Mists of Pandaria: All Specs - Page 34

Kaljurei · 03/05/13, 3:32 AM

Apparently GC made a mistake on the Meta Gem numbers. These below values should be the correct spec multiplier.

1.339 Arms
1.152 Fury SMF
1.257 Fury TG

On that note I wish people would stop trolling on the PTR thread asking for 500% AP Executes. We aren't going to get it, period. Or 400% AP for that matter.

Diwata · 03/05/13, 4:45 AM

Been real busy lately, but has anybody gone about stack ranking the trinkets vs H Lei Shen/Xuen (2/2)? Just wanted to check before I go about it.

Berthold · 03/05/13, 6:15 AM

There are too many things in flux with rppm thing right now that I wouldn't recommend to do any theorycrafting. Will update as soon as things are looking more stable.
The next release will have a 250ms traveltime for CS which should allow the 3 + 2 abilities in one CS.

Collision · 03/05/13, 12:38 PM

Originally Posted by Kaljurei

Apparently GC made a mistake on the Meta Gem numbers. These below values should be the correct spec multiplier.

1.339 Arms
1.152 Fury SMF
1.257 Fury TG

On that note I wish people would stop trolling on the PTR thread asking for 500% AP Executes. We aren't going to get it, period. Or 400% AP for that matter.

I'll ask for 600% AP Executes as much as I damn well please. It's obviously not a serious request. :P

As far as the RPPM trinkets go, Fury most likely is in a stable place with RPPM mechanics due to a lack of haste-procs (Flurry isn't too significant), and last I checked, 2/2 Upgraded Lei Shen's H was better than all the new 522 trinkets, equal to the thunderforged 528 Gaze trinket, and worse than the heroic Gaze/Primordius trinkets.

Simcraft overvalues the new RPPM trinkets due to the fact that we simulate our characters 50,000 times, which takes out the RNG from the trinket. The new RPPM trinkets (Especially the ones with really low RPPM values) are going to be frustrating as hell because they will not be reliable.

Caltiom · 03/06/13, 2:37 AM

Originally Posted by Collision

Simcraft overvalues the new RPPM trinkets due to the fact that we simulate our characters 50,000 times, which takes out the RNG from the trinket. The new RPPM trinkets (Especially the ones with really low RPPM values) are going to be frustrating as hell because they will not be reliable.

You can checkout the statistical details ( stddev, or the "stable" range ( quantille(0.95) - quantille( 0.05 ) ) ) to get hard numbers on the variation. SimulationCraft doesn't average everything out, it's just that the most interesting number for most people is mean DPS

And your statement is a bit dangerous. Extending it you could also say that SimC overvalues a Fury Warrior compared to a Shadow Priest, because he has more variation. But that's a very slippery statement without having actual numbers for it, and you quickly need to go into boss mechanics to judge whether a potential low damage streak is worse than higher average dps for your boss-kill attempt. Or in other words: It is as so often a interpretation Issue, not a problem with the theorycrafting tool itself.

Berthold · 03/06/13, 3:34 AM

Regarding variation.. I did an extensive comparison during Cataclysm based on simcraft as well as raidbot data and *all* specs have a very similar variance: It is not only Fire Mages and Fury Warriors (allthough it is very obvious there where the RNG comes from).

Collision · 03/06/13, 3:39 AM

Originally Posted by Caltiom

You can checkout the statistical details ( stddev, or the "stable" range ( quantille(0.95) - quantille( 0.05 ) ) ) to get hard numbers on the variation. SimulationCraft doesn't average everything out, it's just that the most interesting number for most people is mean DPS

And your statement is a bit dangerous. Extending it you could also say that SimC overvalues a Fury Warrior compared to a Shadow Priest, because he has more variation. But that's a very slippery statement without having actual numbers for it, and you quickly need to go into boss mechanics to judge whether a potential low damage streak is worse than higher average dps for your boss-kill attempt. Or in other words: It is as so often a interpretation Issue, not a problem with the theorycrafting tool itself.

I suppose that's a better way to word it. A person's interpretation of these RPPM trinkets should take into account that they are highly variable in their uptimes one pull to the next.

Super-late edit: But a good example of the lower-RPPM trinkets and how they are highly variable is with monks. If you load up the BIS 1H monk and run a 50k interation on it, the difference between 95% and 5% is 35k, compared to the BIS SMF profile that only has a 17k difference between 95% and 5%.

