Jump to content


Photo

SimulationCraft - for fur and feather wearers


  • This topic is locked This topic is locked
269 replies to this topic

#21 dedmonwakeen

dedmonwakeen

    Bald Bull

  • Members
  • 1,302 posts

Posted 09 March 2009 - 05:23 PM

Nate implemented a energycap uptime:

    Druid_T8_00_55_16_M:
        32.1% : berserking_mh
         4.0% : energy_cap
       100.0% : replenishment
        99.0% : savage_roar

    Druid_T8_00_55_16:
        31.7% : berserking_mh
         3.4% : energy_cap
       100.0% : replenishment
        99.3% : savage_roar
so 3-4% of the time the druids actually are energy capped


Most of capping occurs in the first 15sec. Very very occasionally we hit 100 energy when TF and OOC happen close to each other.

If you run with log=1 you can look at the opening and see if we can solve the energy waste problem.

#22 Allev

Allev

    King Hippo

  • Members
  • 545 posts

Posted 09 March 2009 - 05:25 PM

Edit: To the poster above, it is perfectly fine to use Ferocious Bite during Berserk if you meet certain criteria that are adjusted accordingly. Nightcrowler theorycrafted this list a bit ago. It probably needs to be readjusted for the new 3.1 numbers, but here is the jist of it.



Rip ticks is how many ticks have happened. Rip debuff counter is how many seconds of rip are left. Given that, Ferocious Bite should be used when the rip conditional is met and current energy levels are less than the RIPvsFB value. Of course, these numbers need to be readjusted for the 3.1 patch numbers and for the reduced energy cost of FB during Berserk.


The problem with FB during berserk isn't that it isn't worth it based on Rip-- in fact, if it was guaranteed 35 energy consumed (17.5 during berserk) it would be great all the time.

There are two problems with FB. The first problem comes in during berserk in that latency guarantees you'll consume extra energy. Let's say your latency is 200ms. So when you are able to click the FB button, you lose the base cost (35/17.5) plus energy regenerated in the 200ms. Which makes its damage per energy worth even less.

The second problem is that you generally don't press the button at the right energy amount unless you're REALLY focusing on it, and you're really good. I tend to spam the button, which happens at a frequency of about 5 presses a second-- which means I get an even distribution of from exactly-right to 200-ms-late FBs.

When you're outside of berserk, these things aren't a big deal: you live with wasting under 5 energy. But during berserk, that loss is effectively doubled because of FB's mechanic: wasting 6-10 (or more) energy becomes a poor decision, especially when you can simply FB right after berserk ends and not waste that energy.

#23 ramenchef

ramenchef

    Piston Honda

  • Members
  • 166 posts

Posted 09 March 2009 - 05:49 PM

Ahh, I see the T7 bonus now. Looks like the only thing that seems slightly off is the rake uptime, which seems a tad low.

Something has been bothering me every time I look at the breakdown. I finally figured it out though. Didn't realize to just check the numbers. When comparing the average hit to the average crit, the numbers are off by ~10%. It ranges from a 2.14% crit to a 2.18% crit modifier with the median being 2.166% Am I missing something or shouldn't it be ~2.266% crit modifier? normal damage X 2.0 melee crit modifier X 1.1 crit modifier from talents X 1.03 crit modifier from the meta gem = 2.266% crit modifier.

In regards to the FB during berserk. I have run into situations where, though often rare but very lucky, I get many crits/ooc procs right before I pop TF/Berserk and have SR with 20+ second duration, rip freshly applied, and 5CPs. In this situation, I believe that as long as you follow the guidelines of RipvsFB DPE, FB should be used as long as you have burned off a lot of the energy with shreds, but I digress.

I will definitely look at why energy seems to be capping early on when I get home from work. I can only see it happening, as you said, when OOC procs right as you pop Tiger's Fury.

