Cataclysm Mage Simulators and Formulators - Page 21

Althor · 06/18/11, 10:03 PM

Originally Posted by Maje

I'm not sure as to why it was implemented the way it was (in simcraft), to me the simplest interpertation of the proc seems the most reasonable as well.

That is, assume some [p] proc chance, the trigger any damage [d] done by the source, the proc itself hits for [d] with no additional benefit from debuffs or buffs. Which leads to a simple [p] percent increase in dps for all classes assuming they tailored [p] for warlocks and frost mages properly.

That is exactly how it is implemented in SimulationCraft unless I've made a mistake.
1) It has a given proc chance for each spec.
2) It performs Arcane damage.
3) It does 100% of the damage that procced it.
4) It cannot crit.
5) It gains nothing from player buffs.
6) It can't double dip on things like Curse of Elements.

Small differences in gain are likely explained by not enough iterations (unless I've got a bug in the code).
Obviously there will be differences for pet classes.
I'll run my own tests with it to make sure the code is okay.

Lhivera · 06/18/11, 11:50 PM

Tested a profile without Improved Hot Streak, and the proc rate was as expected: 187.2 Fireballs at a 46.3% crit rate = 86.7 crits * 0.3 = 26 expected hot streaks. Actual hot streaks = 26.2.

Adding Improved Hot Streak increases procs to 39.4. Seems reasonable. I don't think there's anything wrong here. Remember that some of the "new" single-crit Hot Streaks will be replacing old double-crit Hot Streaks, since they reset the counter (IIRC), so the gain may not be as large as expected.

Althor · 06/18/11, 11:54 PM

I've run my own tests on the legendary proc and it looks like I have some debugging to do.
In the meantime, to make testing easier I have added a player option: dtr_proc_chance=0.0 -> 1.0. If set this will override the hardcoded values. This still requires the legendary to be equipped of course.

With regards to Hot Streak, I have added some proc_t's to keep track of various Hot Streak related values.
Running the Mage_Fire_T12H.simc profile I got the following results in a 450second average duration fight.

test_for_crit_hotstreak: 180.0
crit_test_hotstreak: 83.5
normal_hotstreak: 25.2
improved_hotstreak: 14.2

The Hot Streak buff itself was started 35.5 times and refreshed 3.9 times.
Pyroblast! was cast 35.3 times.

The number of casts of Pyroblast! pretty much matches the number of times the buff was started. You wouldn't expect to see more casts than the number of starts as refreshes to the buff don't add anything and are in effect wasted.

trigger_hot_streak() was called 180 times when may_hot_streak for the spell was set.
180.0 was also the number of Fireballs cast on average which makes sense.

crit_test_hotstreak was 83.5 which means it crit 83.5 out of 180.0 times or a 46.4% crit rate. This matches the crit rate shown for Fireball.

normal_hotstreak procced 25.2 times out of 83.5 crits. This is a 30.2% proc rate which seems close enough to the 30% of the 4pc Tier 12 bonus.

Once normal_hostreak is removed that leaves 58.3 crits. that reach the test for improved_hotstreak.
These need 2 crits in a row that make it this far. We see 14.2 procs of improved_hotstreak meaning that the simulator saw only 14.2 instances of those 58.3 remaining crits that arrived back to back giving 2 in a row and then resetting the stacks to 0.

The questions I have are:
1) If we get 2 crits in a row (that make it to the Improved Hotstreak test) and we only have 1 point in Improved Hotstreak and it fails to proc, should we be resetting our counter of crits-in-a-row back to 0? Or should it be that if the next spell is also a crit that makes it this far that it should try to trigger Improved Hotstreak again?

2) Should we be resetting the number of stacks in general once we trigger either normal Hotstreak or Improved Hotstreak?
i.e. If we got 3 crits in a row followed by a non-crit would that mean a possible 1 proc of Improved Hotstreak or 2?

I hope this information added to the Mage module makes it a bit easier to test out theories etc.

I'll keep looking into the legendary proc issue.

Althor · 06/19/11, 1:43 AM

I've discovered the problem with SimulationCraft's modeling of the legendary proc.
The proc was having a chance to proc itself again. I have removed this.
Now no discharge procs can trigger the legendary proc. Discharge procs are basically all of the items that have a chance to proc an effect. For example the VPLC and of course the legendary itself.

I'm also removing SCHOOL_PHYSICAL and SCHOOL_BLEED from the types of damage that have a chance to proc it.

From my testing I'm seeing Arcane Mages with a 35% proc chance gaining around a 33.6% DPS increase from it etc. Which seems in line when you factor out Mirror Images etc.

Althor · 06/19/11, 1:56 AM

I've just checked in a change to the Legendary proc rates where I've adjusted each *default* chance to proc based off my testing using Tier 11 372 profiles (I will re-do them with Tier 12 profiles once they all become available) to compensate for the damage done by pets etc.

