I fixed a small error that in some cases resulted in #NA errors if a garrote was dodged during the last 20 seconds of the fight. The link above should have an updated version.
I did not make much progress analyzing envenom timing, but at least in response to Loot's example, I can say that you can construct another example. Suppose you don't have energy when DP ticks, and when you finally have enough for an envenom, it's 2 seconds after. If you don't time envenoms, you can use it right away, which will give you an earlier 1-stack DP tick. If you wait and time it, the 1-stack tick will occur later. I know it does not explain the phenomenon, but at least it's not all that clearcut as in your nicely constructed example.
I did find that if you turn off energy pooling, timing envenoms does slightly more dps than not timing. However both scenarios are a slight dps decrease compared to pooling energy and not timing envenoms. So I will continue digging. I should also say that all of these differences are minimal, not even statistically significant at 10% (about 20 dps).
The truth is, I enjoy timing the stuff, especially in a battle you have to watch everything around, add to that envenom buff and dp ticks so I'm trying to find arguments to continue doing so :p
DPS gain for disabled pooling and timing envenoms sounds logical tho.
I pretty much linked the loss of dps from timing envenoms to pooling energy. I played with the settings and DPS rules looking for best setup for mutilate, and absolute highest dps so far is when you time the envenoms and disable energy pooling. So basically you can delay an envenom waiting for a DP tick, just don't delay it simply to reach some energy threshold because if you do both, you will be running into energy cap situations which will lower your dps.
While playing with the setting, I also created a simplified version for combat and mutilate only, which does not have mongoose and executioner. I will place the link to simplified file in the main thread. It runs about 15% faster.
I'd like to update the DPS table I prepared when 3.1 came out. The current DPS estimates are based on the simulation spreadsheet 3.1i. I have done extensive testing to determine best settings and best in slot gear for each spec. HaT DPS was calculated for two cases, 1 crit/second and 1.4 crit/second. Due to lack of recent testing of crit/second rates, I do not know which rate best describes current raid situation. Based on last available estimates, I suspect that if several group members die (which can happen on a progression fight), the crit/sec rate can drop even below 1 c/s, however if noone dies and stays on same target all the time, the rate can be even higher than the 1.4 crit/second rate.
Notes about best in slot gear:
- vast majority of best in slot items are the same for all specs, in particular:
Bladetwister offhand, T8 helm, Pendulum of Infinity, T8 chest, Fluxing Energy Coils, T8 gloves, Soul-Devouring Cinch, and Mjolnir's Runestone
- things that differ: all specs except for nvHaT use T8 helm, chest, shoulders, and gloves along with Legguards of Cunning deception, nvHaT uses Shoulderpads of the Intruder and rest T8
- mutilate uses Fang of Oblivion, rest use Golden Saronite Dragon
- most specs use Thirling Blades, mutilate uses Siren's Cry
- drape of Icy intent is preferable for nvHaT if you are below poison cap, otherwise Drape of the Faceless General is BiS
- Footpads of Silence is BiS for nvHaT and combat, Runed Ironhide Boots are better for vHaT and mutilate (basically same as Footpads of Silence for mutilate)
- double Brann's rings BiS for vHaT and combat (15/51/5), Brann's Signet Ring and Godbane Signet BiS for mutilate, Cindershard and Brann's Signet Ring BiS for nvHaT
- Mutilate and nvHaT use Blood of the Old God as second trinket, combat and vHaT use Grim Toll
You can also see that I calculated dps under the assumption that armor penetration is capped at 100%. This assumption does not change the choice of gems for vHaT and combat, but it makes Blood of the Old God the trinket of choice in place of Grim Toll for combat and vHaT.
Before the cap is put in place, combat dps in BiS gear is above mutilate without murder and below mutilate with murder. After the cap is implemented, combat should be around mutilate without murder on a single target.
A new version is out, updated for 3.1.2 changes. Armor penetration is capped at 100%, armor penetration gemming is still ahead of agility gemming by a fair margin for combat spec; however Grim Toll is no longer best second trinket to accompany Mjolnir Runestone. Added an option to calculate double damage (excluding dots) during blade flurry, to better compare combat spec on fights with adds.
I also added the new trinket Dark Matter. It is now best in slot for combat. I did some ranking of trinkets for Combat 15/51/5 after the 100% cap:
Mjolnir + Dark Matter (rest BiS items) = 7604 dps
Mjolnir + DMC: Greatness = 7575 dps
Mjolnir + Blood of the Old God = 7541 dps
Mjolnir + Grim Toll = 7437 dps
On the topic of interaction of different procs, if you plug in Dark Matter as a combination of 222 extra attack power and 612 crit rating with 21% estimated uptime, then instead of 7604 dps, you only get 7588. So the value of synergy is about 16 dps.