Rallik · 03/07/13, 5:35 PM

I put together an Arms action priority list using the T15H Arms profile and swapping to a 2 OP+3 GCDs CS rotation. I also tweaked reck/BB logic. DPS went up about 3000 and comes out to about 174500. I left out impending victory, heroic throw, and storm bolt from the actions for now to keep it simple. I plan to work on it a bit more and re-include those abilities if any are worth it and then make a pass on the gear selections. I'm also seeing some other issues I can't explain though:

Dragon Roar is hitting weaker than I expect by a significant margin. Is this benefiting from all of the physical damage mods that it should? Specifically, are Enrage, Seasoned Soldier, and Skull Banner affecting it like they should?

Opportunity Strike proc rate is lower than I expect by a fair amount(about 14% lower). With 38.04% mastery and 592.5 uses of abilities that can proc it, we would expect 225.4 OS procs, and there are only 193.7. Is everything that's capable of proccing it doing so?(CS, Dragon Roar, Execute, Heroic Leap, HS, Autos, MS, OP, Slam).

I know there is also a very small ICD on the "last chance to proc" OS that prevents multiple proc chances if more than one ability lands in a small window. This can be seen by using any AoE with a lot of htis or by macroing a GCD ability with HS and hitting something. You will only ever get a single proc, and that proc rate is the same as if you only landed a single hit(demonstrating you only got 1 chance at a roll, an ICD on last chance to proc rather than the proc itself). However, in real world testing of a normal rotation, I usually see about a 90-95% rate of procs compared to the expected value, rather than 86%.

Bloodbath damage also looks lower than I expect. Is every ability that ought to contribute contributing?(CS, DR, Execute, HL, HS, MS, Opportunity Strikes, OP, Slam) It also shouldn't munch or roll any damage like old Deep Wounds did since this kind of mechanic has apparently been fixed to be more reliable this expansion, if that's what's going on.

action priority list:

actions=auto_attack
actions+=/recklessness,use_off_gcd=1,if=target.time_to_die<=18|((target.time_to_die>=186|(target.health.pct<20&(debuff.colossus_smash.remains>=5|cooldown.colossus_smash.remains<=1.5)))&(!talent.bloodbath.enabled|!cooldown.bloodbath.remains))
actions+=/mogu_power_potion,if=(target.health.pct<20&buff.recklessness.up)|target.time_to_die<=25
actions+=/bloodbath,use_off_gcd=1,if=talent.bloodbath.enabled&(target.time_to_die<=18|buff.recklessness.up|target.time_to_die>=75)
actions+=/skull_banner,use_off_gcd=1,if=buff.recklessness.up
actions+=/avatar,use_off_gcd=1,if=talent.avatar.enabled&(buff.recklessness.up|target.time_to_die<=24)
actions+=/use_item,name=gauntlets_of_the_last_mogu,use_off_gcd=1,if=talent.bloodbath.enabled&buff.bloodbath.up
actions+=/berserker_rage,use_off_gcd=1,if=buff.enrage.down
actions+=/heroic_leap,use_off_gcd=1,if=debuff.colossus_smash.up
actions+=/heroic_strike,use_off_gcd=1,if=(((debuff.colossus_smash.up&rage>=70))&target.health.pct>=20)|rage>=rage.max-10
actions+=/execute,if=target.time_to_die<1.5
actions+=/dragon_roar,if=talent.dragon_roar.enabled&talent.bloodbath.enabled&buff.bloodbath.up&buff.bloodbath.remains<1.5&target.health.pct>=20
actions+=/mortal_strike
actions+=/dragon_roar,if=talent.dragon_roar.enabled&talent.bloodbath.enabled&buff.bloodbath.up&debuff.colossus_smash.down&target.health.pct>=20
actions+=/colossus_smash,if=debuff.colossus_smash.remains<1
actions+=/execute,if=debuff.colossus_smash.up|buff.recklessness.up|rage=rage.max-25
actions+=/dragon_roar,if=talent.dragon_roar.enabled&(talent.bloodbath.enabled&buff.bloodbath.up&target.health.pct>=20)|(debuff.colossus_smash.down&target.health.pct<20)
actions+=/slam,if=debuff.colossus_smash.up&(debuff.colossus_smash.remains<1|buff.recklessness.up)&target.health.pct>=20
actions+=/overpower,if=buff.taste_for_blood.stack>=3&target.health.pct>=20
actions+=/slam,if=debuff.colossus_smash.up&debuff.colossus_smash.remains<2.5&target.health.pct>=20
actions+=/execute,if=buff.sudden_execute.down
actions+=/overpower,if=target.health.pct>=20|buff.sudden_execute.up
actions+=/slam,if=rage>=40&target.health.pct>=20
actions+=/battle_shout

Berthold · 03/08/13, 12:03 AM

DR wasn't ignoring armor for a strange reason. now fixed.

