Rogue DPS Simulation Spreadsheet - Page 11

• Vulajin · 06/03/09, 3:47 PM

Or the screenshot was taken with an old version of Recount, prior to the change to make it register Sword Specialization procs correctly. In any case, using Recount as a method of determining how Sword Specialization works is fundamentally foolish, because the only way Recount knows how to calculate Sword Specialization hits is by applying an interpretation of the combat log.

If you want to contest the known explanations of how Sword Specialization works, stop clutching at questionable proof methods (Recount? seriously?) and go test it yourself. The testing method is fully laid out in my post from the old thread, which I linked above.

Mavanas · 06/03/09, 7:59 PM

Ok I think I got to the bottom of the problem with Recount. The problem in fact appears to lie with Recount's methodology to flag sword spec procs, and Sword Spec works exactly as explained by Aldriana, at least for autoattacks that were my focus.

I did my own testing as well with two swords: [Razorscale Talon] and [Crystalblade of the Draenei]. I autoattacked boss-level dummy for slightly less than an hour and in that time observed 3006 melee swings and 99 sword spec procs as categorized by Recount. The main hand hit for about 240-380 and the low level and faster offhand hit for about 60-80, so it was easy to distinguish the hits. There were some glancing hits, but no crit because I was below 4.8% crit reduction.

Same as in the wow forum thread, I observed that Sword Spec proc damage ranged from 66 to 388, indicating that both mainhand and offhand damage was classified as sword spec (see the screenshot of Recount).

Next I looked at the combatlog, and found three cases where Recount incorrectly identified regular offhand attacks as Sword Spec procs.

But before that I matched Recount's numbers. They seem to follow the method suggested by Vulajin in his analysis: they treat the first line after "Sword Specialization" as the attack that procced it, and the next line as the damage of the proc itself. They do not however, investigate the timing between mainhand and offhand swings to determine which swing falls out of line. This method normally provides correct identification, but in rare cases when a scheduled OH attack occurs at the same time as a schedule MH attack, it can happen that wrong damage is classified as sword spec.

Consider for instance these lines from the combatlog:

00:13:14.468	SWING_DAMAGE	0x0280000000137335	Mavanas	0x511	0xF1300079AA0018E1	Heroic Training Dummy	0x10a28	61	0	1	0	0	0	nil	1	nil
00:13:15.093	SWING_DAMAGE	0x0280000000137335	Mavanas	0x511	0xF1300079AA0018E1	Heroic Training Dummy	0x10a28	238	0	1	0	0	0	nil	1	nil
00:13:15.968	SWING_DAMAGE	0x0280000000137335	Mavanas	0x511	0xF1300079AA0018E1	Heroic Training Dummy	0x10a28	49	0	1	0	0	0	nil	1	nil
00:13:17.171	SPELL_AURA_REFRESH	0x0280000000137335	Mavanas	0x511	0xF1300079AA0018E1	Heroic Training Dummy	0x10a28	51585	Blade Twisting	0x1	DEBUFF					
00:13:17.500	SPELL_EXTRA_ATTACKS	0x0280000000137335	Mavanas	0x511	0x0280000000137335	Mavanas	0x511	16459	Sword Specialization	0x1	1					
00:13:17.500	SWING_DAMAGE	0x0280000000137335	Mavanas	0x511	0xF1300079AA0018E1	Heroic Training Dummy	0x10a28	346	0	1	0	0	0	nil	nil	nil
00:13:17.500	SWING_DAMAGE	0x0280000000137335	Mavanas	0x511	0xF1300079AA0018E1	Heroic Training Dummy	0x10a28	77	0	1	0	0	0	nil	nil	nil
00:13:17.515	SWING_DAMAGE	0x0280000000137335	Mavanas	0x511	0xF1300079AA0018E1	Heroic Training Dummy	0x10a28	388	0	1	0	0	0	nil	nil	nil
00:13:18.390	SPELL_AURA_REFRESH	0x0280000000137335	Mavanas	0x511	0xF1300079AA0018E1	Heroic Training Dummy	0x10a28	51585	Blade Twisting	0x1	DEBUFF					
00:13:18.390	SPELL_AURA_REFRESH	0x0280000000137335	Mavanas	0x511	0xF1300079AA0018E1	Heroic Training Dummy	0x10a28	51585	Blade Twisting	0x1	DEBUFF					
00:13:19.000	SWING_DAMAGE	0x0280000000137335	Mavanas	0x511	0xF1300079AA0018E1	Heroic Training Dummy	0x10a28	61	0	1	0	0	0	nil	nil	nil

-3.032 OH attack
-2.407 MH attack
-1.532 OH attack
0 Sword spec indicator
0 MH attack
0 OH attack <= flagged by Recount following their methodology
0.015 MH attack <= actual sword spec proc
1.5 OH attack

The expected time between attacks is around 2.35 between mainhand attacks and 1.45 between offhand attacks. If you look at it closely, then the offhand attack that appears two lines below the sword spec is 1.5 seconds after the previous attack. And the sword spec damage is actually on the third line below the sword spec announcement, it occurs 0.015 seconds after the previous mainhand attack.

