Jump to content


Photo

SimulationCraft for Hunters (Panda edition)


  • Please log in to reply
63 replies to this topic

#21 Lokrick

Lokrick

    Piston Honda

  • Members
  • 121 posts

Posted 16 October 2012 - 12:57 AM

Is there any reason as to why simulationcraft doesn't apply BM mastery to MoC? ... The DPS I get from it in game is also higher than the DPS simulated while my gear is lower.


It should be (and it used to). Now the code is not being called, so I'm investigating. If you are referring to other abilities than MoC (as having lower-than-expected dps), please let me know specifics?

#22 NoGoal

NoGoal

    Von Kaiser

  • Members
  • 61 posts

Posted 16 October 2012 - 07:00 AM

I only checked MoC as it sims lower than Lynx Rush for simulationCraft. Tested with versions 505-1 && 505-2, increasing mastery (by 5k, all other stats unchanged) didn't significantly influence MoC DPS.

Edit: Just to make sure I did nothing wrong..

  • download new simc-505-2
  • Options -> Iterations == 10000
  • Import -> BiS -> Hunter_BM_T14N.simc
  • Overrides -> talents=http://us.battle.net/wow/en/tool/talent-calculator#Ya!...100
  • Results -> MoC 277331 dmg per execute time
    Hunter_BM_T14N 83451
    Mastery: 44.50% 5552
[TABLE]Damage Stats|DPS|DPS%|Count|Interval|DPE|DPET|Ticks|T-Hit|T-Crit|T-Avg|T-Crit%|T-Avoid%|Up%
a_murder_of_crows|3289|3.9%|5.1|96.22sec|287419|277331|147.8|8290|17638|10459|23.2%|0.0%|32.8%
kill_command|0 (13408)|0.0% (16.1%)|65.5|6.90sec|92098|88854|0|0.0|0|0|0|0.0%|0.0%|0.0%
[/TABLE]

  • Overrides -> gear_mastery_rating=10552
  • Results -> MoC 277379 dmg per execute time
    Hunter_BM_T14N 88357
    Mastery: 61.18% 10552
[TABLE]Damage Stats|DPS|DPS%|Count|Interval|DPE|DPET|Ticks|T-Hit|T-Crit|T-Avg|T-Crit%|T-Avoid%|Up%
a_murder_of_crows|3292|3.7%|5.1|96.22sec|287456|277379|147.7|8289|17642|10466|23.3%|0.0%|32.8%
kill_command|0 (14955)|0.0% (16.9%)|65.5|6.90sec|102704|99087|0.0|0|0|0|0.0%|0.0%|0.0%
[/TABLE]

#23 Lokrick

Lokrick

    Piston Honda

  • Members
  • 121 posts

Posted 16 October 2012 - 07:43 PM

I only checked MoC

There was definitely a bug in Murder of Crows. It will be fixed in the next release. In the meantime, below are comparisons for the level 75 talents. These all assume dire beast and glaive toss (because it was easier to pick those, not because I necessarily advocate them).

Spec|Ability|DPS
bm|crows|86,715
bm|lynx|86,115
bm|blink|85,459
sv|crows|84,633
sv|lynx|84,483
mm|crows|84,293
mm|lynx|83,901
sv|blink|83,592
mm|blink|82,701


Fixing the bug put crows marginally on top for raw single-target damage for all specs. The default MM profile has crows too low in the, uh, pecking order and so focus starves it. In that case it's below lynx. The above ranking is achieved by prioritizing crows above chimera shot (up where lynx rush gets prioritized).

#24 NoGoal

NoGoal

    Von Kaiser

  • Members
  • 61 posts

Posted 17 October 2012 - 09:53 AM

I also noticed that all BM profiles have actions+=/readiness,wait_for_rapid_fire=1 which is a DPS loss over actions+=/readiness,wait_for_rapid_fire=0. You gain ~17 MoC ticks for ~5 less auto+AS+CS shots.

#25 Nooska

Nooska

    King Hippo

  • Members
  • 540 posts

Posted 17 October 2012 - 11:34 AM

If not waiting for readiness to be cast before readiness is a DPS gain over waiting for it, something is wonky with the simulation; RF and BW should be cast together, and readiness should thus always be used after RF (and KC and T75 and T90 and DB).

