The added utility of a sheep in ZA helps immensely on a timed clear.

Ideally, though, you'd have 1 lock and 1 mage for the Dragonhawk boss.

Many thanks. Missed it for lack of being on firstpage, lesson learned.

I agree the short fights do make frost rather viable, although i hadn't looked deeply enough into the pushback frequency, which may or may not be enough to matter.

The live game will not allow you to chain cast faster than the global cooldown.

Bear has none, DH is rather pathetic on it as well. Lynx I guess has a little (Totems) but mostly doesn't. Eagle has a significant amount, though, with the flying eagles. IB Will get eaten rather quickly, and depending on your luck, this could be a bad fight.

[Edit: Bear has a charge and an AoE Silence, both of which are not "pushback" as such. DH Flamebreath is really the only thing on him, once the adds are downed, which for us at least is AoE, so insta-casts, no pushback.]

Hex Lord also has a ton.

[e] I guess he's not part of the bear timer though

The simulator you selected seems... incomplete. There seem to be a lot of factors missing from the calculation: shaman buffs (wrath of air/mana stream/mana tide/bloodlust), meta gem choices, destruction pots, flame caps, and trinkets. I'm not sure if there's an additional option to enable these variables, but I didn't see it when I visited the site. In addition, do you know if the program simulate proper cooldown stacking? The greatest advantage fire has is the enormous amount of cooldown stacking that can be employed in the last 20%. If a simulator is unable to account for this, then the simulated damage for fire will be far lower than the actual output. I think the generally accepted simulator is vontre's magegraf.

Finally, I don't think the Kael fight is the best assessment of DPS. There are two intense DPS phases in the fight (phase 2 on weapons and phase 3 on thaladrad). Other than those two instances, the rest of the fight is more about coordination and control. Depending on what roles each of the players have, it can dramatically influence DPS.

I've done all the ZA bear runs in my guild as fire with 3/3 imp. flamestrike. I'm just a big fan of aoe, and we have a beast of a prot pally who leads the runs. I mostly aoe trash. Our dps is me, shadowpriest, ele shammy, and a mix of rogues/hunters (at least 1 of each). Feral offtank who can dps on bosses and after his mark is killed for trash.

Guess I just like the dps I can do to the bosses as fire for these runs. I can pop cooldowns on bear boss after scorches are stacked, inivis, get a bloodlust when he goes human again, and catch back up to the tanks when the boss is <10%. Invis is harder to pull off for eagle so I just hold back a bit. Too much other stuff on dragonhawk for me to catch the MT. I should try to invis on lynx, but I haven't noticed me catching the tanks so bad on that fight.

I would think dps casters would all have conc aura for hex lord.

Ours certainly don't. We generally don't run with more than one paladin. That paladin also happens to be our MT. Thus he very rarely goes in the caster group. Sometimes when we've run with 2 then yes split them up. But not all guilds run ZA with multiple paladins.

What prevents you from putting the paladin tank in the caster group?

We run a fairly caster heavy group. Generally we run prot pally, feral druid, hunter, ele shaman, sp, frost mage, destro lock, holy priest, resto druid and arms warrior. Since we run 2 healers, they need the SP. That leaves 2 spots left in the caster group. They generally go to the ele shaman and mage since they need the mana. Especially me since I am the .5 healer in this fight and I do throw out some heals on certain fights to help out during burst damage. I suppose you could switch in the pally for one of us 2, but thats not going to really help the interrupt issues for hex lord. Yes I know ideally I'd be enhance and would be in the physical dps group but I'm not. It's not like the pally's aura helps the druid either so we just leave them where they are.

Theory: Passive haste rating is a bad itemization choice for builds/spell rotations that use stacked arcane blasts to DPS, despite the changes to the GCD in 2.4.

Argument outline: Builds using Arcane Blast are mana-limited, not throughput-limited. Their damage output is a function of how much mana is available during the fight. Haste just exhausts your mana faster.

Caveat: Haste does help cooldown stacking, though more for bloodlust/heroism/icy veins than passive haste. There may also be other fringe benefits.


Is this interesting/controversial enough to explore in more detail?

Remember that if you're going for AM-AB for OO5SR ticks, then too much passive haste will be detrimental, as you'll no longer be well, out of the five second rule.

This is pretty much spot on, there's not much to be discussed really. =p

That's a dumb argument, really. Saying "+haste is useless because AB spam is mana limited" is like saying "+healing is useless beacause it increases overheal."


I'll try an analogue:

You want to build a one week and have a group of 5 workers. They cost 100$ per hour normal working (=frostbolt spam), and 300$ per hour overtime (=AB spam). You won't make it in time with only regular work (=FB spam won't be enough damage), so you'll have to pay a lot of money (=mana) on overtime to get the workload done.

Now you have the idea of adding a 6th worker to your group. So, your group works 20% faster, but also costs 20% more money (=20% haste).
That means you can get the same work done with a lot less overtime (=same damage with a lot less AB spam), which will save you a noticable amout of money.