Edit: To reiterate about the level based AP, not only do you get 150% of your level as AP from Predatory Strikes, you also get double your level as AP from simply shifting into cat form, or in this case, 160 AP.

Also, does adding Mark of the Wild do anything when added to the initial set of actions when taking consumables? I noticed it was in the Moonkin Simcraft file. Are all raid buffs assumed as well? (mark, kings, might, UR, SoE, etc) Finally, I was looking at the wiki and saw an option to set the GCD. Is this assumed 1.0 for rogues/feral druids or do I need to edit the configuration file to set that? I didn't see any major change in DPS when set to 1.0 or 1.5.

#24 Starfox

Starfox

    King Hippo

  • Members
  • 578 posts

Posted 09 March 2009 - 06:57 PM

Ahh, I see the T7 bonus now. Looks like the only thing that seems slightly off is the rake uptime, which seems a tad low.

Something has been bothering me every time I look at the breakdown. I finally figured it out though. Didn't realize to just check the numbers. When comparing the average hit to the average crit, the numbers are off by ~10%. It ranges from a 2.14% crit to a 2.18% crit modifier with the median being 2.166% Am I missing something or shouldn't it be ~2.266% crit modifier? normal damage X 2.0 melee crit modifier X 1.1 crit modifier from talents X 1.03 crit modifier from the meta gem = 2.266% crit modifier.

Yea, i think the sim only applies the 10% to the crit modifier
Actually it seems to be 1.17 crit bonus currently because Predatory Strikes gets only applied to crit bonus
Edit: Ok, just had to apply to another multiplier, r1744 fixed this :)
Druid_T8_00_55_16 crit_bonus for rake: cb=1.266 b_cb=1.00 b_cm=1.10 p_cm=1.03 t_cm=1.00 b_cbm=1.00 p_cbm=1.00 t_cbm=1.00
cb = crit bonus, so a crit has a 2.266 multiplier now :)

In regards to the FB during berserk. I have run into situations where, though often rare but very lucky, I get many crits/ooc procs right before I pop TF/Berserk and have SR with 20+ second duration, rip freshly applied, and 5CPs. In this situation, I believe that as long as you follow the guidelines of RipvsFB DPE, FB should be used as long as you have burned off a lot of the energy with shreds, but I digress.
I will definitely look at why energy seems to be capping early on when I get home from work. I can only see it happening, as you said, when OOC procs right as you pop Tiger's Fury.

Edit: To reiterate about the level based AP, not only do you get 150% of your level as AP from Predatory Strikes, you also get double your level as AP from simply shifting into cat form, or in this case, 160 AP.

This is already in trunk.

Also, does adding Mark of the Wild do anything when added to the initial set of actions when taking consumables? I noticed it was in the Moonkin Simcraft file. Are all raid buffs assumed as well? (mark, kings, might, UR, SoE, etc) Finally, I was looking at the wiki and saw an option to set the GCD. Is this assumed 1.0 for rogues/feral druids or do I need to edit the configuration file to set that? I didn't see any major change in DPS when set to 1.0 or 1.5.

All those simulations run with optimal_raid=1, this is to easily run simulations with only a few players, they all get all buffs automatically applied.
I think there are even some overwrites for every specific buff.
Hello.
Light the fuse.
For all my homies.
Do not run, we are your friends.
SimulationCraft Druid Guy

#25 ramenchef

ramenchef

    Piston Honda

  • Members
  • 166 posts

Posted 09 March 2009 - 09:44 PM

From what I can tell, feral dps is well over the top with the latest changes. I would estimate that the latest build of simulationcraft is showing near 7k dps for ferals. The question then becomes if ferals fall back to early TBC era where they were good early on but other classes/specs scale ahead in future tiers.

#26 Erdluf

Erdluf

    Great Tiger

  • Members
  • 968 posts

Posted 10 March 2009 - 05:28 PM

Just did some simulationcraft r1754 runs, using the druid_t8 file (this is actually t7 gear with 3.1 rules)

SF rotation means Wrath just to proc Eclipse.
Wr rotation means SF just to proc Wrath. No extra SF casts to extend MF.