#26 NoGoal

NoGoal

    Von Kaiser

  • Members
  • 61 posts

Posted 17 October 2012 - 12:41 PM

In either case you trade 1 CD for another. Be it MoC or RF. Knowing that RF isn't a high DPS increase for BM, it can make sense.

It would also probably be a good idea to include a setting to delay the 2nd Stampede until we can stack it with RF.

#27 Lokrick

Lokrick

    Piston Honda

  • Members
  • 121 posts

Posted 18 October 2012 - 07:10 AM

I am working on an improved MM profile. Some of the improvements are for clarity and some are for improved DPS (e.g., following Whitefyst's recommendations).
DPS|Profile
85,100 |new Hunter_MM_T14N
84,412 |current Hunter_MM_T14N


I was surprised to discover that "aimed shot during rapid fire outside of CA" is consistently a 100dps increase, and so I kept it in the current profile (resulting in 17 hard-cast aimed shots vs. 8 aimed shots). The in progress actionlist is shown below. Key elements include:
  • There is a separate "actions.CA" sub-profile to make it easier change the CA behavior. Everything before the "/run_action_list,name=CA" will also happen during CA.
  • glaive toss is no longer included during CA
  • serpent sting and chimera are included during CA
  • steady focus handling is much improved, so it is now up 95% of the time.
  • stampede waits for rapid fire
  • readiness delays for some big CDs. This the primary twiddle that needs work before incorporating into the standard action list, since it's currently very specific to talent choices.

As I understand it, steady focus only needs to be up for the start of the second steady shot, but this is a little more conservative than that (e.g., it does a steady shot pair more eagerly). In particular,
/steady_shot,if=buff.steady_focus.remains<(action.steady_shot.cast_time+1)&!in_flight
tries to start renewing steady_focus as late as possible, just in case. Typically, focus demands start the process earlier. After the first steady shot, there is a much higher priority action to complete the steady shot pair:
/steady_shot,if=buff.pre_steady_focus.up&buff.steady_focus.remains<=5

Note that having one action list in the middle of another doesn't matter; actions lists are grouped by the actual list. All the testing was done with dire beast, murder, and glaive. Other talents are placed logically, but not measured.
Spoiler

This was a fair bit of surgery to the action list and the resulting action sequence seems much improved. And yet the net result is <1k dps difference, or more clearly, < 0.8% difference. I was going to suggest trying to sync up stampede and rapid fire, and then realized I had already included that:
0.000 step_Hunter_MM_T14N performs rapid_fire (100)
0.000 step_Hunter_MM_T14N performs stampede (100)
5.191 step_Hunter_MM_T14N performs readiness (10)
15.619 step_Hunter_MM_T14N performs rapid_fire (69)
196.138 step_Hunter_MM_T14N performs rapid_fire (86)
319.054 step_Hunter_MM_T14N performs readiness (62)
319.054 step_Hunter_MM_T14N performs rapid_fire (62)
319.054 step_Hunter_MM_T14N performs stampede (62)
Are there currently proposed priority improvements that this action list doesn't include?

#28 NoGoal

NoGoal

    Von Kaiser

  • Members
  • 61 posts

Posted 18 October 2012 - 09:44 AM

I'm glad to see MoC fixed in version 505-3!

That said, Lynx Rush is still used in default BM profiles while using MoC is a dps increase. They also include actions+=/readiness,wait_for_rapid_fire=1 which results in a dps loss. Using gear from the best possible dps thread is also a DPS increase.

