[Feral] Mew Simulator - Page 2

Zarc- · 01/08/11, 4:38 PM

Originally Posted by tangedyn

Were you using Mew Formulation or Simulation?
Mew Simulation should give you the most accurate result. Mew Formulation is known to have results deviating from the Simulation if your gear level is not high enough to ensure ~100% uptime on Rip and Savage Roar (in which case Haste will be better than Crit)
AFAIK Rawr.Cat's model was forked off from an old version of Toskk's calculator and has been maintained by someone else. Toskk is currently still actively maintaining the Mew Formulation and has done a lot of good work fixing stuff to be more accurately matching the results of the Simulation. I cannot comment on how accurate Rawr.Cat is because I have not reviewed that model. Mew's model is actively peer-reviewed between Toskk, Yawning and myself.

Anyway, Haste is generally better than Crit at lower levels of gear, but Crit quickly overtakes Haste at some point... maybe ilevel ~346.
The 4.0.6 will probably make Haste the better stat though.

Thank you for your reply. The simulation (not the formulation) told me crit was better with my crit->mastery reforged gear, with which I got the best DPS in-game while testing on a dummy. However, now I did another simulation where I added 160 crit rating and removed 160 haste rating from the stats I used for the simulation (aka aprox what I had when I had my gear haste->mastery reforged). Then the simulation displayed haste as slighly better than crit at a value of 0.65 for haste vs. 0.63 for crit. So I suppose I am at some borderline here gearwise. So it seems Rawr and Mew agree with eachother more than I thought, at least at my gearlevel.

One can only hope that the patch makes it easier to determine which stat is better, and that we don't end up with 3 stats that are super equal. Keep up the good work on Mew!

Windchilla · 01/12/11, 11:36 AM

Originally Posted by tangedyn

Anyway, Haste is generally better than Crit at lower levels of gear, but Crit quickly overtakes Haste at some point... maybe ilevel ~346.
The 4.0.6 will probably make Haste the better stat though.

I'm actually simming the opposite. Crit was slightly better on my gear before hitting full raid epics and now Haste has pulled ahead slightly. This is using all of the raid buffs available however, which is the situation I'm raiding with, so I'm sure those things have an impact on relative stat values.

In short Mastery is much better than both Haste and Crit for me and Haste slightly edges out Crit, but they are very close.

eXcel905 · 01/12/11, 2:39 PM

Hey, great simcraft tool so far helps a lot especially with the pawn scales.

A question I had is how do I compile a version for myself, using a more recent snapshot, after I've done the SVN checkout? I have the JDK installed as well as the JRE.

tangedyn · 01/13/11, 9:24 PM

Originally Posted by eXcel905

Hey, great simcraft tool so far helps a lot especially with the pawn scales.

A question I had is how do I compile a version for myself, using a more recent snapshot, after I've done the SVN checkout? I have the JDK installed as well as the JRE.

Generally we build using Eclipse. Haven't really tried building from command line, I'll have to get back to you on that later.

dedmonwakeen · 01/20/11, 9:51 AM

Originally Posted by zydd

Anyone compared Simulationcraft results (latest beta) with Mew? The difference I got for my feral cat is almost 2k dps (Mew is higher)...not sure why...

Provide links to both outputs and the teams can take a look. Being able to compare multiple tools is crucial to ensure modeling accuracy.

Leafkiller · 01/20/11, 7:20 PM

Based on a question/suggestion from Alaron, I tested a rotation with FB removed until the Blood in the Water phase. I tested two profiles against the default Mew script and four profiles against my script and found that removing FB was either dps neutral or a slight (10-15dps) gain. The worst I saw was a 1 dps loss. The profiles included the premade tier-11 and pre-heroic ones that ship with Mew along with a current profile of my toon (about 2/3rds 359 gear) and a premade tier-11 Worgen I made during beta. While testing my script I also separately tested removing FB during and outside of Berserk and saw the same results for both. I did not test this with a 372 gear profile.

If this proves out to be the case, then our rotation will become simpler, with FB only being used for the "execute" phase in place of Rip. With this change, glyph of FB is a dps loss and should not be used.

indiechixor · 01/21/11, 1:16 AM

Low post count, because I lurk moar.

I recently converted over to feral, so please bear with me (pun intended). So far I am loving this sim, but having difficulties rationalizing the readout.

What is the best way to read the output, and use it towards tweaking your current gear/jem/reforging stats?
i.e. is there an ideal "master" list of what percentages and numbers we should be aiming for? Or is this too subjective?

Thanks for your hard work!

Pyrates · 01/21/11, 4:30 AM

Leafkiller, would you mind sharing the code for your rotation?

Gnorme · 01/21/11, 11:26 AM

Originally Posted by Leafkiller

Based on a question/suggestion from Alaron, I tested a rotation with FB removed until the Blood in the Water phase. I tested two profiles against the default Mew script and four profiles against my script and found that removing FB was either dps neutral or a slight (10-15dps) gain. The worst I saw was a 1 dps loss. The profiles included the premade tier-11 and pre-heroic ones that ship with Mew along with a current profile of my toon (about 2/3rds 359 gear) and a premade tier-11 Worgen I made during beta. While testing my script I also separately tested removing FB during and outside of Berserk and saw the same results for both. I did not test this with a 372 gear profile.

If this proves out to be the case, then our rotation will become simpler, with FB only being used for the "execute" phase in place of Rip. With this change, glyph of FB is a dps loss and should not be used.

I don't see this possible, seeing as using FB is only when ideal situation has arrived where nothing will fall off between that point and when you can build enough combo points to refresh.

Interested to see the code as well.

Is there anyway to account for Dark Intent? I may have missed it but I know thats a tough one, but seems as though I have 3 stack almost 100% of the time.

