Jump to content


Photo

[5.4 Combat] I'm Not Dead Yet


  • Please log in to reply
174 replies to this topic

#121 shadowboy813

shadowboy813

    Von Kaiser

  • Members
  • 50 posts

Posted 12 October 2013 - 01:39 PM

The BF-Evis vs CT discussion is not as straightforward as it seems. I've done extensive testing in-game and Blade flurry does not behave like it has in the past.

In the current game, BF copies unmitigated damage which is then mitigated by the secondary target's armor. If your primary dps target is level 93 and you're cleaving/aoe'ing adds that are level 91-92, blade flurry will not necessarily do 40% damage to secondary targets. This also means that blade flurry benefits from glyph of sharpened knives to spread sunder.

Here's a combat log from when I was testing it. My primary autoattack target was the one sundered. You can clearly see my autoattack damage decrease without any corresponding change in the blade flurry damage once weakened armor expires.

1447 * 0.4 = 578.8 expected damage while sundered, but blade flurry was hitting for 565 damage.
After weakened armor faded, 1415 * 0.4 = 566 damage, which is about what blade flurry was hitting for the whole time (the secondary target was not sundered at any point).

I've done the test the other way where I left the primary target unsundered and sundered the BF target and the damage of BF did decrease once sunder expired on the BF target. BF did more damage than expected while the primary target was unsundered and secondary was sundered.

17:10:58> Shadowvenom's melee swing hits Training Dummy for 1 Physical.(1447 Overkill)
17:10:59> Shadowvenom's Blade Flurry hits Training Dummy for 1 Physical.(564 Overkill)
17:10:59> Shadowvenom's melee swing hits Training Dummy for 1 Physical.(1446 Overkill)
17:11:00> Shadowvenom's Blade Flurry hits Training Dummy for 1 Physical.(564 Overkill)
17:11:01> Shadowvenom's Weakened Armor dissipates from Training Dummy.
17:11:01> Shadowvenom's melee swing hits Training Dummy for 1 Physical.(1414 Overkill)
17:11:02> Shadowvenom's Blade Flurry hits Training Dummy for 1 Physical.(565 Overkill)
17:11:02> Shadowvenom's melee swing hits Training Dummy for 1 Physical.(1415 Overkill)
17:11:03> Shadowvenom's Blade Flurry hits Training Dummy for 1 Physical.(566 Overkill)

The long and short of it is that treating blade flurry as copying 40% of the damage exactly without considering sunder is going to favor blade flurry and raise the target cap needed for CT to be worthwhile. That said, don't think it will change things much as the cutoff point might be one target less. The problem with CT is that most of the damage is in the bleed.

The other (minor) thing I noticed with Pathal's update to the spreadsheet is that it is not considering the DP-I proc from MG on the eviscerate side. This favors CT so between the armor and this it's probably not a very different conclusion.

#122 Zeemus

Zeemus

    Glass Joe

  • Members
  • 1 posts

Posted 14 October 2013 - 02:29 PM

edit: Found my answer.

#123 thottstation

thottstation

    Glass Joe

  • Members
  • 9 posts

Posted 18 October 2013 - 10:14 PM

I've run some numbers and turns out it might not be "a lot of targets" after all. If my calculations are correct, Crimson Tempest precedes Eviscerate as a finisher already with 2 adds:


The problem with that (as I see it) is that Crimson tempest is a DoT. With Eviscerate, it's instant, high damage on everyone. Withg Crimson tempest, a DoT is applied... on addas that are usually dead 4-5 seconds later, meaning the DoT doesn't get to inflict it's full damage, rendereing any "increase" rrelevant, since it doesn't get to inflict it's total damage.

ie.
Good: 100% damage instant
vs
Bad: 200% damage spread over 10secs, but adds die 3secs later.

#124 phantom57

phantom57

    Glass Joe

  • Members
  • 3 posts

Posted 23 October 2013 - 06:10 PM

1 short question - why is seismic bore above shao-tien saber. both give haste, but as we lack expertise on gear it will be way more usefull than the crit from seismic bore - it shares the same racial for humans and has a bit more haste.



#125 Pathal

Pathal

    Don Flamenco

  • • Guide Author
  • 342 posts

Posted 23 October 2013 - 06:35 PM

Long story short, it's because reforges work out that way.

Reforging means items don't have static values, but a range of value depending on how it's reforged. You can get the expertise from somewhere else, and the value of crit equals mastery at some points in high gear sets.

#126 MasterofVirus

MasterofVirus

    Glass Joe

  • Members
  • 2 posts

Posted 25 October 2013 - 03:10 PM

Question: What gives more DPS. Combine Synapse Springs with KS or with AR + SB?

