Well, this is not directly related to WotLK, but more to casting mechanics in general.

For 2-3 months now, I had the impression, that Blizzard might have added a queue for casting spells . Therefore I did some testing today to check, whether this might be true or whether its just some imagination of mine.

So what I did was creating a macro for my G15 that spammed Fireball every 2950 ms and one every 2900 ms. I had zero haste equipped during the test, therefore my Fireball had a casting time of 3,00 sec.

When I used the macro with 2950 ms, I was able to cast between 9 and 12 (10 in average) Fireballs before I got the message „Another action is in progress.“

When I used the macro with 2900 ms, I was able to cast between 3 and 5 (4 in average) Fireballs.

When I used a macro with 3000 ms, I was able to chain-cast Fireballs until I went oom.

When I used a macro with 1500 ms and Scorch, I was able to chain-cast Scorch until I went oom.
When I used a macro with 1495 ms, I was not able to pull off more than one Scorch in a row, due to the Global Cooldown.

So here is my conclusion:

When you use the macro with 2950 ms, that means, when you give the command to cast your second Fireball, the first one is not yet finished by 50 ms. The second Fireball however won't start casting until the first Fireball is finished. That means when you cast your third Fiba the second one is not yet finished by 100 ms. And so on...

When I am able to cast 10 Fibas in a row with the 2950 ms macro, I have a delay between the keystroke and the beginning of the cast of (10-1)*50 ms = 450 ms for the last Fiba.

With the 2900 ms macro the delay is: (4-1)*100 ms = 300 ms.

I did this test with a latency between 50 and 70 ms.

So I think there might indeed be something like a spell casting queue, when you hit your key too early by a margin of somewhat between 300 to 500 ms.

I uploaded a Video (6.5 MB; Xvid-Codec) of one test with the 2950 ms macro on filefront, so you can better understand what I am talking about. Also notice, that the latency bar of quartz increases by roughly 50 ms with each cast.

2950ms_fireball.avi - FileFront.com

I would appreciate, if some of the more experienced theorycrafters could take a look at this and tell me, whether this is a viable test to check for spell casting queues, or whether I am drawing wrong conclusions.

This is comparing apples with oranges, i.e. better gear with worse gear. I would rather compare the following, keeping constant iLvl in mind:

Case1 as above: 12/min

Case2 with 25% haste = 390 haste rating and 50%-390 critrating = 32.74%crit = 9.8 crit/min
+ you cast 6 additional frostbolts, out of which each one will have a 10% chance to apply FoF = 0.6 FoF procs
which means you get 1.2 casts out of those 30 with an additional critchance of 50%, means 0.6 crits more for the whole minute.

critrate = 10.4 crits/min

note: I neglected the effect of FoF procs for case1 and case2 up to the point, where they are equal.

bottom line: if you look only at the critrate, haste does less for the critrate, i.e. you get 1.6 crits less per minute. However, 1.6 crits is worth 1.6 additional casts (100% +damage), meaning you get for case 1 the damage equivalent of 36 casts and for case 2 the damage equivalent of 40.4 casts.
This shows that haste is better than crit, if you compare with the same iLvl. Furthermore, haste favours more BF procs.

I added a dps-distribution chart to simulationcraft output.......





I'm not sure what I expected....... but I was hoping for a steeper bell curve over a 5min fight......

I must admit I'm feeling a bit discouraged....... Heh...... Why bother with all this analysis if RNG just obliterates everything? I mean.... really...... How large is our sample size? I mean..... exactly how many bosses a night does one fight even in a hard-core guild?

EDIT: Ugh. Just realized I was using "average_dmg=1" which averages the min/max of direct dmg spells....... so the curve may flatten even further...... not sure..... I'm too tired to think.

This is against Blizzard's ToS for WoW, and they've stated this many times, that macros created in this fashion are illegal. (It goes without saying then that this sort of post is not appropriate for these forums, and furthermore, I highly doubt any respectable theorycrafter would be caught dead using macros for casting in the manner you've described. If you can't make a macro work using Blizzard's macro feature, it's probably illegal.)

