Elitist Jerks [Feral] Mew Simulator

 03/08/11, 2:26 AM #61 a civilian Piston Honda   Stenhaldi Worgen Druid   Sargeras Let me clarify the math a little. When I write .15*rip-f(1.15*30+12)/40*shred, that is my estimate for the damage gained by refreshing rip during TF when rip still has f fraction of its time remaining. You gain .15*rip damage due to TF's effect, but you lose f rip, worth f*30 energy, which would otherwise be spent on a TF'd shred doing 1.15*shred damage for 40 energy. Thus you lose f*1.15*30/40*shred damage. You also lose CP, but assuming you aren't refreshing rip just as savage roar is about to fall off, the worst case here is that you have to use a 1 CP roar later. This lasts a little over half as long as a full CP roar, so we can say you lose 12 energy, a little under half the cost of savage roar, in the worst case. This energy would otherwise be spent on non-TF'd shreds, so you lose 12/40*shred damage. If f is small, you've probably already refreshed savage roar so this loss won't occur; if f is large, this loss will tend to approach its maximum value, since you'll be short on CP after applying the rip in the first place. Thus as a rough estimate I multiply this number by f, yielding a loss of f*12/40*shred damage. (This factor isn't really significant to my results in the previous post since f is close to 1 in most of the figures there. It just yields a more accurate value for the damage gain when f is small.) This yields a damage gain of .15*rip-f(1.15*30+12)/40*shred for refreshing rip during TF. A similar calculation (but without the CP complication) yields .15*rake-g*1.15*35/40*shred damage for refreshing rake during TF when rake still has g fraction of its time remaining.
 03/08/11, 4:35 AM #62 Leafkiller Piston Honda   Leafkiller Worgen Druid   Stormrage Perhaps we can understand this better by looking at some of the data from the Mew output. I suspect we need to have more data, but this is a start. Using my sim script and the non-hitcap profile without the "C" statement I see the following: Number Mangles: 5.7883 Mangle Uptime (%): 98.94664 Number Shreds: 94.7792 Number Rakes: 24.8578 Rake Uptime (%): 97.75557 Number Rips: 11.4866 Rip Uptime (%): 96.62631 Number Savage Roars: 9.8142 Savage Roar Uptime (%): 92.99265 Savage Roar CP(s): 2.1991 Adding the "C" statement just before the filler shred code gives: Number Mangles: 5.7749 Mangle Uptime (%): 98.97488 Number Shreds: 93.3718 Number Rakes: 24.8272 Rake Uptime (%): 97.9173 Number Rips: 14.1193 Rip Uptime (%): 96.692 Number Savage Roars: 9.7514 Savage Roar Uptime (%): 92.89413 Savage Roar CP(s): 2.2263 There are a lot of small impacts here. Note that we are seeing an average of about 2.6 clipped Rips per run. Also the number of clipped Rakes goes slightly down. Keep in mind that some Rips are already being cast during TF, and some Rips will not be clipped because they occur close enough to the start of TF, that 5 combo points will not be reached in time to clip, and in some cases SR will be refreshed during TF using up the combo points. Here is some DPE data: Without "C": Mangle DPE: 394.34863 Shred DPE: 527.9686 Rake DPE: 1678.81946 Rip DPE: 4143.96923 Bite DPE: 685.54647 With "C": Mangle DPE: 394.91729 Shred DPE: 524.7775 Rake DPE: 1682.56594 Rip DPE: 3530.56237 Bite DPE: 685.70999 You can see the impact on the overall Rip DPE from clipping - more Rips, but around the same number of ticks. And the damage percents: