I noticed that in the "most promising" Frost DW tree you linked to didn't have Deathchill. I always thought that being guaranteed a critical strike every 2 minutes was something that was too good to pass up.

I'm not trying to sound like a jerk, but could I please be given an explanation why this was ruled out? Is it lacking in dps or is it a waste of a point, etc.?

Glancing blows are fairly easy; the best understanding (to my knowledge) is that 24% of your white hits glance against bosses and each glancing blow does 70% normal damage, which means, for the purposes of this sort of math, your white damage is reduced by 16.8%.

ArP is quite a bit harder to include, since it varies from person to person or even during a fight thanks to effects like mjolnir, time spent stacking sunders, or shattering throw, but here are a couple numbers assuming fully debuffed boss (sunder/ff): at no armor pen, mitigation is 35%; at 30% ArP, mitigatin is about 27%, and at 100% ArP, mitigation is 2%. Also, don't forget about FS partial resists, which account for ~5% damage loss, iirc.

One last note, you mentioned 30% personal haste, but because of multiplicative haste buff stacking, non-IIT spec only needs 5.2% haste on gear (not 7%) to break 30% melee haste with, and IIT spec has nearly that with zero haste gear (7% haste on an IIT spec has almost 39% melee haste).

@direddyre

It is fantastic to have 'clear cut' theorycrafting and thought process behind the numbers. But you are forgetting quite possibly the most important aspect of being a DK. Our damage is not based on weapon speed, but rather individual strike damage. The argument isn't so "clear cut" when you think of weapon DPS in this manor.

Take a look at base rank 4 Obliterate for a second:

Obliterate: A brutal instant attack that deals 80% weapon damage plus 467.2, total damage increased 12.5% per each of your diseases on the target, but consumes the diseases.

The modifier that is used, is not the "average dps", but rather the roll of that individual strike.

In the case of the items you used for comparison:

[Lionhead Slasher] strike damage is: 206-384
Average strike = (206+384) / 2 = 295

[Aledar's Battlestar] strike damage is 297-552
Average Strike = (297+552) / 2 = 424.5

295 / 424.5 * 100
= 69%

Meaning that your -average- strike with [Lionhead Slasher] will be 31% less than those of the [Aledar's Battlestar]. In fact, the [Lionhead Slasher] is so inferior, that its -average- strike damage is in fact, less than the MINIMUM strike damage of the [Aledar's Battlestar].

Hopefully this helps to iron out your confusion, and prevent the question from being asked again. Then again, that is what we thought last time the discussion came and passed.

To put it briefly, it is a low value of DPS for the point.

Think of it this way, if you have a 30% crit chance in raids, and you also have Killing Machine procs, this devalues additional crit. The same goes for a button you can press every 2 minutes to guarantee a crit.

If it is used when you have a KM crit on a KM eligible attack, you would have crit anyway.

If you have a 30% crit rate, then you are gaining 70% crit chance on one attack every 2 minutes.

I believe if you ponder on this you will see it as a low value talent, at least in terms of PvE, with no mathematical figuring necessary.

The interesting question would be how it interacts with crit. Does crit push glancing off the table like it does with hits? Or can a crit be converted to a glancing blow? Depending on that it could have a much lower effect.
Also it would probably not affect BCB since BCB cannot glance, but it would affect Necrosis.

(I think you calculated 70% of 24%, you probably mean 30% of 24%, making the dps loss quite a bit lower at 7.2%)

Good point about FS partial resists
Due to the way haste works (Increasing amount of attacks in the same time frame instead of reducing attack speed) I don't think it can actually stack multiplicatively, it's like crit in that it just adds on top. I did forget the 5% personal haste from IIT though.

Overall I'm a bit too lazy to redo calculations, especially since the point still stands/is even stronger than before.

Sorry for multipost, this reply wasn't up when I did the previous one.

As I said in the intro, all instant strike scaling is normalized (has been for a very long time) so the modifier for attack power (which is basically what you're referring to) is the same for any weapon speed, it is in fact just the average damage on the weapon that makes all the difference for us when it comes to weapon speed.

I have found the real culprit of the random blood/death rune if you blood strike 2 times in a row too fast it will convert one blood rune into a death rune twice. Try to stagger the second blood strike by a split second, you can wait till the rune converts to a death rune and usually you wont lose too much but that is for sure why it happens.

You should start by reading the first post in this thread. This, along with a lot of other information, is already laid out.

I am still unsure that you are calculating this correctly. Perhaps you could explain this to me more clearly.

