SimulationCraft - for fur and feather wearers - Page 5

vbdotnetrulz · 03/19/09, 1:59 PM

Originally Posted by folderol

I just checked on the latest PTR build and primal gore DoT crits seem to have a 1.5 base crit multiplier instead of the SimulationCraft assumed 2.0 multiplier.

In Cat DPS Spec, I get on a dummy with no RED meta:
rake Tick=2525 CritTick=3872
rip Tick=2444 CritTick=3594.

SimulationCraft predicts:
rake Tick=2486 CritTick=5640|5899|48.9%
rip Tick=2406 CritTick=5454|5591|49.4%

I didn't spend a lot of time testing so a second confirmation could be useful.

EDIT: I've done a second test without trying to maintain SR so just rip/rake/mangle unbuffed on a dummy and the results are more in line with what SC predicts.
Wow Web Stats
So base 2.0 multiplier seems ok, maybe I should upgrade my Recount on PTR

I was so concerned about this when I read your post that I went and tested too hehe.

I have primal gore and Relentless Earthsiege meta. I came to the same results as you did when you retested - the 2.0 multiplier is still there

Starfox · 03/19/09, 2:13 PM

Oh, did anyone verify if primal gore crit% is reduced like other melee crit vs bossmobs?

Starfox · 03/20/09, 4:38 AM

Idol of the Corruptor is a 100% proc chance: Item - Druid T8 Feral Relic - Spell - World of Warcraft, fixed it accordingly
Feral T8 2pc is 2% per tick (Item - Druid T8 Feral 2P Bonus - Spell - World of Warcraft), it shares the same buff with OoC, is now in trunk.
For balance, wrath and starfire now use their real level 80 base damage, as they gain ~4-5 damage per level up.

Allev · 03/20/09, 11:56 AM

So I was looking over code...

It looks like you are treating every attack as if it's on a 1-roll system. However, "yellow" attacks follow a 2-roll system-- first it's hit or miss, then a second roll occurs to verify chance to crit. The original testing for this occurred way back in TBC, by rogues, and it was verified to be consistent with feral druids.

I tested this by setting hit rating and expertise rating to 0; if I do this, I should get a lower crit rating on shreds than I do on melee autoattacks, but they're the same.

Edit: This isn't a game-changer in the sim as at least ferals are close to both the hit and expertise caps through gear. The closer you are to the caps, the less this bug impacts the final numbers.

Edit2: Reference here.

Edit3: So, Primal Precision is triggered only for finishers-- this returns 80% of a finishing move's energy if it misses. However, this mechanic is in effect for all CP-generating attacks: we get 80% of our energy back on shreds, mangles, and rakes if they miss. I did a search in the druid file for "energy" and didn't find any further references to energy refunds on misses. (I only started looking for this mechanic because the other druid sim recently had a bug which overrated hit/expertise in its energy refunds, and if you didn't know the 2-roll mechanic, maybe you didn't know this one.)

Allev · 03/20/09, 12:37 PM

Also, in the Druid_T8_00_55_16 model, you really should change the last action line:

action+=/shred

to

actions+=/shred,rake>=2
actions+=/shred,energy>=60

Doing this guarantees you have more energy to rake when its timer is about to expire, leading to higher rake uptime.

I'm still working on best-case behavior for the mangle profile-- I'm guessing because it is more difficult to land 3 shreds in a cycle if you are worrying about rake uptime, making the glyph worth less and less.

dedmonwakeen · 03/20/09, 12:48 PM

Originally Posted by Allev

So I was looking over code...

It looks like you are treating every attack as if it's on a 1-roll system. However, "yellow" attacks follow a 2-roll system-- first it's hit or miss, then a second roll occurs to verify chance to crit. The original testing for this occurred way back in TBC, by rogues, and it was verified to be consistent with feral druids.

I tested this by setting hit rating and expertise rating to 0; if I do this, I should get a lower crit rating on shreds than I do on melee autoattacks, but they're the same.

Edit: This isn't a game-changer in the sim as at least ferals are close to both the hit and expertise caps through gear. The closer you are to the caps, the less this bug impacts the final numbers.

Edit2: Reference here.

