(DW Frost)So I've found another .4% DPS gain in simcraft and another 3-5% cleaving

[Gendori]

So I posted some of this in the Mastersimple thread yesterday, but I've now went a lot deeper and changed some things and with the results I got I felt it deserved a new thread. I managed to put together a DW setup for my character last week (http://us.battle.net/wow/en/characte...engos/advanced) and was researching the rotation and stuff and saw On Mendenbarr's blog about the Mastersimple rotation.

Now, I've played DW before this expansion and I wasn't a huge fan of the rotation because obliterate just felt awkward to me, so I was intrigued, but I didn't want to give up too much DPS so I played with simcraft to find something that works. My first attempt which I posted yesterday was to remove the obliterate Line and add in a line for Plague strike if Unholy>0 just above Plague leech and empower runic weapon. Apparently, with mastersimple, people have been simming it by just replacing the obliterate line with the PS Line, and results are not good. By Prioritizing plague strike LAST below everything except PL and ERW, I can still take obliterate out and have it be only a .6% DPS loss. Well, .6 is nice But I want better, So I added in Death and decay Above PS, So when theres nothing left to do I cast DnD if its off CD then PS, With that extra line now I'm only at a .3% DPS loss.

After that I did some thinking and some testing on a dummy to figure out exactly WHY this would be the case, and I've concluded that with even less than 4k Haste, by taking out obliterate we are still so GCD capped, that 75% of the time we can just ignore unholy runes entirely, and that using PL and ERW to regenerate runes will still be favorable Plague Striking, and Indeed, after doing some LFR's I concluded that when I now add in AMS soaking, that I have enough resources to ignore UH runes most of the time but not enough where diseases fall off. I then went back to Simcraft and Instead of adding the DnD and PL lines just above Plague leech, I put them last below PL and ERW, the results? A .25% DPS GAIN over the default Obliterate line priority. I then thought that it may just be a specific Character thing So I loaded up the default T16H Profile and heres what I got using the following Action Priority List.

actions.single_target=blood_tap,if=talent.blood_tap.enabled&(buff.blood_charge.stack>10&(r unic_power>76|(runic_power>=20&buff.killing_machine.react)))
actions.single_target+=/frost_strike,if=buff.killing_machine.react|runic_power>88
actions.single_target+=/howling_blast,if=death>1|frost>1
actions.single_target+=/unholy_blight,if=talent.unholy_blight.enabled&((dot.frost_fever.remains<3|dot.blood_plague .remains<3))
actions.single_target+=/soul_reaper,if=target.health.pct-3*(target.health.pct%target.time_to_die)<=35
actions.single_target+=/blood_tap,if=talent.blood_tap.enabled&((target.health.pct-3*(target.health.pct%target.time_to_die)<=35&cooldown.soul_reaper.remains=0))
actions.single_target+=/howling_blast,if=!dot.frost_fever.ticking
actions.single_target+=/plague_strike,if=!dot.blood_plague.ticking&unholy>0
actions.single_target+=/howling_blast,if=buff.rime.react
actions.single_target+=/frost_strike,if=runic_power>76
actions.single_target+=/howling_blast
actions.single_target+=/frost_strike,if=talent.runic_empowerment.enabled&unholy=1
actions.single_target+=/blood_tap,if=talent.blood_tap.enabled&(target.health.pct-3*(target.health.pct%target.time_to_die)>35|buff.blood_charge.stack>=8)
actions.single_target+=/frost_strike,if=runic_power>=40
actions.single_target+=/horn_of_winter
actions.single_target+=/blood_tap,if=talent.blood_tap.enabled
actions.single_target+=/plague_leech,if=talent.plague_leech.enabled
actions.single_target+=/empower_rune_weapon
actions.single_target+=/death_and_decay,if=unholy>0
actions.single_target+=/plague_strike,if=unholy>0

Light Movement-Single Target
Default-370103
Modified-371323

a .32% DPS gain over the default priority!

Lets add More Targets

Light Movement-2 Targets
Default-432980
Modified-444370

or a 2.6% DPS gain over default, Clearly when you add more targets it pays to just ignore UH runes entirely.

Light Movement-3 targets

Default-495301
Modified-516784

A 4.3% DPS gain, as we add more targets the gap widens

Light Movement-4 Targets

Default-557751
Modified-589202

A 5.6% DPS gain!