The numbers I've come up with from my testing are:

    case TREE_BALANCE:      chance *= 1.004; break;
    case TREE_ARCANE:       chance *= 1.009; break;
    case TREE_FIRE:         chance *= 1.020; break;
    case TREE_FROST:        chance *= 1.063; break;
    case TREE_SHADOW:       chance *= 1.010; break;
    case TREE_ELEMENTAL:    chance *= 1.029; break;
    case TREE_AFFLICTION:   chance *= 1.061; break;
    case TREE_DEMONOLOGY:   chance *= 1.089; break;
    case TREE_DESTRUCTION:  chance *= 1.066; break;

Note that if you manually provide a proc chance for a player with dtr_proc_chance=<X> it will use that number directly without any adjustment.

*edit* I've also just added a sim-wide default_dtr_proc_chance=<X> (where X is between 0.0 and 1.0). Setting this value override the default value of a 1% proc chance and *will* be adjusted by spec. Again though the player option dtr_proc_chance will override this value for the player in question.

*edit2* And just to round off the options I've added a player-wide dtr_base_proc_chance=<X>. This option if specified will override the sim-wide default_dtr_proc_chance but will be overridden in turn by dtr_proc_chance if given. This option *will* be adjusted by spec (at least until we start hardcoding the actual proc chances after testing it out on Live).

So, to for example sim 4 Arcane Mages, 3 with the legendary staff name in their main_hand slot with different proc chances and 1 without the staff:

ptr=1
Mage_Arcane_T11H.simc
Mage_Arcane_T11H_DTR.simc name=Mage_Arcane_T11H_DTR_1   dtr_base_proc_chance=0.01
Mage_Arcane_T11H_DTR.simc name=Mage_Arcane_T11H_DTR_10 dtr_base_proc_chance=0.10
Mage_Arcane_T11H_DTR.simc name=Mage_Arcane_T11H_DTR_35 dtr_base_proc_chance=0.35

That will give a base proc chance of 1%, 10% and 35% to each of the _DTR profiles and then multiply each by 1.009. So, 1.009%, 10.09% and 35.315%.

Maje · 06/19/11, 7:23 AM

Yes, I botched up the HS formula, the correct one is:
HS = (c*k(c_)+c^2*(1-k(c_)))/(1+c*(1-k(c_)))

Which at .465 crit rate reduces to .2194 chance to proc. per fireball cast ie. 39.45 hotstreaks for 180 fireballs which fits the numbers in Simcraft.

Originally Posted by Althor

The questions I have are:
1) If we get 2 crits in a row (that make it to the Improved Hotstreak test) and we only have 1 point in Improved Hotstreak and it fails to proc, should we be resetting our counter of crits-in-a-row back to 0? Or should it be that if the next spell is also a crit that makes it this far that it should try to trigger Improved Hotstreak again?

I don't think anyone tested how it works, the most reasonable assumption is probably not to reset the counter untill HS actually procs.

Originally Posted by Althor

2) Should we be resetting the number of stacks in general once we trigger either normal Hotstreak or Improved Hotstreak?
i.e. If we got 3 crits in a row followed by a non-crit would that mean a possible 1 proc of Improved Hotstreak or 2?

Yes the counter is reset.

Regarding the Legendary, why are all the different multipliers for classes, from what I understood they only tailored the proc% specifically for pet classes (warlock, ?frost mage) I doubt they actually recalculated it assuming Mirror Image/Shadow Beast/Treants.

On a different topic, the Mirror Image (normal spell) should be revisited, specifically the spellpower coefficients don't fit with Simcraft numbers. I did some tests without/with glyph (frost/fire) and the numbers didn't match, I have the coefficients I came up with on a different machine (I'll update the thread once I get to it).

EDIT: the test was done here: Cataclysm Mage Simulators and Formulators

I did a few more tests and the numbers I came up with are 0.1 for frost bolt, 0.15 for fireball and 0.064 for fireblast. Simcraft seems to model it as 0.75 of the mages spell power multiplied by 0.25 for frost bolt and so on. Those numbers just do not fit with what I'm seeing in testing and on WoL.

The coef. I came up with do seem to fit but they work directly of off 100% of the mage's spellpower, another tidbit that doesn't seem to add up is the base damage for Fire Blast the datamined spell lists it as 88-98 which is far from the real case which seems to be around 140 base damage.

Althor · 06/19/11, 9:08 AM

With regards to the Legendary's proc rates, I went with what I did so that each spec would see the same benefit from the proc as every other spec. It's quite likely that Blizzard won't go to quite that level of compensation for some classes but as their only announced information was that they want it to be equal for all specs I figured this would satisfy the most people.

Once 4.2.0 goes live and people start getting their DTR's we'd of course *greatly* appreciate proc rates from all the caster specs.

As for my questions, we already implement #2 as you've answered. For #1 however we're currently resetting the stacks if IHS fails to proc. It's likely something that could be tested by someone with enough time on their hand. But it's not a major concern at present as most (all?) Fire Mages spec 2/2 IHS.

amped · 06/26/11, 8:06 PM

May be a silly question, but looking at SimulationCraft results I noticed that only the use effect on Shard of Woe is showing up under the gear section. Is it being modeled correctly with 205 mana cost reduction? Arcane seems to be modeling a lot lower with SimulationCraft than with Rawr and other models, and I think SoW might be the problem.