SF rotation: 5805 DPS
Wr rotation: 5692 DPS

I removed 200 haste from each:

SF: 5630
Wr: 5724

I believe simulationcraft uses a 0.1s lag, with double-lag charged for GCD clipping. Still, I was a bit surprised to see Wr DPS actually go up.

With the original gear modified -200 Haste, +200 Crit (which seems plausible. I don't pretend to know where that gear is though):

SF: 5791
Wr: 5905

which would confirm that SF scales better with haste than with crit. It does appear that with low-haste, high-crit gear, you could make a strong argument for a Wrath rotation.

In all of the runs (fully raid buffed), SF mana was fairly stable. Wrath mana always lasted the entire five minutes, but it was being depleted.

#27 dedmonwakeen

dedmonwakeen

    Bald Bull

  • Members
  • 1,302 posts

Posted 10 March 2009 - 06:55 PM

From what I can tell, feral dps is well over the top with the latest changes. I would estimate that the latest build of simulationcraft is showing near 7k dps for ferals. The question then becomes if ferals fall back to early TBC era where they were good early on but other classes/specs scale ahead in future tiers.


Kitty dps has crossed 7k in simulationcraft now...... I just fixed a bug related to the gcd while in cat form......

#28 charriu

charriu

    Piston Honda

  • Members
  • 147 posts

Posted 10 March 2009 - 08:03 PM

Kitty dps has crossed 7k in simulationcraft now...... I just fixed a bug related to the gcd while in cat form......


At least it's still behind the HaT-Rogue... We'll see some nerfs in the next patch, I guess.

#29 Starfox

Starfox

    King Hippo

  • Members
  • 578 posts

Posted 10 March 2009 - 09:17 PM

Kitty dps has crossed 7k in simulationcraft now...... I just fixed a bug related to the gcd while in cat form......

It was crossing 7k for me even before that that gcd fix, but now it is at 7.3k, only beaten by the HaT+Envenomglyph rogue.
I don't know, but the damage of rip is a bit ridiculous from my point of view.
Rip: ~24% of total damage, but only used every ~30sec
When I have it with rip, a question to ferals: Is the critchance on dots reduced vs bossmobs? Is there any testing on this? Because right now rip/rake dot-crit% do not suffer from this reduction.

Here are the numbers with the fixed gcd for catform
SimulationCraft for World of Warcraft build 3.1.0 (iterations=6666, max_time=300, optimal_raid=1)


DPS Ranking:
  14549 100.0%  Raid
   7333  50.4%  Druid_T8_00_55_16
   7225  49.6%  Druid_T8_00_55_16_M

Player=Druid_T8_00_55_16  DPS=7332.6 (Error=+/-5.1 Range=+/-752)  DPR=598.4  RPS=12.3/12.0  (energy)
  Core Stats:  strength=81  agility=1165  stamina=1085  intellect=115  spirit=135  health=18472  mana=3746
  Spell Stats:  power=0  hit=10.1%  crit=16.7%  penetration=0  haste=10.7%  mp5=0
  Attack Stats:  power=6771  hit=8.1%  crit=29.1%  expertise=6.2  penetration=0  haste=13.9%
  Actions:
    ferocious_bite        Count=  8.5|34.4sec  DPE=15732| 6%  DPET=13678  DPR= 257.5  Miss=0.0%  Hit=6840  Crit=15968|18761|97.6%  Dodge=0.3%
    melee                 Count=434.7| 0.7sec  DPE= 1372|27%  DPET= 2009  DPR=   1.$  Miss=0.0%  Hit= 890  Crit= 2021| 2274|47.6%  Glance= 668|24.0%  Dodge=0.3%
    rake                  Count= 27.3|11.1sec  DPE=13857|17%  DPET=12049  DPR= 422.3  Miss=0.0%  Hit= 791  Crit= 1796| 2182|47.4%  Dodge=0.3%  TickCount=80  Tick=2576  CritTick=5845|6109|52.4%
    rip                   Count=  9.7|30.6sec  DPE=51915|23%  DPET=45132  DPR=1882.3  Miss=0.0%  Dodge=0.3%  TickCount=123  Tick=2469  CritTick=5597|5737|52.7%
    shred                 Count= 82.0| 3.6sec  DPE= 6919|26%  DPET= 6016  DPR= 265.3  Miss=0.0%  Hit=4318  Crit= 9793|10350|47.8%  Dodge=0.3%