Lets try other fight models

HecticAddCleave-1 Target

Default-511993
Modified 504458

This shows to be a DPS Loss, I am not sure why exactly but it does, maybe someone could chime in and Figure it out

HelterSkelter-1 Target

Default-329839
Modified-332803

Interesting its almost a 1% DPS gain over default

Heavy Movement-1 Target

Default-312424
Modified-313364

A .3% gain

Lets add in some heavier AMS soaking (200k soak every 40 seconds)

Light Movement-1 Target

Default-380873
Modified-381791

Roughly in Line with the default AMS soaking

Lets Add in even More AMS soaking (300k every 40 seconds)

Light Movement-1 Target

Default-381878
Modified-382877

Again In line with previous results, the Amount of AMS soaking seems to have no bearing on the results

How about No AMS soaking?

Light Movement-1 Target

Default-358643
Modified-359205

Still a Single target gain without AMS soaking, although less than with

Conclusion-In every conceiveable instance except for a HecticAddCleave fight (a model which I dont know enough of to figure out why), It is a single and cleave DPS gain to ignore Unholy runes entirely runes EXCEPT when

1)You have 20-40 Runic power and PS and ERW isnt up- In that case you DnD, if that isnt up you PS
2)Your Blood Plague is about to fall off (highly unlikely)

Thoughts?

[Shiira]

Nice work, I was doing something similar already but I'll give this a go tonight.

[Taninsam]

I'm not surprised, I do this kind of rotation during lust and heavy ams-soaking situations. Sometimes you'll get so overflowed with resources, it's best to leave your UH runes like if you were gaming them with RE.

[Darkfriend]

The problem w/ a .4% gain is it is well within the margin of error for sim-c. It may/may not actually occur in a real situation (and even if it does, it'll be hard to tell, sample bias and small sample size, not to mention RNG and other variables.)

[Gendori]

The problem w/ a .4% gain is it is well within the margin of error for sim-c. It may/may not actually occur in a real situation (and even if it does, it'll be hard to tell, sample bias and small sample size, not to mention RNG and other variables.)

You can say this for the standard DW rotation as well, who is to say that THAT'S optimal? We can only trust the modeling that the simcraft guys gave us, and knowing the quality of the guys that did the DK Model, Im putting faith in these numbers

As far as RNG goes. There is actually LESS RNG involved in this rotation because you are not relying on Obliterate procs to prop up your DPS to the desired level. I have tested this extensively in simcraft on max iterations several times, with the default profile never coming out ahead or even close not once. 150000 Iterations is not a small sample size

[Deleted]

''We have to go deeper'' nice work!

[Minoan]

The problem w/ a .4% gain is it is well within the margin of error for sim-c. It may/may not actually occur in a real situation (and even if it does, it'll be hard to tell, sample bias and small sample size, not to mention RNG and other variables.)

We were kinda ignoring 1 UH rune in t14(not really, but yea, it was making life easier)
we already discussed that Obliterate with his 6 grinds is like a Russian Roulette with 6 slots for bullet. Every time something crits, you are taking 1 bullet, ending with 1/6 chance to get Crit Obli and Rime without wasting KM on it.

Interesting topic, looking for more feedback now, was playing like that part of the time. Sometimes pressing Obli was making it clunky, for real.

[Nanaboostme]

Wow i dont even know... is that another language

[Deleted]

Absolutely fantastic work. Thanks.

If anyone's interested in a sim profile of how I actually play (way fewer lines/parameters but absolutely same dps), here it comes:

actions.precombat=flask,type=winters_bite
actions.precombat+=/food,type=black_pepper_ribs_and_shrimp
actions.precombat+=/frost_presence
actions.precombat+=/snapshot_stats
actions.precombat+=/army_of_the_dead
actions.precombat+=/raise_dead
actions.precombat+=/mogu_power_potion

actions=auto_attack
actions+=/antimagic_shell,damage=100000
actions+=/raise_dead
actions+=/mogu_power_potion,if=target.time_to_die<=30|(target.time_to_die<=60&buff.pillar_of_frost.u p)
actions+=/pillar_of_frost
actions+=/blood_fury
actions+=/blood_tap,if=buff.blood_charge.stack>10
actions+=/frost_strike,if=buff.killing_machine.react|runic_power>88
actions+=/soul_reaper,if=target.health.pct<=35
actions+=/howling_blast,if=!dot.frost_fever.ticking
actions+=/plague_strike,if=!dot.blood_plague.ticking&unholy>0
actions+=/howling_blast
actions+=/blood_tap
actions+=/frost_strike,if=runic_power>=40
actions+=/horn_of_winter
actions+=/plague_leech
actions+=/empower_rune_weapon
actions+=/death_and_decay
actions+=/plague_strike

