Sancus

Caster haste is significantly weakened by the fact that it doesn't increase DPM like every other dps boost in the game(ie, haste increases your dps at the cost of consuming more mana) and due to the GCD cap.

The only raiding caster/spec in the game that spends 100% of its time casting a spell that isn't severely GCD capped is the Fire mage... and they're hardly competitive dps in the first place.

(and well, it's actually 95% due to scorch, but whatever :P)

<Vontre> I removed the cooldown on evo
<sancus> and what happened?
<Vontre> DPS went down rofl

 

◊ Anias

If you are going to speculate wildly in my thread, I would like to provide you with a tool to use for purposes of basic analysis. It's called a timeline.


Start - - - - - 5seconds - - - - - 10seconds - - - - - 15 seconds - - - - - 20 seconds - - - - - 25 seconds - - - - - 30 seconds

Map your casts/attacks out on a timeline for whatever period of time you feel like analyzing, I find 30 seconds to be a useful period.
Then map your casts/attacks while hasted on a duplicate timeline.

Compare the two timelines side by side to examine the gains you pick up from haste. In particular note when spells/instants come sooner, as that will give you a meaningful idea of how haste benefits you.

Most of your claims/speculation/gross-misinformation can be analyzed in this fashion, and answers arrived at accordingly.

Caster's would do well to remember that the server has some built in packet/spellcast overlap as a means of handling shifting latency conditions, and mellee would do well to remember that haste generally decreases your instant/auto attack ratio by giving you more auto attacks with the same number of instants.

 

Amonra

The theorycraft seems fine, but the problem arises when this is translated into real conditions (i.e. latency and response time).

This is because mindflay is bottom of the priority list of spells to be cast on most fights. So whenever latency or response time cause a delay in spellcasting, mindflay the spell that gets dropped from the rotation while the others are maintained.

For myself, mind flay typically accounts for about 25% of my overall dps, meaning that I only get a quarter of the full potential benefits from haste.

 

Roywyn

I'm risking to beat an old horse, but Heroism/Bloodlust and the GCD has been bugging me a lot recently.

I'm using /stopcasting in all my casts, and use Quartz as cast bar addon, and use the option to show the global cooldown. The GCS is a little light that runs from the left end to the right end of my cast, right under the cast bar.

When I cast a 1.5s cast time spell, I see the cast bar move from the left to the right at exactly the same speed as the the GCD light moves from the left to the right.
I also see very easily if a spell is cast faster than the GCD, because then the cast bar starts behind the GCD light (casting lag) and when it gets closer to the GCD timer or even reaches/overtakes it, then the spell casts noticably faster than the GCD.


That being said, a 1.5s cast time spell (Arcane Blast, 1.5s cast time with 3 debuffs) under Heroism (=> 1.15s cast time) is cast significantly faster then the GCD. I have seen that for a while in several raids, and decided to copy a premade Draenei shaman to the PTR.

I specced him for faster Healing waves (1.0s, 1.5s, 2.0s for rank 1, 2, 3), made /stopcasting macros for those 3 spells installed Quartz and enabled the GCD light under the cast bar.
I played with those 3 spells/macros to get a feeling for timing.
Rank 2 is very nice to use, because at 1.5s cast time, the cast bar and the GCD move at the same speed.

Then I used Heroism.
Rank 2 was now casting significantly faster than the GCD completed (like it felt from raiding on the live servers).
Rank 3 now had 1.54s cast time, and it's cast bar moved about as fast as the GCD timer.

I repeated that 3 times, and everytime it felt exactly the same. Like that the global cooldown is still 1.5 seconds while under the effect of Heroism.


I've read a lot of things about the GCD being shorter under Heroism/Bloodlust, people claiming the opposite. I don't want to say that anyone is wrong. But I'm sitting there with my testing and it seems to contradict the general consensus.

I couldn't find how others got their results of how it reduced. I heard rumours that Bloodlust reduces the GCD but Heroism doesn't. Then again, usually horde players defend that it reduces the GCD while the doubt comes from alliance.

So, how did people find that it reduced the GCD? Which settings/spells did you use?
And those who found it reduces the GCD, could you try to repeat your tests with Quartz, GCD timer and /stopcasting macro for a 1.5s cast time spell and confirm that you can chain them under Heroism or Bloodlust and see the GCD pass at the same speed as the spell is cast?

I used up my PTR premade copies, so I sadly can't copy a Horde Shaman to try Bloodlust.

 

◊ manly

It's been a while I had a sneaking suspicion that AB spam during bloodlust was not working as it should (ie: general conscencus being that GCD goes to 1.15s). I chalked it up for the longest time to lag, but its been 2 weeks I got the impression as well that bloodlust was not lowering the GCD. Since this week I stopped using AB spam during bloodlust, instead I do 2x AB, AM, fireball as bloodlust rotation, and it seems like it works a lot better. Of course, this is nothing more than anecdotal evidence, but I definately got the same impression as horde.

