Moonkin TTT Article pre-publish draft
This is a compendium of information on raiding as a Balance Druid. It contains everything a beginner needs to achieve basic competence at raid DPS, as well more detailed discussion for people more interested in understanding the theory or more advanced play. The article is supplemented by the attached spreadsheet.
[top]Talents and Glyphs
The core of the spec will include the following talents:
This leaves 5 points. Fill them in the following order:
[top]Resulting Spell Coefficients and Damage Multipliers
Some useful numbers to have, when fully talented and raid buffed.
Overall damage multiplier: 1.13 (Earth and Moon) * 1.04 (Master Shapeshifter) * 1.03 (Earth and Moon) * 1.03 (Sanctified Retribution) = 1.247.
Haste multiplier, before haste rating: 1.05 (Wrath of Air) * 1.03 (Celestial Focus) * 1.03 (Improved Moonkin Form) = 1.114. Multiply this by (1 + HasteRating/3279) to obtain final haste ratio.
Crit Chance, before Intellect and crit rating: 5% (Moonkin Form) + 5% (Improved Scorch) 3% (Totem of Wrath) + 3% (Improved Faerie Fire) = 16%. Adding in 4% (Nature's Majesty), 3% (Improved Insect Swarm) and 10% (Improved Moonfire) gives 20% for Wrath, 23% for Starfire, and 26% for Moonfire, added to whatever you see reported in your character sheet when out of form.
Effective Starfire coefficient: (1 (base) + 0.2 (Wrath of Cenarius)) * 1.247 (damage multiplier) * 1.1 (Moonfury) = 1.646.
Effective Wrath coefficient: (2/3.5 (base) + 0.1 (Wrath of Cenarius)) * 1.247 (damage multiplier) * 1.1 (Moonfury) * 1.03 (Improved Insect Swarm) = 0.949.
Effective Insect Swarm coefficient, per tick: 0.2 (base) * 1.247 (damage multiplier) * 1.3 (Glyph of Insect Swarm) = 0.324.
Effective Moonfire coefficient, direct damage: 1.5/3.5 * (1.5/3.5 / (1.5/3.5 + 12/15)) (base) * 1.247 (damage multiplier) * (1 + 0.1 (Moonfury) + 0.1 (Improved Moonfire) - 0.9 (Glyph of Moonfire)) = 0.056.
Effective Moonfire coefficient, per tick: 12/15*(12/15 / (1.5/3.5 + 12/15) / 4) (base) * 1.247 (damage multiplier) * (1 + 0.1 (Moonfury) + 0.1 (Improved Moonfire) + 0.75 (Glyph of Moonfire)) = 0.317.
Major Glyphs: the best three for single-target DPS are [Glyph of Insect Swarm], [Glyph of Moonfire], and [Glyph of Starfire]. Even in AoE situations, [Glyph of Starfall] generally doesn't pull its weight.
Note that [Glyph of Insect Swarm] removes the 3% miss debuff, which may make it worthwhile for your raid to have you unglyph it at some fights. If you need to do so, replace it with Starfall.
Minor Glyphs: none of these add any DPS, so it's entirely up to personal taste. [Glyph of Typhoon] is the only one that affects combat mechanics, but it's still up to you whether you prefer it or not. [Glyph of the Wild] can save a tiny bit of mana if you ever have to rebuff people during a fight. It has one other obscure use--if you're desperate for mana to Hurricane during a 25-man fight, you can cast Gift of the Wild to force a Clearcast.
In roughly descending order of importance.
26.23 hit rating gives 1% to hit with spells. After Balance of Power and Improved Faerie Fire, we need 263 hit rating (10%) to cap at 100% (Alliance with a Draenei in their party need 237 hit rating (9%)). If you are below the hit cap, hit rating is the strongest DPS stat per point by a wide margin. As a result, you can expect that your optimal gear/gem setup will always place you at hit cap.
Improves all of our damaging spells. See the coefficients in the above section on spells.
32.79 haste rating gives 1% spell haste. There is a threshold at which haste loses some value; this occurs when the GCD hits 1.0s with Nature's Grace active. This requires 1.5/(1.03*1.03*1.05*1.2) - 1 = 12.21% haste, or 401 haste rating. If you are not under the effect of Wrath of Air Totem, the cap becomes 585 haste rating. Below this cap, all of our spells feel the full value of haste. Above it, Wrath and instants no longer benefit from haste when Nature's Grace is active, but all other spells still benefit fully (see "Casting Mechanics" below for more detail). Haste is a very strong stat under the cap, and even above it continues to be one of our primary DPS stats.