With the following changes, I'm reaching ~109.3k dps instead of ~106.4k dps - both having optimal_raid=1 - for Hunter_BM_T14H.
optimal_raid=1
professions=leatherworking=600/enchanting=600
talents=http://us.battle.net/wow/en/tool/talent-calculator#Ya!...100
actions+=/readiness,wait_for_rapid_fire=0
head=yaungol_slayers_headguard,id=87004,gems=agile_primal_80agi_160hit_180agi,reforge=exp_crit
neck=choker_of_the_unleashed_storm,id=86953,reforge=mastery_hit
shoulders=waterborne_shoulderguards,id=90505,gems=160agi_60agi,enchant=200agi_100crit,reforge=exp_haste
back=legbreaker_greatcloak,id=86963,enchant=180hit,reforge=mastery_exp
chest=yaungol_slayers_tunic,id=87002,gems=160agi_160agi,enchant=80all,reforge=mastery_haste
wrists=jagged_hornet_bracers,id=86997,enchant=500agi,reforge=mastery_crit
hands=yaungol_slayers_gloves,id=87003,enchant=170exp,reforge=hit_exp
waist=rangers_chain_of_unending_summer,id=87182,gems=160agi_160agi,reforge=exp_hit
legs=yaungol_slayers_legguards,id=87005,gems=160agi_60agi,enchant=285agi_165crit,reforge=haste_crit
feet=monstrous_stompers,id=86985,gems=80agi_160haste_120haste,enchant=140agi,reforge=crit_exp
finger1=painful_thorned_ring,id=86974,enchant=160agi,reforge=mastery_crit
finger2=regails_band_of_the_endless,id=90503,enchant=160agi,reforge=haste_hit
trinket1=relic_of_xuen,id=79328
trinket2=terror_in_the_mists,id=87167
main_hand=taoren_the_soul_burner,id=87168,gems=500agi,enchant=lord_blastingtons_scope_of_doom,reforge=mastery_crit


#29 Whitefyst

Whitefyst

    Great Tiger

  • Members
  • 771 posts

Posted 18 October 2012 - 06:07 PM

I was surprised to discover that "aimed shot during rapid fire outside of CA" is consistently a 100dps increase, and so I kept it in the current profile (resulting in 17 hard-cast aimed shots vs. 8 aimed shots). The in progress actionlist is shown below.


I just want to make a clarification on this statement from my calculations and observations since the statement is not true for all cases and all haste levels. Whether casting AI or not during RFs outisde the CA phase to be beneficial depends on your haste levels and your talents.

With MoC, DB, and GT, I found the tradeoff to be around 1.3s AI cast time before the AI buff. I now find it to be closer to the 1.4s AI cast time. These numbers do not take into account the additional focus cost of using an AI rotation, which reduces the critical cast time some.

Thus, if you have sufficient haste from gear to make the AI cast time to be below about 1.3s, AI should be superior. If your cast time is greater than 1.4s, casting AS should still be superior. The cutoff point should be somewhere between these two levels, where using either should provide similar DPS, with me usually choosing to use AS since it is the safer of the two choices to use.

#30 Lokrick

Lokrick

    Piston Honda

  • Members
  • 121 posts

Posted 18 October 2012 - 08:23 PM

I'm glad to see MoC fixed in version 505-3!

That said, Lynx Rush is still used in default BM profiles while using MoC is a dps increase. They also include actions+=/readiness,wait_for_rapid_fire=1 which results in a dps loss. Using gear from the best possible dps thread is also a DPS increase.


Thanks for the input! I applied everything but the readiness change. I'll have to experiment with that since I noticed a high variance. At 25k iterations, there's only a difference of 20dps, so at the moment it's in the noise. However that reminded me to copy over the stampede rule to sync up with rapid_fire. The results are below (including most of the improvements I discussed elsewhere for MM) and will be in the next release.

DPS|Spec|Talent
107,843|Hunter_BM_T14H|crows
106,222|old_Hunter_BM_T14H|lynx
102,694|Hunter_SV_T14H|crows
102,646|Hunter_SV_T14H|lynx
102,537|Hunter_MM_T14H|crows
102,074|Hunter_MM_T14H|lynx


Spoiler


#31 Lokrick

Lokrick

    Piston Honda

  • Members
  • 121 posts

Posted 18 October 2012 - 09:01 PM

I just want to make a clarification ... With MoC, DB, and GT, I found the tradeoff to be around 1.3s AI cast time before the AI buff. I now find it to be closer to the 1.4s AI cast time. ... Thus, if you have sufficient haste from gear to make the AI cast time to be below about 1.3s, AI should be superior. If your cast time is greater than 1.4s, casting AS should still be superior. The cutoff point should be somewhere between these two levels, where using either should provide similar DPS, with me usually choosing to use AS since it is the safer of the two choices to use.


I ran a version that directly tested against cast_time. I don't incorporate that into the profile because it does not correctly include human reaction time, but it was worth the experiment (note: these are still with dire beast, not murder).