"shard_of_woe,heroic=1,ilevel=379,quality=epic,use=1935haste_10dur_60cd", don't see any Mana Cost reduction included.

Zimeron · 06/26/11, 10:49 PM

Originally Posted by amped;1956099"

shard_of_woe,heroic=1,ilevel=379,quality=epic,use=1935haste_10dur_60cd", don't see any Mana Cost reduction included.

In the output at the bottom you can see flags for unique gear, like set bonuses, certain trinkets, etc. In the arcane profile you will see the below line, telling you that the item has a unique bonus (the mana reduction cost) that's not explicitly stated in the profile, but it is included.

# trinket2=shard_of_woe,heroic=1

When comparing SimC's output to Rawr or other DPS tools, please ensure all settings match up, including fight length, buffs, etc. If there is a large disparity please open a ticket once you have found what is most likely causing the disparity.

Chaosandfury · 07/01/11, 4:46 PM

So I can see Simcraft is modeling arcane on gear that doesn't exist.

The upgraded versions of the valor point gear save the wand, aren't in the game as far as I know. I'm pretty sure they were data-mined, but maybe pulled at the last second?

I imagine this would affect some gear choices are far as the obvious ones go Wrist Wraps of Arrogant doom, instead of Emberlfame Bracers (sorry no links), the bracers were selected because of hit rating, that loss of hit rating effects the whole set.

Thanks.

PterOs · 07/01/11, 9:39 PM

Not only that, but it also has heroic version of reputation cloak(lol). So, in total, 3 items at the moment of writing my post are not available to players on Live:
VP Ring [Crystalline Brimstone Ring], VP Bracers [Emberflame Bracers] and Reputation Cloak [Rippling Flamewrath Drape].

Hinalover · 07/02/11, 12:56 PM

Originally Posted by Chaosandfury

So I can see Simcraft is modeling arcane on gear that doesn't exist.

The upgraded versions of the valor point gear save the wand, aren't in the game as far as I know. I'm pretty sure they were data-mined, but maybe pulled at the last second?

I imagine this would affect some gear choices are far as the obvious ones go Wrist Wraps of Arrogant doom, instead of Emberlfame Bracers (sorry no links), the bracers were selected because of hit rating, that loss of hit rating effects the whole set.

Thanks.

I actually made a comment about this to Althor a few weeks ago indicating that the above mentioned items were not on the PTR's vender lists. He replied "As for upgrading items, I thought the BoE's were going to be able to be upgraded as well? If it turns out they can't we'll adjust the profiles of course."

Morthiras · 07/02/11, 5:12 PM

Today I sat down to perform a few tests with Simcraft and Rawr for my own mage.

First I used Rawr to determine the best sets for each spec. I have considered all the relevant items from a level of 359 to a level of 379 (normal) or 404 (heroic). I ignored items with spirit, because I wanted to make full use of the budget of the items. I've also checked if the heroic version of an item exists on the live servers. I determined lists with [Dragonwrath, Tarecgosa's Rest] separately. I've done the optimizations for my Horde mage with enchanting and tailoring. I left all buffs and settings at the default settings when importing a character.

Furthermore I have adjusted the talent specs a little:

Arcane: Talent Calculator - World of Warcraft

Fire: Talent Calculator - World of Warcraft

Frost: Talent Calculator - World of Warcraft

What surprised me:
- Both for arcane and fire Rawr preferred the 4 set bonus of T11 over T12 for normal modes. T12 seems to be better at a later stage.
- I didn't expect the trinket choices Rawr made, for example Necromantic Focus for frost.

Additionally I tried a different approach in finding an alternative set which is close to BiS: maximizing the average item level. Luckily these do not differ from the first ones very much, except for arcane as [Shard of Woe] makes a huge difference.

You can download the xml files here: Mage_T12.zip

I then used Simcraft to convert the xml files into simc profiles and ran simulations for a 5 minute fight.
My results can be found on the following pages:
Patchwerk: Simulationcraft Results
Helter Skelter: Simulationcraft Results
Patchwerk (no FM): Simulationcraft Results
Helter Skelter (no FM): Simulationcraft Results

In these reports, you can see the profiles that Rawr optimized for me.

Nathyiel · 07/03/11, 3:20 AM

Just a one note, all your profile have Focus Magic, even Arcane. It's true that mage often share it with each other. For comparison purpose, can you remake your test without it

Morthiras · 07/03/11, 9:27 AM

Originally Posted by Nathyiel

Just a one note, all your profile have Focus Magic, even Arcane. It's true that mage often share it with each other. For comparison purpose, can you remake your test without it

I hope this helps:

Patchwerk: Simulationcraft Results
Helter Skelter: Simulationcraft Results