I swept it under the rug as part of my many conservative pro-wand assumptions. When I stack the deck in favor of wanding and it still loses I still get a valid conclusion.

By not including the extra mana regen during wanding I made an anti-wand assumption that breaks with the spirit of the analysis (moreover it's probably the worst mistake I could have made). Thankfully Wizeowel logged in his pve gear/spec so I get a nice estimate of 495-292 = 195 mp5 gained in solid post-Illidan gear. Using wand_mps = -39, the tradeoff coefficient in my conservative estimate improves to 2.97, which is still insufficient to beat frostbolt spam's 1.97.

Even the ~10% improved wand dps from the holy [Wand of Cleansing Light] (thank you again for your counterintuitive upgrade lists, Rawr) is not going to tip the scales.

Can you explain what you mean by this?

Mana that you have in an entire fight is essentially finite. Based on constant 60% regen, evoc, pots, gems, etc. If you have 20k mana in a 4 min fight, that is it.

DPM stays constant when you add haste rating. But DPM goes up when you add +int, +dmg, and even +crit. So yes, you will run out faster on a stat by stat comparision basis. In a situation where mana is finite for a fight, haste makes you reach that point where you "lapse" into your high DPM, lower DPS rotations. In contrast, +dmg, +int, +crit does not make you reach that stage sooner. They add more damage without the cost of having to use more mana. And to an arcane mage, mana = damage.

Taking this to an extreme, if you add enough haste, then you won't be regenning mana even during your AB rotations because you cast them so fast they use up more mana than you regen.

On throwing AB rotations off. It depends on the rotations. But various rotations are already set based on the inherent cast times of the spells such that they take advantage of the third debuff on AB. All the 3xAB rotation variations are meant to sqeeze in DPS and then start the cycle again with an AB just before the debuff runs out. Add haste in any increasing amount and you will soon find that you have to change the rotations because you are casting spells so fast you start the AB too early, yet you can't quite squeeze in another spell, etc.

In the end, it boils down to mana and sustainability. If you can stack more haste and still sustain your 100% AB spam, by all means. But if you can't stack on more haste without running into serious mana problems, then stack +dmg, int and crit which doesn't cause such problems. It is raid comp dependent. Not every arcane mage has such luxuries of super heavy mana support. As long as you are having to resort to rotations in a raid boss fight, then there is much more leeway to add +int, +dmg, and +crit because if you are in a situation to get more mana, you can easily transform all that into more ABs to use it up and obtain higher DPS anyway.

Its only when you are already at 100% AB spam, with mana to spare. Then you want to add on haste because you have excess mana that you can't use. If the goal of every aspiring arcane mage is to reach 100% AB spam in a raid fight. Then +int really helps, +dmg and +crit is just more damage but doesn't help, and +haste actually puts you further away from this goal.

Imagine the following situation:
You do a fight in which you are able to spam 100% AB and your mana reaches zero exactly when the fight is over. Now add haste to your gear which, for the same fightlength, allows the additional use of one frostbolt and substract an equal amount of crit.
I'm quite sure that in this case the damage won by the frostbolt is bigger than the damage lost by the missing crit rating. That's also what Rawr says, if I understand correctly.

No, the goal is to maximise damage, just like any other dps caster class. If that means popping your Hero+IV+Skull with Frostbolt instead of AB, then so be it.

The best advice about haste is to work it out for yourself, depending on your gear and the fight you are doing. Instead you seem to want to come up with some generalised theory based on assumptions which don't stand up to deeper scrutiny. You can't say "haste is bad if X" because X is actually a combination of 50 other variables. For instance you said, "if you add enough haste, then you won't be regenning mana even during your AB rotations because you cast them so fast they use up more mana than you regen" - show me the gear where this is true. Even with full Sunwell gear, over 400 haste I'm seeing the opportunity to cast 2900-3000dps frostbolts which still return 64-75 mp5.

The bolded part is actually wrong when you look at it from an overall perspective.

Your DPM for the overall fight actually increases with haste.
You're right that haste doesn't increase the DPM of AB spam, nor the DPM of FB (or whatever youre filler is) spam.
Haste however does make you reduce your AB spam time and increase your FB spam time, which is an overall DPM increase because FB is better DPM.