Rule|Aimed Count| DPS
buff.rapid_fire.up|17.8|102,602 or 102681
cast_time<=1.8|17.8|102,610
cast_time<=1.4|15.3|102,677
cast_time<=1.3|7.0|102,619


So the rule does have an affect on teh number of aimed shots, however the dps differences are well within the noise (note the difference between two runs with the current buff-based rule). I threw in the 1.8s case just to compare, and it ends up with the same number of aimed shots as just "when rapid fire is up". I think this all validates your primary rule which is to avoid aimed outside CA except during very high haste (e.g., if your raid uses Bloodlust outside CA, stack rapid fire on it), and suggests that at the current haste levels, aimed usage is indeed a wash and so stick with the safer arcane usage.

#32 NoGoal

NoGoal

    Von Kaiser

  • Members
  • 61 posts

Posted 19 October 2012 - 10:06 AM

There's something odd with scopes too.

Mirror Scope sims higher than LBSoD.

Buffs section:
Dynamic Buffs|Start|Refresh|Interval|Trigger|Up-Time|Benefit
lord_blastingtons_scope_of_doom|17.0|14.0|24.9sec|13.4sec|53.60%|55.60%
mirror_scope|9.0|0.0|51.9sec|51.9sec|20.00%|22.24%


Considering 1800 agi vs 900 crit rating and uptimes, I would expect mirror_scope to be worse.

Now looking at Stat Buff details:
mirror_scope|stat:crit_rating|amount:900.00
buff_lord_blastingtons_scope_of_doom|nothing|nothing


#33 Lokrick

Lokrick

    Piston Honda

  • Members
  • 121 posts

Posted 19 October 2012 - 05:05 PM

There's something odd with scopes too.

Most definitely; thank you! There was a general problem with parsing stat buffs (as opposed to ratings buffs) from the spell data. The fix will be in the next release. Basically, we were missing 3.7k from the scope. There may be comparable chunks of damage missing from other classes (e.g., if they have a pure Int or Agi enchant) MM T14N comparison:

lord_blastingtons_scope_of_doom| 88,712
mirror_scope| 85,651
no_scope|85,014


Edit: I re-ran the full normal mode raid, and only hunters are impacted by the change. The other possible buffs that I spot checked were all manually implemented, and therefore not affected by spell-data parsing.

Edit: Zeherah spent some time with a real (virtual?) mirror scope, and determined that it has no ICD (that had already been confirmed for lbs). Removing the ICD for mirror scope raised its dps by 500. The table above is updated.

#34 Lokrick

Lokrick

    Piston Honda

  • Members
  • 121 posts

Posted 23 February 2013 - 05:59 PM

Thanks in large measure to damien, T15 support is in the dev release of simc. Use ptr=1 to access it. New versions are generated shortly after the PTR client releases, so for example, the current version has the the +50% to chimera but not the adjustment back down to +25%. We are currently incorporating thunderforged items into the T15H profilesbut have only just started playing with gearsets. Thus, there's a single unoptimized gearset used for all the specs:
professions=blacksmithing=600/engineering=600
talents=000211
head=saurok_stalkers_headguard,id=96628,gems=capacitive_primal_80agi_160crit_180crit,reforge=hit_crit
neck=necklace_of_the_terracotta_archer,id=96702,gems=160agi_60exp
shoulders=saurok_stalkers_spaulders,id=96630,gems=160agi_80agi_160crit_120haste,enchant=200agi_100crit,reforge=hit_haste
back=grey_wind_mistcloak,id=95015,gems=80agi_160crit_60haste,enchant=180crit
chest=scales_of_shaped_flesh,id=95034,gems=160agi_160agi_160agi_180agi,enchant=80all,reforge=hit_exp
wrists=bracers_of_mutagenic_fervor,id=96494,gems=160agi,enchant=180agi,reforge=mastery_exp
hands=saurok_stalkers_gloves,id=96627,gems=160agi_160agi_60agi,enchant=170haste,addon=synapse_springs_mark_ii
waist=jingling_fetishgirdle,id=95000,gems=80agi_160hit_80agi_160hit_120agi,reforge=mastery_crit
legs=saurok_stalkers_legguards,id=96629,gems=80agi_160crit_80agi_160hit_120agi,enchant=285agi_165crit,reforge=mastery_exp
feet=treads_of_the_sanguine_volley,id=95009,gems=80agi_160crit_80agi_160hit_120agi,enchant=140agi,reforge=mastery_haste
finger1=radens_swift_seal,id=95021,gems=160agi,reforge=hit_crit
finger2=goresoaked_gear,id=96513,gems=80agi_160crit_60crit,reforge=crit_exp
trinket1=bad_juju,id=96409,reforge=mastery_crit
trinket2=talisman_of_bloodlust,id=96492
main_hand=durumus_baleful_gaze,id=96472,gems=160agi_160agi_60agi,enchant=lord_blastingtons_scope_of_doom,reforge=haste_exp