I am sure something similar happened to the guy on wow forums who tested this before.

P.S. I used this excel file to match recount's methodology and identify all incorrectly flagged sword spec procs.

Kildrazien · 06/05/09, 10:49 AM

Mavanas, to put it short, could you please tell me how i can input my own trinkets into the spreadsheet? I was thinking the Mark of Norgannon might be a good choice because of being able to activate with the Wrathstone and Blade Flurry multiplicatively. I would really appreciate it.

Mavanas · 06/09/09, 6:23 PM

I'd like to do an update on best in slot items according to the simulation and a comparison of DPS among different specs in best in slot gear. A few things changed since last time, 100% armor penetration cap was put in the game and I spent more time analyzing what's best after the change. Dark Matter (DM), Comet's Trail (CT) and Solar Bindings were discovered and changed several best in slot lists. It's become common to hit the crit cap for white autoattacks, which changed the EP weights of hit, crit and agility. Because crit and armor penetration caps (during trinket procs) became so important, I added the two caps to the Gear tab, so you can see them alongside with expertise and various hit caps. The new version of the spreadsheet 3.1k also has a properly working sword spec, new HAT proc timing simulation, a couple bug fixes and new Overkill as in 3.1.3.

Tree		Spec	DPS	Special Notes
vHaT (1 c/s)	8/20/43	7795	ArPen gems, AG food, MR+CT
nvHaT (1 c/s)	23/5/43	7410	AG gems, Exp and hit capped (from gear), DMC:G+CT
Combat		15/51/5	7708	ArPen gems and food, CT+MR
Combat		7/51/13	7694	ArPen gems and food, CT+MR
Combat		18/51/2	7708	ArPen gems and food, eviscerate finisher, CT+MR
Mutilate	51/7/13	7690	Exp capped (from gear), hit cap, then AP gems, MP (3/3), DM+CT
Mutilate	51/13/7	7699	Exp capped (from gear), hit cap, then AP gems, MP (3/3), DM+CT
Mutilate	51/13/7	8202	Murder

Main highlights:
1. Armor penetration is still best gemming choice for combat and vHaT; due to DM, mutilate does more damage with AP gems.
2. Going above ArPen cap is normally a dps decrease. If you end up being above ArPen cap, it's normally worthwhile to swich to some AG gems and/or AG food.
3. Choosing trinkets is a tricky matter, due to hit, crit and ArPen caps. DMC:G is still the best for HaT specs, and DM just does not work there because of the crit cap. MR+CT combo is best for combat. DM+CT is best for mutilate.
4. In BiS gear, mutilate without Murder is on par with combat; however, the discovery of CT gave combat a slight edge over mutilate.
5. It's hard to judge what the crit/sec rate for HaT is going to be when entire group is in BiS gear. Right now 1c/s is a bit generous, but it could be a better estimate in BiS gear.

BiS Gear:

Combat				Mutilate
[Golden Saronite Dragon]		[Fang of Oblivion]
[Bladetwister]				[Bladetwister]
[Twirling Blades]			[Skyforge Crossbow]
[Conqueror's Terrorblade Helmet]	[Conqueror's Terrorblade Helmet]
[Pendulum of Infinity]			[Pendulum of Infinity]
[Conqueror's Terrorblade Pauldrons]	[Shoulderpads of the Intruder]
[Drape of the Faceless General]		[Drape of the Faceless General]
[Embrace of the Gladiator]		[Conqueror's Terrorblade Breastplate]
Solar Bindings				Solar Bindings
[Conqueror's Terrorblade Gauntlets]	[Conqueror's Terrorblade Gauntlets]
[Soul-Devouring Cinch]			[Soul-Devouring Cinch]
[Conqueror's Terrorblade Legplates]	[Conqueror's Terrorblade Legplates]
[Footpads of Silence]			[Footpads of Silence]
[Band of Lights]			[Band of Lights]
[Brann's Signet Ring]			[Brann's Signet Ring]
[Mjolnir Runestone]			Comet's Trail
Comet's Trail				Dark Matter

• Aldriana · 06/09/09, 7:16 PM

I might note that Solar Bindings puts you way, way over the expertise cap for combat; I strongly suggest using Fluxing Energy Coils instead. Also, in my analysis, Band of Lights is a better second ring than Cindershard. So I might try making those changes, which should close the gap between the two specs somewhat.

Second, I confess to a certain degree of skepticism over those Mutilate numbers - they're a good 5% higher than anything anyone has suggested previously. Could you confirm what cycle and/or priority system you're using to generate that sort of DPS output?