Logic would dictate that haste does not scale with crit due to the fact that more haste does not imply that you will crit more often. Your chance to crit does not then scale with the number of casts you've made. In other words, all of the extra casts you are gaining from stacking haste could result in all non-crits. That's simply how RNG works.

And as far as gear goes, if you look at the stat choice on WoTLK gear, you'll see that Blizzard has a certain fondness for either haste+damage or crit+damage. This is in line with their change to +damage/+healing. IE, they need gear that appeals to classes that need haste and crit. Unfortunately, you'll also find that the gear with spirit on it is the one with haste on it. Apparently casters that want haste also want spirit. While that may be true, it doesn't imply that all specs that want crit don't want spirit.

Nonetheless, as others have mentioned, you can't look at crit rate per minute. Instead, you have to look at a value of how much 1 haste and 1 crit is worth in terms of spell power. That is to say, is it better to take 1 haste or 1 crit (or 1 spell power).

What it will come down to for fire spec is the difference between gemming for reckless (haste) or for potent (crit). But usually this only applied when there was a socket bonus worth obtaining, since in either case (up to a certain point, anywhere, somewhere between 1400 and 1600 spell damage), it was always better to gem for 12 damage.

I'll wait until I actually hit level 80 and can properly judge my character using Rawr to determine whether I need haste or crit. I'm thinking crit will be more desirable, but Rawr seems to disagree (and that's with an 18.53 build). We'll see how things tune out though, once Rawr is properly tuned for 80.

The term "crit rate" is simply not used in this manner; crits per minute isn't a very useful number to consider unless there is a specific talent or ability that depends on it, and often when there are such abilities, it's useful to convert such effects to an equivalent increase in chance to crit, for the properties of additional crit chance are well known and understood.

In the context of the Fingers of Frost question, the point is that additional haste will not increase the proportion of casts that occur while under FoF. It simply makes them go faster. That is, there is negligible interaction between haste and FoF, relevant only if there is a case where the buff would fall off without haste but would not with it. This is a small fraction, at best, of all FoF procs, only meaningful when movement is extreme (yet if you must move and the risk of losing FoF altogether is high, burning the charges on Ice Lance would not be unreasonable).


In a general sense, haste scales crit, but crit scales haste as well. This has to do with the structure of the general average DPS formula.



Where D is the average DPS, and the variables are hit chance, damage multipliers, base damage, coefficient, spell power, crit bonus, crit chance, haste, and base casting time.

If we increase crit by an amount ∆c, then we have...



This is the DPS increase from adding some abstract amount of crit for a generic spell. As haste increases, this value increases (remember the 1+z term); thus, crit scales with haste.

Conversely, haste scales with crit.



This is the DPS increase from adding haste. It increases as 1+bc (and hence, crit) increases. Hence, haste scales with crit.

These relationships become more clear when we consider the percent DPS increase, rather than the absolute DPS increase.



We see that the percent DPS increase for crit is constant for changes in all other variables except crit, which scales it down. The same goes for haste with haste. This is a natural consequence of linear scaling.


Thus, we also see that haste is/was generally more valuable than crit had to do with (a) the rating conversion (15.76 haste per 1% haste beats 22.08 crit per 1% crit) and (b) the fact that we start with a lot more crit than we do haste (hence making the 1/b+c factor larger than the 1+z factor).

Yes, this is a simple model. It doesn't account for a lot of things. This is why we use more advanced tools like Rawr and sims; WoW is fairly unfriendly toward abstract manipulation for all but the most simple cases.


Something to note on the way to level 80 is that, compared to our level 70 stat equivalences, spell power has gained significantly. Rating conversions inflated by a comparatively large factor (see the combat ratings thread), but base damage inflated rather minimally, and as a result, this skews the stat equivalences in the direction of spell power, at least until spell power levels inflate (disproportionately so) to compensate.