Using -average- calculations:

How is 80% of 295 + 467.2 (or 703.2) going to do more damage than 80% of 424.5 +467.2 (or 806.8)?

Lets ignore averages, and assume max strike modifiers

80% of 384 + 467.2
equals 774.4

80% of 552 + 467.2
equals 908.8

This is -without- a contribution of AP. To my understanding, Our strikes are all based off your damage on the initial strike which is then modified by AP. My numbers are based on individual weapon strike damage without that modification. Speed and AP have no impact on this comparison at all apart from the fact that we are comparing a faster weapon with a lower strike damage to a slower weapon with a higher strike damage. But in the end, all that matter is what those strike damages are.

Is there something I am missing? Please help me understand what you mean by normalization of the AP modifier (when I am not using an AP modifier at all). To be quite honest, my thought process reflects my in game results. Has your math applied in game confirmed your conclusions? If you are correct (which I am not convinced of yet), then we have been misleading people in this thread for quite some time and I think as a community we should ensure that is -not- the scenario.

Note that I'm not saying that our strikes will do more damage with a faster weapon, a weapon with less average damage will always do less damage on instant strikes, I was merely exploring how that relates to the damage you gain through auto attacks.

How scaling on instant Strikes works is simple, 14 AP represents 1dps. If your weapon has a speed of 2.5, then that means 14 AP will add 2.5 damage to each strike. This was a problem for instant strikes, since that made weapon speed an incredibly powerful stat (many rogues were using a blue 2.0 speed dagger instead of the epic 1.8 speed ones for example) so they changed it that every weapon type has a static speed modifier. I'm not 100% sure what those modifiers are, but I think I remember it was 3.4 for 2h, 2.7 for 1h and 1.8 for daggers (don't quote me on this, it's not even important whether those numbers are correct, it's just important that they exist). That is what I mean by normalization.

So regardless of speed, 14 AP add 2.7 damage to an instant strike if you use a 1h non dagger weapon (assuming 100% weapon damage etc).

So, the damage on the weapon itself is just the base damage, for strikes it has no relevance whatsoever concerning the scaling via AP. If one weapon adds a static 50 damage to each strike, but loses 50 dps to auto attacks, which one is better? That was the question I was trying to find the answer to.

The way I understand what you're saying is, that if one weapon hits for 10% as much as another weapon, the strikes will also hit for 10% as much? Simple way to test that, equip a white lvl1 weapon and keep the rest of the gear, according to that theory you should now hit for less than 1% of the damage you usually do.

The main point you're missing is that if you add in the AP the base weapon difference becomes realtively small.

Both fast and slow 1H get the same contribution from AP to their strikes (which is only fair since the amount of strikes you can do is independant from weapon speed).
So while it may be true that without AP the slower weapon does about 125 damage more (about 15% more).
When you add AP the numbers get upped with a static amount of damage, lets say you get 1500 damage extra from AP.
Suddenly the numbers are ~2250 for fast vs 2375 for slow. Still about 125 damage difference, but no longer a 15%, but rather a 5% increase.

If you add on top that white dps (and necrosis) is still a large factor and the gap between slow and fast is smaller than you might think at first.

However, slow is still WAY better. I haven't really read the theorycrafting of the last page yet (i need some sleep now, so maybe i will tomorrow).
But you (poster of the theorycraft) can not claim that we should ignore simulator results. If we do that we might as well abandon this thread since 95% of all data processed here comes directly from the simulator.
I'm personally extremely confident in its abilities, and is less likely to make calculation errors or something than theorycrafting. Granted all the number crunching happens non-transparant, but you could skim through the code.

The simulator indicated that for each 0.1 speed equals about 5 weapon dps. (Personally not very fond if this equation as weapon speed is not a "dps stat"; average damage and dps are. But it's easier to work with, that is true)
So according to that data if you go for a 1.5 speed weapon instead of a 2.6 one, it has to do 5x11=55 more dps.
So the fast CC has that over the rare rep weapons, ill give you that. But the AT weapon (from the crusader tokens), should be better.

Not necessarily a good thing.
Simulators are needed when too many things interact with each other, but you cannot derive rules from a simulator that you haven't already fed it, you can only make assumptions.

Qualified decisions in gear should be based on a basic understanding of how things work, a simulator is not the best substitute for that. We can present the results of our research and simulator work and others can make their decisions based on that, but if we leave the theorycrafting just to the guys "skimming through the code", then we are just throwing different setups at the simulator and present our results here without really knowing how they came to be.