Player=Druid_T8_00_55_16_M  DPS=7224.8 (Error=+/-5.0 Range=+/-777)  DPR=589.7  RPS=12.3/12.0  (energy)
  Core Stats:  strength=81  agility=1165  stamina=1085  intellect=115  spirit=135  health=18472  mana=3746
  Spell Stats:  power=0  hit=10.1%  crit=16.7%  penetration=0  haste=10.7%  mp5=0
  Attack Stats:  power=6771  hit=8.1%  crit=29.1%  expertise=6.2  penetration=0  haste=13.9%
  Actions:
    ferocious_bite        Count=  9.0|32.6sec  DPE=15712| 7%  DPET=13663  DPR= 257.0  Miss=0.0%  Hit=6845  Crit=15943|18761|97.7%  Dodge=0.3%
    mangle_cat            Count= 26.1|11.6sec  DPE= 4601| 6%  DPET= 4001  DPR= 142.5  Miss=0.0%  Hit=2880  Crit= 6537| 7029|47.3%  Dodge=0.3%
    melee                 Count=434.7| 0.7sec  DPE= 1373|28%  DPET= 2011  DPR=   1.$  Miss=0.0%  Hit= 890  Crit= 2021| 2274|47.7%  Glance= 669|24.0%  Dodge=0.3%
    rake                  Count= 25.1|12.0sec  DPE=14003|16%  DPET=12177  DPR= 421.7  Miss=0.0%  Hit= 791  Crit= 1793| 2182|47.6%  Dodge=0.3%  TickCount=74  Tick=2606  CritTick=5910|6109|52.6%
    rip                   Count=  9.9|30.2sec  DPE=51551|24%  DPET=44832  DPR=1852.7  Miss=0.0%  Dodge=0.3%  TickCount=124  Tick=2469  CritTick=5597|5737|52.5%
    shred                 Count= 62.0| 4.7sec  DPE= 6934|20%  DPET= 6030  DPR= 323.2  Miss=0.0%  Hit=4324  Crit= 9806|10350|47.9%  Dodge=0.3%

Gains:

    Druid_T8_00_55_16_M:
        energy_regen=2961.8
        omen_of_clarity=1051.2
        primal_precision=1.4
        tigers_fury=599.9

    Druid_T8_00_55_16:
        energy_regen=2963.0
        omen_of_clarity=1047.3
        primal_precision=1.4
        tigers_fury=600.0

Procs:

    Druid_T8_00_55_16_M:
        combo_points=142.1|2.1sec
        darkmoon_greatness=7.0|42.5sec
        omen_of_clarity=25.4|11.7sec
        primal_fury=54.0|5.5sec

    Druid_T8_00_55_16:
        combo_points=138.7|2.1sec
        darkmoon_greatness=7.0|42.5sec
        omen_of_clarity=25.3|11.8sec
        primal_fury=52.1|5.7sec

Up-Times:

    Global:
        100.0% : blood_frenzy
        100.0% : mangle
        100.0% : master_poisoner
        100.0% : savage_combat
        100.0% : totem_of_wrath

    Druid_T8_00_55_16_M:
        32.4% : berserking_mh
         0.5% : energy_cap
        100.0% : replenishment
        99.5% : savage_roar