BB: taking your horridon kill (gz on the wol :p), I see 4.6% BB. simc is reporting 4.8%. How much would you expect? It is not munching any damage.

In the same log you have ~1.3*#melee_attacks as amount of OS. Simming your character gives me 1.4*melee_attacks. How much would you expect?
Opportunity Strike is proccing from every warrior attack that hits and and is not a proc itself, but the ICD might be wrong. We have a 0.5, but I found several 0.4 in your log. Maybe somebody can provide more data here.( I am bad at scripting the wol output and it takes forever to do manually)

Thanks for the detailed feedback so far!

Rallik · 03/08/13, 2:14 PM

Originally Posted by Berthold

DR wasn't ignoring armor for a strange reason. now fixed.

BB: taking your horridon kill (gz on the wol :p), I see 4.6% BB. simc is reporting 4.8%. How much would you expect? It is not munching any damage.

In the same log you have ~1.3*#melee_attacks as amount of OS. Simming your character gives me 1.4*melee_attacks. How much would you expect?
Opportunity Strike is proccing from every warrior attack that hits and and is not a proc itself, but the ICD might be wrong. We have a 0.5, but I found several 0.4 in your log. Maybe somebody can provide more data here.( I am bad at scripting the wol output and it takes forever to do manually)

Thanks for the detailed feedback so far!

Thanks for the DR fix!

I think my issue with BB was just related to my action list. My logic is a bit bad and BB is getting used early while Reck is still waiting on a CS to trigger, so they don't overlap completely. Fixing that most of the way is showing the higher BB damage I was expecting.

As for OS, I don't think Horridon is too good of a fight to look at since all the AoEs are hitting different numbers of targets all fight, and it's only the first hit of each that can count towards a proc. That could be parsed out to find just spell casts that hit targets or something, which I'll try eventually. I looked at my Jin'rokh parse though and saw 86% OS procs as compared to expected ideal, which is the same rate SC is showing. But, in much longer duration dummy testing, I never saw a proc rate that low and usually saw an average of about 92% of ideal proc rate. That's not too large of a difference, but it affects the value of HS/haste/mastery fairly significantly since HS/haste cause more clashes with the ICD and mastery is obviously directly impacted by diminished proc rates. I'll collect some more raid data from various logs to look into it more.

Seerow · 03/08/13, 9:19 PM

Looking at the priority list:
With the t15 4piece, is it worth breaking up skull banner and recklessness? Right now you have them set to go off together.
If not, is it worth changing the logic to have it save up rage and spamming Slam during the recklessness period to take advantage of the ~95% crit rate you'll have during that time frame? Right now it's set to only Slam during a CS window with recklessness up, it might be worth it even outside the CS window during Reck+Banner.

cuppicake · 03/09/13, 6:51 PM

Are we sure on the BiS trinkets for Arms?

I'm seeing a horrifically low uptime on the Spark buff.

In practice I see the Feather around 18-20%.

Is the crit-based proc and its interaction with Overpower's increased crit chance and relatively high number of uses actually better than we'd expect? Is the relative value of the crit proc better because of the higher uptime on CS we have? Not to mention of course the passive Strength.

EDIT: I should point out my last comment was in reference to Gaze...might not make sense otherwise.

Berthold · 03/10/13, 1:53 AM

in BiS gear, spark also has ~20% uptime, but after the +20% hotfix feathers might be slightly better if you can reforge the hit away. Care to create a BiS set with it and somebody with access to the latest simcraft (like me) can test it.

cuppicake · 03/10/13, 2:46 PM

Originally Posted by Berthold

in BiS gear, spark also has ~20% uptime, but after the +20% hotfix feathers might be slightly better if you can reforge the hit away. Care to create a BiS set with it and somebody with access to the latest simcraft (like me) can test it.

That's the crux of my question...

In practice I've seen under 10% uptime on the Spark buff.

I don't know if people are perennially unlucky or if it's not stimulated correctly.

Berthold · 03/10/13, 2:58 PM

could you provide some logs of longer fights ?