[Mendenbarr]

Well, this complicates things a little.
Nice detective work Gendori!
It's not a gain for everyone, still requires a certain amount of mastery rating, and depends on your weapon ilevel, but I've ballparked 14k mastery as the breakpoint.
Would you mind if I featured your work here in a post on my blog?

[Gendori]

Well, this complicates things a little.
Nice detective work Gendori!
It's not a gain for everyone, still requires a certain amount of mastery rating, and depends on your weapon ilevel, but I've ballparked 14k mastery as the breakpoint.
Would you mind if I featured your work here in a post on my blog?

Go right ahead, glad I could help, you can Use My DK's character name Kengos, as Gendori is My shaman

Also I did more work yesterday and added a line that allows you to Frost strike with 28 Runic Power, as long as you have an unholy rune up, and there was no appreciable loss or gain in DPS

[RedFlame]

Well, this complicates things a little.
Nice detective work Gendori!
It's not a gain for everyone, still requires a certain amount of mastery rating, and depends on your weapon ilevel, but I've ballparked 14k mastery as the breakpoint.
Would you mind if I featured your work here in a post on my blog?

question, does that 14k breakpoint require 4p/gal/thok? anyway to ballpark diff combinations of lacking those (I'm missing thoks and 4p, so im not sure if i should go with mastersimple, but i do like the idea of skipping obl, it feels clunky)

[Schizoide]

The problem w/ a .4% gain is it is well within the margin of error for sim-c.

Assuming he used a reasonable number of iterations, that is not correct. 10k iterations gives a ~0.05% error margin.

Of course the other question is whether the 0.4% is achievable in the actual game, rather than something that only shows up in a sim playing perfectly on patchwerk.

[Nangz]

Assuming he used a reasonable number of iterations, that is not correct. 10k iterations gives a ~0.05% error margin.

Of course the other question is whether the 0.4% is achievable in the actual game, rather than something that only shows up in a sim playing perfectly on patchwerk.

A margin of error works both ways, whos to say the masterfrost priority isn't too high? Its almost completely irrelevant when comparing the two. Mastersimple is theoretically easier in practice as well so it has that going for it.

[Gendori]

Assuming he used a reasonable number of iterations, that is not correct. 10k iterations gives a ~0.05% error margin.

Of course the other question is whether the 0.4% is achievable in the actual game, rather than something that only shows up in a sim playing perfectly on patchwerk.

Yes, all of My Sims were done with at least 25k Iterations, some several times

[Gorfion]

I'm currently in close to full normal gear (http://us.battle.net/wow/en/characte...Gorfion/simple --4 piece, amp and cd reduction trinkets, double normal 1Hs), and this is simming as a very slight dps loss (3k) for me still. I'm sitting at just under 16k mastery. I currently use a similar priority to this sometimes anyway, so I'm going to play around with it a bit more. Thanks for bringing this up!

[Skillslam]

Did something happen to simcraft today? I'm simming haste > mastery > crit for dw frost w/ 4pc and evil eye lolwut

[Darkfriend]

You can say this for the standard DW rotation as well, who is to say that THAT'S optimal? We can only trust the modeling that the simcraft guys gave us, and knowing the quality of the guys that did the DK Model, Im putting faith in these numbers

As far as RNG goes. There is actually LESS RNG involved in this rotation because you are not relying on Obliterate procs to prop up your DPS to the desired level. I have tested this extensively in simcraft on max iterations several times, with the default profile never coming out ahead or even close not once. 150000 Iterations is not a small sample size

No offense, but 150k is rather small considering most other sims are run 50k times many many times over.

Also, the small sample size was related to real world testing, not sims.

- - - Updated - - -

Assuming he used a reasonable number of iterations, that is not correct. 10k iterations gives a ~0.05% error margin.