Log on with different model:
1- Create a character of the desired model. Log on/off.
2- At character selection screen, select your actual character; mouseover the new, desired model character, and hold down left click; hit enter and release left click at the same time.
bug Arcane Potency only applies to the first Arcane Missile bolt.

 

• Intermission

This thread seems to imply otherwise.

The initial graphs are a bit odd, read down a little for a further explanation.

 

◊ Anias

That's new, once the data set gets a bit larger I'll update individual procs accordingly. It looks like there's different behavior between individual procs now, but I want a chance to go stand beating on a blasted land servant to confirm differing behavior and that's not happening till the weekend. Stay tuned faithful readers.

 

Boop

We know that the amount of benefit that a given spell gets from + dmg is related to its cast time.

If I remember correctly, the formula is:

Actual Spell Damage = (Base Spell Cast time x +dmg stat)/3.5

I'm curious if anyone has investigated whether or not the use of spell haste would further diminish the affects of + dmg due to lowered cast time. My gut instinct says no, as the coefficient is not affected by talents that reduce cast time, but I have nothing other than what I think it "should" be, and that's far from fact.

 

• Chicken

Spell haste rating does not affect the coefficient in any way. A shadow bolt cast while your [Quagmirran's Eye] had procced will do the exact same damage as one without. I'm taking this trinket as example mostly since it has a noticeable amount of spell haste going on, and the fact that I know a Warlock that used it and, indeed, had his shadow bolts continue doing the same damage.

 

Daidalos

You are correct Roywyn.
All the testing I've done and all the evidence from people who actually tested instead of assuming has shown bloodlust does NOT lower GCD.

Also it is mentioned here: Working Theories of Theorycrafting as of 2.1
I find it really annoying that so many people on EJ forums keep repeating that it does without actually knowing anything about it.

 

Nite_Moogle

You replied to a month old post. When it was posted, that thread did in fact say that BL/H affected the global cooldown. Just because it's on the last page doesn't mean it's current.

 

Daidalos

Oh I wasn't looking at the dates.. however even a month ago I thought it should have been well known. 2.1 has been out for a very long time.

 

Angrenous

How much do you value haste in PVP? My friend thinks it's crap, but I like my haste-based openers that result in 5 seconds of very fast weapon speed. The two main benefits are probably combat potency so I can mash sinister strike again and additional attacks which result in more proc rates of weapons and abilities. Still, the top rogues don't seem to care much about haste - sure they might use slice and dice on top of blade flurry, but that's about as far as it goes. Is haste for rogues purely a PVE Thing?

 

Fenz

How good is Blade of Wizardry? Is it better now then weapons with crit on it?

 

Anaxo

Blade of Wizardry is not that good. The haste proc has a hidden 45s cooldown, and you are give up far too much spell damage and raw stats in exchange for the chance at haste. You could work toward a [Stormcaller] (Thrallmar) or [Blade of the Archmage] (Honor Hold) at the equivalent level.

 

yaya@EU-hyjal

Haste is very comparable to heal power because they both increase hps (Heal Per Second). One difference though is that heal power increases additively hps and haste multiplicatively. This means that the more heal power you have, the more beneficial haste is.

Let us compare haste rating and +heal with respects to hps.

The benefit of haste depends on the healpower of the caster and the hps of the spell used, therefore we have to make some asumptions before going further. A resto shaman will mainly use the stwo spells: healing wave and healing salve. Lesser healing wave is not considered here because it does not benefit from haste due to its 1.5s casting time. I am also assuming the standard resto shaman spec (i.e. purification, improved healing salve, etc...). The value of haste as compared to +heal is given in the two following tables (for the two different spells) and the formulas are in the annex at the end of the post.

The following table shows the value of 1% haste when casting chain heal max rank. How to read the table : if you are standing at 2200 heal power, your hps without haste is 2269 hp/s and 1% haste is equivalent to +34 heal.

This is the same table but for healing wave (max rank) which has a different hps value. At 2200 healpower, 1% haste is worth 48 +heal to a shaman casting healing wave.

Conclusion : haste is VERY powerful for a T6 resto shaman. To illustrate this, let us take a look at this item which is a crafted BoE:

Living Earth Bindings
Binds when equipped
Wrist Mail
432 Armor
+27 Stamina
+20 Intellect
Equip: Improves spell haste rating by 28.
Equip: Increases healing done by up to 64 and damage done by up to 22 for all magical spells and effects.

