SimulationCraft for Rogues - Page 12

dedmonwakeen · 03/09/09, 4:05 PM

Originally Posted by folderol

With the current SVN revision (r1746), I still get a raid DPS increase of about 2000 whenever I switch to Hemorrhage. Have you committed your changes yet ?

The change is there..... It was in target_t::assess_damage() in sc_target.cpp.

Change event_t::cancel() to event_t::early().

Can you rebuild just to make sure......

Originally Posted by bural

I'm currently reading up on the specifics on SimulationCraft, so please excuse me if I missed something, but your 23-5-43 spec with hemo has Puncturing Wounds instead of Ruthlessness. While I see the point of keeping the development of the spreadsheet introduced in the Free Hat: Deep Subtlety thread and this stricly seperated from this simulation, I gotta admit I'm in general somewhat surprised all the previous backstab builds take Lethaligy over Ruthlessness.

edit: additional comments on Ruthlessness; using a combo point builder over a low combo point finisher is in almost all cases (except for less than 2 combo points and capped out energy) a dps loss. I'm wondering if you already tried optimizing the talents and simply found improving the combo point builder to be better than ruthlessness due to the relatively conservative HAT events/sec

This was an oversight..... I was experimenting back-and-forth between Fist/Hemo and Dagger/Backstab......

I'll re-run.

EDIT: The post has been updated with Ruthlessness over Puncturing Wounds.

PessimiStick · 03/09/09, 4:16 PM

Originally Posted by bural

I'm currently reading up on the specifics on SimulationCraft, so please excuse me if I missed something, but your 23-5-43 spec with hemo has Puncturing Wounds instead of Ruthlessness. While I see the point of keeping the development of the spreadsheet introduced in the Free Hat: Deep Subtlety thread and this stricly seperated from this simulation, I gotta admit I'm in general somewhat surprised all the previous backstab builds take Lethaligy over Ruthlessness.

edit: additional comments on Ruthlessness; using a combo point builder over a low combo point finisher is in almost all cases (except for less than 2 combo points and capped out energy) a dps loss. I'm wondering if you already tried optimizing the talents and simply found improving the combo point builder to be better than ruthlessness due to the relatively conservative HAT events/sec

These are both good points. I'm sure Puncturing Wounds was just an oversight in switching from BS to Hemo. I agree with the logic change for using combo moves though -- in playing with Simulationcraft quite a bit a few weeks ago, I got the best results from:

actions+=/auto_attack/pool_energy/slice_and_dice,min_combo_points=4,refresh_at=3
actions+=/hemorrhage,max_combo_points=2,energy>=80
actions+=/eviscerate,min_combo_points=5

The Rogue_T8_23_05_43 is set to hemo with 3 CP and 80+ energy right now, which I was showing as 70-80 DPS behind max_combo_points=2 using 3.0.9 mechanics.

Edit: Looking at the 4:16 pm update, I think changing the builder logic to only hemo with 2 CP might push it over 8k.

• Maestroquark · 03/09/09, 5:47 PM

Originally Posted by Neto-

I don't think that's what they intended to do, ultimately, with the addition of the Envenom glyph - as much as we like it, it really should be in a deep Assassination tree talent, as the only spec that truly needs the glyph is Mutilate.

Master Poisoner would be the logical place for this, as Envenom stripping Deadly Poisons works directly against the debuff uptime. I could see removes 1/3/5 or even 1/2/3 less DP stacks as the way to go.

dedmonwakeen · 03/09/09, 11:43 PM

SimulationCraft r1754 is available for download..... including all the feedback I've gotten so far. The rapid progress in the last couple of days is completely due to the detailed analysis provided by the community. Thanks again!