    Druid_T8_00_55_16:
        31.8% : berserking_mh
         0.4% : energy_cap
        100.0% : replenishment
        99.5% : savage_roar

Waiting:
    42.8% : Druid_T8_00_55_16_M
    44.6% : Druid_T8_00_55_16

Waiting skyrocketed from 24% to over 40%, but dps also increased by ~280.
I set regen_periodicity=0.2 for this simulation, pretty weird that with it set to 0.2 the average fight length is very close to 300sec, but when i let it at the default (1.0), the average fightlength is about 280sec.
Hello.
Light the fuse.
For all my homies.
Do not run, we are your friends.
SimulationCraft Druid Guy

#30 ramenchef

ramenchef

    Piston Honda

  • Members
  • 166 posts

Posted 11 March 2009 - 05:26 AM

What is periodicity? if you mean time between energy ticks, it should probably be set to 0.1 due to rogues/druids regenerating energy at a rate of 10 energy/1 second in 1 energy intervals. I'd actually be interested in seeing if that changes anything.

#31 charriu

charriu

    Piston Honda

  • Members
  • 147 posts

Posted 11 March 2009 - 11:44 AM

I don't know, but the damage of rip is a bit ridiculous from my point of view.
Rip: ~24% of total damage, but only used every ~30sec


Rip seems a bit high, indeed. But not much...

Rip with all glyphs (shred, rip, sr), 2t7 and :

Per Tick: tickDmg = (39+99*CP [base dmg]+AP*0.06+21*CP [idol])*1.36 [glyphed sr]*1.1 [naturalist]*1.3 [mangle]
Total: ripDmg = tickDmg * (6 [base] + 2 [rip glyph] + 2 [2t7] + 3 [shred glyph]) * (1 + CRIT * 1.1 [pred. instincts] * 1.03 [RED])

So, at 7k AP with 50% crit and 5CP, thats nearly 42k damage over 26s.

#32 dedmonwakeen

dedmonwakeen

    Bald Bull

  • Members
  • 1,302 posts

Posted 11 March 2009 - 11:55 AM

What is periodicity? if you mean time between energy ticks, it should probably be set to 0.1 due to rogues/druids regenerating energy at a rate of 10 energy/1 second in 1 energy intervals. I'd actually be interested in seeing if that changes anything.


Changing periodicity will not change the rate of energy regen...... just how smooth/choppy it is.

The reason it defaults to 1.0sec is for performance reasons...... although I haven't experimented in a while.

#33 Starfox

Starfox

    King Hippo

  • Members
  • 578 posts

Posted 11 March 2009 - 05:00 PM

Changing periodicity will not change the rate of energy regen...... just how smooth/choppy it is.

The reason it defaults to 1.0sec is for performance reasons...... although I haven't experimented in a while.


The weird thing i don't quite get.
regen_periodicity=1.0: fights length is about 284sec
regen_periodicity=0.2: fights length is right at 300sec
Hello.
Light the fuse.
For all my homies.
Do not run, we are your friends.
SimulationCraft Druid Guy

#34 dedmonwakeen

dedmonwakeen

    Bald Bull

  • Members
  • 1,302 posts

Posted 11 March 2009 - 05:51 PM

The weird thing i don't quite get.
regen_periodicity=1.0: fights length is about 284sec
regen_periodicity=0.2: fights length is right at 300sec


Experimenting across a range of values (multiple instances at each value) gave a "reasonably" random spread for fight length...... but the distribution was a bit spiky.....

All iterations are really health-limited despite the max_time setting. The boss health is calculated half-way through the very first iteration. Depending upon how lucky/unlucky the players are during that first half the fight time will be somewhere between 285sec and 300sec. It is never greater than max_time because boss health is calculated by damage done outside of BL and the Execute abilities/talents.

#35 Starfox

Starfox

    King Hippo

  • Members
  • 578 posts

Posted 11 March 2009 - 11:10 PM

Experimenting across a range of values (multiple instances at each value) gave a "reasonably" random spread for fight length...... but the distribution was a bit spiky.....

