Death Knight: Simple Questions/Simple Answers - Page 67

Chairon · 02/03/09, 1:06 AM

I have been cooking up a hypothesis on an alternate way of calculating the DK stat weights on how they affect DPS. My goal is to get some insight on how viable this new method would be.

Let's say one was to take all the significant stats a DK has and statistically model them on how they will affect output DPS.

1) Strength
2) Attack Power
3) Critical Hit Rating
4) Agility
5) Armor Penetration Rating
6) Haste Rating
7) Weapon DPS

Once gathered, could you use all these as predictors in a logistic regression to determine an outcome variable, namely DPS?

I propose collecting multiple sustained DPS parses, then separating each predictor into 5 different dichotomous bands (0 if the DPS predictor does not fall in the range, 1 if it does).

Once each predictor is defined by one of these bands, we could then place them in a regression to model the continuous variable, DPS.

A model summary would generate respective beta coefficients for each of the prediction variables. The relative weight for each predictor would be the individual standardized beta coefficient divided by the sum of all the beta coefficients.

Areas of concern with this method:

1) I'm not quite sure what type of regression model I should use. I initially thought I could use a binary logistic regression, but since DPS is a continuous variable, would a linear regression be better suited?

2) Data collection. The DPS parses would all have to be under the same conditions. Perhaps a target dummy?

Any feedback would be appreciated. I have been using the commonly accepted stat weights up to this point, and will continue to do so. I'm just wondering if there are alternative ways to look at the core of our DPS.

Edit: Also added weapon DPS

foolish_fool · 02/03/09, 1:28 AM

I am quite certain that you will find that type of modeling to be a statistical nightmare, once you begin. There are so many variables to take into account even in the setting up of such an experiment, not to mention the complex interaction between factors that do not act linearly. You could make simplifying assumptions, but that would greatly decrease the validity of what you are actually trying to achieve. You would likely end up having to write a simulator just to get sufficient consistent data: in which case you may as well just use said sim to calculate stat weights.

Kithic · 02/03/09, 1:46 AM

Ok, I need help.

Ive been using a 2h frost build 21/50 with a Titansteel destroyed, and as my gear has improved ive been seeing great numbers using the plaguestrike-less rotation.

TAP
>IT>BS>FS>OB>FS>OB>DMP

Today thought, I got hold of a Hatestrike and went and bought a Grasscutter with saved up badges. I use frost since I raid with a 2h 17/0/54 DK, and I have heard good things about 0/44/27 DW build, so I went, found the spec, hit the dummy and sadface.

I know the limitations of using a dummy. I know the dps results arent as authentic as a raid environment, but using it as a control to compare 2 specs shouldnt be overly problematic. I also have read that if anything, it should BENEFIT the dw spec due to HB not needing sunders to be at its best, which only increases my concern.

with 2h I avg 2700 dps on dummy, sans pet.
with 44/27 I havent come close to that, even with a perma pet.

IT PS HB BS BS DMP
IT PS HB IT IT DMP x repeat

is what I have been using, im not sure if Im doing it wrong or what, but the best I can manage in the exact same gear is 2300 (WITH GHOUL) and this doesnt seem right.

What I really need is some help getting the optimal rotation, ( I think what my friend told me to use, isnt quite right)

Thanks for your help!

One last thing, Im using IT, FS, Ghoul glyphs, and IT Sigil. FC on Hatestrike (MH) CG on Grasscutter (OH)

edit : forgot to post the rotation ive been using for DW 44/27

Kashir · 02/03/09, 1:53 AM

Originally Posted by Lazare

However, once you include IBF scaling, it's hard to see how you can avoid the conclusion that defence wins. You'd have to assume that 1% increased damage reduction on IBF was worth only 0.11% avoidance for dodge to still win. Or to put it another way, IBF would have to be up for only 11% of all incoming damage - when just hitting the button every cooldown should average 20%, worst case. Unless I'm missing something basic, I don't see how dodge can still win.

Death Knight tanking is reliant on cooldowns. The most dangerous period is when our cooldowns are down, while the safest period is when IBF is up (~35% mitigation is better than anything but Lichborne).

Taking your figures, we could still make the argument that a constant 0.11% avoidance (during both the safe and dangerous periods) is worth more than 1% mitigation during the safe period. In other words, IBF is "good enough" already; what we need is a safety-net when IBF, UA, BA, VB etc. are not active.

Dodge and defense are both excellent stats, it's very rare that you'll have to make a 1:1 decision between the two, and even if you make the "wrong" decision it's not going to matter much.

JALbert · 02/03/09, 2:13 AM

Originally Posted by Kashir

Death Knight tanking is reliant on cooldowns. The most dangerous period is when our cooldowns are down, while the safest period is when IBF is up (~35% mitigation is better than anything but Lichborne).

Taking your figures, we could still make the argument that a constant 0.11% avoidance (during both the safe and dangerous periods) is worth more than 1% mitigation during the safe period. In other words, IBF is "good enough" already; what we need is a safety-net when IBF, UA, BA, VB etc. are not active.

Dodge and defense are both excellent stats, it's very rare that you'll have to make a 1:1 decision between the two, and even if you make the "wrong" decision it's not going to matter much.

If you're worried about dying to a spike while off cooldowns, then avoidance isn't going to help you either. At that point you're looking at Stam being the best option.

Chairon · 02/03/09, 2:21 AM

Originally Posted by foolish_fool

I am quite certain that you will find that type of modeling to be a statistical nightmare, once you begin. There are so many variables to take into account even in the setting up of such an experiment, not to mention the complex interaction between factors that do not act linearly. You could make simplifying assumptions, but that would greatly decrease the validity of what you are actually trying to achieve. You would likely end up having to write a simulator just to get sufficient consistent data: in which case you may as well just use said sim to calculate stat weights.