Edit3: So, Primal Precision is triggered only for finishers-- this returns 80% of a finishing move's energy if it misses. However, this mechanic is in effect for all CP-generating attacks: we get 80% of our energy back on shreds, mangles, and rakes if they miss. I did a search in the druid file for "energy" and didn't find any further references to energy refunds on misses. (I only started looking for this mechanic because the other druid sim recently had a bug which overrated hit/expertise in its energy refunds, and if you didn't know the 2-roll mechanic, maybe you didn't know this one.)

Thanks for the heads-up. I was a little disconcerted by two things in the article:
(1) It has not been updated for 3.0
(2) It refers to "some" specials as opposed to "all".
Given it is not a huge game-changer (your Edit1) I'm inclined not to change anything until new testing is done.

Regarding Primal Precision:

static void trigger_primal_precision( druid_attack_t* a )
{
  druid_t* p = a -> player -> cast_druid();
  
  if ( ! p -> talents.primal_precision )
    return;

  if ( ! a -> requires_combo_points )
    return;

  if( a -> result_is_hit() )
    return;

  double energy_restored = a -> resource_consumed * p -> talents.primal_precision * 0.40;

  p -> resource_gain( RESOURCE_ENERGY, energy_restored, p -> gains_primal_precision );
}

So.... No energy back if
(1) You don't have the talent
(2) The attack does not require CPs (ie: not a finisher)
(3) Your attack connected
Looks good, I think.

Originally Posted by Allev

actions+=/shred,rake>=2
actions+=/shred,energy>=60

Doing this guarantees you have more energy to rake when its timer is about to expire, leading to higher rake uptime.

I'm still working on best-case behavior for the mangle profile-- I'm guessing because it is more difficult to land 3 shreds in a cycle if you are worrying about rake uptime, making the glyph worth less and less.

Thanks for the investigation. Always looking to tweak the default profiles......

a civilian · 03/20/09, 1:14 PM

Originally Posted by dedmonwakeen

So.... No energy back if
(1) You don't have the talent
(2) The attack does not require CPs (ie: not a finisher)
(3) Your attack connected
Looks good, I think.

You should get energy back if the attack wasn't a finisher, regardless of whether you have the talent. Any ability that (a) costs energy and (b) does not cost combo points, refunds 80% of its energy cost if it fails to hit. This is a base effect on all non-finisher abilities; the talent only extends the effect to finishers as well.

(Actually I think it only applies to abilities that generate combo points (so no swipe), but that's not really relevant here.)

dedmonwakeen · 03/20/09, 1:39 PM

Originally Posted by a civilian

You should get energy back if the attack wasn't a finisher, regardless of whether you have the talent. Any ability that (a) costs energy and (b) does not cost combo points, refunds 80% of its energy cost if it fails to hit. This is a base effect on all non-finisher abilities; the talent only extends the effect to finishers as well.

(Actually I think it only applies to abilities that generate combo points (so no swipe), but that's not really relevant here.)

Thanks for the clarification! Fixed in r1935.

Allev · 03/20/09, 3:13 PM

Originally Posted by dedmonwakeen

Thanks for the heads-up. I was a little disconcerted by two things in the article:
(1) It has not been updated for 3.0
(2) It refers to "some" specials as opposed to "all".
Given it is not a huge game-changer (your Edit1) I'm inclined not to change anything until new testing is done.

I'll revise what I said there in that it's not a huge game changer for final DPS numbers given the gear you have set up there; however, it undervalues hit rating with regards to scale factors. Essentially, every point of hit rating also increases your crit rating towards yellow attacks.

While the article hasn't changed for 3.0, every piece of rogue/druid theorycraft (to my knowledge) has been working off of the 2-roll system for a long time.

I'll get back here if I can find better references to the exact tests done, with less vague claims of accuracy.

dedmonwakeen · 03/20/09, 3:43 PM

Originally Posted by Allev

I'll revise what I said there in that it's not a huge game changer for final DPS numbers given the gear you have set up there; however, it undervalues hit rating with regards to scale factors. Essentially, every point of hit rating also increases your crit rating towards yellow attacks.

While the article hasn't changed for 3.0, every piece of rogue/druid theorycraft (to my knowledge) has been working off of the 2-roll system for a long time.

I'll get back here if I can find better references to the exact tests done, with less vague claims of accuracy.

Thanks, Allev! I don't want to come across as intransigent. It's just that we have so much work to do and this opens a whole can of worms across the board, so I want to be careful before I dive in......