So, you the work done with a lot less overtime, have a decent amount of money left, which you can now spend on more overtime to have a terrace built!


This is the analogue. I'll put some numbers into Vontre's to get some values to compare it.


Edit:

Okay, now to properly model it. Take a 5 minute fight, 300sec.
15k base mana, 11k from Evocation, 9k from 3 gems, 8k from 3 pots. 400mp5 while casting, 350 from an SP, 50 from BoW should be very very generous. 800mp5 means 48k mana over 5 minutes.
Totals at 91k mana. Spread over 5 minutes, we can burn 300 mana per second.

Lets now look at the spells, assuming 40/0/21 with winter's chill, no cooldowns used at all.
Frostbolt spam - 1714 DPS, 92 mps.
AB(3) spam - 2292 DPS, 476 mps. +578 DPS, +384 mps compared to FB spam, or 1.505 damage per mana.
AB3/FB3 - 1789 DPS, 137 mps. +75 DPS, +45 mps compared to FB spam, or 1.667 damage per mana.

This means, to get the most damage, we use AB3FB3 as our efficient rotation, and AB3 as spam to burn excessive mana.

That means 147s of AB spam and 153s of rotating burns 90.933 mana for 610.641 damage.

Let's now add 5% haste to our gear. For those who want the numbers for that:
// Frostbolt spam - 1800 DPS, 97 mps.
// AB(3) spam - 2407 DPS, 476 mps. +607 DPS, +403 mps compared to FB spam, or 1.505 damage per mana.
// AB3/FB3 - 1878 DPS, 144 mps. +79 DPS, +47 mps compared to FB spam, or 1.667 damage per mana.

With this, we get 134s AB spam, 166s rotating, burning 90.852 mana for 634.307 damage.

Which is 23.666 more damage, or 3.88% more damage.
I didn't include any cooldowns. I didn't do any fancy graphing. I probably have a few rounding errors.
AB3/FB3 might not work due to the known client/server AB debuff delay issues, and we'll have to fall back to AB spam and FB spam.
That will probably shift numbers a bit, but not make them completely worthless.
Edit: Just checked it with only AB3 spam and FB spam. Slightly lower numbers, but still a ~3.9% increase.


3.88% more damage from 5% haste on gear. This is very far from being a useless stat.


I'll try another split of mana and damage for the cycles, to illustrate where your mana comes from.

An AB3/FB3 rotation for 5 minutes will do 536.700 damage for 41.100 mana, at 13 DPM.
That leaves us with 49.900 mana to spare, that we burn by spamming AB instead of rotating for a while.
But changing from rotation to spam, we burn the 49.900 mana into 73.900 damage, at 1.5 DPM.

With 5% haste, we spend 5% more on our rotation for 5% more damage.
That means we spend another 2055 mana for 26.500 more damage in our rotation.
That means can spam less AB, as we have less excessive mana. With a 1.5 DPM conversion on AB spam, our missing 2055 mana translates into 3.100 less damage from AB spam.
Notice that since AB casts faster, it will be even less AB spam time, since we have less mana and spam consumes it even faster.

So, haste allows us to change how we spend our mana. It allows to change from spending 2k mana on AB into spending it on rotations which are nearly 10 times as efficent to convert mana into damage.

Edit: The following was written before I saw Roywyn's model. It's interesting, and haste looks more useful than I'd imagined. I don't trust myself to check that kind of math on a good day much less at 6am, but I have no particularly compelling reason to believe it's incorrect. Interesting that with more haste you want to cast fewer arcane blasts.

=============================

Yes, William Gibson FTW.

I'm working from the old two-cycle theory of max DPS for arcane. From my understanding, what that TC means in real life is that for any fight an arcane mage will run some ratio of two cycles, one being fully stacked AB and one being filler (some combination of AM/scorch/fireball/frostbolt). The two-cycle theory states that if you pick the correct filler cycle & the correct ratio between AB spam and filler, your damage is maximized and you end the fight with zero mana.

I'm also basically ignoring the effect haste has on OO5SR vs I5SR ticks. Partially because the math's ugly, partially because arcane mages already have 45% regen I5SR, and partially because it's so dependent on which filler cycle you pick. Full disclosure: my thoughts are so far 100% napkin math, but I still think spell haste is a bad itemization choice _regardless_ of your filler cycle.

The reasoning (in slightly more ugly detail) is as follows: Take any given fight duration - say 3 minutes. Given the DPS/DPM of your AB and filler rotations, along with your total mana pool over the duration (initial + all forms of regen), work out the optimal ratio between AB spam and filler - by the two-cycle theory, if you chose the correct cycles you're doing max DPS. Now maintaining that rotation, add enough passive haste to go OOM at the 2:40 mark.