If something behaves differently than anticipated then I want to know why, I can use a simulator to "prove" that it's the case, but I still want to know the reason and the simulator cannot provide that in many cases.

Take for instance a couple pages back the simulation results that suggested that Grim Toll is more than 4 times as good as Comet from Algalon25, that just can't be right, a little mistrust should be in order me thinks.

edit:
Only white attacks obviously.

I probably used the wrong wording, I don't care that much whether crit is capped through glancing, since that realistically never occurs, but rather:
If I have a 30% crit rate, will the glancing hits reduce that to ~22% by cutting off 27%? If I interpret your results correctly, the answer would be yes, but just for confirmation since it didn't seem 100% clear.

Sorry about the silly miscalc; of course you're right that it's 7.2% reduction. As you note, BCB causes yellow hits which cannot glance and have yellow miss rate, too, so the precise calculation is quite a bit more complicated than 15% of your pre-crit white dps, though I think for your purposes that approximation is Good Enough(TM).

Haste from percent-based buffs does stack multiplicatively with each other and with total haste from gear, so if you have 20% haste on gear, IIT, and raid buffs, you have 1.2*1.05*1.2*1.03 for a total of 55.7% haste. This is quite easy to see since such buffs are reflected on your paper doll swing speed (though I admit I've never done or seen any tests to verify that the paper doll is correct).

Rurahk has the glancing mechanics essentially correct. White swings, where glancing is relevant, are on a single roll system, and crit/hit cannot push glancing blows off the table; there is no way to mitigate glances, although they don't reduce the number of white crits you should get unless you are the crit cap:

A DW DK with no hit or expertise would need something like 53% white crit in order to hit the crit cap--and expertise cap along with a bit of hit on gear brings the cap over 65% easily--so it's not really a concern for us like it can be for rogues with certain gear setups.

@direddyre: Firstly, normalization only affects AP scaling, IE the forumla to calculate an Obliterate's damage is (AP/14*2.4+weapon damage)*25% assuming two diseases. The 2.4 in the parenthesis is the normalized number, instead of the weapon's actual speed.

Secondly, looking at current parses, strike based mechanics fill up roughly 50-60% of our damage (Obliterate, Blood Strike, Plague Strike, BCB, Frost Strike) while white attacks do around 25%. I don't see how it would be a DPS upgrade to boost white dps when strikes do twice as much of our damage and are heavily reliant on weapon speed due to the multipliers involved.

The haste formula is
multiplier=1/(1+percent/100)

I have 6.11% personal haste, with 5% from IIT that would be 11.11% haste if it stacked additive, or 1.0611 x 1.05 = 11.41% if it stacked multiplicative.

1/(1 + (11.11/100)) = 0.9000
1/(1 + (11.41/100)) = 0.8975

applied to a 2.6 speed weapon, that's

2.34 or 2.3335, my listed weapon speed 2.33, so you are apparently correct, thanks.

I have redone some calculations with the following corrections: added increased haste, added glancing blows, assuming that crits can become glancing and crit chance is thus reduced by glancing chance and I added 3% crit from Annihilation to specials that I simply forgot.

I don't want to edit the main post so people know what the things written here are about, so I'll just write the new results:

Assuming 40% haste (for simplicity) we get 1.79 dps per weapon dps (without BCB, 2.01 with BCB) and assuming full subversion we get new values of 2.72 for obliterate, 1.06 for BS, 0.90 for PS and 1.76 assuming now 60% for FS, I still don't know if that's a reasonable number, but it's higher than the previous one at least, even if you consider Annihilation. Totalling that makes 1.05 instead of 1.02 per avg weapon dmg.

For the example in question, [Lionhead Slasher] vs [Grimhorn Crusher], 78 avg dmg vs 53.4 dps, we now get 81.9 dps increase vs 95.59 dps loss, [Lionhead Slasher] still has a comfortable lead.
That is correct, did I state anything different?
Going by percentages of damage is not an adequate way of understanding this.

If you want an "easy" way of understanding, consider this:
On a very basic level, in order for 1 point of weapon damage to be equal to 1 point of dps, you would need to strike every second, but you don't. The higher percentages of strikes has 2 reasons:
1. They scale better with AP, you strike more than once every 2.4 secs, but AP treats it as a 2.4 sec hit, this has absolutely no impact on the calculation per weapon dmg point.
2. Strikes have static modifiers, people tend to underestimate those since they don't scale, but let me assure you that +1000 static damage to obliterate is very significant especially for DW, again this does not influence the calculation.