Allev · 03/20/09, 4:12 PM

A reference from WoWWiki's combat table page:

Melee special attacks by players

Melee special attacks, also called "Yellow-damage" melee attacks due to the color of their on-screen damage numbers, are not resolved the same way as white-damage melee attacks. Yellow damage attacks include all instant attacks (such as Overpower and Backstab) and all on-next-swing special attacks (such as Heroic Strike and Cleave).
The exact mechanism by which special attacks are resolved is not yet known. What is known, however, is that a special attack made against a mob or a shield-carrying player can be blocked and be a crit at the same time. This result is impossible under the mutually-exclusive attack results table described in the preceding section.
Two-roll theory for melee special attacks
In a discussion thread on ElitistJerks, a user named Vulajin kept track of his critical hit rate over many many Backstabs.
He corrected for all possible factors he could think of, and in the end discovered that the rate he got was consistent with his attack being resolved using two die rolls: A first roll to determine whether the attack missed (using the assumed miss chance for the mob targets he was attacking), and a second roll to determine whether an attack that didn't miss was a critical hit (using his tooltip Crit chance adjusted for the mob target's level).
If his data are accurate, and if the game mechanics for yellow-damage attacks haven't changed since the time that discussion thread was written, then there are at least two random numbers generated to determine the outcome of special attacks. It is unfortunate, though, that these tests were done with Backstab and not with a special attack that can be made against a mob from the front, e.g. Sinister Strike or Heroic Strike. Such data would be more useful, because attacks from the front can be parried and blocked.

Link to the wowwiki page is here.

Link to the initial EJ discussion here.

So far this is research just for backstab-- it's hard to sort the "too old" to the "taken as an assumption" jump, but let me be clear: this is the way the math has been discussed for a long time.

Sadly a simple search for "2-roll" turns up too many results to find the original research.

dedmonwakeen · 03/20/09, 4:31 PM

Originally Posted by Allev

So far this is research just for backstab-- it's hard to sort the "too old" to the "taken as an assumption" jump, but let me be clear: this is the way the math has been discussed for a long time.

Sadly a simple search for "2-roll" turns up too many results to find the original research.

Thanks for doing the research. I'll take a look at the architecture and see how we can cleanly make this configurable.

Allev · 03/20/09, 4:41 PM

Yeah, I'm still not convinced by what I've found, although I remember being fully convinced before-- although for reference, here's some research on other research into 2-roll mechanics (not sure what you're doing right now with that):

Spellcasting 2-roll?

Evidence of hunter 2-roll

Starfox · 03/21/09, 10:05 AM

dedmonwakeen pushed a new executable for windows: simcraft-r1937-windows.zip - simulationcraft - windows executable
Regarding druid:

All T8 bonuses available
2pc_balance: +15% (World of Warcraft - English (NA) Forums -> Vanquish the evil!--More Moonkin stuff)
2pc_feral: This actually triggers the same Clearcasting buff as OoC, so the energy gains also use OoC but the proc report lists t8_2pc seperately from OoC
Feral cp-moves refund 80% energy on miss/dodge/etc. (Thanks @ Allev for pointing it out)
Feral/Balance T8 Idols are in
- idol_of_the_corruptor=1
- idol_of_the_crying_wind=1 (i cryed when i saw it)
Starfire: instant=1 to only use it when T8 4pc proced

Hope that list is complete.
Feedback (especially feral, fulltime moonkin myself) is alway appreciated

Wulfmanjack · 03/21/09, 1:12 PM

A question from me about bleed uptime.

   Druid_T8_00_55_16:
        31.1% : berserking_mh
         0.7% : energy_cap
        78.1% : rake
        100.0% : replenishment
        81.6% : rip
        99.5% : savage_roar

That result i get with the last build .

There is a over 4 second gap between those Rips and nearly 2 second gap between each Rake.

   rake                  Count= 27.2|11.2sec  
    rip                   Count=  9.9|30.7sec

I´ve excluded FB out of the rotation and its nearly the same.

   Druid_T8_00_55_16:
        32.3% : berserking_mh
         0.7% : energy_cap
        81.0% : rake
        100.0% : replenishment
        82.2% : rip
        99.6% : savage_roar

Why are these gaps so long?