But I'm not going to beat the dead horse. Every stat has their benefits, stick your gear/setup in rawr to find out what's best for you.


Back on topic - Wanding.
Using Leulier's sheet with 3 warlocks (2 destro, 1 aff, sunwell gear) and 2 shadow priests, we get ~65% ISB uptime and 10.025 shadow DPS total.

If we now add a shadow wand with 1.5s shoot time, our ISB uptime drops to 51%.
That's a 14% uptime loss, causing up to lose ~280 DPS.

Sunwell wands have 198 DPS.
Wand spec +25%, Misery +5%, MCoE +13%, crits +15%, shadow priest shadow vulnerability +10%, misses -4%, partial resists -6%. ISB uptime adds another +10% on average.
Spell haste doesn't affect wand. Normal haste might affect it, but stats are not before the WotLK pre-patch.

So, that's a total of 369 DPS. Our wanding still increases net damage, but 3/4 of its damage are eaten up by the ISB.
It's even more if you add the lost VE/VT mana.


So, don't use your shadow wands then. What else can we do?
Fire wands - all arena wands are fire, 198 DPS, 1.9s shoot time.
[Sporeling's Firestick] - 97 DPS, 1.3s cast time.
Holy wands - [Wand of Cleansing Light] - 198 DPS, 1.5s shoot time.
I didn't remember that this one was not shadow! You'll have to see a lot of those drop to get one though.
[Consecrated Wand] should be mentioned for its 1.2s speed, but it's nature damage eating stormstrike charges, it's an alliance quest reward and the damage is really bad.


So, what are our options for wands then, assuming JoW is up?

Shadow wands deals little to negative damage due to ISB.
Arena wands are 1.9s, 351 DPS, 19.5 m/s from JoW.
The sporeling wand is 1.3s, 172 DPS, 28.5 m/s from JoW.
The sunwell wand is 1.5s, 270 DPS, 24.5 m/s from JoW.

Wands gain another 39 m/s from 40% spirit regen outside 5SR.


That gives us a tradeoff for 2.0 for frostbolt, 2.9 for the arena wand, 3.1 for the holy wand and 3.3 for the sporeling wand with JoW.

So, frostbolt as filler is still 50% better than wanding, and fast wands for better JoW are not worth the damage loss from lower DPS.
Also, Tizzlewump's DPS values are really low, missing debuffs and probably some more. So, frostbolt will pull ahead quite a bit more there.

3700 Frostbolts?

Well that let me rethink my spec, I m currently 40/21, using scorch as filter spell as (criticize me for that) as I dont really care for IV at all (cd stacking in start of fight might even get me close to agro so I rather just do a plain AB spam and do my ap Pyro when agro is save and I got a clearcast).

My gear lvl is 4T5, havent had much luck with T6 drops (like anatherion s belt)

So the question for me isnt if I should spec 21 points in Frost instead of Fire to have IV instead of Pyro, but if Frostbolting has higher dps or better dpm as scorch.
I m currently even thinking about respecing the fire tree to use Fireballs instead scorch.

Can someone help me in theoriecraft which of scorch, Frostbolt or Fireball got best dps/dpm - assuming sunfire enchant and scorch spec without "the mana return when crit" talent and Fireball spec with it.

Sry for my bad english, I m no native speaker.

Yes sorry 3700 dps was blatantly exaggerating. Had that value in my head somehow. Still, it was around 3k dps with 75 mp5 when stacking hero+IV+sliver.

You are also quite welcome to download Rawr - Release: Rawr b14.1 and find the answer for yourself, based on your own gear and tailored to your situation. There is also a thread called http://elitistjerks.com/f31/t16781-mage_help_me_please/ so that you know for the next time.

Well, I have done that and my conclusion is that highest filter spell dps is with Fireball while all 3 mentioned options got nearly same mps - and thats the reason why I wonder why so many prefer this Icy veins build when the Fire tree got a Pyro and a Blastwave to offer.

While fireball is (very) slightly higher dps than frostbolt, it uses a lot more mana. IV and cold snap are also a lot more powerful than pyroblast.