All numbers are subject to standard error due to RNG of about 5 dps.
armor penetration gemming is still ahead of agility gemming by a fair margin for combat spec
How much, exactly? I realize the spreadsheet vs simulation will result in different numbers but the last numbers I saw were aldriana's here, which seemed to imply roughly a 0.25% DPS difference between AGI and ArP gemming.
I plugged in exact same gear as in that Aldriana's post into my simulation and the results are:
Combat (armor penetration setup) - 7551 (7577 with Dark Matter instead of DMC:G)
Combat 2 (agility setup) - 7498 (7481 with Dark Matter instead of DMC:G)
The reason you actually see a drop switching to Dark Matter here is because under the agility setup, during Dark Matter proc, crit chance is clipped, i.e. miss rate + dodge rate + glancing chance + crit chance is more than 107%. Which means half of the proc crit rating is wasted.
So switching from armor penetration to agility is a loss of about 50 dps (0.7%), plus you have to watch out not to cap your crit chance by using Dark Matter.
If anyone's interested, I've rewritten my Recount HAT extension entirely so that it's standalone and tracks crits per second more accurately. It will also allow you to see in real-time, what the maximum potential crits per second you can get is, and what your current crits per second is in your group. (excluding yourself)
NOTE: Alot of research remains to be done as to which ability crits actually proc HAT, and how they proc HAT (e.g. AoE).
There's a file 'TopHAT_Abilities.lua' that contains a table of all these abilities. Feel free to contribute your own findings.
Are the Attack Power gems not good with ulduar gear at all?
I am 51/13/7 specced with AP gems and i get more dps in the simulation than agility and armor pen. gems.
Last edited by SalmorIsBack : 05/28/09 at 12:35 PM.
Are the Attack Power gems not good with ulduar gear at all?
I am 51/13/7 specced with AP gems and i get more dps in the simulation than agility and armor pen. gems.
Based on an answer in Simple Questions thread and additional research, I realized the sword spec was not modeled correctly in my Simulation. Previously, it did not proc off yellow damage and when it procced off offhand it caused an extra offhand attack. Also in my previous implementation, sword spec procs were following the white damage hit table, so they could glance and had a higher chance to miss.
All in all, the value of sword spec was lower than it should have been.
So I fixed the Simulation and want to list the characteristics of the Sword Spec the way I implemented it so that everyone can see it and point out if there are still problems with it.
1. Sword Spec is a special yellow-damage attack that follows the hit table for special attacks. It is thus based on a two-roll system, where first roll determines whether it's a miss, dodge, parry or a hit, and second roll determines whether it's a crit or not.
2. Provided mainhand weapon is a sword, sword spec can be triggered by a special attack, such as a Sinister Strike, Eviscerate, Envenom, and initial application of Rupture.
3. Provided mainhand weapon is a sword, sword spec can be triggered by a main hand autoattack swing.
4. Provided offhand weapon is a sword, sword spec can be triggered by an offhand autoattack swing.
5. Sword spec proc does damage equal to a main hand autoattack regardless of what procced it.
6. Sword spec proc can itself trigger a mainhand poison attack, any trinkets that proc off mainhand attacks, such as DMC:G, any trinkets that proc of mainhand crits, such as Pyrite Infuser, provided the sword spec proc was a crit, and finally it can trigger mainhand weapon enchant procs, such as Berserking.
7. Sword spec proc does not reset swing timers
8. Sword spec proc does not proc off itself.
With these rules, I have tested the value of sword spec again. In BiS gear, it came out to be around 7%, holding all other stats and parameters constant. This number seems a bit high to me, so if you see any problems with the rules above or have any other thoughts that may shed light on this problem, let me know.
The direct effect from extra attacks is about 4.7%. So the rest 2.3% comes from extra mainhand poison damage and potentially higher uptime of Berserking. I should note that since sword spec is a special attack done with a slow main hand weapon, the chance to apply wound poison is close to 100% (96.43% to be exact).
For comparison, I am getting only about 3.5% increase from CQC in same gear, just changing the type of the weapons and the corresponding talents.
I don't know where you're getting the notion that Sword Spec procs can't glance, because they clearly can - see the attached image.