Of course the other question is whether the 0.4% is achievable in the actual game, rather than something that only shows up in a sim playing perfectly on patchwerk.

Theoretical margin of error =/= actual margin of error. Sim-c is talking about under those exact circumstances. Which never are actually exactly mirrored in any real fight. This drastically increases the margin of error (even though it's still small.)

Sim-C margin is "Under these exact circumstances given 50k iterations there is a .05% differential between results" VS "In a real life scenario where fight length can change by 5% in a single pull and RNG with trinkets, procs, positioning, etc etc etc the differential is X" (I have no clue what the exact or average number would be.)

[Nangz]

No offense, but 150k is rather small considering most other sims are run 50k times many many times over.

Also, the small sample size was related to real world testing, not sims.

- - - Updated - - -



Theoretical margin of error =/= actual margin of error. Sim-c is talking about under those exact circumstances. Which never are actually exactly mirrored in any real fight. This drastically increases the margin of error (even though it's still small.)

Sim-C margin is "Under these exact circumstances given 50k iterations there is a .05% differential between results" VS "In a real life scenario where fight length can change by 5% in a single pull and RNG with trinkets, procs, positioning, etc etc etc the differential is X" (I have no clue what the exact or average number would be.)

I feel you have very little idea as to the actual number of iterations that are simmed when it comes to theorycrafting. Mend has said quotes in the past like 10k for numbers, 25k for stat weights. (http://www.destinysoftworks.com/2013...tioncraft.html) Anything more is really just overkill. This point is further evidenced by Mend's complete 5.4 round up which was done ENTIRELY with 25k iteration sims.

Real world testing doesn't even reach such numbers for comparisons of this nature. For instance, simming festerblight as superior was done with sims and corroborated with a few parses that performed well not anywhere near the thousands you seem to imply is necessary.

The entire discussion about margin of error is irrelevant, it works both ways and sims that run anything more than 25k are almost completely overkill.


For reference, a basic import>sim with varying iterations gives these margins of errors.

100 iterations- .62%
1000 iterations- .2%
10000 iterations- .06%

[Gendori]

Did something happen to simcraft today? I'm simming haste > mastery > crit for dw frost w/ 4pc and evil eye lolwut

Indeed, I am not getting the results I have gotten for the past 3 days either, for my character now it is showing a tiny (.1%) DPS loss and has Haste as my second best stat .3 ahead of crit, whereas before haste has always been my worst stat.

However, I am still showing a decent DPS gain over default for 2-4 targets as usual

- - - Updated - - -

I'm currently in close to full normal gear (http://us.battle.net/wow/en/characte...Gorfion/simple --4 piece, amp and cd reduction trinkets, double normal 1Hs), and this is simming as a very slight dps loss (3k) for me still. I'm sitting at just under 16k mastery. I currently use a similar priority to this sometimes anyway, so I'm going to play around with it a bit more. Thanks for bringing this up!

Indeed, I am seeing the same results as you, so I did some more work today and came up with a new priority that closes the gap for you. Now keep in Mind that Mendenbarr did say that taking obliterate out is dependant on your mastery level and your weapon's item level etc. I changed your Mainhand to Warforged to test this and indeed I came out with a slight DPS GAIN with the priority below. It also squeezes some extra DPS out of the default profile as well, and should be a gain for anyone that sims this priority over my old one. With this Priority, if you are capped on UH runes you will spend a single rune before using HOW,ERW, or PL. By the way speaking of, I have also changed the priority to put ERW, under the HoW line, as I have seen this to be a DPS gain, I have tested every combination of HoW,ERW and PL. So here is my new priority

You can Plug this into simcraft by taking out the Obliterate Line and replacing the priority List below the Frost strike with >40 Runic Power line with the following list

actions.single_target+=/death_and_decay,if=unholy=2
actions.single_target+=/plague_strike,if=unholy=2
actions.single_target+=/horn_of_winter
actions.single_target+=/empower_rune_weapon
actions.single_target+=/blood_tap,if=talent.blood_tap.enabled
actions.single_target+=/plague_leech,if=talent.plague_leech.enabled
actions.single_target+=/death_and_decay,if=unholy>0
actions.single_target+=/plague_strike,if=unholy>0