Back on the subject of haste - 100% AB spam is not sustainable in a fight such as Brutallus, without extraordinary support such as multiple shadowpriests or multiple innervates. Aiming for 100% AB time is not always the right approach. In the case of haste vs crit, haste will cause your AB spam to run you oom faster causing you to switch to filler earlier. This is not a dps loss as the AB spam time lost due to haste is still being used for dps via filler, as well as regenning mana.

I'm not sure how coherent that explanation was, so to clarify i'll give a simplified example.
With 0 haste you cast 100 AB's in 150 seconds.
With 10% cast time reduction you cast those same 100 AB's in 135 seconds.
The remaining 15 seconds are used for 6.6 frostbolts, during which you regen mana to be used on more AB later.

As you can see, 10% cast time reduction is not a 10% damage increase due to mana concerns, but the damage increase is still significant. The reason why crit is not strictly better despite not affecting mana usage is the same as for any other spec. It simply doesnt give enough of a damage increase for the item budget it uses, and AB's small 175% crits only amplify this issue.

Until recently (actually until i sat down and spent a few hours reading over this thread) I was also under the impression that haste was BAD for an arcane mage but I think I finally understand why haste is actually a decent stat to acquire.

First off, to get rid of the obvious, dps = damage/time. The goal is to find out what does the most amount of damage over a fixed amount of time.

Let's assume that someway, somehow, you managed to get a passive 20% haste on your gear, and that we're looking at a 5 minute fight. If you were to use the exact same cast sequence in your 20% haste gear as you did in your non haste gear, you'd go oom in about 3 minutes, 50 seconds (trying to account for the mana lost from regen and spriest); you'd have delivered roughly the same amount of damage but in a little less than four minutes instead of five.

That extra minute you just acquired is not dead time because you can still cast your filler spell and have a net gain in mana.

Ex) assuming that you cast AB 90% of the time, that translates (before haste gear) into roughly 180 ABs and 12 FrBs.
If you add in that extra haste gear, you'd get something like
First four minutes: 175ABs and 12 FrBs
Last minute: 6 ABs and 20FrBs.
TOTAL: 181 ABs and 32 FrBs.

180ABs and 12 FrBs < 181 Abs and 32 FrBS
Being able to cast more spells in the same interval of time is a dps increase even if the ratio of Arcane Blast to Frostbolts decreases.

This has interesting implications that I, for one at least, had not realized. The Arcane spec is more complicated than assuming that higher AB:FrB ratio means higher dps, you really do have to take the whole picture into account. Yes, in the short term, your run OOM faster BUT it also allows you to squeeze in additional filler FrBs; essentially, though it may be counterintuitive, haste is actually increasing your mana pool.

The question then becomes, what is better? That massive passive haste in the example above came at the expense of something, would more dmg be better? more crit? int and spirit?

Rawr will help give you an idea of what to expect, but there really isn't an answer that can be posted here that will serve everyone well for every encounter.

At first, I belong to the school of thought similar to Zinaida and Elysianfield:

With haste you go OOM sooner, buying you time to go into mana regen phase i.e. FrB, thus gaining you more mana for more AB spam, which threfore results in a total dps increase.

Then Alvira struck me with this:

"Mana that you have in an entire fight is essentially finite. Based on constant 60% regen, evoc, pots, gems, etc. If you have 20k mana in a 4 min fight, that is it."

Which I thought was a powerful argument. Casting FrB may seem like it's giving you a +ve mana return, but whether you cast AB or FrB, your 60% mana regen remains constant. So you don't really get 'extra' mana from FrB regen by going OOM earlier, more like you are proportionating more mana into FrB than AB.

Then again, haste is not necessary a bad stat. As just a few posts earlier there were discussion on why not wand instead of FrB? Perhaps stacking haste then go into wanding instead of FrB'ing is an idea?

However the point where the balance is tipped is really, haste is often stacked at the expense of spell damage, crit and int. So for shorter fights, I believe haste might just increase dps, but for longer fights spell damage, crit and int may be the way to go. In other words, I believe haste means different thing to mages in T5, T6 and Sunwell content.