FrostBergs · 03/10/09, 11:18 AM

Can you verify that Savage Combat talent is adjusting the attack power correctly because I am not sure it is. I presume there should be a composite_attack_power_multiplier() function in the sc_rogue.cpp containing something like this:

double rogue_t::composite_attack_power_multiplier()
{
  double apm = player_t::composite_attack_power_multiplier();

  apm *= 1.0 + talents.savage_combat * 0.02;

  return apm;
}

Lokkun · 03/10/09, 11:47 AM

Originally Posted by FrostBergs

Can you verify that Savage Combat talent is adjusting the attack power correctly because I am not sure it is. I presume there should be a composite_attack_power_multiplier() function in the sc_rogue.cpp containing something like this:

[code]

I just checked this and it looks ok for me. Savage Combat and Deadliness are applied to the Attack Power. It is covered in the following line of code, you can find it under "void rogue_t::init_base()":

  initial_attack_power_multiplier  *= 1.0 + ( talents.savage_combat * 0.02 +
                                              talents.deadliness    * 0.02 );

FrostBergs · 03/10/09, 11:57 AM

Originally Posted by Lokkun

I just checked this and it looks ok for me. Savage Combat and Deadliness are applied to the Attack Power. It is covered in the following line of code, you can find it under "void rogue_t::init_base()":

  initial_attack_power_multiplier  *= 1.0 + ( talents.savage_combat * 0.02 +
                                              talents.deadliness    * 0.02 );

Yep, ok, I found it.

However...

Hard coding the values here rather than overriding the virtual will cause other methods to miss the "updated" AP. Such is the case with reporting. It will show an incorrect value for the AP there as the multiplier is not being taken into account (the report is simply calling composite_attack_power()).

Changing the report to:

report_tag ? "power="       : "", p -> composite_attack_power() * p -> composite_attack_power_multiplier(),

and overriding the method would be cleaner. This, of course, will not affect the final DPS values and is probably not a huge concern.

dedmonwakeen · 03/10/09, 3:07 PM

Originally Posted by FrostBergs

Hard coding the values here rather than overriding the virtual will cause other methods to miss the "updated" AP. Such is the case with reporting. It will show an incorrect value for the AP there as the multiplier is not being taken into account (the report is simply calling composite_attack_power()).

Changing the report to:

report_tag ? "power="       : "", p -> composite_attack_power() * p -> composite_attack_power_multiplier(),

and overriding the method would be cleaner. This, of course, will not affect the final DPS values and is probably not a huge concern.

Thanks for the review. Setting initial_attack_power_multiplier in init_base will affect the base version of composite_attack_power_multiplier(). In general, the purpose of the composite_xxx() methods are to provide the class module a way of dynamically changing the results. Where possible, I try to pre-calculate...... but I must admit that doesn't happen as much as I like.

Excellent catch on the reporting problem. As far as I'm concerned, that is a bug. I'll make the change right now.

I'd like to remind everyone that the Rogue module is being supported by bits and pieces of a couple people..... If someone wwants to jump in and make that their focus, just send me an email. Provided you can live within a few simple rules, I'm very open about extending commit access.....

Todesunke · 03/11/09, 11:36 AM

Player=Rogue_T8_51_13_07 DPS=6642.0 (Error=+/-9.5 Range=+/-535) DPR=408.9 RPS=16.2/16.0 (energy)
Core Stats: strength=113 agility=1289 stamina=1105 intellect=37 spirit=65 health=18616 mana=0
Spell Stats: power=0 hit=10.1% crit=14.2% penetration=0 haste=10.7% mp5=0
Attack Stats: power=3742 hit=8.1% crit=29.3% expertise=6.1 penetration=0 haste=10.7%
Actions:
deadly_poison Count= 0.0| 0.0sec DPE= inf|12% DPET= inf DPR= inf Miss=nan% TickCount=95 Tick=2350
envenom Count= 20.9|13.4sec DPE= 7163| 8% DPET= 6230 DPR= 204.6 Miss=0.0% Hit=4851 Crit=10378|15112|42.2% Dodge=0.4%
instant_poison Count=162.7| 1.8sec DPE= 2106|18% DPET= inf DPR= inf Miss=0.0% Hit=1762 Crit= 2830| 3448|32.2%
melee_main_hand Count=408.9| 0.7sec DPE= 1010|22% DPET= 1437 DPR= inf Miss=12.9% Hit= 761 Crit= 1635| 2108|45.3% Glance= 572|24.0% Dodge=0.4%
melee_off_hand Count=408.9| 0.7sec DPE= 762|16% DPET= 1083 DPR= inf Miss=12.9% Hit= 572 Crit= 1230| 1609|45.6% Glance= 430|23.9% Dodge=0.4%
mutilate Count=119.4| 4.9sec DPE= 2893|18% DPET= 5023 DPR= 110.5 Miss=0.0% Hit=1457 Crit= 3626| 4880|66.5% Dodge=0.4%
rupture Count= 13.7|20.7sec DPE= 9262| 7% DPET= 8053 DPR= 370.5 Miss=0.0% Dodge=0.4% TickCount=106 Tick=1203

