I've checked that.
I'm still suspicious this is on of those clear cases of theorycraft being an imperfect tool - assuming you don't have Desecration for the entirety of the rotation is unrealistical, because you would have it up until the refresh window, so a comparison between a no Desecration + BB vs a Desecration + BS rotation is sort of irrelevant; we should probably look at a 2 rotations interval, one with BB and Desecreation for the first half and one that refreshes it on the second part sacrificing a Blood Boil.

Theorycrafting AoE dps rotations is very hard because the dps windows are very hard to quantify.
On a single target like a boss it's safe to assume an extremely high number of rotations will be required before the kill, and thus average dps is a consistent theorycrafting tool.
On the other hand, I think it's extremely wrong to assume that the average AoE window even last for 20 seconds, and I'm also confident that in the much more commong instance of very small AoE dps windows squeezing in as many BB as you can will give you the best dps output, as long as you don't end up into a case of "Desecration just expired".

When I made the calculations for BS in AoE-sitations on the last page I was aware of the fact that the modelling is far from optimal (not everything taken into account, just one parse, Anub not optimal fight to take numbers from (which begs the question, what is?)), however, the numbers turned out so much in favor of BS that all these things can be ignored.

I was thinking about the AoE-window as well, though. As you've said, it's very hard to calculate a baseline-rotation and for that you pretty much have to assume 20 second windows, else you would have to do it on a case by case basis. Also, if you only have say 10 seconds of AoE you usually have a boss or something to hit after that is done. As the DPS-requirements on the boss are often times stricter than on things that may be AoEd on the way to killing the boss, you don't want to drop the Desolation buff. I guess my point is, even if you have very small AoE-windows, there is usually something going on before and after that recommends keeping Desolation up.

Edit: The fact that we now have Reaping back in spec also causes us to have the second rune set be used for 3xSS. So we will only use BS during the first set of the single target rotation, which very much increases the chance of Desolation dropping off on the start of your AoE-rotation.

Do notice that the 10k rule only applies when the other option is to do 4 seconds of uninterrupted dps at full capacity. If there is, for example, 1 second window, where you can't do any dps (an example of this would be in ToC the point when Anub'arak comes out from burrow, there is a small, 1-2 second, window where he can not be attacked and you have to wait for tank to pick him up) the "value" of the AotD goes up to 13-15k dps.

Also, comparing AotD against our full DPS isn't really a valid comparison, since unless we've done something terribly wrong beforehand our both diseases and most likely unholy blight would be ticking during the AotD summoning and our ghoul (and maybe even gargoyle if you are aiming for full burst) would still be happily bashing away. The value of the AotD shouldn't be compared against full dps but dps from melee and strikes only.

At this moment my diseases and unholy blight logged a total of 22% of my dps and my ghoul was almost 10%. Even with Unholy Blight and 4*T9 nerfed I'd say the "tipping point" would be closer to 13k than 10k. Even with ICC gear I don't think our dps will go over that anytime soon.

Looking at the stat weights - does the sim not give a value for weapon speed? The reason I ask is because after looking at the Quel'Dalar chain rewards, it seems that if you only base your evaluations on the stat weights in the first post (using the original low value of haste at .9 or whatever), you'd find that the order from best to worst is the 2hand agi mace, 2hand agi sword, and then 2hand str sword.

However, that's not taking into account weapon speed differences, which are actually relatively significant and in reverse order - str sword is 3.6, agi sword is 3.4, and agi mace is 3.2. I know different classes and specs have different value of weaponspeed - some may generally prefer slower but it's really not that important, while others completely depend on it - and was just curious where we lie.

Weapon speed is an irrelevant stat for nearly everything except tanking. What matters is weapon DPS and average weapon damage and as a general rule, the weapon with the highest DPS will almost always be better. In the case of a tie, the weapon with the highest average damage wins every time unless there is something horribly, horribly wrong with the stat itemization.

You realize you just contradicted yourself, right? The speed alone is irrelevant, but it is an easy way to measure average weapon damage. All weapons at the same ilvl generally have the same dps, with some minor tenths-place variation. Therefore weapon speed can be used to calculate average damage by multiplying dps x speed, and becomes an easier way of measuring average damage value, especially in terms of stat weights using a resource like the Wowhead database.

