Blood - Stat Weights, Damage Taken vs. DPS

[Storm 426]

Introduction

---

Pretty much since 5.4 dropped we've had a general consensus that:

 

Mastery = best for survival

Haste/Crit = highest tank dps, if survival isn't a problem

Dodge/Parry = a good middle ground of Damage taken and damage done

 

However I haven't been able to find any sim's that really look hard at damage taken versus damage done for a Blood DK. So, I decided to go ahead and do some sims and see what I can come up with.

 

 

The Sims

---

All of these sims are done based on my character, currently item level 570, so a mix of normal and heroic gear with pretty much everything fully upgraded. My first sim is done with Elite skill level on a patchwork boss, and default TMI standard Boss.

 

Blood DPS Stat Weights

Crit - 1.54

Parry - 1.13

Haste - 1.04

Strength - 1.00

Dodge - 0.98

Mastery - 0

 

Apparently Crit is king right now, even for Blood DK's, and it's almost 50% better than Parry and Dodge. What I do find interesting though is that Haste is far less valuable than I expected. You'll actually gain at least as much dps from Parry/Dodge as you will from Haste!

 

Blood Damage Taken Stat Weights

Mastery - 2.03

Parry - 1.35

Dodge - 1.41

Strength - 1.00

Haste - 0.06

Crit - 0.05

 

This is about what I expected to see. Mastery is roughly 66% better than dodge/parry when it comes to reducing your damage taken per second. Str still makes a showing on this chart since it provides some parry. Unexpectedly Haste is a 0 for all practical purposes. I thought that haste would have at least some small benefit in the sims from faster rune regen, but apparently that benefit is negligible. Crit's non 0 value I'm going to just call a glitch, but I'll leave it in there because that's the data the sim gave me.

 

So that leads me to this list of Hybrid stat weights which average the DPS and DTPS gains from each stat:

Hybrid Average Stat Weights

Parry - 1.24

Dodge - 1.20

Mastery - 1.02

Strength - 1.00

Crit - 0.80

Haste - 0.55

 

 

Conclusions

---

 

For pure Survival: Mastery is way better than Dodge and Parry from a purely survival standpoint, 66% better. So if staying alive is a concern there's really no replacement for more mastery.

 

For Pure DPS: I just want to throw these number out there. If a piece of gear has Crit/Haste, and another piece of gear has an equal amount of Dodge/Parry. The stat weights from a purely tank DPS standpoint are:

 

Crit/Haste gear: 1.29

Dodge/Parry gear: 1.06

 

That's only a 20% dps difference between a 100% pure dps piece, and a 100% pure tank piece, and on the survival end of things dodge/parry is 100% better.

 

Overall Conclusions:

- Haste is trash now, avoid it at all costs. even if it's haste/mastery

- Crit/Mastery gear is slightly more acceptable

- Dodge/Mastery or Parry/Mastery gear is great

- Dodge/Parry gear is our best gear overall. (Except when you need more Mastery to survive)

 

Other Thoughts

---

I realize that DTPS isn't the true measure of a tanks survivability since it doesn't take into account spike damage. So you could argue that DPS gains from dodge/parry are irrelevant if you die due to them being weaker than mastery, or because of the RNG nature of avoidance stats. One could also argue that once you reach that certain plateau where death is no longer a threat then any further survival stats are wasted (this was our argument for Haste tanking in ToT). In either case I was just looking for a more quantifiable comparison between tanking and tank-dps'ing, assuming that most advanced tanks are concerned with both minimizing damage and maximizing their dps.

 

Also from an extremely advanced 'meter whore' perspective you could also argue that you could use the extra blood shields from higher mastery to take more damage on purpose and increase your vengeance. It's hard to simulate how much dps you can gain in this manner though, and it would be very situational.

[Tengenstein 11]

Sim with TMI as your scale factor, it's a metric specifically designed for measuring smoothness of damage intake. Also Tanks don't gain vengeance form environmental effects (shit on the floor), so if your meter whore idea is blood shield>stand in shit, it won't work

[Storm 426]

Standing in stuff works sometimes. for example if I pop cooldowns and stand in Garrosh's annihilate I can hit 400k vengeance by the second annihilate. but standing in a desecrated weapon doesn't generate vengeance. that's why I say its situational.

[Reniat 23]

How many iterations did you use to make these weights? 