Without "C": Mangle Damage (%): 1.07319 Shred Damage (%): 26.96929 Rake Damage (%): 19.67437 Rip Damage (%): 26.35181 Bite Damage (%): 2.90876 With "C": Mangle Damage (%): 1.07613 Shred Damage (%): 26.39717 Rake Damage (%): 19.698 Rip Damage (%): 27.14795 Bite Damage (%): 2.68013 The dps for the non "C" run is DPS: 22592.52336 +/- 10.45928 StdDev: 533.63694 DPS (> 25%): 21582.07916 DPS (< 25%): 25623.85593 while the dps for the "C" run is DPS: 22595.7482 +/- 10.28701 StdDev: 524.84768 DPS (> 25%): 21695.58801 DPS (< 25%): 25296.22874 I am not sure why the dps is lower in the BiTW phase in the run with the "C" statement. It is set to be 25% of the time of the fight - so the time spent in it is the same for both test runs. In the BiTW phase the code for FBing on 5 combo points occurs very early in the script - so the Rip clipping should not be happening - which means that it would have to be a condition upon entering the phase that is different.
 03/08/11, 8:32 AM #64 Junlex Piston Honda   Rhiannon Worgen Warrior   Anachronos (EU) To try to identify some sort of pattern as to when rip would last 11 ticks vs 12 ticks I ran a short log which I have uploaded: World of Logs - Real Time Raid Analysis I ran hit/exp capped with no trinket procs, no auto-attacks, no abilities used apart from mangle, shred and rip, to try to reduce the number of potential variables but no obvious pattern emerged. From this short run of 15 rips, 13 had 12 ticks, 1 had 11 ticks (starting at 13:03:22.303) and 1 had 13 ticks (starting at 13:06:16.292). A lot more data's required though, as when I did 6 rips earlier, 2 had 12 ticks while 4 had 11 ticks. Last edited by Junlex : 03/08/11 at 8:44 AM.
 03/08/11, 1:24 PM #65 Leafkiller Piston Honda   Leafkiller Worgen Druid   Stormrage A couple of additional thoughts. The check I added to test this does not check for the "A more powerful spell is already in effect" error. However, this should not be an issue in the current script since it is not pre-potting, and it is using the Potion of Tol'vir during Heroism which happens in the last 40 seconds of the fight where Rip is no longer being cast. The trinkets used are Unheeded Warning and Fluid Death so they are not a factor. I can't think of any other procs that affect AP outside of the potion. This is a condition that would have to be dealt with if I ever tried to implement this in Ovale though - or if I turned on pre-potting. Currently there is no call in the Mew script API to see if a Rip cannot be overwritten - but it would be very easy to add one (unfortunately I never made my Mew Eclipse environment capable of compiling the code). Coming back to the 2.6 number. The current approach I have been testing is to put all abilities outside of Shred ahead of the Rip refresh. The first test was ahead of most abilities - except for one that would impact this - the three rip shreds. I will tinker with the script to selectively prioritize the Rip refresh higher and see if I can increase the 2.6 number and along with the overall dps. Edit: testing with Rip refresh ahead of SR loses 15dps. Putting a check into the Rip Shred makes no difference at all. This code is never hit when we already have 5 combo points (not too surprising). I have been able to push the 2.6 number to over 3, but the net result was the 15dps decrease due to lower uptime on SR. Last edited by Leafkiller : 03/08/11 at 1:46 PM.
03/10/11, 7:09 PM   #66
Leafkiller
Piston Honda