The mace and str sword are both 267.2, and the agi sword is 267.4. Simply multiplying dps by speed puts your average damage at 855.04 for the mace, 961.92 for the str sword, and 909.16 for the agi sword.

There has to be some way to put that into a value that can be compared to other stat weights. Is it really so overpowering that a 3.6 weapon will always beat a 3.5 weapon for DKs at the same ilvl? I know that used to be the case for a lot of classes, but then they introduced normalization because they don't want weapons to have one completely overpowering feature. So I was just trying to figure out if we know how significant the difference is or not.

The problem with using weapon speed to weight items is that it is a dummy stat with no intrinsic raw value. By that I mean that since weapon speed is no more than a ratio of two other stats, it can only be used to make a valid comparison for items of the same ilvl. You cannot create stat weights for it because of this.

A simple examination of the meaning behind stat weighting reveals why this is true: Stat weights tell us how much we are required to increase one stat to get a similar gain from a different stat. These are normally normalized to attack power i.e. in the current table of weights we see that we need to use 3.1 AP in order to get the same damage increase we would see from 1 point of strength. You cannot do the same with weapon speed. A gain of 1 weapon speed does not correlate to any increase in damage and therefore can have no equivalent in attack power. How could it? Weapon speed by itself is meaningless.

If that's the case, then weapon dps shouldn't really be used in stat weights either, because its weight is heavily reliant on weapon speed. Sure, it contributes at a flat rate to melee attacks. But its contribution to strikes will vary on weapon speed, which especially in 3.3 will contribute to a great part of our dps.

For example, a 200 dps weapon with 2.0 speed (ignoring normalization) will be doing less strike damage than a 150 dps weapon with 4.0 speed even though the 200 dps weapon would be doing more melee damage. But if they were both 4.0 speed, the 200 dps weapon would do more strike damage and melee damage.

So you can't really say weapon dps contributes in a predictable fashion like strength or AP, either. You have to calculate in weapon speed.

At least for enhancement shamans, this difference is huge. It's my understanding (admittedly I'm not super familiar with the class) that for enh shammies, a 2.6 speed Naxx weapon is better than a 1.5 TOC weapon, regardless of stats.

So if what you say is true, then at least for the purposes of using a spreadsheet or other tool to calculate BIS weapons, the stat weights alone are useless I guess. We need to be able to know how much of a difference speed makes when the weapons are all the same ilvl - could stats ever be enough to overcome? And is there a point where the speed difference is enough for a weapon to not be an upgrade (like say going from a 3.5 ilvl 232 weapon to a 3.2 ilvl 245)?

It'd be nice to know what the breaking point for unholy DKs is - I'm tempted to say weapon speed matters more the higher ilvl you go, but I don't know enough of the formulae to calculate it out myself.

Weapon DPS absolutely contributes in a predictable fashion. A 10 DPS upgrade on a weapon will, suprise, give you 10 more white DPS. This is always true. It is true that it won't contribute to strike damage in a consistent manner, but noone is asserting that it will. This is why you need both DPS and average damage numbers in order to determine the worth of a weapon.

Additionally, Enhancement shaman value 2.6 speed weapons so high because of the scaling properties of their flametongue enchant. The damage they do with flametongue weapon is directly proportional to the speed of the weapon, regardless of itemlevel or overall damage. Death Knights do not have an analogue for this behavior.

Finally, you missed the entire point of my discussion. Weapon speed CANNOT be used in stat weights, the reasons for which I have already given. It is a fairly simple affair to calculate stat weights for weapon speed and DPS and use these values in conjunction with the stats on a weapon to determine which weapon is the best of the choices available to you.

I think it's a bit premature to say that the glyph of disease is only useful for single target fights which are long.

The following are damage breakdowns for a glyph of disease based rotation versus a glyph of the ghoul based rotation. They're 1 000 000 second runs resetting every 3 minutes, with an initial rotation to get up procs and diseases, but then no further disease optimisation. This means that the GoD rotation doesn't necessarily always maximise diseases but the sim doesn't have to worry about the use of the blood runes. I used the new 2H Toc unholy set, many 180s fights and turned off wait for procs for Gargoyle, but otherwise it was the defaults.