Note that for reasons related to latency, the change in value at the soft cap occurs over a small range rather than instantaneously. Don't obsess over hitting 401 exactly.
Critical strike rating:
45.9 crit rating gives 1% to crit. With Vengeance and a Chaotic Skyflare Diamond, crits with our main nukes do 2.09 times the damage of non crits. In addition to increasing the average damage of our spells due to more frequent crits, it provides two other DPS benefits: it increases the uptime of Nature's Grace, and shorten the average time required to proc an Eclipse. Both of these turn out to be minor terms in the marginal value of crit rating, however. With enough crit rating and the 2T8 set bonus, your crit chance during Lunar Eclipse will hit 100%, reducing the value of subsequent crit rating. Will all talents/buffs except Focus Magic, this occurs at 1346 - Int/3, where Int is your Armory value.
Crit is typically stronger than haste provided that you are over the soft haste cap, but still weaker than spellpower. Above the Lunar crit cap, crit becomes roughly equal in value to post-cap haste.
In addition, crit provides a mana benefit due to increased proc rate of the Moonkin Form regen.
Intellect provides 0.006% to crit per point, and 0.12 spellpower per point with Lunar Guidance. With Blessing of Kings, Improved Mark of the Wild, and Furor, it provides 0.0074% (135 points per 1% crit, roughly 1/3 of a crit rating) and 0.148 spellpower. This makes Int a weak, but not insignficant, DPS stat.
Similarly, with these talents, one point of Intellect increases your maximum mana by 18.5, your regen from Replenishment by 0.185 MP5, and your regen from 3/3 Dreamstate by 0.123 MP5
Spirit provides 0.3 spellpower per point with Improved Moonkin Form. With Blessing of Kings and Improved Mark of the Wild, it provides 0.337 spellpower. This makes is a weak DPS stat, very similar to Int.
Spirit and Int will both increase your natural regen, which scales like (0.016725 * sqrt(Int) * Spi) MP5. Int will increase the magnitude of your Moonkin Form proc by 0.37/point.
Stat priority for DPS will almost always run in the following order:
Hit rating (to cap) >> Haste rating (to soft cap) = Spellpower > Crit rating > Haste rating = Crit rating (above Lunar cap) >> Intellect > Spirit.
For exact numerical comparisons, plug your setup into Wrathcalcs.
In general, gem [Runed Cardinal Ruby] in all sockets (or [Veiled Ametrine]/[Rigid King's Amber] to make the hit cap), save for two [Purified Dreadstone] wherever you can get the best socket bonuses to meet your meta requirement. You can potentially use a [Potent Ametrine] or a [Purified Dreadstone] if a piece has a strong socket bonus.
Below the haste softcap, [Runed Cardinal Ruby] and [Quick King's Amber] are nearly identical in value. I'd recommend sticking with Runed to avoid constant regemming as your haste value changes. Definitely use [Reckless Ametrine] to pick up any socket bonus.
If you're looking to increase your HP for certain raid encounters, [Glowing Dreadstone] are a good way to do so without a large DPS loss (either by replacing Purified gems, or Runed gems in pieces with good socket bonuses).
In your meta socket (only use hats with meta sockets), use [Chaotic Skyflare Diamond].
If you have access to the T8 2-piece, use it. It's not worth removing for any currently available gear. The T9 2-piece is also quite good, making 2T8/2T9 the best possible option currently. 4T9 with ilvl 258 pieces is nearly as good.
The T8 4-piece isn't currently modeled in the spreadsheet, but it's worth using in lieu of any stat upgrades you might otherwise get at the Ulduar level. The benefit is most significant in multi-target and movement situations, rather than stationary burns. Once you have access to Coliseum-level gear, upgrade to 2T8/2T9 as soon as you can.
If you don't have access to Ulduar gear, 4T7 is a very good bonus.
Buy [Idol of Lunar Fury] for 25 Emblems of Triumph. It has no ICD and a 70% proc rate, giving it essentially 100% uptime.
If that's not available to you, buy [Idol of the Shooting Star] for 25 Emblems of Valor.
Whenever possible, evaluate trinkets using the same metric you would any other item. Compute the average uptime of the proc or use effect to obtain an average stat value.