Without optimization, this puts us at:
185847  33.8%  Hunter_BM_T15H
185337  33.7%  Hunter_MM_T15H
178521  32.5%  Hunter_SV_T15H

Please post here new gearsets, reforgings for different specs, better professions combinations, etc.

#35 Lokrick

Lokrick

    Piston Honda

  • Members
  • 121 posts

Posted 23 February 2013 - 06:57 PM

During PTR and releases, there will often be minor tweaks that won't be reflected into simc until the next client deploy (because that's when they update the client spell data). This is to show how to make such tweaks from the command line so that we can analyze specs in the light of upcoming changes. First, get the current spell data to tweak (note the ptr=1 to get the ptr data):

> simc ptr=1 spell_query=class_spell.name=chimera_shot
SimulationCraft 510-11 for World of Warcraft 5.2.0 PTR (build level 16618)
Name          : Chimera Shot (id=53209) [Scaling Spell (3)]
Class         : Marksmanship Hunter
Resource      : 45 Focus
Spell Level   : 60
Range         : 40 yards
Cast Time     : Ranged Shot
GCD           : 1 seconds
Velocity      : 40 yards/sec
Cooldown      : 9 seconds
Effects       :
#1 (id=45831) : Server Side Script (77)
                Base Value: 0 | Scaled Value: 0
#2 (id=45832) : Normalized Weapon Damage (121)
                Base Value: 0 | Scaled Value: 1869.45 (avg=1.5)
#3 (id=139158): Weapon Damage% (31)
                Base Value: 315 | Scaled Value: 3.15 (avg=0)
Description   : An instant shot that causes $s3% ranged weapon Nature damage plus $<damage>, refreshing the duration of your Serpent Sting and healing you for $?s119447[${$53353m1+$119447m1}][$53353m1]% of your total health. Applies the Hunter's Mark effect.
Variables     : $damage=${$m2}
For damage, we need to override the values for effects #2 and #3. Weapon damage is always computed in terms of the average (which gets multiplied internally for level, etc.), so we will change that for #2. #3 i a constant additional amount, so we will change the base value for it. The attributes are specified by their individual IDs, and are placed after ptr=1 and before the profiles to be simulated. Thus we can resimulate MM with:
> simc.exe ptr=1 override.spell_data=effect.45832.average=1.125 override.spell_data=effect.139158.base_value=236.25 Hunter_MM_T15H.simc

Similarly for the explosive shot and black arrows changes:
simc ptr=1 override.spell_data=effect.45937.average=0.308 override.spell_data=effect.170217.average=0.308 override.spell_data=effect.1289.average=0.1265 Hunter_SV_T15H.simc

These all easily combine into a single run, as long as all the overrides are before the profiles:
> simc.exe ptr=1 override.spell_data=effect.45832.average=1.125 override.spell_data=effect.139158.base_value=236.25 override.spell_data=effect.45937.average=0.308 override.spell_data=effect.170217.average=0.308 override.spell_data=effect.1289.average=0.1265 Hunter_SV_T15H.simc Hunter_MM_T15H.simc Hunter_BM_T15H.simc

 184903  34.0%  Hunter_BM_T15H
 180104  33.1%  Hunter_MM_T15H
 178912  32.9%  Hunter_SV_T15H
Note that this is still with unoptimized gear.

Edit: My estimates of their corrections weren't quite right, and I missed one effect. The current dev version has the correct numbers, resulting in:

> simc.exe ptr=1 Hunter_T14H.simc
126633  33.7%  Hunter_BM_T14H
125272  33.3%  Hunter_MM_T14H
124125  33.0%  Hunter_SV_T14H

> simc.exe ptr=1 Hunter_BM_T15H.simc Hunter_MM_T15H.simc Hunter_SV_T15H.simc
543163 100.0%  Raid
184603  34.0%  Hunter_BM_T15H
181897  33.5%  Hunter_MM_T15H
176662  32.5%  Hunter_SV_T15H

Note that this profile includes some Thunderforged items. Since those are expected to be fairly rare, we plan to remove those from the T15 profiles.