Firstly the GoD run, at 9884 dps

Ability	Total	%	Landed	Hit%	Crit%	Miss%	Average
ScourgeStrikeMagical	1961020407	19.8	250000	51.2	48.8	0	7844.1
MainHand	1774548660	18	394445	66.3	33.7	0	4498.8
ScourgeStrike	1148703179	11.6	250000	51	49	0	4594.8
Ghoul	944687391	9.6	950003	87	13	0	994.4
DeathCoil	890021618	9	144442	65.1	34.9	0	6161.8
BloodPlague	805663808	8.2	327777	66.1	33.9	0	2458
FrostFever	634943805	6.4	322221	100	0	0	1970.5
Gargoyle	582643592	5.9	122232	87.2	12.8	0	4766.7
WanderingPlague	487340564	4.9	219779	100	0	0	2217.4
BloodCakedBlade	295736558	3	118642	100	0	0	2492.7
BloodStrike	194112791	2	55556	57.3	42.7	0	3494
UnholyBlight	88980385	0.9	144442	100	0	0	616
PlagueStrike	42627697	0.4	11112	60	40	0	3836.2
IcyTouch	32853897	0.3	11112	65.1	34.9	0	2956.6
Horn	0	0	33334	100	0	0	0
Pestilence	0	0	44444	100	0	0	0

Secondly the GotG run at 9703 dps

Ability	Total	%	Landed	Hit%	Crit%	Miss%	Average
MainHand	1772988683	18.3	394445	66.3	33.7	0	4494.9
ScourgeStrikeMagical	1608996399	16.6	205556	51.1	48.9	0	7827.5
Ghoul	1099580268	11.3	950003	87	13	0	1157.4
ScourgeStrike	941975171	9.7	205556	51	49	0	4582.6
DeathCoil	887386086	9.1	144442	65.1	34.9	0	6143.5
BloodPlague	660044817	6.8	294444	66.1	33.9	0	2241.7
Gargoyle	583495678	6	122232	87	13	0	4773.7
FrostFever	581315920	6	294444	100	0	0	1974.3
WanderingPlague	420051765	4.3	199211	100	0	0	2108.6
BloodStrike	362671066	3.7	100000	57	43	0	3626.7
BloodCakedBlade	295449938	3	118642	100	0	0	2490.3
PlagueStrike	228619027	2.4	55556	59.9	40.1	0	4115.1
IcyTouch	171845066	1.8	55556	65.1	34.9	0	3093.2
UnholyBlight	88713223	0.9	144442	100	0	0	614.2
Horn	0	0	33334	100	0	0	0

For some reason with these short runs I don't get an extra DC in the GotG run.

But if it is contributing to strike damage in an inconsistent manner, you can't determine it's value to overall damage. We care about white dps and yellow dps, not just white. Predictable variable X (rate of white increases) + unpredictable variable Y (rate of yellow increases/decreases) absolutely does not produce a predictable variable Z (rate of overall increase.

.

My point then is that DPS of a weapon can't be used in stat weights then, either. Which means that:
I would like to see your alternative, fairly simple method of calculating weapon choice, so I can easily pick out whether (Edit: bah, they're not on normal wowhead yet. These two handers) is better. Just using the 3.3 stat weights, the Spire is better than the Quel'Dalar, Might of the Faithful, but again, that's not taking into account the fact that Spire is 3.2 and Might is 3.6 in any fashion whatsoever.

A few moments of testing confirms your findings Consider, I removed all gear except a basic weapon (no runeforge) and a greatness card. I waited for a proc, applied both diseases and then refreshed once greatness was down. Every time without fail the diseases used my current AP when I refreshed. It would appear that the functionality of GoD has changed.

Wow that's big news. It's actually good news in a way. Sure it's a nerf to your potential disease damage, but it makes it much easier to use. While we aren't gcd constrained anymore, running the GoD rotation optimally is actually quite difficult, and having to figure out whether its worth reapplying your diseases is rather difficult in practice (other than when you have all procs up).

Before I used the latest version of the simulator it didn't reset trinkets between runs and so there were some parses where the average disease tick was pretty much the same for both glyphs and the dps was still competitive (at least with either the 4t9 or 2t10 set bonuses).

Using the numbers in my last post we can theoretically compare the dps (after a short start up phase.)

Edit: Sorry, didn't post these the first time around Consider, and then I left this in edit while trying to calculate a proc uptime.