The main exception is haste. Remember that even if you're under the soft cap, haste rating that's applied in a short burst (such as the [Embrace of the Spider] effect) will be mostly over the softcap. Don't make the mistake of averaging out the haste value from a trinket and then applying it all at a sub-softcap value. If you have control over the haste effect ([Scale of Fates] or [Talisman of Volatile Power]), you can mitigate this somewhat by activating the trinket only at the beginning of Lunar Eclipse. Even this is of only marginal value however--you sacrifice uptime by waiting to use it, and the Eclipse is only ~13 seconds of casting out of a 20 second buff. The result is that these trinkets are still disproportionately weak for us. [Talisman of Volatile Power] is especially bad in this way, as you lose the highest-stacked part of the buff when Eclipse ends.
Finally, the [Reign of the Unliving] proc is extremely strong. The two versions of it are the two best available trinkets in the game, by a wide margin.
Use [Flask of the Frost Wyrm] and a 46 spellpower food (such as a Feast).
The best DPS potion is [Potion of Speed] (during a Lunar Eclipse), if you don't need to use a [Runic Mana Potion] during the fight. [Potion of Wild Magic] is nearly as good, and is useful to have around if for some reason you need the DPS burst during your Solar phase.
Excluding profession bonuses.
Tailoring gives the strongest DPS benefit. Lightweave Embroidery gives 295 spellpower for 15 seconds, 35% proc on spellcast, 60 second cooldown. It can proc off DoT ticks, making the average proc time around 2 seconds. 70 spellpower is a decent estimate of the average benefit, minus the 23 haste you'd normally have on your cloak, so the total benefit here is worth roughly 55 spellpower (somewhat less if you're below the haste softcap, potentially slightly weaker than the other professions). Amusingly, the enchant also gives 1 Spirit for some reason.
A minor convenience of Tailoring is that for the cost of one [Eternium Thread], you can swap your leg enchant between 20 Spirit and 30 Stamina for fights where a little extra HP is helpful.
Beyond that, Alchemy, Blacksmithing, Enchanting, Inscription, Jewelcrafting, and Leatherworking all provide a roughly equal gain of 46-48 spellpower.
Alchemy: At any time, either [Flask of the North] or Mixology (with your [Flask of the Frost Wyrm]) will give you 47 spellpower.
A number of small conveniences: [Endless Mana Potion], [Crazy Alchemist's Potion], double-duration Flasks, and the ability to make your own [Cardinal Ruby].
Blacksmithing: An extra socket each in your wrists and gloves, each with a [Runed Cardinal Ruby], gives 46 spellpower.
Enchanting: 23 spellpower to each ring gives 46 spellpower.
Inscription: 70 spellpower/15 crit to shoulders in place of the Sons of Hodir enchant gives 46 spellpower.
Jewelcrafting: 3 [Runed Dragon's Eye] in place of 3 [Runed Cardinal Ruby] gives 48 spellpower.
Leatherworking: 76 spellpower to bracers in place of the usual 30 spellpower gives 46 spellpower.
Engineering and Skinning are weaker, both giving less benefit than the 46 spellpower.
Engineering gives Hyperspeed Accelerators (340 haste for 12s every minute = 68 haste on average) and 27 spellpower to cloak. The net gain is 45 haste - 1 spellpower. This can be improved slightly by syncing the Accelerators with Lunar Eclipses.
Skinning gives 40 crit rating.
Mining and Herbalism provide minor survivability benefits, but no DPS gain.
[top]General Gear Selection Principles
Caster epics have Intellect, Stamina, spellpower, and 2 out of the following: crit rating, hit rating, haste rating, or Spirit. Keep in mind three rules of thumb, which are enough to get a quick estimate of the value of any piece:
There are two main ways to judge gear in more detail and plan your upgrades:
1) Use Rawr. This has the advantage of ease of use, and automating the tedious arithmetic. The disadvantage is that it's a bit of black box, so a) you're simply trusting the model it uses, and b) you don't learn as much in the process. (a) is generally not too big a deal, since Rawr has been based on the WrathCalcs model recently and so should give very similar results.
2) Use WrathCalcs to determine marginal values of your stats, and then plug those into a service like Lootrank or the Wowhead item comparator which will use the stat weights to sort items. This would be a very quick and easy way to get quite accurate results except for one complication: the hit cap. The best way to handle the hit cap is to:
--Set the stat weight of hit rating in your item comparison tool to 0.
--Whenever you have multiple options for reaching max hit, compare the computed value of the hit item and the non-hit item in the slots you're considering. This tells you how much potential DPS you're giving up in that slot to acquire your hit.
--Equip the items with the best ratio of hit rating gained to DPS value lost.