 

greetings



#127 phup

phup

    Von Kaiser

  • Members
  • 40 posts

Posted 25 October 2013 - 05:21 PM

So, to update the CT vs Eviscerate discussion, I tweaked the calculations a bit.

https://docs.google....drive_web#gid=0

I'm getting something a little different than you, but still reasonably low. I don't think the breakpoint to start casting CT is until BF is maxed out on targets. And that's just the breakpoint in damage assuming it lasts the full DoT before a refresh. Even if I add support for things like "refresh in the last <3s", the breakpoint doesn't move at all since you're relying on a low chance for DP refreshes.

There probably isn't any reason to start casting CT until you're close to 10 targets or more. This isn't quite an exact answer still, but it's closer to what we need.

 

I was looking at that and I can't remember why Eviscerate would have a lower net energy cost than Crimson Tempest.



#128 Pathal

Pathal

    Don Flamenco

  • • Guide Author
  • 342 posts

Posted 25 October 2013 - 06:08 PM

Eviscerate can trigger Main Gauche, which can proc Combat Potency 20% of the time. At 50% mastery proc chance, that's 10% of the time it costs 15 energy less. Or a 1.5 energy reduction on average.

http://wowpedia.org/Combat_Potency

#129 phup

phup

    Von Kaiser

  • Members
  • 40 posts

Posted 25 October 2013 - 06:19 PM

Gotcha.  Fierydemise pointed that out over on the wow forums as well.  However, I still have one gripe with your spreadsheet.

 
Is it right to use net energy to make the comparison?  if you compare based on 33.5 energy and 35 energy instead of 8.5 and 10, then crimson tempest is better at just 4 targets instead of 6.
 
I really think that is the more valid comparison.  Shouldn't relentless strikes just be considered part of your overall energy regen (every combo point spent = 5 energy regen on average, regardless of how you spend them) and not really used in determining the efficiency of a finishing move?
 
For example, pretend there was a finisher that cost 25.00000001 energy and did 25 damage. Nobody in their right mind would ever use this ability.  And its DPE the way I look at it is 25/25.00000001 which is roughly 1 dpe.  However, If you use net energy the way your spreadhseet does, it's 25 / 0.00000001 which is like 2.5 billion DPE which is clearly ridiculous.  Including relentless strikes energy return gives false and/or absurd additional value to finishers the less they cost.  You should really just use the base energy cost when trying to determine relative values of finishers, which means 4+ targets is good for CT.

Edited by phup, 25 October 2013 - 06:19 PM.


#130 Pathal

Pathal

    Don Flamenco

  • • Guide Author
  • 342 posts

Posted 25 October 2013 - 06:47 PM

I'll break this into two parts, since it looks like you actually have two concerns.

Relentless Strikes depends on how abstractly you think about the calculations. In general, it's best to attach energy mechanics to things that trigger them when possible (Venomous Wounds is an exception, but that's another topic). The reason you do this is because the net result is the same, and you need to account for that energy generated somehow. If we have a finisher that costs 35 energy, but generates 25, it's a lot easier to think of it as a 10 energy finisher, than it is to think of it as damage plus energy (which becomes damage). Either case, we're 10 energy lower than we would have been, so this simplicity allows for easier computation for the same final result. Treating RS as energy regen that increases with haste is pretty complicated and requires iteration to convergence to get any meaningful result. We can't really sit here and do iteration to convergence on a simple spreadsheet. If I don't account for the energy from these in the cost, how do you intend to compute the value of the energy? It's not ideal when calculating damage from finishers in this isolated scenario, but it's better than ignoring it. What would be ideal, would be the full energy cost of the build up to a CT, include the damage from CT, and then compare the full cycle DPE. But I'm just trying to improve what was already done. Not create the end all version.

 

Your second point is about the validity of DPEnergy in this case. DPE isn't always the best. As you pointed out, the 25 damage, >billion DPE is broken. Someone could do 26/5 and it would still be better at >5 e/s and 1s GCD. The problem is that its entirely dependent on resources. The moment you are no longer resource constrained and become GCD constrained is when DPE stops mattering. The value of energy also depends on the rotation being executed, so you can't pretend they're a constant value and ignore them in simplistic comparisons. But it also requires what I said earlier, it's best used with a full cycle.

 

Lastly, I want to remind everyone this is done assuming the full DoT damage is dealt. If the targets aren't alive for the full DoT, the calculations change quite a bit, and casting CT before the DoT has <3s left also has a large impact on performance. Practical gains aren't until some point after the breakpoint, for many reasons including some mobs dying quicker than others.


Edited by Pathal, 25 October 2013 - 06:49 PM.