I have two questions.
1.) Wouldn't it be better to play a rupture free cycle due to the evenom glyph (assuming there is a bleeding effect from another partymember) and take the two points out of blood splatter into quick recovery?
2.) Wouldn't it be better to use SR instead of WD (mainhand) due to the PPM change?
I've never worked with SimulationCraft before, that's why I'm afraid of making mistakes and haven't tested it by myself.

dedmonwakeen · 03/11/09, 12:26 PM

Originally Posted by Todesunke

I have two questions.
1.) Wouldn't it be better to play a rupture free cycle due to the evenom glyph (assuming there is a bleeding effect from another partymember) and take the two points out of blood splatter into quick recovery?
2.) Wouldn't it be better to use SR instead of WD (mainhand) due to the PPM change?
I've never worked with SimulationCraft before, that's why I'm afraid of making mistakes and haven't tested it by myself.

#1: While there are some specs that can drop Rupture, the damage-per-energy and damger-per-gcd of Rupture is too strong to overcome in this spec. Caveat: Someone more experienced than I might be able to tweak the action list conditionals to get more out of Envenom.

#2: Good call. This definitely a win. I'll make the change now.

PessimiStick · 03/11/09, 5:19 PM

I found a possible bug in r1765. It seems like Envenom is not proccing MH poisons.

    envenom               Count= 99.9| 2.8sec  DPE= 7634|34%  DPET= 6107  DPR= 218.1  Miss=0.0%  Hit=5351  Crit=11466|15451|37.9%  Dodge=0.6%
    hemorrhage            Count= 46.3| 6.1sec  DPE= 2616| 5%  DPET= 2093  DPR=  74.7  Miss=0.0%  Hit=1689  Crit= 4199| 6510|37.3%  Dodge=0.6%
    instant_poison        Count=229.5| 1.2sec  DPE= 1763|18%  DPET=    1  DPR=   1.$  Miss=0.0%  Hit=1496  Crit= 2403| 2864|29.5%
    melee_main_hand       Count=212.2| 1.3sec  DPE= 1822|17%  DPET= 1373  DPR=   1.$  Miss=14.8%  Hit=1503  Crit= 3221| 4455|37.3%  Glance=1128|23.9%  Dodge=0.6%
    rupture               Count= 13.5|21.4sec  DPE=12438| 7%  DPET= 9952  DPR= 497.5  Miss=0.0%  Dodge=0.6%  TickCount=129  Tick=1301

        96.9% : envenom
        99.5% : slice_and_dice

With Imp. Poisons and 96.9% Envenom uptime, IP procrate should be in the neighborhood of 93.4%
(100 + 46 + 212 + 13) * .934 = 346.5 IP procs expected.
(46 + 212 + 13) * .934 = 253 IP procs expected if Envenom doesn't count.
Actual procs of 229.5 is low for both estimates, but it's substantially off from the real expectation. In fact, throwing another 100 IP procs into the results picks up nearly 650 DPS, which would put this spec firmly into the realm of "hotfix worthy", IMO.

nuoHep · 03/11/09, 5:27 PM

Misses & dodges. The sim (as I understand) reports actual swings made in the Count field.
You can actually check it by creating a profile with capped hit and exp and a 3.2 mainhand to see if there is something fishy with poison procs. They will line up perfectly.