There has been a discussion about button spamming earlier on in this thread (page 116). I was under the impression after reading these earlier entries in this thread that the mechanism had been changed and button spamming was now OK. I am using Quartz and have fixed my casts to my mouse wheel (which as far as I know isn't against regulations), with the various casts tied into different modifiers in a macro, ie arcane barrage is "mouse wheel down", arcane blast is "ctrl+mouse wheel down", etc. So I'm giving a quick spin to my wheel just before Quartz is indicating the cast is free. I have had no problems and feel my casting time is quicker than most other players in my guild not using the system (I get off more casts compared to similarly-specced players), but I have no empirical evidence to support this. I also feel this allows me to concentrate just that little bit less on button timing and more on, for instance, getting out of that green patch of area damage or on my cooldowns.
Is there any way that we can get clarity on this issue? It seems such a fundamental piece of theorycrafting, that I think we really need to decide where things stand.

It's not really about the DOT when using firestarter, it's about the burst.

I just use max rank both times. With blast wave the second flamestrike isn't always placed exactly where the first one is anyway. Also you may want to then follow with cone of cold to begin an arcane explosion kiting sequence, if the mobs are especially durable. That'll have them moving off the DOT anyway. So the dot doesn't really figure into my equation most times when I'm doing the firestarter sequence, at least one that involves blast wave.

flamestrike/dragon breath/flamestrike/blizzard-flamestrike is how I'd do more of a Hyjal-type senario. Downranking might or might not have some value there, I'd have to do the math, and there isn't a lot of poit in that until I'm doing the math for 80, as the L70 situation is not one where I really need sustained AOE. Burst is getting the job done really well.

I had done the same thing right after the patch where they changed the casting system. Basically, if you just turn the mouse wheel all the time, you will always get the best possible result in chain casting, it is as efficient as a G15 button macro because of the server-side spell queue of a few hundred milliseconds. (The server will always start the next cast as soon as possible.) Sometimes, when chaincasting under bloodlust and IV my 1.3s frostbolts, I just continuously turned the wheel.
Since it is a built-in option in the game to set the mouse wheel to a button, this cannot be against the regulations.

However, I changed it back a while ago because it was putting a bit too much strain on my right hand/arm

As far as I am aware, so long as you are doing something using the standard tools available with the game, it's ok. It's when people use special hardware (like the G15 keyboards) to introduce scripts and macros that are otherwise not available in the game.

Am I too late? Is it possible to add some comment about 4T6 to the OP? I remember that somewhere in the thread mentioned 4T6 is about 150 spell power, and somewhere else a beta tester said only 4T6 and skull last till naxx. Which should be the 4th piece to wear?

First off, he is using it to get an accurate number for a test, not exploiting in any form.

Second, there is no difference between you hitting a key every 3s and your g15 hitting a key every 3s in terms of what blizzard sees.

Third, they don't care about g15's that do simple rotations or eek out 1% more dps, the intent of the rule is to stop automation of farming.

This effect has been around since 2.3, here's some details on how it works and what you can expect from it:

World of Warcraft - English (NA) Forums -> 2.3 "Quartzless" spellcasting mechanic

Summary:

Quartz is still better than spamming
Spamming will work reasonably well, but has the potential of being bad



Side note for others, I've got a working version of my rehash of Lhiveras old TCoM up, Theory Craft-o-Matic , if anyone would like to give it a gander any commentary would be welcome in the thread on the US Mage forums World of Warcraft - English (NA) Forums -> Lhivera's TCoM, redux! (Updated 10 Nov) or PMs here.

You mentioned in that thread that you don't trust Rawr.Mage. I'd be very interested in finding out what about it you find problematic as I'm trying my best to have it as accurate as possible.

Shortly after Wrath comes out, I'll be bringing my mage up to level 70. With my healing priest I immediately went into PMC and was Kara ready shortly after turning 70. I'm wondering what everything thinks about the spellfire tailoring now. My mage is currently Enchant/Skinning, so I'd have to drop skinning, but is it even worth it to craft the SF set with everything so nerfed?