Okay looking firstly at the ghoul dps numbers for both specs
The ghoul glyph is is worth 1099.6 - 944.7 = 154.9 dps
Gaining a SS (and using a Pest) in place of an IT, PS and BS and 1/8th DC is actually only about 730 extra damage, and we get to do this once every 20 seconds so its only 36.5 dps.
The main advantage is actually with the extra disease ticks, we get two extra ticks of each disease every minutes and about a third of the disease ticks proc wandering plague. So that's about 9838 extra damage a minute, or 164.0 dps
so your talking about 200.5 dps for the glyph of disease.

That's assuming you actually can use all the extra RP which isn't always the case.

So yeah in theory with 4 piece t9 the GoD will win on most fights. Although there will be slightly lower uptimes on some procs. The 2 piece t9 is probably the worst, but I can actually model that and it comes out as 17.3% uptime for a 2 BS cycle (assuming they're applied back to back), versus 15% for a one BS cycle due to the 45 second internal cooldown.

Without 4 piece t9 it'd be much closer.

Edit 2: Okay with 4 piece t9 the disease damage drops down to about 140.7 dps. Still the disease glyph wins out, and it gets a slight extra boost when you get 2 piece t10.
Anyway here's a nice table

Approximate DPS boosts

Glyph	DPS
Ghoul	155
Disease	177
Dis+4T9	200
Dis+2T10	181

The two simulations from your post have huge gaps in disease damage done and, as such, the overall dps of each can no longer be compared.

With GoD no longer increasing disease damage (technically, it will still give you an ever so slight increase in number of disease ticks, but not enough to sway the following), the glyph basically grants you a dps increase equal to (1 SS - 1 BS - 1 PS - 1 IT) / 20 seconds. Using the average number of those four aforemtnioned abilities from your simulation, that's (7844 + 4595 - 3494 - 3836 - 2957) / 20 = 107.7 dps.

Glyph of the Ghoul increased the damage your Ghoul did by (1099580268 - 944687391) / 1000000 = 154.9 dps.

Thus, the latter will win out, and pretty solidly at that. GoD will be a worthwhile choice for Deathbringer Saurfang, if nothing else.

Edit: Actually, the increase in number of disease ticks might be somewhat more significant than I gave it credit for. Still not sure if it would push things over.

Edit 2: Trying to figure out why there's an increase of approximately 200 in BP average tick with and without GoD in your sims, but FF is almost exactly the same in average tick. Something odd there.

You are misreading me. I never asserted that weapon DPS was the end all and be all of stats on a weapon. In fact, I said that you need to use both a weapons listed DPS and its average damage to determine its overall value. This is because, like you said, you can only determine a weapon's effect on your white DPS if you only use the DPS listing; likewise, you can only determine a weapon's contribution to your yellow DPS if you only consider its average damage output. Note that weapon speed is never mentioned when considering these two values.

Furthermore, you seem to not understand what I said earlier when I attempted to explain that weapon speed is meaningless when looking at stat weights. Allow me to try again:

Let's first look at what a stat weighting tells us. Simply looking at the table, it appears to tell us the value of each different stat as a ratio when compared to AP. For instance, the tables tells use that strength is worth 3.1 AP. This means that if I were to somehow gain 3.1 attack power, I would see the exact same damage increase over the duration of a bossfight as if I had gained 1 strength. The same logic can be applied to weapon DPS. We know that it takes 14 Attack Power to increase the DPS of a weapon by 1. Therefore, to get the overall weighting of DPS as related to AP, we simply take 14 and multiply it by the total percentage of our damage that is white. This is slightly complicated by things like BCB and necrosis, but you get my point.
The same is true of average weapon damage, though it is a more complicated calculation since weapon damage does not contribute equally to each remaining portion of a Death Knight's damage. You must calculate its contribution to SS, BS etc. individually and then sum the results to get your final weighting for the average weapon damage. A mathematically simple task, but a somewhat arduous one. From the results of these calculations, we can get the weighting of weapon damage with respect to attack power.

If you have been paying attention, you will notice that nowhere in these calculations was weapon speed brought up. That is because an increase or decrease in weapon speed by itself does not contribute to your damage output in any meaningful fashion; it is a mathematically irrelevant number. That does not mean that it is sometimes useful tool to use when comparing items of identical itemlevel (for then it is directly proportional to average damage), just that it cannot be used when compiling a set of stat weightings for a given character, since you will not always be comparing weapons of identical ilvl.