#36 cmwilli

cmwilli

    Glass Joe

  • Members
  • 11 posts

Posted 08 April 2013 - 06:02 AM

I ran a reforge plot comparing crit/haste/mastery and received the following picture. I am unsure how to read it. I was unable to find anything in these forums or the Simulation Craft wiki. From my knowledge about the class, I'm assuming red is higher dps and blue is lower dps. The shape however confuses me. I would have expected it to be a triangle with each point representing all reforging to a single stat. With six points, I would guess that 3 points would represent a single stat and the remaining 3 points would be two stats and none of the 3rd but the graph seems to indicate each pure single stat as a side not a point.

To help me understand could you point out the location of some test points. For example
100% Crit
50% Crit 50% Haste
50% Crit 25% Haste 25% Mastery
33% Crit 33% Haste 33% Mastery
Posted Image

#37 Lokrick

Lokrick

    Piston Honda

  • Members
  • 121 posts

Posted 08 April 2013 - 03:57 PM

I ran a reforge plot comparing crit/haste/mastery and received the following picture. I am unsure how to read it. I was unable to find anything in these forums or the Simulation Craft wiki. From my knowledge about the class, I'm assuming red is higher dps and blue is lower dps. The shape however confuses me. I would have expected it to be a triangle ...


I primarily use the command line so I'm less familiar with the GUI. That's a 2-d representation of a N-d result, and so is necessarily not going to be obvious. However, the triangle vs. hexagon may actually be a presentation bug; can you file an issue for simulationcraft with the details of your sim?

You can find the raw data in the generated CSV file (typically named reforge_chart.csv) and examine it yourself with a spreadsheet. Each variant that gets simulated will be represented in a single row, as in:

crit_rating| haste_rating| mastery_rating| DPS| DPS-Error
-3200| 0 |3200| 113798.9027| 169.919018
-3200| 320 |2880 |113750.723 |169.151617...


Finding the max DPS rows in any spreadsheet is pretty straightforward (e.g., sort or color the table by the dps row).

#38 Lokrick

Lokrick

    Piston Honda

  • Members
  • 121 posts

Posted 13 April 2013 - 07:38 PM

I've updated the hunter module for MoP Patch 5.3 PTR – Build 16837: Hunter Buffs, which includes the increase to Aspect of the Hawk and the initial Blink Strikes spec (conveniently, Blizzard named it slightly differently from Blink Strike). Below is a comparison of Murder with Blink Strikes for the gear set which is reforged Crit>Haste>Mastery.

> simc.exe ptr=1 threads=7 iterations=10000 Hunter_BM_T15H.simc Hunter_MM_T15H.simc Hunter_SV_T15H.simc

208681  34.5%  Hunter_BM_T15H
200393  33.2%  Hunter_SV_T15H
195033  32.3%  Hunter_MM_T15H

> simc.exe ptr=1 threads=7 iterations=10000 Hunter_BM_T15H.simc talents=000221 Hunter_MM_T15H.simc talents=000221 Hunter_SV_T15H.simc  talents=000221

 210071  34.9%  Hunter_BM_T15H
 199016  33.1%  Hunter_SV_T15H
 192809  32.0%  Hunter_MM_T15H


#39 Lokrick

Lokrick

    Piston Honda

  • Members
  • 121 posts

Posted 14 April 2013 - 05:06 AM

Based on the recent discussion of reforging priorities, I tested the haste > crit > mastery reforging (it's currently crit > haste > mastery), and it's a dps improvement for all profiles.

Edit: I added correct ring enchant for the C>H>M numbers for BM and MM, which brought them a little closer. But I also noticed some problems with the profiles, so I'll update when I fix that.

#40 NoGoal

NoGoal

    Von Kaiser

  • Members
  • 61 posts

Posted 14 April 2013 - 09:16 AM

I've noticed that Beast Cleave isn't implemented for BM Hunters. Could you add it?




0 user(s) are reading this topic

0 members, 0 guests, 0 anonymous users