Mana throughput isn't "finite" though. I think a better word in this case is "limited". Anyway, although it would seem that mana throughput is fixed by the fight length, there are some other factors like JoW and clearcasting which depend on the number of spells you cast. Furthermore, since evocation is usually a big timeout during any fight, you also have to consider how much extra casting time you get from a hasted evocation. Note that Alvira mentioned having "20k mana", actually your mana throughput is much higher on a six-minute fight. If you have a 14k pool and evocate 12k of it back, 25k from VT, 8k from JoW, 8k from pots, 9k from gems, plus another 8k from totems. That's 84k mana already without considering your 60% regen, so in reality you are approaching a mana throughput of 150k. Not to mention clearcasting. In this sense, the mana spent on frostbolts is really very trivial, maybe 10k-15k. So claiming that extra mana spent frostbolting is cutting into your AB is factually true but kind of misleading since it's not a 1-to-1 relationship. AB costs much more per cast and per second than frostbolt.

Look at the itemisation please. Show us items which have haste but no spell damage, or haste but no int. In fact, it's hard to find any caster pieces with haste that are worse than an equivalent iLevel piece that doesn't have haste.

I wouldn't take the 20k mana Alvira mentioned literally. I think in fact the 'finite' mana pool that he mentioned includes all other mana regen factors that you said (JoW, clearcast etc.). The point is, after you put in every possible positive mana regen factor, that mana pool is still 'finite'; i.e. there's no other way to increase this total pool of mana except lengthening the fight. Optimal dps output is when you use every drop of mana from this 'finite' mana pool for AB, then isn't it true that for every FrB that you cast you are moving away from optimal dps? (Because the mana used for FrB has to come from this 'finite' mana pool and no where else, it's a bit like the first law of thermodynamics - and not only that, it also eats into the time, because time is also finite - 1 AB is 1.5s whereas 1 FrB is 2.5s)

As for haste stacking at the expense of other stats, I was referring more to gemming. But it's certainly possible that one might stack haste at the expense of other stats, e.g. picking [Bracers of Nimble Thought] over [Cuffs of Devastation].

Just curious, are there any WWS logs of an arcane mage pulling 2k+dps at Brutallus? Couldn't find any. In theory, they can, of course, but I would really like to see a log.

Yesterday I specced arcane (weird 47/0/14 spec) for ZA timerun because we had 1 shadowpriest and 1 warlock, so CoS was used most of the time, and I thougt I would perform better with arcane + cos than with fire and no coe. Besides that, I just wanted to play a litte bit around with it.

I was quite suprised how good it actually turned out to be (warlock used CoR on some bosses), considering I didn't had a shamy nor did I have 2 pally buffs or improved divine spirit. Since I'm fire specced normally, my gear is also far from perfect for arcane. It's gemmed for 2/48/11 and I don't even have SCB.

Here is the log: Wow Web Stats

Was arcane just so good for me because the fights in ZA are rather short compared to Sunwell bosses or could I sustain it for a fight like Brutallus and deal even more damage with all the missing buffs I would normally have? I know 1900 isn't really impressive for bosses, fire will do this with ease, but like I said, I was suprised.

So I would really appreciate a log of an arcane mage pulling 2k+ (preferable 2,2k-2,3k, so that's it's comparable to fire dps) dps on Brutallus to check his armory in order to find out, if I could give arcane a shot next week at Brutallus.

The thing that people talking about haste seem to not be considering is that frostbolt is better DPM than arcane blast. So take a fight where you cast 100 ABs (ignoring ramp time, 150 seconds) and go OOM exactly at the end. If you get 2.3%, you could give yourself 153.5 seconds of effective cast time during that fight, and cast 99 ABs and 2 Frostbolts. This uses the same amount of mana, but does more damage in the same time. Our goal is always to maximize dps, which does not have to be the same thing as maximizing AB percent (if it did, we'd spam AB until OOM and never cast frostbolt).

Edit (because I realized I had more to say): This doesn't mean that haste is good, but it should not be totally ignored given the choice. Rawr in the item budget category for me values 10 Int ~= 11.7 spell damage > 10 spell haste > 10 spell crit, so [Shadowcaster's Drape] ends up better than [Brute Cloak of the Ogre-Magi], for example. In terms of gemming, spell damage or int will almost certainly end up better than haste, but depending on what gear you have access to, the best piece for some slots may have haste on it.