Obviously, trying to find the exact numbers after considering this is exactly what I wanted to do.

While the 1001 number may be static, the 25% bonus from two diseases isn't, and also you have several talents increasing ability damage by varying percentages (Black Ice, Glacier Rot, Blood of the North, Tundra Stalker and Merciless Combat to some extent). Nearly all stike mechanics also have a 2.45 crit modifier due to Guile of Gorefiend, as well as Obliterate having 15% extra crit chance from Rime (Obliterate 24% and Blood Strike 9% extra if you have Subversion also).

The discussion isn't which scales better with AP, you come off as if white attacks scale better with weapon DPS than strikes scale with weapon damage.

Okay, I've read through your post and started working on some math on my own to double check, but halfway through I noticed something. Have you taken the fact that all strikes swing with both hands in your calculations?

Edit: You also missed Glacier Rot modifier for Frost Strike.

I have not, just as I have not considered offhand for auto attacking, as I said in the intro it's consistently 50% and thus irrelevant for the calculation.

Just as a side note, I forgot Obliterate glyph and I didn't include Merciless Combat (just like I didn't include Blood Lust). I did some rough numbers and the result is still the same.
That one I missed indeed. Probably getting rather close now with the 245 weap vs the 200 weap, maybe it's time to assume BCB!

Did you remember to add subversion crit to blood strike as well? I was just looking over your first page of calcs and you didn't mention a different number set for blood strike w/ and w/o subversion.

Looks like your bottom line is about four tiers, or a little over 50 weapon dps, will put a 1.5 weapon ahead of a 2.6 one in a 3/53/15 build in terms of strike vs. white damage. Of course, this ignores the (very large) stat differential as well, though it's nontrivial to put an actual DPS number to that stat increase.

Following calculations assume 35% crit, 30% haste, 15% miss and the weapons Reaper of Dark Souls and Lionhead Slasher. Numbers with a following ~ are rounded down. Armor, ArP, AP and Glancing Blows have been ignored.

Average damage
R: 281+422/2=351,5
L: 206+384/2=295

Obliterate
((AvgDmg+1001)*1,25*1,15)+(((AvgDmg+1001)*1,25*1,15)*(0,5*1,45))
((AvgDmg+1001)*1,25*1,15)+(((AvgDmg+1001)*1,25*1,15)*(0,59*1,45)) (Subversion)

R: 1944,21875+(1944,21875*0,725)= 3353,77~
1944,21875+(1944,21875*0,8555)= 3607,595 (Sub)

L: 1863+(1863*0,725)= 3213,675
1863+(1863*0,85555)= 3456,89

Obliterate both strikes
R: 3353,77+(3353,7*0,575)= 5282,18775
3607,595+(3607,595*0,575)= 5681,962~ (Sub)
L: 3213,675+(3213,675*0,575)= 5061,538
3456,89+(3456,89*0,575)= 5444,60175 (Sub)

Blood Strike
((AvgDmg*0,44+370)*1,25*1,15)+(((AvgDmg*0,44+370)*1,25*1,15)*(0,35*1,45))
((AvgDmg*0,44+370)*1,25*1,15)+(((AvgDmg*0,44+370)*1,25*1,15)*(0,41*1,45)) (Subversion)

R: 754,19875+(754,19875*0,5075)= 1136,954~
754,19874+(754,19875*0,638)= 1235,377~ (Sub)

L: 718,4625+(718,4625*0,5075)= 1083,08~
718,4625+(718,4625*0,638)= 1176,841~ (Sub)

Blood Strike both strikes
R: 1136,954+(1136,954*0,575)= 1790,7~
1235,177+(1235,377*0,575)= 1945,718~ (Sub)
L: 1083,08+(1083,08*0,575)= 1705,851
1083,08+(1083,08*0,575)= 1853,525~ (Sub)

Plague Strike
(AvgDmg*0,5*1,15)+((AvgDmg*0,5*1,15)*(0,41*1,3))

R: 202,1125+(202,1125*0,533)= 309,838~
L: 169,625+(169,625*0,533)= 260,035~

Plague Strike both strikes
R: 309,835+(305,835*0,575)= 487,995~
L: 260,03+(206,035*0,575)= 409,555~

Frost Strike
((AvgDmg*0,6+165)*1,15*1,1*1,1*1,13*1,2)+(((AvgDmg*0,6+165)*1,15*1,1*1,1*1,13*1,2)*(0 ,55*1,45))