#131 phup

phup

    Von Kaiser

  • Members
  • 40 posts

Posted 25 October 2013 - 07:28 PM

I'm assuming we are still energy constrained as we still are 99.9999% of the time when blade flurry is turned on, and I'm not saying DPE isn't a good comparison, i think it's the best comparison in this situation.  I'm saying that your calculation of DPE that includes relentless strikes (RS) isn't the right way to determine DPE.  Both of those finishers trigger RS, so throw it out, it doesn't impact which is stronger in an energy constrained situation. What matters is how much energy they actually cost, not how much energy you end up with afterwards due to other things triggering.  You can't do an eviscerate at 8.5 energy.

 

Here is another hypothetical example.  Assume we are still energy constrained.  Say you had a choice of two eviscerates.  One that cost 50 energy and did 50 damage, and one that cost 75 energy and did 75 damage.  By your calculation, the 50 energy evisc with a net cost of 25 is 2 dpe, but I say it's 1 dpe.  And you say the 75 energy one at 50 net is 1.5 dpe.  Clearly worse by your standard but I still say it's 1dpe and they are the same as far as energy efficiency goes.  And in the long run, using the 75 energy eviscerate will be better because while its DPE is the same it's more efficient in using combo points.

 

Or for a more practical hypothetical demonstration of why including relentless strikes is wrong, pretend you reduce the total damage you get from using CT by 20% but you also reduce its energy cost by 20% from 35 to 28.  According to me, it's DPE shouldn't change.  You're spending 20% less energy and getting 20% less damage.  However, if you make those changes in your spreadsheet, CT becomes better than eviscerate at just one target because its net energy is 3.  But there is no way we magically made CT more energy efficient when we reduced it's damage by the exact percentage that we reduced its cost.   But if you use a proper comparison of 33.5 for eviscerate and 28 for crimson tempest it remains only better at 4+ targets.   I hope that makes it clear that you really should not include RS energy return when trying to compare the DPE of two finishers.


Edited by phup, 25 October 2013 - 07:30 PM.


#132 Pathal

Pathal

    Don Flamenco

  • • Guide Author
  • 342 posts

Posted 25 October 2013 - 08:10 PM

You can ignore it if you want, I don't really care. Since the start I've said it's not the final exact answer, and I've argued that calculating the DPE of the full cycle is better than what's currently shown. Even Shadowboy's bug mention causes variation in the damage calculations alone. I'm not going to meticulously recreate the wheel with all the inputs that impact this issue. Just bringing something simple like MG and auras on secondary targets required like double the calculations. DPE is what we need to calculate, it's just that this doesn't really handle the breadth of the situation.

 

If you were to calculate the full cycle's DPE, as you should be if you wanted to do this right, you would still have to treat RS as a net energy cost, instead of a haste inflated source of energy generation. You can't pretend 1e is worth x across all rotations. It has to be dependent on the cycles in the calculations. RS procs with with the rate of finishers, which is dependent on energy generation. It's a feedback loop that is severed (that also immediately reaches the limit value) by reducing the net energy consumption of the cycle.



#133 phup

phup

    Von Kaiser

  • Members
  • 40 posts

Posted 25 October 2013 - 09:26 PM

What if CT happened to cost 25 energy, how would your version of the comparison work?

 

Don't get me wrong, the spreadsheet is awesome, it takes into account about a dozen things i would have forgotten in trying to compare the two. It's just that if you take RS out of the cost before you calculate the value of the finisher, you are inflating the value of that 1.5 energy you get back from main gauche combat potency.  The way you do it, it makes it seem like a 15% improvement to eviscerate's DPE when it's not.  My method of comparing 33.5 to 35 isn't actually correct either. It's making the same mistake, just on a smaller scale.

 

Overall, you are right.  You would have to calculate the impact on the whole system to truly compare the two.  It would be great if we could put a damage value on one point of energy in a particular rotation.  CT and Evisc do x amount of damage for y amount of net resources.  The only difference in resources is 1.5 energy points.  The damage is calculated down to the last detail in your spreadsheet.  So once your spreadsheet get to 4 targets and the damage from CT is more than Eviscerate's damage the question is...  what's worth more,  the extra damage from CT or the extra 1.5 energy from eviscerate.

 

Running SimCraft does attempt to do just that, and they've been trying to fix the CT and blade flurry modeling.  I don't know how accurate it is, but with SimC in BiS gear, at 6 targets they are essentially even, and at 7 targets CT pulls ahead

 

Edit: Also, I just like to argue.  I think your general conclusion is spot on.  It's going to be the rare situation where 6 or even 4 mobs stay alive long enough and in adequate numbers for CT to be a (usually small) gain.  And even then it's risky because clipping it too early or mobs dying before you expected could turn that CT into a mistake.


Edited by phup, 25 October 2013 - 09:35 PM.