The only WWS for an arcane mage on Brutullas that I've seen would be Fireangel with around 2400-2500 DPS. Unfortunately, the only reports I had links to have expired... but someone here probably has one, or knows of other arcane mages doing really well on Brut.

No, it does not. You're treating AB uptime like it's fixed value and everything else adapts to that. That might be how you plan a fight given gear choices that you locked in at the Aldor bank. But varying your gear choice -- and indeed, even speccing arcane in the first place -- is based on the notion that you can cast a filler cycle for ever and ever and the extra mana can be dumped into AB cycles to improve dps. The viability of the filler cycle is the groundwork for the build, while constructing a setup where the AB spam brings your dps to a competitive level is where all the work goes and has gone.

Let's say you have enough haste to increase your dps by a factor of k (e.g. under icy veins with no passive haste we have k=1.20). Call t_1 your filler uptime with no passive haste and t_k your filler uptime with haste factor k. As long as we don't mind an oversimplified model of clearcasting, we can say mps* and dps all scale by exactly this factor, k. If t_1 is the same quantity as t in the quote from Kavan, we have

t_k = (M - k*T*abmps) / (k*cyclemps - k*abmps)
.= k * (M/k - T * abmps) / (cyclemps - abmps)
.< k * t_1.

It is intuitively obvious that t_1 < t_k ("more haste means less AB uptime means more cycle uptime"), but the preceding quantifies this. Below, the first inequality is the intuitively obvious one and the second one follows from M/k < M:

t_1 < t_k < k*t_1.

As an example, if we have 90% AB uptime with no passive haste then adding 5% passive haste puts our new cycle uptime somewhere between 10% and 10.5%, or an AB uptime between 89.5% and 90%. But remember, we're not trying to optimize AB uptime, we're trying to optimize aggregate damage. Apply the preceding double inequality to Kavan's forumal for totaldamage_k (our total damage done with haste factor k, of course):

totaldamage_k = t_k * k * cycledps + (T-t_k) * k * abdps
.> t_1 * k * cycledps + (T - k * t_1)*k*abdps
.= k * (t_1 * cycledps) + k * ((T - k*t_1) * abdps)
.= k * (t_1 * cycledps) + k * ((T-t_1) * abdps + (t_1 - k*t_1) * abdps)
.= k * (t_1 * cycledps) + k * (T-t_1) * abdps - k * (r*t_1*abdps).

The surprise quantity r is just k-1, the haste percentage that appears on our character sheet. The inequality at the top says that this is only an estimate -- a lower bound on our dps gain from haste factor k. The last equality says that the gain is almost a factor of k, but we have to pay a tax based on how much filler we were using before we added the spell haste. That tax is slightly more than r percent of our AB spam dps, applied on our filler rotation. Thus

totaldamage_k > k*t_1 * (cycledps - r*abdps) + k*(T-t_1)*abdps.

Let's pick some numbers to illustrate why this is good. If we take k=1.05 so that r=.05 and the very rough estimate that abdps = 1.5*cycledps (throw better numbers at me and I'll happily edit this part later) then our filler cycle dps is only suffering by about 7.5%. If your no-haste setup had the benchmark 90% AB uptime, the weighted average tells us that our total dps has gone up by .9*.05 - .1*.075 = 3.75%.

Compare that to going from 30% crit to 35% crit on AB which can only boast a 3.2% boost to your AB dps (those damned 1.75 crits), or 2.9% of your total dps. To finish the crit comparison I can't neglect the contribution to filler cycle dps, but even generously giving the whole cycle FrB's 225% modifer with the same crit rate (unfairly penalizing arcane impact) produces a 4.5% increase on 10% of your total cast time, or a grand total of 3.45% increase.

So gearing an extra 5% haste gets you at least 3.75% more dps but gearing an extra 5% crit is pretty close to 3.45% more dps. Factor in that you need 5*22=110 critical strike rating to get that 5% but only 5*15.75=79 haste rating, and the picture for crit rating is starting to look very bleak indeed.