I also wanted to say that the report on 'penetration' in the melee section is broken. To actually change it from 0 to 1 I had to put 1k+ arpen rating on the gear.

Ticia · 03/11/09, 6:40 PM

Originally Posted by nuoHep

Misses & dodges. The sim (as I understand) reports actual swings made in the Count field.
You can actually check it by creating a profile with capped hit and exp and a 3.2 mainhand to see if there is something fishy with poison procs. They will line up perfectly.

I also wanted to say that the report on 'penetration' in the melee section is broken. To actually change it from 0 to 1 I had to put 1k+ arpen rating on the gear.

Misses and dodges actually makes it even more clear that Envenom isn't proccing poisons. Using the numbers from PessimiStick's run, we show:

99.3 Envenom Hits
46.0 Hemo Hits
179.6 Mainhand Hits
13.4 Rupture Hits

335.3 Total Hits

If the IP Poison Proc Rate is 93.4%, there should be 313.2 Poison Procs.

However, if we assume Envenom is failing to proc poisons, then we end up with:

46.0 Hemo Hits
179.6 Mainhand Hits
13.4 Rupture Hits

239 Total Hits
Again, at 93.4% Proc Rate, we get 223.2 Poison Procs, which is very close to the observed 229.5 Procs.

Edit: I actually calculate the IP Proc Rate at 94.5, which gets us to 316.9 Procs Ideal and 225.8 Procs if Envenom is not proccing MH poison.

nuoHep · 03/11/09, 8:08 PM

My bad, yeah. I was actully testing it on Wound Poisons several days ago (because I was similarly puzzled by the numbers).

Looking through the generated log for Rogue_23_05_43 it looks like Envenom is not proccing instant poison at all.

32.79    Rogue_T8_23_05_43 schedules execute for envenom
32.79    Rogue_T8_23_05_43 performs envenom
32.79    Rogue_T8_23_05_43 envenom hits Fluffy Pillow for 10943 nature damage (crit)
32.79    Rogue_T8_23_05_43 gains 25 (25) energy from relentless_strikes (35)
32.79    Player Rogue_T8_23_05_43 loses Combo Points (5)
32.79    Player Rogue_T8_23_05_43 gains Combo Points (1)
32.95    Rogue_T8_23_05_43 performs melee_main_hand

The last envenom was performed at 31.49 so the buff is there.
SnD is also not proccing it (though I'm not really sure if it actully does, I believe it does)

22.30    Rogue_T8_23_05_43 schedules execute for slice_and_dice
22.30    Rogue_T8_23_05_43 performs slice_and_dice
22.30    Rogue_T8_23_05_43 gains 25 (25) energy from relentless_strikes (60)
22.30    Player Rogue_T8_23_05_43 loses Combo Points (5)
22.49    Rogue_T8_23_05_43 performs melee_main_hand

The last envenom was performed at 19.54
Though it does proc off rupture

47.31    Rogue_T8_23_05_43 schedules execute for rupture
47.31    Rogue_T8_23_05_43 performs rupture
47.31    Rogue_T8_23_05_43 gains 25 (25) energy from relentless_strikes (75)
47.31    Player Rogue_T8_23_05_43 loses Combo Points (5)
47.31    Rogue_T8_23_05_43 performs instant_poison

Also, the first envenom is performed 6.97 seconds into simualation. That's basically ~2 hemos, 2 finishers and a couple melee swings without the buff.

As a diclaimer: the log was done with a 3.2 speed mainhand, 1500 hit rating and 500 expertise rating. So basically each swing/ability under envenom buff should've procced mh poison.

• Vulajin · 03/11/09, 8:29 PM

SnD does not proc poisons. Envenom and Rupture definitely proc poisons.