As per our formula, we see that 28 spell haste rating (which represents 28/15.7 = 1.8% haste) is equivalent to +61 heal for a 2200 +heal shaman casting healing salve, in terms of hps. It is of course huge because it's like if the bracers had +125 heal. It makes it the best piece of equipment for short fights, or if you are in a shadow priest group.



Annex : formulas

What is the gain in hps when you gain x +heal with no haste ?

If you gain x +heal the hps increase depends on the spell cast. Let hc be the healing coefficient (i.e. the coeffient of +heal added to the heal) and ct the casting time of the spell. The hps increase is (hc/ct)x. With purification this becomes 1.1 * (hc/ct)x

For healing salve we have hc = 1.25 (i.e. healing salve benefits 125% of your +heal) and ct = 2.5 s. Therefore the hps increase is 1.1 * (1.25/2.5) * x. With the talent improved salve, it becomes 1.2 * 1.1 * (1.25/2.5) * x which is 0.66x

For healing wave we have hc = 0.86 and ct = 2.5 s therefore the hps increase is (with 3 healing way stacks and purification) 1.1 * 1.18 * (0.86/2.5)x which is 0.446x


What is the gain in hps when you gain 1% spell haste ?

If 'a' is your hps without haste, then the hps increase is a/100


When is +x heal equivalent to 1% haste ?

+x heal is equivalent to 1% haste when the hps increases (calculated above) are equal.

That is for healing salve x = a / 66 and for healing wave x = a / 44.6


Computation of hps

Let x be the +healing value of the character

For healing salve the average raw healing is H = (1547 + 1.25x) * 1.1 * 1.2

For healing wave the average raw healing (3 stacks of healing way) is H = (2285 + 0.86x) * 1.1 * 1.18
For both spells the hps is then of course H/2.5

 

Daidalos

I am getting different numbers when I do my conversions.
I go step by step below see if you can find my or the above poster's error.

chain heal rank 5: avg base heal = 884
with 2200 total plus healing chain heal rank 5 heals for
(884+ (2.5/3.5)(2200)) *1.2 *1.1 = 3241 single target.
for 3 jumps it is 3241 *1.75 = 5672.04

5672.04 healing / 2.5 s = 2268.816hps

with 1% haste
5672.04 healing / (2.5s *(1/1.01) ) = 2291.548 hps

2291.548-2268.816 = 22.732hps
so to calculate the amount of healing done we mul by unhasted cast time.
22.732*2.5 = 56.83
Now we convert to amount of plus healing.
First we calculate how much it would be on a single hit of chain heal as opposed to 3.
56.83/1.75 = 32.47
then solve using CH coefficients with imp CH and purification.
32.47 / (2.5/3.5 +.2 +.1) = 32 +healing equiv for 1% haste on chain heal rank 5 with 2200 +healing.

 

yaya@EU-hyjal

You are right Daidalos, I forgot to include the talents in the calculation (shame on me) and I made another mistake more subtle. I will correct my post in a moment (I can't right now) and we'll see if I come up with the same number as yours.

Anyway thank you for pointing that out

 

Daidalos

Ah yeah that would def throw things off. I didn't check the healing wave numbers but did you assume 3 stack of healing way or not? Might be good to note which way you calculated.

The point still stands that haste is a great way to inrease HPS though. With mana not looking to be a problem I'm stacking +heal and haste as much as I can.

 

yaya@EU-hyjal

I have updated my post (cf. post #91). There is still a small discrepancy with your computation but I cant' figure out why yet. As you said, the same conclusion holds true, haste is very interesting especially after the patch 2.3 regen buff for chamans.

 

yaya@EU-hyjal

When chain casting a spell, latency adds a flat idle duration in between each cast. For instance when chain-casting healin salve with a 0.5s latency, the spell becomes actually a 3s cast spell, in terms of hps produced.

The value of spell haste is reduced in that case because haste applies to the 2.5s of the cast but not to the 0.5s of latency (obviously).

The following table shows the value of 1% haste for a shaman with +2200 healpower chain-casting healing salve, with varying values of latency (expresssed in second). The computation is at the end of the post.


Conclusion

Latency has definitely a negative impact on the benefits from spell haste. However it's not as bad as I thought initially. The bottom line is that 1% spell haste (i.e. 15.7 spell haste rating) is roughly equivalent to 30 healpower for a T6 shaman with reasonnable latency, chain-casting salve.



Annex: computation

Damn was I bored today to go into so much detail

Let us consider a healing spell.
Let base be its base healing with no healing (e.g. for healing wave it's 1547)
Let cast_time be its cast time (2,5s for the salve)
Let hc be its coefficient received from +heal (1,25 for the salve)
Let alpha be the healing multiplier due to talents (e.g. alpha=1.1*1.2 when considering a salve with purification and improved salve)
Let raw_heal the raw healing produced by the spell.