 

also, Don't bother with damage taken stat weights, or TMI. TMI (for blood dk only. TMI is fine for other tanks afaik) is inherently inaccurate due to HOW blood dks achieve stability, which is to say it doesn't model remotely accurate DS timing. with the absence of healers it is fundamentally flawed, since achieving stability (the thing TMI aims to quantify) involves working with incoming heals to make a steady damage flow, or rhythm, which requires 1) a reactionary force and 2) healers, neither of which are accurately present in the sim for TMI to even try to work with. 

 

Your conclusions are fine but i would avoid saying avoidance is "best" in a broad stroke since it's only the best option if you are trying for a damage oriented build in the first place. Rather it is best for a damage build without scumbagging and stacking crit. 

[Storm 426]

I did 10,000 iterations.

 

If DTPS, and TMI are all inaccurate for DK's, is there a good way to find stat weights for defensive stats?

 

Also, what are your thoughts on my sim saying dodge and parry are roughly equal to haste and strength for dps gains as a tank?

[Reniat 23]

I did 10,000 iterations.

 

If DTPS, and TMI are all inaccurate for DK's, is there a good way to find stat weights for defensive stats?

 

Also, what are your thoughts on my sim saying dodge and parry are roughly equal to haste and strength for dps gains as a tank?

If you're going to make static weights go for 25,000 at least. 10k is still small enough that you could likely see noticeable RNG deviations between two sims.

 

There isn't a defensive set of stat weights for blood because there really isn't a point to defensive stat weights for blood, in it's current iteration. We choose stats based on HOW they work, not how potent. Avoidance is roughly even with mastery for total damage reduction, and can actually surpass in many situations, but because of it's fundamental RNG it will nearly always be inferior to mastery (in it's current form anyway) for raw survivability. We are just too prone to burst to rely on avoidance and have it come out as a net gain, and mastery is 100% reliable (because you control it), so regardless of the actual "weight" between the two, mastery>avoidance. 

 

Dodge/parry are good offensive stats, so they should be at least comparable to the rest. Haste should be slightly ahead until about 4-8k (depending on a lot of things), which is basically when you are unable to spend your resources. This is mostly because dodge/parry are only dps gains if you are actively tanking (and less so for 15-20 seconds afterwords). Haste gives its damage without needing to be actively tanking, which makes a difference. The difference is slight however, and since avoidance is better for survival I would say avoid haste unless you are for whatever reason partial to the style, in which case as long as you arent resource capped there isn't anything inherently wrong with a haste(to resourcecap)>avoidance build for say 10m content. 

 

Strength is a bit different. It's nearly impossible to put str in a stat weight for dps as a tank right now because of drastically variable AP. Str will be very good as a dps stat in low vengeance situations, and very bad as a dps stat in very high vengeance situations. This is because it doesn't scale with AP, whereas haste/crit/avoidance all do. So your weight will only be 100% accurate for the exact vengeance you got in the sim you made the weights with. 

[Tengenstein 11]

Reniat, you can add healers and then base TMIs based off just your own effect or off the combined effects of healers and yourself. And you can add reactionary force via the damage taken/health percentage lost in the last X seconds or if you drop below a static health threshold conditionals.

 

I haven't looked at the defualt Action list for Blood DKs, its' probably slightly off, but i'm fairly sure you could build a fairly accurate simulated priority with the tools SimC has.

Edited November 13, 2013 by Tengenstein

[Reniat 23]

The problem is you can't just add HoTs and say "you added healers". You'd need a healer sim that casted in reaction to your damage intake. 

 

Also there is an issue that bothers me with TMI where it chunks the data. Small chunks, but chunks nonetheless. Instead of one long string of damage/heals/mitigation it has lots of very small windows, and if you have a burst right before the end of a window that's going to reflect in the TMI result, because blood does have to dip down before we go back up (and no the death strike at the immediate front of the next window doesn't even it out. that's not how stability works). It makes sense from a computation standpoint to chunk it in this manner, and for the other tanking classes this really doesn't have much affect since they aren't NEARLY as reactionary, but it does make a difference for us. 

 

EDIT: just to be clear i dont think this is a fundamental flaw with the math or concepts behind TMI, and am a fan of theck and his work. It's just I don't think there has been remotely accurate enough implementation of TMI to make the data useful for blood. Case in point, people misusing simc and "finding" that spamming death strike off CD was better using TMI as proof, which is just poor use of well intended even if not useful data. 