Thinking of the way we evaluate our value for raids, namely in term of dps (which is a timely rate) it is pretty irrelevant how many of your casts will crit in total, if you can achieve more damage with less crits. Thus I think the rate discussion is not totally out of the world. Modelling FoF charges as extra crits is pretty straightforward when it comes to FrB dps, which is pretty much all there is about frost. WE will give a % increase to dps as will IV which can be modeled as a flat modifier.
As a side note: BF scales with haste and not with crit.

Slow slaves won't happen. Very few raids run with 4-5 fire mages which is how many is needed to make a Slow slave worthwhile. I can't imagine many mages volunteering for the role either unless they switch who the Slow slave is around for each raid.

I do believe there were some charts a while back showing Slow slave dps and naturally it was lower than any other mage spec and just generally most other dps specs.

I can't speak for Zaldinar, but many random people I talked to are put off by the way Rawr values intellect for fights where mana is limited. They are just not aware of how accurate Rawr calculates mana consumption and values each stat accordingly. You might consider putting a simple check box of some sorts where you can choose between mana limited and mana unlimited.

Either way, Rawr is hands down the best tool out there for mages looking for some real TC action.

Any macro that automates the game for you is illegal. A simple rotation like a hunter rotation macro is obviously acceptable. It's not automating anything. You are sending commands every time. However, setting up intervals between spell casts is illegal in that you are not in control of your actions. Your spells are automatically being sent after the set duration. I don't know Blizzard's stance on the difference between "testing" and using said macros for grinding/raiding. I wouldn't advocate it though. There is simply no reason to even use such a macro for testing purposes.

It's trivial to assume what Blizzard deems to be appropriate useage of such macros. Claiming they don't care that you use it for "this" reason and not another reason is assinine. That's like saying it's okay to cheat on a practice test, just as long as you don't cheat on the real test.

It depends mostly on your intentions.

In an ideal world, blizzard doesn't want you to set up macros that make you run in circles and shoot things; effectively leveling for you with you doing no actions. As such, spamming a single spell is 'fair' in the sense that it doesn't goes over the lines. Problem is, if blizzard said that spamming a single key is fine, then people would use it to stay out of afk in AV or something, which gets us back to square one. As a result of this, even spamming one key repeatedly is not fine, mostly because of the afk scenario.

In any case, its mostly a matter of being in a grey area. You might get into troubles because it depends on the interpretation of the GM. You most likely won't. If you do use it to do unattended stuff (ie: it plays for you), then yes, thats clearly outsides of the lines.

My post in the thread was a bit misleading, I'll be rectifying that soon...

My basic overall issue with Rawr is not really that I don't trust it, it's that I won't trust it until I can verify it, and I don't like verifying through digging through code. Not that my TCoM does it any better at the moment, but having a transparent set of derivations of how the numbers were created is very important in my view.

That and no working Mac version (BOO for non-portable programs!).

The mana concern is one I have heard a lot as well, I'm torn between thinking "Valuing mana for a DPS calculation is silly, isolate the DPS value of a rotation/gearset/spec, then identify if its mana sustainable, don't trade off gear for mana right off" and "Well, that is one of the logical next steps of the TC process".

That mana concern, I think, is really a users decision, and agree a very obvious checkbox for it would be good.

I don't really see why anyone would model FoF as extra crits. It's much cleaner to model it as extra crit chance, where the actual extra crit chance is dependent on the exact rotation used (granted with the proc-based nature of Frost now, I won't say that's a trivial task, but it's doable in principle).

I don't quite understand what you mean when you say Brain Freeze scales with haste but not crit. Haste will speed up the GCD consumed by the Brain Freeze Fireball, and crit will increase the chance of that Fireball to crit, but Brain Freeze still procs on 10% of all casts and thus makes up a fixed percentage of all your casts (in a long-run scenario). Haste doesn't change that because you're casting proportionally more Frostbolts as well.