Worgen Druid

Stormrage
 Originally Posted by Junlex To try to identify some sort of pattern as to when rip would last 11 ticks vs 12 ticks I ran a short log which I have uploaded: World of Logs - Real Time Raid Analysis I ran hit/exp capped with no trinket procs, no auto-attacks, no abilities used apart from mangle, shred and rip, to try to reduce the number of potential variables but no obvious pattern emerged. From this short run of 15 rips, 13 had 12 ticks, 1 had 11 ticks (starting at 13:03:22.303) and 1 had 13 ticks (starting at 13:06:16.292). A lot more data's required though, as when I did 6 rips earlier, 2 had 12 ticks while 4 had 11 ticks.
Interesting but also frustrating data. I certainly do not see a pattern in there - other than the durations of the Rips matched the number of ticks - 22, 24 and 26 seconds. I don't know how we model that correctly.

03/11/11, 7:26 AM   #67
Helistar
Von Kaiser

Night Elf Druid

Dalaran (EU)
 Originally Posted by Leafkiller I am not sure why the dps is lower in the BiTW phase in the run with the "C" statement. [....]
My guess is that the lack of rip refresh causes the "without C" script to enter the <25% state with 5 CPs, which means one additional FB during the <25% phase, which would match the damage% difference for FB that you report (without C: 2.9%, with C: 2.68%). You didn't paste the Number of Bites, but I guess it's higher in the "without C" case.

A simple solution would be to fudge the 25% BitW threshold so that it doesn't happen always at the same time. In reality what should be done is to change the script so that if it sees that, upon approaching 25%, by not refreshing rip early it can cast one additional FB, then the early rip refresh is ignored. The calculation of the gain in Rip damage is based on the assumption that at that time CPs have no other use, which becomes false as we cross the 25% threshold.