A prime example of this would be to compare the following weapons: [Dual-blade Butcher] and Quel'Delar, Ferocity of the Scorned (for the sake of this argument, I will ignore weapon stats/sockets). If one were to weight weapon speed and use that value in their determination of weapon value, you would likely conclude that the Dual-Bladed Butcher is the superior weapon of choice. However, looking deeper we see that both weapons have nearly identical average damage output (~1% difference) and the Quel'Dalar Sword has a higher listed DPS. As such, it is a superior weapon in all regards, despite possessing a speed 0.2s faster than DBB.

hey consider, i saw your spec 17/0/54 having 1 point in bladed armor instead of morbidity, why is it that you don't put one in there for an extra 5% death coil dmg when bladed armor isn't nearly as strong? Plus 5 seconds off dnd cd is helpful in many situations.

Weapon DPS * Weapon speed = Average Damage

There are two degrees of freedom here. You can therefore choose any two of these variables and get a basis for the space of all (white damage, yellow damage) possibilities. You only lose independence when you add the third. Therefore it is reasonable to apply stat weights to any two of these variables.

For any function you compute that uses average damage, f(dps, dmg, ap, str, crit, ...), you can make the substitution dmg = dps * speed, and thus compute g(dps, speed, ap, str, crit, ...) = f(dps, dps*speed, ap, str, crit, ...).

You only get in trouble if you try to use stat weights from all three. (Though there's no reason not to compute weights for all three variables and then just look at any pair.)

Edit: DPS obviously isn't actually linear combinations of stats, but the whole idea of stat weights assumes it is in some local neighborhood as an approximation. For DKs this seems to be acceptable, but to be truly accurate it would be best to simulate the different gear/talent combinations of interest and forget about stat weights entirely. For some classes the weights would vary so much depending on the neighborhood that stat weights are useless.

I'm not sure you would necessarily conclude DBB would be superior - it would depend on the weights, wouldn't it?

But this comparison completely misses what I am trying to find out. I understand that it's a somewhat safe assumption that a higher dps weapon will generally win regardless of speed because, let's face it, speeds are not over that large a range, and even if they were, higher dps also correspond to higher ilvl and accompanying stats (the exception being these new statless wonders Blizzard is keen on right now and now we have to evaluate procs).

However, I want to compare (for example) the Quel'Dalar quest rewards. There are three that are worth looking at (I'm not sure how to link them as you did but my previous link can take you to them): Lightborn Spire, Quel'Dalar, Ferocity of the Scorned, and Quel'Dalar, Might of the Faithful.

I know you keep saying "use average damage, weapon speed is meaningless" but even ignoring that you can calculate average damage using the formula Qira posted, we have no stat weight for average damage either, and therefore you can't really judge which is best.

Furthermore, going back to your own example - what if you were trying to compare Spire to Butcher instead of Ferocity? Using the current stat weights, Spire is supposedly the best of the bunch anyway (if you're not already expertise capped). But! as it is 3.2 speed, it has the lowest average damage - lower even than DBB. Depending on how significant that is, it's possible that Spire is worse than DBB . . . without a stat weight for avg damage, how can you tell?

And yes, I agree with Qira that stat weights are just a usefully hasty approximation. If it's not really that accurate, we're going to have to make like ret pallies, use RAWR to tell us upgrades based on our current gear, and hope for the best.

You both are missing the point that Darkside is trying to bring across. I have tried to say the same in the past, but people generally dont understand or listen.
The thing is, weapon speed absolutely says NOTHING about dps for death knights (it does for classes with unnormalized attacks and some weapon buffs).

Yes you can calculate average weapon damage with having both speed and dps, but that still means that weapon speed itself tells you nothing.
Weapon dps says something important, because it is a contributor to white damage. More dps gives higher white dps, and vice versa.
Average weapon damage says something important too, more weapon damage means more damage for yellow strikes.
But weapon speed tells you nothing. What does going from 3.0 to 3.4 weapon speed tell you ?
NOTHING; unless you give one of the other variables and you can say something. But weapon speed is a dead stat when it comes to measuring dps or AEP.

Why then do we use weapon speed ?