Let lat be the latency
Let h_power the total +heal of the character (typically 2200 for a shaman in BT)
We have raw_heal = alpha * ( base + hc * h_power )

Let x be the +heal equivalent to 1% haste for the healing spell considered, in terms of hps.

We are looking for 'x'

Let hps_heal be the increase in hps due to gaining +x heal with no haste
let hps_haste be the increase in hps due to gaining 1% spell haste

Due to the definition of 'x' we have:

(1) hps_heal = hps_haste

Besides, the raw healing gained with +x heal is: alpha * hc * x
Therefore if we divide by the cast time we get the hps gain:

(2) hps_heal = ( alpha * hc * x ) / ( cast_time + lat )

Let us now compute the hps gained with 1% spell haste.
First let's notice that the casting time with 1% haste is ( cast_time/1.01 + lat )
The hps increase is the difference between the hps with haste and the hps without:

hps_haste = raw_heal / ( cast_time/1.01 + lat )
- raw_heal / ( cast_time + lat )

or (by replacing raw_heal):

(3) hps_haste = alpha * ( base + hc * h_power ) * ( 1/(cast_time/1.01 + lat) - 1/(cast_time + lat) )

From (1), (2) and (3) we have:

( alpha * hc * x ) / ( cast_time + lat ) =
alpha * ( base + hc * h_power ) * ( 1/(cast_time/1.01 + lat) - 1/(cast_time + lat) )

We can divide by (alpha * hc) on both sides:

x / ( cast_time + lat ) = ( base/hc + h_power ) * ( 1/(cast_time/1.01 + lat) - 1/(cast_time + lat) )

We multiply both side by ( cast_time + lat ) so we get:

x = ( base/hc + h_power ) * ( (cast_time + lat)/(cast_time/1.01 + lat) - 1 )

This is what we wanted.

(It's interesting to note that the value of spell haste is not affected by talents)

 

Daidalos

I would also point out that with the server side 2.3 changes for spells I think most of the time the average time time between casts will be low. Basically the server allows the client to send requests to cast another spell at anytime. The server then checks if there has been enough time to start the next cast.

I believe this is how current communication works.

For example. in 2.2 mechanics With 400ms lag. spamming 2.5s cast.
time: 0.000: client sends request to begin cast.
time: 0.001: server has not recieved anything yet. nothing happens
...........
time: 0.400: server accepts request. You begin to cast.
time: 0.401-2.899: attempting to cast does nothing. Client has not recieved the ok from server to allow next cast yet.
time: 2.900: cast completes. Server sends response to client notifying it can cast the next spell
time: 2.901: client still cannot cast next spell. notification from server not recieved yet.
.............
time: 3.300: client recieves ok to being cast. client sends request to server to begin cast.
time: 3.700: server gets request to begin cast.
etc

now this is why stopcasting improved caster dps so much. Stop casting basically said well the server knows the cast completed at time 2.9 so any time after 2.9-(.400 latency) in this case after 2.500s the client can cancel that spell and begin the next spell. However since latency varys if you cut it exactly at 2.5s and your latency had dropped to 390ms it would actually cancel the spell. Since the server would get the cancel notification before the spell completed.


My understanding of how 2.3 works is the following again assuming 400 ms 2.5s cast
time: 0.000: client sends request to begin cast.
time: 0.001: server has not recieved anything yet. nothing happens
...........
time: 0.400: server accepts request. You begin to cast.
time: 0.401-2.499: while casting client can spam casting attempts. Server receives these attempts then checks if the spell has completed. Spell has not completed and does not begin the next cast.
time 2.501: client sends spell cast attempt. (client is super good at spamming his buttons )
time: 2.900: cast completes.
time: 2.901: server recieves the attempt that was sent at 2.501. Server begins next cast.
etc

So in theory chain casting even with latency you can get full effect from haste benefits without adding to your average cast time. Now in practice lag will still affect things and there is only so fast a person can hit a button not to mention there is only so fast the client can send requests (due to keyboard, operating system etc etc).


I'd say it would be hard to get below 50ms between casts just due to real world restrictions but this time between casts is not related to latency only the time between button mashes.

 

yaya@EU-hyjal

Very interesting, thanks for the info

 

Kiirby

Is Spell haste better for-
1. Spriest
2. Warlock (destro or affliction)

If someone knows, i would like to know. Also, if you have any Number crunches or DPS gains.. etc (even from a perfect situation of standing still doing DPS) I would greatly appreciate it. THank you

 

Aluscia

It is probably a stupid question, but I'll ask anyway...

Does your personal spell haste rating affect your Water Elemental's cast speed? I know that they scale with +dmg, so I was curious if this also is the case.