R: 709,2759366+(709,2759366*0,7975)= 1274,923~
L: 645,310908+(645,310908*0,7975)= 1159,946~

Frost Strike both strikes
R: 1274,923+(1274,923*0,575)= 2008~
L: 1159,946+(1159,946*0,575)= 1826,914~

Blood Caked Blade
AvgDmg*0,25*1,25*0,3

R: 351,5*0,25*1,25*0,3= 32,953~
L: 295*0,25*1,25*0,3= 27,656~

White Hits
(AvgDmg+(AvgDmg*0,35))*10,582*1,2
(AvgDmg+(AvgDmg*0,35)+BCB)*10,582+Necrosis (BCB, need a calculated value of Necrosis shown later)

Speed
R: 2,7*0,7= 1,89
L: 1,5*0,7= 1,05

Attacks/second
R: 20/1,89= 10,582
L: 20/1,05= 19,05

Necrosis
((AvgDmg+(AvgDmg*0,35))*10,582*1,2)-((AvgDmg+(AvgDmg*0,35))*10,582)
R: 1581,7482
L:2389,799

Damage
R: (351,5+(351,5*0,35))*10,582*1,2= 9490,4892
(351,5+(351,5*0,35)+32,953)*10,582+1581,7482= 10039,5982 (BCB)
L: (295+(295*0,35))*19,05*1,2= 14338,993~
(295+(295*0,35))*19,05+2384,799= 15168,696 (BCB)

Sorry for all the rough math, see the next post for a conclusion.

Subversion

Over 20 seconds; 4 Obliterate, 1 Plague Strike, 2 Blood Strike, 4 Frost Strike.

R: (4*5681,962)+487,995+(2*1945,718)+(4*2008)= 35139,279
L: (4*5444,60175)+409,555+(2*1853,525)+(4*1826,914)= 33202,668

DPS
R: 35139,279/20= 1757
L: 33232,905/20= 1660,1

1757-1660,1= 96,9 DPS lost if you equip Lionhead Slasher.

Autoattacks
R: 9490,4892*0,85= 8066,915
L: 14338,792*0,85= 12187,9732

DPS
R: 8533,66/20= 426,7 DPS
L: 12187,9732/20= 609,4~ DPS

426,7-609,4= -182,7 (182,7 DPS gained by equipping Lionhead Slasher)

BCB

Over 20 seconds; 4 Obliterate, 1 Plague Strike, 2 Blood Strike, 4 Frost Strike.

R: (4*5282,5775)+487,995+(2*1790,7)+(4*2008)= 33232,905
L: (4*5061,538)+409,555+(2*1705,851)+(4*1826,914)= 31375,065

DPS
R: 33232,905/20= 1661,6
L: 31375,065/20= 1568,75

1661,6-1568,75= 92,85 DPS lost from equipping Lionhead Slasher.

Autoattacks
R: 10039,5982*0,85= 8533,66~
L: 15168,696*0,85= 12892,8

DPS
R: 8533,66/20= 426,7 DPS
L: 12892,8/20= 644,6 DPS

426,7-644,6=-217,9 (218 DPS gained by equipping Lionhead Slasher

Conclusion

It seems I owe Direddyre an apology, unless my math is broken somewhere (which could be very likely seeing how much if it there is). Equipping two Lionhead Slashers will decrease your strike DPS by 96,9 DPS but increase your autoattack DPS by 182,7 if specced for Subversion, for BCB it will decrease your Strike DPS by 92,85 and increase your autoattack DPS by 218.

This means that Subversion specs gain 85 DPS from using Lionhead Slasher over Reaper of Dark Souls and BCB builds gain 125 DPS. However, this is based on ignoring glancing blows, which would swing more toward Strike DPS.

Q u o t e:
Frost
Runes - Fallen Crusader / Razorice
15/53/3 - Talent Calculator - World of Warcraft

Cycle;
HB - Ob - BS - BS - FSx
Ob - Ob - Ob - FSx

Priority system;
Frost Fever HB > rime HB with KM > FS with KM > Ob > BS > rime HB > FS until runes refresh

Presence: Blood

Glyphs;
* Obliterate
* Frost Strike
* Howling Blast

Additional Notes;
* Sigil: Awareness
* for aoe fights HB with KM replaces FS with KM in the priority system
* spell hit cap (approx 289) has a higher statweight than most DW builds