--The same process can be used for set bonuses.
The first rule is to always be casting (or waiting out a GCD after an instant). Anytime a spell ends, you should already be pressing the key trying to start your next one (see below). Don't delay a cast to make a decision or react to proc--train yourself to start another spell regardless of what's going on, and change the subsequent spellcast if necessary after you've had another second to think.
As a preliminary note, this macro might be useful for automatically assisting with your nukes when you have a raid member targeted:
The theorycrafting reflected in this article and in the associated spreadsheet uses the following understanding of spellcasting rules:
GCD: whenever you attempt any GCD-incurring action (including any spellcast), a client-side GCD begins to run immediately. If the action fails, the client cancels the GCD when it finds out about the failure from the server. If the action succeeds, the GCD runs to completion. The client will not transmit any action to the server while a GCD is running.
Queueing: when a spellcast ends on the server, if the server has recently (within 200-300 milliseconds) received a cast command from the client, it will begin that cast immediately.
[top]How to Time Your Next Spell
First, recall that the client processes events when you release the key. Keep that in mind for learning your timing.
After a Starfire: You want to press the next spell such that it arrives at the server in the "queueing window" as your Starfire is ending. In constant latency, you'd press it just under one cast time from when you pressed the previous one (i.e. when Quartz hits the red part of the cast bar). But since you can't count on latency to be constant, you should start spamming your key a few tenths of a second before that point, and continue until you see the next cast begin.
After an instant: No queueing here, and the ability to cast the next spell is determined by the client-side GCD. Watch your Quartz GCD spark and press your key right as it ends. Pressing it early is a waste and will return a "spell not ready" error in your client.
After a Wrath, <1s cast time: Identical to an instant cast. In highly variable latency, you may want to spam your spell a few times to make sure it fires.
After a Wrath, >1s cast time: Since the Wrath cast time is equal to the GCD, you can't start spamming your next spell early like you do after a Starfire. Start pressing your key as the GCD ends (as with an instant), and continue pressing it until you see that the server has begun your next cast (as with a Starfire).
The above does mean that Wrath is more "lossy" than Starfire in actual use, since the client-side GCD prevents you from using the queueing window optimally.
[top]Value of Haste
As a result of the above, Starfire always feels the full effect of haste. No plausible amount of haste brings it near a 1s cast time.
Instants feel the full effect of haste until the softcap, where they become fixed at 1s and feel zero effect from haste.
Wrath behaves mostly like an instant (since it's typically GCD-limited). But due to variable latency and Wrath's transition from cast-like to instant-like behavior at the softcap, there's a bit of a fuzzy zone where haste decays to 0 in value. The theorycraft here treats it identically to an instant.
[top]The Eclipse Rotation
You will generally operate in a four-step cycle.
Basically, you cast Wrath until Lunar procs and cast Starfire until Solar procs. But once DoTs and other instants are involved, thinking in terms of the four-step cycle helps you plan your casts well.
Lunar is slightly stronger than Solar (slightly faster to proc and slightly more damage per proc, in addition to extending Moonfire and benefiting from 4T8 if you have it). So in a situation where you can proc either (the beginning of the fight, or if you've failed to proc Solar for a few seconds and Lunar is coming off of ICD), cast Wrath for Lunar.
[top]Subtleties of Transitions
Solar->Lunar: Wrath procs Nature's Grace when the cast completes, but procs Eclipse when it actually hits the target. This allows a sort of clairvoyance regarding Lunar Eclipses: if you see Nature's Grace refresh when you cast a Wrath, you know that that cast is a crit, and therefore that it has a 60% chance of proccing an Eclipse when it reaches the target (but be careful to remember that this doesn't work if Starfall is up, or if you've just refreshed Moonfire in Pre-Lunar phase).
If you switch to Starfire as soon as you see this happen (not interrupting a cast, merely swapping after the next Wrath), you have a 60% chance of effectively lengthening a Lunar Eclipse by starting to cast Starfire before it even procs. This is not in the spreadsheet yet, but some brief math indicates that it's worth trying to do, even though you waste a Starfire cast the other 40% of the time.
Lunar->Solar: Due to lag, it's possible for a Starfire which lands during the final split-second of Lunar Eclipse to proc a Solar (thus skipping pre-Solar phase entirely). This occurs if the Lunar buff fades late enough that it's still active on the server when the cast completes, but early enough that it's gone by the time the crit is returned to the client. You can force this to happen by canceling Eclipse when the final Lunar Starfire is in the "red zone" on Quartz.