Basically, sword spec works thusly: regardless of which hand triggers the sword spec proc, the resulting attack is made with the main hand, and is in all respects equivalent to a white attack made with that weapon except for the fact that it can't, itself, proc another sword spec attack. It can proc everything else, be dodged, miss, glance, and anything else - it's just a regular MH white attack.
Sword spec can proc off any attack made with the weapon. For the main hand, this notably include Sinister Strike, Eviscerate, Envenom, and Rupture (though not Slice and Dice).
That may be a bit high, but not out of line it has always been good. Through out BC, and even back as far as vanilla WoW, Swords were the superior spec because of its mechanics. I am probably over simplifying things, but there have been only 2 things that have changed this. The first is the fact that so far sword itemization has been horrible in Wrath. With weapon speed playing such a major part in combat dps, the lack of a 2.5 or slower mainhand and lack of 1.5 or faster offhand has pretty much doomed the spec. Until the recent changes to some swords, the only way to get an offhand weapon faster than 1.6 was to use a dagger. Add in the fact that CQC is higher in the tree giving you more flexibility in placing filler points made it a much more attractive spec.
I don't know where you're getting the notion that Sword Spec procs can't glance, because they clearly can - see the attached image.
Basically, sword spec works thusly: regardless of which hand triggers the sword spec proc, the resulting attack is made with the main hand, and is in all respects equivalent to a white attack made with that weapon except for the fact that it can't, itself, proc another sword spec attack. It can proc everything else, be dodged, miss, glance, and anything else - it's just a regular MH white attack.
Sword spec can proc off any attack made with the weapon. For the main hand, this notably include Sinister Strike, Eviscerate, Envenom, and Rupture (though not Slice and Dice).
Thanks for your answer and the proof. I found the source of my confusion. There were two changes two sword spec, which were related to this issue.
Patch 2.1.0 said: "Sword Specialization: This ability now grants extra yellow attacks instead of extra white attacks. This change will make Sword Specialization no longer reset the weapon swing timer when it triggers off special attacks, and results in a net increase in damage dealt."
Patch 2.1.3 reverted part of this change: "Sword Specialization: The change to Sword Specialization making its extra attacks appear in yellow has been reverted. Extra attacks will appear in white and act like any auto-attack. They will no longer reset the swing time of your weapon."
The discussion of the first change is what caught my attention.
After switching sword spec procs to white damage attacks, its value lined up much closer with CQC (4% value).
the resulting attack is made with the main hand, and is in all respects equivalent to a white attack made with that weapon except for the fact that it can't, itself, proc another sword spec attack.
How long ago was that mechanic tested? I ask because this topic came up on the official forums not too long ago in this thread:
In particular, I'd draw your attention to post #37 in which the screenshot posted seems to clearly indicate a successful (non-glancing) sword spec extra attack from the off-hand (going by the minimum damage). Is this a combat log parsing error by recount, or are we making incorrect assumptions about sword spec?
That thread also involved some testing by people to see the effect of Sword Spec if a non-sword (or less relevantly, no weapon at all) was equipped in the mainhand, in order to see if OH sword spec procs generate OH attacks, which it seems to me they do.
Based on that screenshot it does seem like there are sword spec attacks with OH damage. Also if you read the whole post, there was also a confirmation that if you only have an offhand and it's a sword, you still generate sword spec procs with your offhand.
But all that discussion just makes me want to do more testing. First, in that screenshot, average damage of sword spec hits does not seem to reflect a higher percentage of offhand procs as you would expect with a faster offhand. It can definitely be due to a very small dataset, but it's yet another reason to do more testing. What if they made it random which hand procs an extra attack as long as both are swords.
Second, Antares did not test if sword spec procs from yellow-damage attacks are MH attacks only.
From analysis perspective, this makes me wonder if combat potency procs from sword spec offhand attacks would outweight the difference in damage between offhand damage and mainhand damage.
@Jeppathum, in the Simulation currently Shiv does not proc sword spec. Once we can nail the sword spec mechanic, it will as long as offhand weapon is a sword.
Since you never Shiv, there's no value in coding that.
As for the posted log/thread, the order in the combat log is not to be trusted.
It can vary, but in general, the attack directly before the line indicating the proc is the procced attack.
I guess recount has no other way of knowing what damage was done by a sword spec proc, so it just takes the next line as an estimate. On the other hand, when another person only equipped an offhand sword and no mainhand weapon, he was still getting offhand sword spec procs.
EDIT: Sorry I guess it can still be due to same latency problems with combat log/recount. What that person thought were offhand sword spec procs, could have been regular attacks that appeared in a wrong spot in the combat log.