First of all, its easier. When comparing weapons you generally talk about weapons that have the same or nearly the same item level, or better worded drop in the same tier raid content.
Second of all, it's easier to obtain. Every weapon has dps and speed clearly noted. Average damage however is not on the tooltip, it can only be deducted with calculations (dps x speed ; or ; low end + top end / 2). The only thing noted on the tooltip are the low end and top end damage.

But that doesn't take away it's always used in the wrong way. We always say: "Pick the slow weapon, because it does mroe damage per hit". That is totally wrong of course, what we actually do when we say that is compare the average damage in an abstract way.

But I'm with Darkside, we should technically stop using weapon speed all together.

Weapon speed is not completely redundant for even normalized strikes, as the normalization only occurs for the ap contribution to weapon damage. The "pure" weapon damage (that is the damage in the tooltip of the weapon) is unaffected by normalization, so for strikes slower weapons are technically still slightly better. However, this contribution is comparatively small as the AP contribution usually far outweighs the base damage contribution and especially for Unholy, since SS has a relatively small weapon damage coefficient. So even a 3.6 speed to 3.1 speed change won't do much of a difference.

Edit: Just to give an example, the normalization formula for 2H is
normalized_damage = base_weapon_damage + (3.3 * Attack Power / 14)

Now let's compare Lightborne Spire and Quel'Delar, Might of the Faithful

Let's assume 8000 AP, LS has an average damage of 855, QD an average damage of 962, so plugging in AP values the normalized damage for LS is 2740 and for WD it is 2847. That is a 107 damage difference between a 3.2 and a 3.6 speed (which is almost as far as you can reasonably get). No consider that SS has as far as I know a 48% weapon damage coefficient, we are down to a 50 damage difference. BS and PS have a similarly low coefficient and that's all the strikes we use. Now take in crit modifiers and debuffs on the target that increase SS damage (EP and such) you get perhaps a 100 damage difference per strike, which will be 35 DPS if you assume you use 7 strikes per 20 sec (4x SS, 2xBS, 1xPS). As mentioned, this is with a difference in weapon speed of 0.4, so a 0.1 difference makes up less than 10 DPS.

I understand what you're getting at about weapon speed meaning nothing in a vacuum. But as you said, we're looking at this with weapons of roughly the same level.

It's almost always the case that you should take a higher ilvl weapon. That's doesn't cause me any problems when I'm trying to figure out upgrades. Usually, the trouble comes trying to decide between weapons of the same ilvl. And here's where you can safely no longer consider weapon speed as a meaningless stat - at ilvl 251, 1 speed = 267.2 average damage (ignoring ferocity's 267.4). What I've been wanting is a shorthand way of measuring that on a database like Wowhead.

But fair enough, I understand why you might want to take weapon speed out of the discussion to keep things more clear and mathematically pure. While you both have been arguing as to why weapon speed is meaningless, it still hasn't answered my real question - how do i compare weapons while taking into account how they affect strikes (especially now that SS is so much more important)? The current stat weights are insufficient.

Here's how we both win - determine a stat weight for how average weapon damage contributes to yellow dps.

Once we've got that, I can easily translate that into weapon speed for any set group of weapons of the same ilvl on Wowhead. Or calculate it out individually for each ilvl so I can compare between ilvls when a weapon is painfully fast, like the 3.1 speed Reclaimed Shadowstrike.

You are correct in asserting that you can use weapon speed and weapon DPS to produce identical results to a situation that uses DPS and weapon damage alone. However, the stat weights produced in such a calculation can't be plugged into a gear calculator like wowhead's to produce accurate results, since the overall "value" of the weapon could not be calculated by simply multiplying weaponspeed*speed_weight+weaponDPS*DPS_weight etc. etc. You would need to do something like weaponspeed*DPS*speed_weight+weaponDPS*DPS_weight and what you are really doing in such a situation is computing the stat weighting for average damage, just calling it by a different name. This problem is not encountered when using average weapon damage, the reasons for which I have already demonstrated. In the case of computing weights for weapon damage and DPS, one can simply take weapondamage*damage_weight+weapondps*dps_weight etc. to get and exact value for the weapon as a whole.

Nyth_ has the right idea when he says that weapon speed by itself tells us nothing. It cannot alone be used to determine the value of a weapon and as such is useless when compiling stat weights, since they are meant to be independent values.