PS Great tool, keep up the great work, still got my gm asking me how I do so much dmg over all the classes in which in a perfect world should be blowing me out the water. Rogues, hunters, mages, dk's and they still can't beat me. And I owe that to MEW and Fluiddruid.net. Thank you

Leafkiller · 01/21/11, 3:15 PM

My script is posted at "The Fluid Druid" here: The Fluid Druid - View topic - Leafkiller's 4.01 Feral Ovale Script

The script is broken into three sets of five chunks which includes the default script, my script, and a copy of my script that I can experiment on without worrying about making extra copies all the time. This allows me to edit the script easily without having to select a new script in Mew and jump back and forth between different versions. I suggest the following settings for testing this:

   int prelude = 1;
   int globalBuffs = 0;  // this is shared
   int globalDebuffs = 1;
   int mainRotation = 2;
   int filler = 1;

You can then edit the function "getMainExp()" to comment out the two FB lines. You will need a 4.06 ready version of Mew to test this. I typically disable RSV generation and turn on the high resolution timer (both are settings in Model Parameters).

For Dark Intent, there is a setting in Mew (at least the 4.06 ready version) for that under Buffs/Debuffs.

What we are trading off for with this change is slightly higher Rip uptime (95.5% versus 95.1%) and more Shreds (88.8 versus 83.9). Mostly this is Shred scaling compared to FB scaling.

Husyor · 01/27/11, 1:38 AM

So I assume this is already the case and taken into account, but we enter both the STR value and the AP value in caster form, yet STR buffs AP. For the most part, STR doesn't matter since you would use AGI, but for the STR from the chest enchant and I use the +35 STR enchant on gloves and am just trying to make sure that my numbers are coming out as they should. Like I said, I assume its accounted, but I was just curious.

tangedyn · 01/27/11, 3:30 AM

Originally Posted by Husyor

So I assume this is already the case and taken into account, but we enter both the STR value and the AP value in caster form, yet STR buffs AP. For the most part, STR doesn't matter since you would use AGI, but for the STR from the chest enchant and I use the +35 STR enchant on gloves and am just trying to make sure that my numbers are coming out as they should. Like I said, I assume its accounted, but I was just curious.

Yes, Mew does take STR into account.
As long as you are entering stats from your character sheet into Mew, everything should be accounted for correctly.
Note that if you arbitrary add 35 STR to your character sheet stats and enter that into Mew, it will NOT be correct. By enchanting your gloves with 35 STR, your character sheet AP should increase by 87 too, and this 87 additional AP has to be entered into Mew too.

Alarron · 01/29/11, 4:41 PM

I've been studying the outputs of both Mew and Simulationcraft: currently, they differ significantly on their estimates for feral DPS. One of the big discrepancies I've noted is the estimation of crit%. Using my current gear with all buffs, Mew estimates a 37.43% crit rate for melee, 37.40% crit rate for specials, and 10.15% miss/dodge chance. SC estimates a 36.4% crit rate for melee, 32.9% crit rate for specials, and a 9.6% miss/dodge chance.

Thoughts?

Yawning · 01/29/11, 6:45 PM

Originally Posted by Alarron

I've been studying the outputs of both Mew and Simulationcraft: currently, they differ significantly on their estimates for feral DPS. One of the big discrepancies I've noted is the estimation of crit%. Using my current gear with all buffs, Mew estimates a 37.43% crit rate for melee, 37.40% crit rate for specials, and 10.15% miss/dodge chance. SC estimates a 36.4% crit rate for melee, 32.9% crit rate for specials, and a 9.6% miss/dodge chance.

Thoughts?

Hmmm, checking over my character: (520 hit rating, 188 expertise rating)
Hit: 520 / 120.109 = +4.329401 %
Expertise: 188 / 30.0272 = 6.26099 Expertise = -1.565248% Dodge

Base mob avoidance at 85 vs a raid boss is -8% hit 6.5% dodge.

14.5 - (4.329401 + 1.565248) = 8.605351

Mew Formulation: Chance Miss/Dodge (%): 8.60535
Mew Simulation: Chance Miss/Dodge (%): 8.60535

Taking the lazy way of working out if crit is handled properly:
Character sheet crit rate (w/MotW, in cat form): 41.11%

I'm no where near the white crit cap, so I would expect 41.11 - 4.8 = 36.31% White Crit.

Mew Formulation: Crit % for White: 36.31453
Mew Simulation: Crit % for White: 36.34242 (10k iterations), 36.31799 (50k iterations)

There's actually 2 yellow crit rates to consider. One which factors in yellow attacks being two roll, one which does not (periodic effects, can multiply this by 1 - (pMiss + pDodge) to get the former). For historical reasons Mew reports the latter, thus reported white crit ~= yellow crit (though internally it is doing the right thing). It's fairly obvious that the figure you give for Simcraft's yellow crit is the former (0.364 * (1 - 0.096) = 0.329056).

As of right now I don't think Mew is doing the wrong thing when calculating combat table probabilities (though I only compared things without procs or raid buffs to in game, I *could* be grossly mishandling those, however I don't have any major reasons to suspect so at the moment).

Edit: Bleah. There's a discrepancy in auto attack damage self buffed between a Simcraft profile that only white swings and a Mew one that does the same. So there's differences in any one of armor mitigation (formula or value), deriving cat form AP, or deriving weapon damage. That's about as far as I can go looking into it today.

Alarron · 01/30/11, 12:14 AM

Thanks for the theorycraft refresher, Yawning. Mew "felt" right, I didn't understand why the crit dep wasn't showing up in Mew's output, but that makes sense.

I second the problem with the melee damage; SC is way low. I'll put in a ticket for it.