#134 Dinth

Dinth

    Glass Joe

  • Members
  • 4 posts
  • LocationDallas, TX

Posted 27 November 2013 - 02:53 AM

Haven't seen this asked before, and it's something I've been thinking about with how closely our KS and AR/SB cd's line up with AOC; ignoring the concept of energy capping, which I think is silly anyways, could it technically be a dps increase to use KS with 3 seconds remaining on AR? SB is 3 seconds shorter than AR, and KS lasts 3 seconds, so there's a perfect window there to use KS and benefit from the 20% melee attack speed from AR and the 50% increased damage during KS.

 

It's probably pretty minimal, but I'd be interested to see how much of a gain it would be, if indeed it was.



#135 phup

phup

    Von Kaiser

  • Members
  • 40 posts

Posted 03 December 2013 - 09:38 PM

I can't imagine that's anything but a loss.  It's a 3 second 20% boost to your passive damage (over what you would have gotten doing KS outside of AR) at the cost of a lot of energy and 3 or 4 lowered GCD abilities.

 

Changing the simulationcraft rule for killing spree to allow it in the last 3 seconds of adrenaline rush results in a 1.5% dps loss.



#136 Jarush

Jarush

    Von Kaiser

  • Members
  • 69 posts

Posted 04 December 2013 - 02:35 PM

Is it really a dps increase to always wait adrenaline rush to end before using killing spree? In some cases it's many finishers and loss of much cd reduction to killing spree. I've thought it's best to use kspree when it's off cd (ofc always before using AR if possible) because it's cd is so short with restless blades + assurance. With those reductions kspree always comes from cd during AR so is it really worth it to wait till the end of AR before using it? You will lose many killing sprees by delaying it.



#137 phup

phup

    Von Kaiser

  • Members
  • 40 posts

Posted 04 December 2013 - 08:35 PM

According ot SimC, you are correct.  In SimC, the current rule for KS is basically use when AR is down and energy is below 35.  I tried different variations of that rule.

 

1. Default (AR must be down, energy must be below 35) - 423k dps BiS gear

2. No restrictions, use KS on cooldown regardless of energy or AR status - 427k dps

3. Energy has to be below 35, but using it during AR is fine. - 430k

4. Energy has to be below 45, but using it during AR is fine. - 433k

5. Energy has to be below 60, but using it during AR is fine. - 431k

 

Basically it seems to say get your energy low(ish), but use KS pretty much on cooldown regardless of AR status.

 

Waiting until AR was over has been burned into my muscle memory for so long, I find that surprising, but I'm definitely going to try out more aggressive KS usage during AR.



#138 Dinth

Dinth

    Glass Joe

  • Members
  • 4 posts
  • LocationDallas, TX

Posted 05 December 2013 - 01:11 AM

Do those variations assume that SB is being used in conjunction with AR? I would think the loss of CP generation would have more of an impact than any "lost" energy.



#139 phup

phup

    Von Kaiser

  • Members
  • 40 posts

Posted 05 December 2013 - 05:14 PM

the action list doesn't explicitly force them to happen at the same time, but due to the way it's setup, that's what ends up happening just as if they were macro'd together.  and looking at sample outputs, it does occasionally do KS while both are up, but it seems to think that getting in extra KS makes up for the wasted energy regen / lost cp generation.



#140 Pathal

Pathal

    Don Flamenco

  • • Guide Author
  • 342 posts

Posted 05 December 2013 - 08:51 PM

I started playing around with that last night. I originally thought it might be a bug, where the player could still cast abilities while all 3 were up, but exporting a combat log seemed to show otherwise.

 

This could be a due to a combination of the sheer massive amount of passive damage during AR+SB that gets buffed by the 50% modifier during KS. Shadow Blades ignores armor, the rapid attack speed increase buffs that further, the loss of the DW penalty during SB means more hits land, and they all proc poisons and MG more often and harder. By comparison, you lose 3 seconds of GCDs, which is only ~3 GCDs due to how the .8s GCD lines up with the 3s duration and the constant length of AR+SB.

 

Additionally, this appears to only be an increase during AR+SB, and the damage per execute of various abilities change noticeably from such a small change. Delaying KS until SB is down, but okay when AR is up, is worse than the default on SimC. And it looks like there's a damage increase per execute to KS, the only thing that should increase it's damage would be Bandit's Guile or bad luck prevention for RPPM causing various points in the fight to be more likely to have a proc (think start of the fight). As a result, this might not be true for every gear level.

 

Shadowcraft isn't set up well for this. It would probably be best for people to get Iron Jugg logs (or maybe target dummy logs) to investigate further. It could still be an issue where the default profile for Combat is just under optimised.


Edited by Pathal, 05 December 2013 - 08:52 PM.





0 user(s) are reading this topic

0 members, 0 guests, 0 anonymous users