 03/21/11, 8:04 PM #68 Beanna Glass Joe   Beanna Tauren Druid   Conseil des Ombres (EU) Hello ! I was about to swap my profession in order to get the awesome Synapse Springs and I did some simulations on Mew just to be sure I was not mistaken in forgetting LW to become an engineer. I was rather surprised to see that engineering is about a ~50 DPS loss on my toon according to Mew and I don't find any reason to explain this DPS drop. Why Mew shows Synapse Springs as being under any +80 agi from other professions? (I correctly reduced my agility by 80 when swapping my professions on Mew.) I mean, this great activated +480 agi mini-trinket timed with TF to refresh our dots for full duration with a big AP boost and which produces a big energy pool to get the most out of it should be better than other professions, I am right? Do you think that Synapse Springs may not be modeled correctly or timed with TF in Mew's script? I dare not change my profession as Mew does not confirm to me that's a good decision. I'm not sure about my estimates on Synapse Springs. Rawr seems to consider that both professions are exactly equal, which I do not really think is correct neither... Last edited by Beanna : 03/21/11 at 8:13 PM.
 03/22/11, 12:00 AM #69 tangedyn Piston Honda   tangedyn Tauren Druid   Thaurissan Rawr uses basic formulation, and the obvious easy way to model it would be to simply treat it as 80 agility when averaged over time. Mew Formulation does the same thing. At the moment, Mew Simulation simply uses the Springs every time it's off CD, not attempting to synch it. If you glyph TF, attempting to line synch up Synapse Springs will probably be a DPS loss. If you glyph Berserk instead, the Mew seems to indicate a small DPS increase. Not sure why it is showing up as a DPS loss with Tiger's Fury glyphed, my guess is that it's spending more time desynched that it would normally be averaged over an infinite fight duration.
 03/22/11, 9:08 AM #70 Petitourson Glass Joe   Petitourson Tauren Druid   Conseil des Ombres (EU) I have run 20+ simulations to try and pin down what's causing the discrepancy. With the default script (using my gear for reference) Synapse Springs was always a dps loss (-11.5 < N < -2.6) until I enabled both a 10% duration randomization and the high resolution timer simultaneously (+0.35 < N < +24.1). Synapse Springs would fall behind if only one of the two options was enabled. With Leafkiller's Mew script Synapse Springs is always a dps increase, the gap widening the more advanced options I ticked on (10% duration randomization, high resolution timer) with gains comprised +4.3 < N < +22.6 All sims were run with GoBerserk since it's the only one worth glyphing as an engineer.
 03/22/11, 1:22 PM #71 Beanna Glass Joe   Beanna Tauren Druid   Conseil des Ombres (EU) Of course I plan to use the glyph of Berserker in order to synch SS with TF! My bad, I didn't checked the boxes, especially High Resolution Timer. I've just done the simulations once more and Synapse Springs appears to be a ~+47 DPS increase for my toon so I'll finally go for engineering. Thank you very much for your advices.
 04/12/11, 6:16 AM #72 Yawning Von Kaiser   Yawning Night Elf Druid   No WoW Account Mew-20110412 is now available. Major changes since last release:Dark Intent's periodic damage increase is now 1/2/3% per 4.1 PTR patch notes. Interrupts are reliable per 4.1 patch notes. Simulation is now the default model for all new profiles. The Simulator now allows for casting Rebirth/Tranquility durring the encounter. Furor and Predator's Swiftness are now modeled by the Simulator. Spell Vulnerability/Haste/Crit Taken are specifiable buffs/debuffs (Mostly used for Tranquility/Rebirth, also affects certain trinkets). New Simulator script calls getBerserkBaseDuration(), getEncounterDuration(), getElapsedTime(), and isAutoAttackEnabled() New Simulator script actions AUTOATTACK_START, AUTOATTACK_STOP. The Simulator now expects the script to manage the swing timer. Formulation no longer uses Ferocious Bite at all over 25% mob HP. New option that will cause Mew to model Glyph of Shred as +1/1/2 ticks. BUG: Change potion cooldown. (Was 60 sec is now 120 sec, no DPS impact for encounter durations worth modeling.) BUG: Apply the ToTT multiplier to trinket sourced damage procs. (DMC:H is still sub-par compared to other options.) BUG: Stampede buff duration is now always 8 sec regardless of the number of points in the talent. Numerous other internal changes, please see the SVN commit logs if you are curious. Note: If you have a custom script, please look at the documentation in the google code wiki or the example script to ensure that you start the swing timer. As usual questions/comments/bug reports appreciated.
 04/13/11, 8:02 PM #73 Leafkiller Piston Honda   Leafkiller Worgen Druid   Stormrage I updated my sim script to include the updated sim script that Yawning just released along with modifications to my two scripts that incorporate changes to work with the new Mew release (the external buff code is the same in all three scripts and the interrupt and FC code is close to identical). In my so-called "universal" script, there is a variable, "rotation" that will use the default Mew script if set to "0", my script if set to "1" and an experimental script if set to "2". The experimental rotation currently emulates the Atremedes fight. The script can be found here: The Fluid Druid - View topic - Leafkiller's 4.0.6 Feral Ovale Script
 04/25/11, 11:47 PM #74 tangedyn Piston Honda   tangedyn Tauren Druid   Thaurissan Mew-20110425 is now available. Major changes since last release: - Simulation Progress Bar updates now throttled to reduce system resource overhead - (4.1) Berserk is now off the GCD - Leafkiller's optimized Cat Strategy is now available in Samples/universal1.script - Simulations now generates a Report Output, which can be copy and pasted Notes: - We have created a Google Discussion Group at Mew Theorycraft | Google Groups as the official feedback and support forum for Mew. We will continue to keep an eye on a few other forums, but not as frequently. - When reporting results of your own Simulation runs, please copy and paste the entire Report Output (from the Report tab) in your post. This will allow us to analyze your results better.
04/26/11, 12:21 PM   #75
Leafkiller
Piston Honda

Worgen Druid

Stormrage
For anyone using my script, please keep the following in mind:

 The version of my script that is checked into the Mew tree is set to the Atramedes fight. There is a variable: `int rotation = 2;` that is near the top of the file - that dictates which rotation to use. There are three rotations in the file. If you set the variable to 0, it will use the default rotation (which may be out of date since I updated it on the last public release of Mew). if you set it to 1, you get my version of a "Pathwork" rotation, and if you set it to 2, you will get my version of the Atramedes fight - which is what it was set to when I added it to the Mew tree. The reason I put multiple rotations into the file was due to the slowness of the Mew file I/O. I did not want to wait for that when testing different fight scenarios. At some point I would like to have the fight scenarios added to the drop down list so they can be individually selected without having to walk through the file system.

 Elitist Jerks [Feral] Mew Simulator