13  Haste=0.99  Lag=0.00

Before the haste cap, the value for haste is at ~1.75

I'll try to get a BiS profile for 3.2.0 together and post it here to get some feedback.

Kricketz · 06/26/09, 11:47 PM

Alright, I just started boomkin'ing for my new guild. Herehttp://http://www.wowarmory.com/char...Lair&n=Krickey is a ling to my armory, right now im just below hit cap, with talents, you can view my spec as well. Basically, im pushing around 3.4-3.7k dps and i feel like i could do more. Anyone wanna take a look and see if there is anything i can do to improve? I'm using lunar eclipse procs, have read just about everything i can everywhere and still don't seem to be able to do enough. Also if anyone has some suggestions for the hodir fight, doing hard mode there and barely breaking 5-6k. I wish i had a WWS to post, but my guild doesn't have anyone recording it right now (go figure). But yeah, any suggestions or changes will be appreciated!

• Hamlet · 06/27/09, 12:00 AM

Originally Posted by Kricketz

Alright, I just started boomkin'ing for my new guild. Herehttp://http://www.wowarmory.com/char...Lair&n=Krickey is a ling to my armory, right now im just below hit cap, with talents, you can view my spec as well. Basically, im pushing around 3.4-3.7k dps and i feel like i could do more. Anyone wanna take a look and see if there is anything i can do to improve? I'm using lunar eclipse procs, have read just about everything i can everywhere and still don't seem to be able to do enough. Also if anyone has some suggestions for the hodir fight, doing hard mode there and barely breaking 5-6k. I wish i had a WWS to post, but my guild doesn't have anyone recording it right now (go figure). But yeah, any suggestions or changes will be appreciated!

http://elitistjerks.com/f73/t59457-m...dated_3_1_3_a/

http://elitistjerks.com/f73/a45-here...p_me_not_suck/

-------

I threw up a spreadsheet update with some Vezax changes if anyone wants to tinker with it.

Starfox · 06/27/09, 12:43 AM

Originally Posted by Atheril92

Testing set out now

-50% proc chance on relic (possibly higher)
--Procs so much it's annoying. It's the same animation as the finishing of a healing spell
--Might be bugged? Seems stupidly high proc
-Moonfire crits are 200%
--Do not proc NG
--Do not regen mana from Moonkin Form

The 'new' Eclipse isn't currently implemented

The procchance on the Idol is 70%, so it has no cooldown?
And nice that Vengeance also effects the dot-crits of MF

• Hamlet · 06/29/09, 12:10 PM

SimCraft folks--what are you currently showing as the DPS benefit from 4T8? I haven't gotten around to trying to find a good way to model it in WC, but after playing with it for a weeks, I'm skeptical that it's very strong at all against single targets. I might try the gear setup in my Armory for a bit instead for a significant stat bump instead of the set bonus.

Horao · 06/29/09, 12:50 PM

Originally Posted by Altiris

I was able to log on PTR and test Balance of Power. It is still 4%. Had 8.49% hit from gear, 3% from iFF and 2/2 BoP and got a miss after a few casts. If it were 6%, that wouldn't be possible.

http://img268.imageshack.us/img268/3014/newbop.jpg

The way it's worded currently, it's listed at 6%. Last I checked PTR was no mods, but it'd be good to try 0/2 BoP, throw out 100 casts, do 1/2 BoP, another 100 and finally 2/2 with another 100.

• Hamlet · 06/29/09, 12:53 PM

Originally Posted by Horao

http://img268.imageshack.us/img268/3014/newbop.jpg

The way it's worded currently, it's listed at 6%. Last I checked PTR was no mods, but it'd be good to try 0/2 BoP, throw out 100 casts, do 1/2 BoP, another 100 and finally 2/2 with another 100.

You wouldn't have to do any elaborate statistical test, just gear yourself to 8% from gear + IFF + BoP and see if you miss. Someone tried this above and reported that they did miss, although some confirmation would never hurt.

dukes · 06/29/09, 1:48 PM

Just had a miss on PTR heroic training dummy - faerie fire on the target and 9.07% hit in char sheet (238 hit rating). It's either that they've implemented the text and not the talent (which is possible as Eclipse isn't implemented with the changed version), or they aren't actually changing it.