There are definitely some rounding errors up there, and I would welcome a better estimate on 1.5*cycledps=abdps. It is also worth noting that the more filler you were casting before you geared spell haste, the worse it is, putting it in the same boat as int and spr as opposed to crit and spellpower's advantages over long fights. I have also neglected haste's excellent synergy with 3 minute cooldowns (not to mention its not-so-excellent synergy with gcd clipping). At the very least I hope you can now see that the value of spell haste is very close to the value of spell crit and, like all options to an arcanist, it is highly environment-dependent. Consult a more precise theorycrafting tool.


Edit:
*: Of course mps does not scale at all in this fashion since haste does not affect your spirit or shadow priest regen, but the only place the mps appears in the calculations is as the difference cyclemps - abmps. The nonhasted regens cancel each other out and the difference does scale the way I claim it does.

But realise we are not talking about the same thing here? In your example, if you take into account of the various mana regen factors, you actually have MORE mana than required for 100 ABs in the 150 sec fight, therefore in this scenarios haste is good because it allows your utilise the unused mana.

Whereas in the finite mana theory, it is assumed that, taken into consideration of all mana regen factors, you have exactly 67.2k mana for 100 ABs in a 150sec fight. The catch is you don't have all the 67.2k mana all ready for use, instead much of them comes trickling in. If at any point of the fight your mana on hand is insufficent for an AB cast and you resort to FrB, the end result is always less htan 100 ABs. I liken it to filling a leaked tank with a fixed volume of water; the tank will be able to keep up a certain water level (maintaining AB cast) if the leak is small, but the moment you make the leak wider i.e. haste, your AB up time will reduce.

I don't disagree with what you said about the role of filler cycle in the spec. But in reality, a boss fight don't last forever and ever, it has a finite duration. To maximise dps, you want as many AB casts as possible within this finite duration. In a perfect world, it's 100% AB. Alas, the world is less than perfect so we weave in the filler cycle to achieve a value as close to 100% AB as possible. Isn't that the purpose of the filler cycle - to enable as many AB casts as possible?

Heh, took me a lot pains to digest those math

I could follow half of it until this point:

totaldamage_k = t_k * k * cycledps + (T-t_k) * k * abdps

which was made in reference to Kavan's formula:

M = t * cyclemps + (T-t) * abmps
total damage = t * cycledps + (T-t) * abdps

Yes I agree that - totaldamage_k > total damage

However I'm not convinced or I haven't been able to see your provision that

M = M_k

i.e. totaldamage_k utilises the same pool of mana M as found in Kavan's equations.

Not that I care, but using that parse as an example gives a bleak outcast for arcane. Maybe one with at least 1 innervate on the arcane mage would make it look a lot better.
Even that would be a long shot from what fire can do. Taking this week's parse for example
Wow Web Stats

I'm still considering spec'ing fire just for Brutallus, but here's a WWS where I did relatively well as arcane. Note that this is our guild's first Brutallus kill (so not all the play is as smooth as I hope it will become); I don't have t6 bracers, a Skull of Gul'dan, or Zhar'doom (which I picked up this week); and all innervates at this point are saved for healers.

Also, I hope more people will see the necessity of evaluating haste rating and crit rating more quantitatively and in context, as simplistic rules ranking stat priority, without any reference to existing totals of each, lack any real value (and can easily be wrong for a given setup). I wonder if Blizzard has not done this endgame calculation quite well themselves, or perhaps they've simply gotten lucky with how the values work out, but Rawr shows haste rating and crit rating in equilibrium for 40/0/21 with a good dose of Sunwell gear, with the precise equilibrium point of course shifting slightly one way or the other depending on slight fight parameter changes. The real challenge in exploiting increased haste for its full potential lies in adapting one's use of what have been called "overlap cycles" and the timing of Heroism and Icy Veins. But then, one of the reasons I like playing this spec is that it's not so simple to use.

Huh? You have 67200 mana, and 150 seconds to use it. You can either cast 100 ABs with no haste for 67200 mana, or given some haste, 2 Frostbolts, 99 ABs for 67096 mana and more damage. Obviously you have to give something up for the haste, but the point is that it's a tradeoff you have to consider, not a case of "haste is always bad."