I think that I could do a linear decomposition to determine the interactions between the predictors. Are you saying that gathering accurat data would be next to impossible?

Facemelt · 02/03/09, 9:42 AM

Hey guys,

I was pretty happy with my current spec, but as I started to get to harder content, esp. to bosses who do lots of magic dmg, I though about changing a few things.

My armory: The World of Warcraft Armory
My talents: The World of Warcraft Armory

I thought about getting 3 points in acclimation (instead Rune tap and 2h spec) but I keep hearing different things about the usefulness of this talent...

Ritualist · 02/03/09, 2:39 PM

I am completely underwhelmed with Cinderglacier/Razorice for my offhand Split Greathammer. Is Berserking a viable option, or should I just stick with one of those two? I really don't see any change whatsoever when I alternate between the two. I thought I would mention that our mages are arcane so Razorice doesn't benefit us much as others.

My Armory

♦ Darkside · 02/03/09, 2:44 PM

Originally Posted by Ritualist

I am completely underwhelmed with Cinderglacier/Razorice for my offhand Split Greathammer. Is Berserking a viable option, or should I just stick with one of those two? I really don't see any change whatsoever when I alternate between the two. I thought I would mention that our mages are arcane so Razorice doesn't benefit us much as others.

My Armory

Beserking is NEVER an option for a Deathknight. If you don't like the performance of Cinderglaicer or Razorice, go 2x Fallen Crusader (it will probably increase the uptime by 60% or so), but never ever get a "normal" enchant. Runeforges are infinitely better.

DanoPDX · 02/03/09, 2:54 PM

Simple question ----------------------

I am getting mixed advice on tanking caps for hit and expertise. I have great DPS gear but want to start saving up for tanking spec.

Getting 540 defense is the first part to worry about Which I don't even have that with just blues. (I have Rep rewards, crafted, and some drops that have DEF) I am a blacksmith so I can add 3 gem slots for hit or dodge or whatever.

So if I try to get hit cap and expertise cap I will loose dodge or defense.

Someone said hit was not really as important, about 200 is ok which I think is bull. That's 6.10 (Aprox) hit. Then they said but get about 24-26 expertise is next priority, which I also think is bull. And yes this was with a 2H not a DW...

I have experimented with expertise and find that 8% chance to hit (I almost always have a draenei in my group) and 12ish expertise is fine when I am DPS in a 25 man. Maybe three attacks at most will get dodged in a fight, maybe 1 depending on how long the fight is.

Can I apply the same theory to tanking? I am looking to tank 5 man heroics or 10 man Naxx to start and I don't want to be squishy. Is 26 expertise more important than hit, and is the claim that 6% chance to hit is all that's needed valid in any way?

What should I be going for in tanking stats?

540 Def
??? Avoidance (Total added together from Dodge, Parry, and where ever else since DKs have no shield)
?? Hit
?? Expertise

Where can I go for more info -------------------------------

Can you point me to a thread that has updated (3.0.8) tanking advice? I have seen some threads but it is hard to find the info I want on the first page with latest changes to DKs and hard to wade through 54 pages of info.

Where should I be for these things:

1) Avoidance (Parry and Dodge) and which is easiest to raise up to start off with? Pros and cons of dodge vs parry.
2) Stats on gear. Agility gives Crit and Dodge, Str gives AP and Parry.. Which will get avoidance quickest without lowering threat. (In otherwords having high crit but low AP, or vise versa)
3) Which is better for generating max threat? AP or STR and I say that because some bonuses are calculated off the STR stat or Armor from items but don't calculate a bonus from total AP. Pros and cons of white damage and IT, PS, BS, FS...

rickf · 02/03/09, 4:13 PM

I'm quite familiar with gemming at lvl 80, but while I level (currently 65) my 2H Unholy DK I've come across a couple of items with sockets. On top of that I'm a lvl 450 JC.

While I'm in the 65 - 75 range I'm obviously going to be uncapped on exp and hit. Since I am a JC I have many gems laying around (in other words - money for gems is no object - I can change them every day if I wish.)

So... Should I just pile on the str while leveling, or should I try to get my hit and exp up? I'm pretty sure I could cap my exp, but I won't be fighting many bosses until I approach 80.

What do you suggest? Str? Sta? Hit? Exp?

Thank you.

♦ Darkside · 02/03/09, 4:23 PM

It hardly matters at all. You're going to encounter almost no pieces with gear with sockets on them and what you do get you'll end up replacing in two levels. Leveling as a DK is the easiest thing you'll ever do, just grab the cheapest gem you can find and be on your way.

Pyros · 02/03/09, 4:40 PM

Just gem bold bloodstones in everything if you do run into socketed gear, but while there's a few in BC content, in northrend there's almost no socket gear at all, even when you run instances. You'd probably have enough if you cut like 5 of them.

foolish_fool · 02/03/09, 5:55 PM

Originally Posted by Chairon

I think that I could do a linear decomposition to determine the interactions between the predictors. Are you saying that gathering accurat data would be next to impossible?

Yes. Mainly because test dummy performance is not going to provide accurate enough data for you to calculate stat weights that are accurate in a raid environment, and because actual raid parses will vary immensely, not only on simple things like what raid buffs are present (which you could probably account for), but also vary by things like the other dps in the raid, as well as various "RNG" factors.

Aihime · 02/03/09, 8:13 PM

okay, so i've read the frost and unholy dps thread but frost says 9% for hit cap and for unholy it's 8%
which one is which?