[Tengenstein 11]

You can program the healers to do stuff based on your damage taken over a time window or health loss over a time window, or health thresh holds

 

TMI uses a moving average, so its not so much discrete 6 second periods (6 by defualt) but overlapping periods (one window is 1-6, the next 2-7, the next 3-8, etc). so each spike is averaged out, so aslong as you use a DS sometime  in the 5.99999recurring seconds after the spike event  that makes you want to DS  it will be recognized statistically, and the sooner you react to the spike the bigger the statistical effect will be.

 

I agree that it's very easy to misuse it to show some bullshit, as much like strength weights, it depends very highly on how hard the boss is hitting you, and if you set up a boss who has trouble getting through your blood shield before you refresh it spurts out bullshit

Edited November 14, 2013 by Tengenstein

[Reniat 23]

I was under the impression that there were no healers in TMI outside of hots. Could you let me know how to incorporate healer response into it?

 

Also, there is the other issue of not having great way to algorithmically(?) time death strike to accurately reflect real world timing, since you will react differently to two bursts of the same size depending on a lot of factors, and currently no ones gone through the trouble of making one because of just how ridiculously complicated it can be.

 

Can it get close? More or less, but I have a problem with making survival decisions that can (and imo should) be made via real time analysis as you tank and not with something that still doesn't quite accurately portray reality, and even if it WERE 100% accurate it would ONLY be accurate for the limited conditions of that specific sim. If you find that avoidance created better TMI for lots of adds, that would only matter for the specific number of adds and specific hit range of said adds that you had the sim set to. 

 

So my point i guess is even if you could make the sim work well enough and accurately enough, the results that result from it would be pretty isolated simply because we can tell without any simwork (as long as you have eyes and have your health somewhere in your UI) that mastery is smoother than avoidance for "normal" tanking, and anything else would be extremely specific. this is not even counting the fact that current tank dynamics basically says that for meaningful progression content you should be maxing damage as much as is safe anyway, meaning you won't pick a build based on even the most perfect TMI implementation. 

 

Just so i'm not shooting myself in the foot later, lemme clarify that I realize all of this may change. TMI might be implemented for blood perfectly down the road, and dynamics of tanking could swing back to wanting every scrap of smoothness, and they could change blood to the point where the RNG of avoidance doesn't auto-gg to mastery or they could change how DS works and DS/min becomes relevent again that would make the choice of mastery vs haste vs avoidance not so black and white in terms of survival. That would make TMI valuable for blood because you couldn't just easily figure out what is best because they would be all relatively close in potency, unlike now where even with approaching 60% avoidance I would rather have the mastery levels I did in t13 in terms of smoothing out damage. (but you gotta crank dem deeps right? sigh..looking forward to static AP again)

[Tengenstein 11]

That's true of all sims, not just Blood DKs, All tanks and DPS sims also suffer from the sim is only an accurate model for the fight it's modelling. I can't disagree that what stats you should stack are entirely down to what content you're doing and what (if anything) if causing tank death, but it's nice to have some simulated results to back up your expert knowledge,

 

I totally agree that with now Smoothness isn't the the sole goal for tanks its all bleh(and i'm mastery capping until we hit a progression boss enrage), and i really hope they do away with vengeance too.; it was originally a mechanic so tanks stacking stam, could keep up with the crit stacking fury warriors, but now all tanks either get DPS boosts out of stacking tanks stats or survivability out of stacking DPS stats it's basically a mechanic to fuck tanks over when a raid boss isn't hitting them. Seriously try questing as a prot pally on the timeless isle, you may not die (except to gulp frogs) but it takes an eternity to kill anything.

 

To incorporate healers into your TMI go to options tab and then in Actor only TMI select disable. then import a your healers, and copy paste them into your DKs simulation tab.Make sure their APLs are correct for how your healers heal (might want to make sure they're targeting you). I also tend to write my own bosses, something like the following to try and get a rough approximation of them.

 

enemy=”TMI_Standard_Boss_T15H25″

level=93
race=humanoid
role=tank
position=front
spec=unknown

actions.precombat=snapshot_stats

actions=auto_attack,damage=900000,attack_speed=1.5
actions+=/spell_dot,damage=30000,tick_time=2,num_ticks=15,aoe_tanks=1,if=!ticking
actions+=/melee_nuke,damage=1000000,cooldown=27,attack_speed=3,aoe_tanks=1

 

Thats not an actual boss though.