Mavanas · 06/10/09, 12:45 AM

I have tried other T8 combinations for combat, and the particular one above came out ahead. I tried Fluxing Energy Coils along with Legguards of Cunning Deception and rest t8 (which was the best setup before Solar Bindings were discovered), and it was slightly lower dps. In the setup above, you are not over expertise cap because you use Embrace of the Gladiator. Also I haven't seen the Band of Light before, but now looking at it, it should definitely beat the Cindershard Ring.

You were right about Mutilate, I was wondering about it too, and it turned out an error the way I implemented the new Overkill. The extra energy regen was up for the whole fight, hence a huge dps difference. The new numbers are much more in line with Combat as they were before (7650 without murder). So I will change the previous post.

• Aldriana · 06/10/09, 1:06 AM

Ah, I missed the fact that you're using legs offset rather than chest - you are indeed reasonably places relative to the expertise cap. I did try that combination in my own calculations, and found it to be pretty decisively inferior to fluxing/legs - not by a lot, but pretty clearly behind. I rather imagine it's a difference in modeling methods and/or a bug in one or both of our implementations, but suffice it to say they're reasonably close together and, as such, for the short term people can use whichever they can more easily obtain - given that one takes a piece of Algalon 25 hard and the other a piece off Yogg25 Hard, it's safe to say that most of us won't have either - much less both - for a couple of months yet.

Mavanas · 06/10/09, 9:37 AM

The difference between Solar Bindings/Embrace of the Gladiator and Fluxing Coils/Legguards of Cunning Deception comes out positive, but not statistically significant (10dps on last try). What that means is in 500 boss fights, one will not be able to tell the difference between the two setups because of RNG. By the way, to get the best out of Fluxing Coils/Legguards of Cunning Deception set, I had to switch some gems from ArPen to agility since I was hitting the ArPen cap and it started to lower the value of ArPen.

While the following estimates are not very precise and have 0.05 standard error (10,000 observations), I did calculate that ArPen dps-per-point value drops from 2 to 1.4 when you are 50 points above the soft ArPen cap (getting capped when MR procs). So it drops from the top stat to below attack power and agility at that point. I am also getting higher dps/point values for white hit and expertise than before, which is likely because I am over crit cap and below expertise cap in Solar Bindings/Embrace BiS setup.

impi · 06/10/09, 9:58 AM

Hey Mavanas,
Thanks for updating the spreadsheet! I like the new numbers with BiS Gear, because Mutilate is just more flexible to play for me than Combat.
I just downloaded your new spreadsheet and opened it with Excel, there is a warning like "Excel found unreadable content and would like to repair it." After repairing the file there is a warning that some data is lost.
I sent the download link to few other people but everyone got the same problem.
Could you fix that?

Mavanas · 06/11/09, 11:24 AM

Try this mirror site at filefront. I also reuploaded the file to myotherdrive.

chalon · 06/11/09, 3:28 PM

Minor note, but you should be running Agi gems as the filler ones in your Mutilate setup, not AP.

EDIT: Also, wouldn't Blood of the Old God actually be a better choice over Mjolnir in at least the 51/7/13 setup, simply because (unless I'm missing it), you aren't going to be poison hit capped?

• Aldriana · 06/11/09, 3:44 PM

It's actually possible that in BIS AP gems are superior do to crit-capping. I'm going to assume that Mavanas at least tested that.

Manaba · 06/11/09, 4:19 PM

Originally Posted by Mavanas

Main highlights:
1. Armor penetration is still best gemming choice for combat and vHaT; due to DM, mutilate does more damage with AP gems.

Crit capping due to DM is the reason he switched to AP gems vs Agil gems.

chalon · 06/11/09, 4:52 PM

Hmm, it's actually pretty interesting then that none of the setups yield Blood of the Old God as a better choice. The gear you have listed has a total of 180 hit rating. Now, if you're running 51/7/13, you want 262 hit rating as Alliance in order to be poison hit capped. Thus with Blood of the Old God you would be at 288 hit rating, which isn't over by a tremendous amount. Then combine the fact that the proc is better for you than Mjolnir's proc, and the fact that you would then most likely be free to once again gem Agi without worrying about being crit-capped (especially since you've lost 3% crit, as well).

Now, this is all moot I guess if every setup still shows that 51/13/7 is superior according to your simulation. I'll have to look at it more when I get home I guess.

Mavanas · 06/11/09, 7:06 PM

Chalon, if you look at the table above where I listed several specs and their DPS, next to 51/7/13 in the special notes I do indicate that highest DPS is achieved with BotOG (short for Blood of the Old God). Same thing with 23/5/43. For 51/13/7 BotOG loses because all that hit becomes white hit which has about 40% less dps value.

I did test that changing agility gems to AP gems in BiS gear for mutilate is a modest dps increase.