However you can construct an experiment which over a long period of time should help figure it out:
Ho: mainhand attacks proc mainhand sword spec damage and offhand attacks proc offhand sword spec damage
Experiment:
1. Use slow main hand sword and fast offhand sword, without enchants and trinkets, and record average damage of sword spec hits.
2. Swap the weapons and repeat the expriment.
- If Ho is true, swapping the weapons should not significantly change the average damage of sword spec hits. There can be some noise due to latency though.
- If however sword spec procs are only done with main hand damage, you will see a significant reduction in average damage of sword spec hits. While latency can weave in some wrong damage in that average, you would probably see a drastic difference in the two averages if Ho is false.
The problem is that you can't reliably track "sword spec hits".
A better way is to use a fast, low damage gray sword OH, and a slow, high damage MH with no haste effects, and just track the overall hits for each hand.
I'm still not sold on 15/15/5 being better than 18/51/2. 5/5 Relentless strikes seems to "energy cap" me far too often during AR or random streaks of "combat potency procs". With 18/51/2 I never have issues getting "energy capped" and using WWS reports (and napkin math) the 20% increased dmg ends up being greater than if I had 5/5 relentless strikes for me.
In fact, from the combat log I see main hand sword spec attacks.
Looking at the one at the top you have mains at
30; 32; 34; SS(34) 35; 36; 39; SS(41); 41; 41; 43
Main hands seem to be spaced about 2 seconds apart and there seems to be an extra main hand every time there is a sword spec proc (at 35 and at 41)
----
It should also be noted that I tested sword spec for 3.1 PTR and it was acting as expected.
Namely, sword spec hits proc main hand white attacks. Whether from a main hand or offhand and whether from a special or not (including Shiv).
Its also easy to track them. Look at the actual combat logs and the times between the attacks. Find which attack has an extra attack stuck in between the times it is supposed to proc.
Last edited by Dontmindme : 06/03/09 at 1:07 PM.
Reason: Tracking sword spec attacks
A full explanation of how the Sword Specialization mechanic works can be found here: PvE DPS / WotLK Discussion
The short version is that the combat log linked above does not contradict the explanation Aldriana has given. When you see "You gain 1 extra attacks through Sword Specialization" in your combat log, this line always immediately precedes the swing that causes the Sword Specialization proc. The second swing after the "You gain 1 extra attacks" line is always the Sword Specialization attack itself.
It was recount showing a low damage Sword Spec extra attack that put the doubt in my head actually, but as explained that might be an error with the way it parsed the log. I did a bit of testing with 0 haste rating in a standard 15/51/5 Sword build with [Angry Dread] and [Dalaran Sword] in order to make sure that there's no confusion between OH and MH sword spec procs, and it seems Sword Spec does indeed work as Aldriana explained. I don't know why I even doubted it in the first place to be honest. I did 1500 autoattacks on a heroic dummy, and then a level 60 dummy to eliminate some misses, both tests showed that all sword spec extra attacks come from the mace:
I'm no statistician, so I'm sure someone will tell me if they require bigger sample sizes, but at the very least these results don't show any reason to doubt the established sword spec mechanics.
I reviewed your examples in that old testing in Blasted Lands, and I totally agree with your logic.
It also appears that recount reads combat log according to the same rule: first line after "Sword Specialization" notice is the attack that caused it, and second line is the damage done by sword spec itself. See for instance this screenshot. This test was done with offhand weapon equipped only. There was one sword spec proc, and the damage assigned by recount to that one proc was the second line after the "Sword Specialization" line. By the way note that if there is no mainhand weapon equipped, the sword spec damage is equal to offhand damage, it did not cause a swing with an empty hand (it'd be smaller damage with no weapon).
Now with that in mind consider the recount report on this screenshot. If I understand the conclusions of that wow forum thread correctly, it's not the two sword spec lines on the screenshot that serve as evidence, it is the summary of sword spec damage prepared by recount in the same screenshot. As you can see, the damage done by sword spec hits ranges between 371 and 1215. The tester has Berserking but double proc of that can only explain a range from 371 to 525 (using combat spec multipliers for attack power). Even if you assume a greatness and pyrite infuser procced at the same time as double Berserking, it still cannot explain how the upper bound could be 1215. So they make a conclusion that it was both offhand and mainhand damage represented in sword spec damage summary.
This still does not deny latency between actual combat timing and combatlog representation of it as a potential explanation. But also something could have changed regarding sword spec mechanics since previous tests were done.
05/09/09, 2:32 AM
Similar Threads