All iterations are really health-limited despite the max_time setting. The boss health is calculated half-way through the very first iteration. Depending upon how lucky/unlucky the players are during that first half the fight time will be somewhere between 285sec and 300sec. It is never greater than max_time because boss health is calculated by damage done outside of BL and the Execute abilities/talents.

Hm, I think i got my misinterpretation, was using iterations=6666.
As there is no real report on average fightlength, I used the energy_regen, but with higher periodicity it happens more often to cap out on a regen event, and therefor less gets reported. Most probably in both cases the fights were actually of the same length.
Hello.
Light the fuse.
For all my homies.
Do not run, we are your friends.
SimulationCraft Druid Guy

#36 Allev

Allev

    King Hippo

  • Members
  • 545 posts

Posted 12 March 2009 - 03:22 PM

Time now to figure out why SimulationCraft is overvaluing hit/expertise for ferals compared to all the other theorycraft...

#37 dedmonwakeen

dedmonwakeen

    Bald Bull

  • Members
  • 1,302 posts

Posted 12 March 2009 - 03:42 PM

Time now to figure out why SimulationCraft is overvaluing hit/expertise for ferals compared to all the other theorycraft...


By default, the scale factor generation for hit/expertise SUBTRACTS stat value to determine the slope of the dps curve.

I went back and forth on this one..... Input welcome.

#38 Allev

Allev

    King Hippo

  • Members
  • 545 posts

Posted 12 March 2009 - 04:05 PM

That shouldn't be a big deal, hit rating is fairly consistent from having no hit rating to being capped (according to the other theorycraft).

Also-- any chance an option for rake time can be added? I'd like to set up my profile to not shred unless rake has time before it expires (like mangle in the mangle profile).

#39 Starfox

Starfox

    King Hippo

  • Members
  • 578 posts

Posted 12 March 2009 - 04:37 PM

Also-- any chance an option for rake time can be added? I'd like to set up my profile to not shred unless rake has time before it expires (like mangle in the mangle profile).


r1785: rake>=Y, rake<=Y
Post your profiles if you can squeeze out more dps :)
Hello.
Light the fuse.
For all my homies.
Do not run, we are your friends.
SimulationCraft Druid Guy

#40 Allev

Allev

    King Hippo

  • Members
  • 545 posts

Posted 12 March 2009 - 06:58 PM

Is Omen of Clarity getting a double-bonus?

I'm looking into some unexpected Ferocious Bite behavior (limiting the energy it consumes is decreasing DPS), and I may have found the problem with hit rating while doing so.

The relevant code:

// druid_attack_t::cost ====================================================

double druid_attack_t::cost()
{
  druid_t* p = player -> cast_druid();
  double c = attack_t::cost();
  if ( c == 0 ) return 0;
  if ( p -> _buffs.omen_of_clarity ) return 0;
  if ( p -> _buffs.berserk) c *= 0.5;
  return c;
}

// druid_attack_t::consume_resource ========================================

void druid_attack_t::consume_resource()
{
  druid_t* p = player -> cast_druid();
  attack_t::consume_resource();
  if( p -> _buffs.omen_of_clarity )
  {
    // Treat the savings like a energy gain.
    double amount = attack_t::cost();
    if( amount > 0 )
    {
      p -> gains_omen_of_clarity -> add( amount );
      p -> _buffs.omen_of_clarity = 0;
      p -> aura_loss( "Omen of Clarity" );
    }
  }
}

In cost(), it looks like you make the cost of the attack 0. In consume_resource(), it looks like you're adding energy. If you are doing both, then you're doing something wrong! Although I'm not sure if these are two different implementations which are separate for cat and moonkin.

Edit: Crit ticks are still 5% too high.




0 user(s) are reading this topic

0 members, 0 guests, 0 anonymous users