If you were fire or frost, you'd chuckle gleefully and continue casting for the last 20s, but you're not - you're arcane, so you can do one of two things:

1) DPS for 20s on pure regen. Note this is still an improvement over the baseline - haste will always help _some_
2) Go back in time & run your theorycraft machine again, taking haste into account. You'll come up with a slightly different ratio of cycles that allows you to end the fight at zero mana at the 3 minute mark again. Again, an improvement over baseline.

In both cases there is an improvement, but it's minor compared to frost/fire, and I suspect related in a complicated and fugly way to mana availability & shifting relative efficiencies of cycles.

This is all very, very abstract and makes a number of dangerous assumptions. Do you know many arcane mages who know precisely how long each fight will be, how much total mana (including ALL forms of regen) they'll have at their disposal over that period of time, and use that information to work out exactly how many of each spell to cast? Yeah.

AB3FB3 burns 137 mps. You have up to 800mp5 from spirit, SP, BoW - stuff that works continuously (unlike pots, gems, Evo).
That means your mana will go up while doing your rotation (or spamming FB).

So, spam AB when your cooldowns are up and when you have high mana. Cast rotations when you're low on mana.
Start spamming AB when the mob has less than 90s to live and you should be able to finish with close to zero mana.

It's complicated in sheets, but it's really not rocket surgery in practise.


Edit:

In defense of Kaiida and Vontre - yes, the idea of Arcane specs is that you generate a ton of mana, spend it on some base rotation, and burn all the excessive mana via AB.
As such, using haste is counterintuitive, I agree.

But the issue is that the mana-to-damage conversion of AB is just horrid. "More mana" is simply the least effective way of increasing your total damage.

Look at the numbers of the first calculations.
With 41k mana, you do 537k damage.
You add another 50k mana for 74k more damage.

121% more mana for 14% more damage. Ouch.

Also something must be said of the transitions between "Economy" and "DPS" rotation. Looking at either the Kavan-style (AB/AM)*3/Sc or the conventional AB*3/AM/Sc (or AB*3/Fball*3 or AB*3/Fbolt*3) the exact turn-over point between the two states can be a signifficant difference in Overall DPS.

Take the following wrong scenario:
AB*3/AM/Sc -> AB spam. You gain 1x AB ramp time which effectively, while it looks like a conversion from Economy -> DPS cycle it in fact is one and a half Economy Cycle ( AB*3/Am/Sc/AB*3 -> AB spam). The optimal way to do this would be to forgo Sc and instead AB*3/AM/ -> AB spam, netting a fully ramped AB at the point of cycle-swap.

Likewise, it's possible to mix the cycles for even more confusing but possibly better theorycraft. Eg: AB*5/AM/Sc, or injecting a AM in the AB spam.

I feel there is signifficant error margin (>1%) to be erradicated from correct/effective cycle control. It's easy to say "start AB spam when you think boss will die with you OOM" but if you interrupt your cycle at the wrong point you might end up with 7.5sec worth of AB ramp you wouldn't have if you had skipped that last scorch.

Arcane mages in fact have 60% I5SR. Arcane Meditation was changed to 30% a while back, 2.2 I believe.

You sure it's not 1.5/(1+x/1570)/1.3=1 ??

#2923:
I did not actually test it myself, but it is my understanding that the benefit of haste rating is calculated using base cast time, which in this case is 2.5 seconds. If i am wrong, someone please say so.

I think some people are missing the huge DPS increase haste has on AB(3).
Lets compare some spells and see the effect of 25% haste for example:
1. scortch 1.5 seconds, with 25% hate - 1.5/1.25 = 1.2 seconds
2. fireball 3.5 seconds, with talents 3.0 seconds, with 25% haste - 3.5/1.25 - 0.5 = 2.3 seconds
3. frostbolt 3 seconds, with talents 2.5 seconds, with 25% haste - 3/1.25 - 0.5 = 1.9 seconds
4. AB 2.5 seconds, ramped X3 1.5 seconds, with 25% haste - 2.5/1.25 - 1 = 1.0 seconds

25% haste means that in the time you would cast 100 spells normally, you would accually cast 125.
now lets check what accually happends for each of these spells:

1. instead of 100 scortchs in 1.5*100 = 150 seconds, with 25% hate you would cast 150/1.2 = 125 scortches.
2. instead of 100 fireballs in 3.0*100 = 300 seconds, with 25% haste you would cast 300/2.3 = 130.4 fireballs
3. instead of 100 frostbolts in 2.5*100 = 250 seconds, with 25% haste you would cast 250/1.9 = 131.57 frostbolts.
4. instead of 100 AB(3) in 1.5 * 100 = 150 seconds, with 25% haste you would cast 150/1.0 = 150 AB(3).

So from the 25% haste we got from our gear we effective gain:
1. Scortch gains exsactly 25% effective haste.
2. Fireballs gain 30.4% effective haste.
3. Frostbolts gain 31.57% effective haste.
4. AB(3) gains 50% effective haste.

Notice frosbolts gain more than fireballs, this is because the "-0.5 seconds cast time" talent helps frostbolt more (0.5/3 > 0.5/3.5).

here is the question... I'll try it after work