To aid in this process, here's a macro line which will cancel Eclipse (Lunar only) if the remaining duration is less than the cast time of Starfire:
Alternatively, you can just use WiseEclipse (see below), which performs the same check whenever the even UNIT_SPELLCAST_SUCCEEDED fires, which is slightly more convenient and reliable.
Against a single target, reapply MF and IS whenever they drop. MF should have higher priority for reapplication, but the difference is tiny. During a Wrath-heavy part of the cycle, it may be best to reapply MF last (after IS or other instants) to minimize the chance that it will drop before you cast SF again.
There's currently a slight open question about whether it's worth waiting until Eclipse ends if a DoT expires in the final few seconds Eclipse. This is hard to model and the DPS difference is very small, so consider it a matter of personal taste for the moment.
Under the effect of Bloodlust, Wrath will gain very little DPS since it's usually so close to the 1s GCD minimum in the first place. The goal during Bloodlust is to use Lunar Eclipse as much as possible.
In multi-target situations, you'll mostly be selecting spells by your own judgment rather than any kind of formula. It's important to be familiar with the damage done by each of your spells. Overall, you'll be trying to maintain your usual Eclipse behavior as much as possible. Some things to keep in mind:
Aside from aforementioned Starfall usage, you'll basically AoE by chaining Hurricane. One little nuance is that you can start Hurricanes under the effect of Nature's Grace (procced by Starfall, for example), for significantly higher AoE DPS. Typhoon adds little damage but is another way of "forcing" Nature's Grace (but if you don't have it Glyphed, typically inconveniences everyone else in your group). With Idol of Lunar Fury, keeping one Moonfire ticking will slightly improve the DPS of Hurricane.
Things I'm still pondering or simply haven't gotten around to mathing out in detail--if you're looking to contribute, these are good topics to think about. Added bonus if you also give a suggestion for refining the spreadsheet model.
[top]Known Spreadsheet Issues
[top]Mods & Tools
Squawk and Awe: A convienent way of tracking your DoT's and Eclipse timers, and assorted other class-specific procs.
WiseEclipse: Automatically cancels Lunar eclipse when a Starfire completes, if there's not enough time left for another. Helps chain Lunar into Solar Eclipse. You should use either this or a macro, as mentioned above.
Quartz: as a primary caster, you should have a proper cast bar. This is an excellent one.
Rawr: A good way to do large gear comparisons that would otherwise be tedious. It uses roughly the same calculations as the spreadsheet.
Simulationcraft: A simulation-based approach to theorycrafting.
WrathCalcs (attached to this post). Moonkin theorycraft spreadsheet. Will give you the exact value of stats/talents/glyphs in your current setup, and lets you experiment with different rotations. Originally made by Adoriele here. I've since taken it over and used it as the platform for my own theorycrafting, reflected in the version posted here.
Add a link to Wise Eclipse (much better than macros!).
Very nice. Some things to consider:
1) Introduction, saying that this is targeted at the raiding Moonkin (or perhaps a title of "Moonkin Raiding" would be enough).
2) I believe it is possible to queue during the GCD, just not during the first second of the GCD (of course they can become the same thing).
3) A note that Haste>SP (below soft cap) doesn't occur until Ulduar gear levels, and even then haste usually doesn't pass SP on an itemization-cost basis (or not by much).
4) On multiple targets, a note that Hurricane is most likely your highest personal DPS (and DPM) choice given 4 or more targets, and is most likely not your highest DPS choice for 1 or 2 targets. For three targets, it is a close call, influenced by crit, haste, Gale Winds and Ebon Plague.
5) Mods and Tools: Mention Simulationcraft. I think the most reliable answers to your "open questions" will come from a simulator, and SC is a pretty good simulator.
Essentially: What happens when you weave instant casts into non-instant spam?
He showed that when queuing an instant cast Fireball at the end of a Frostbolt, the gcd actually began at queue-time and not cast-time. This was evidenced by the fact that the time between the instant cast Fireball and the start of the next Frostbolt was consistently less than a full gcd.
While one possible interpretation is that a queuing is perhaps available on the gcd, testing of pure instant cast spam (with an unnamed illegal device) shows that after many casts the average time between casts is slightly over the gcd.
So SimulationCraft will:
(1) Only allow queuing after actions with non-zero cast times >= 1.0sec (i think i could possibly reduce this to 0.8sec, but not enough testing yet)
(2) Start the GCD timer at queue time
I think Tigafin's report implies no particular penalty for an instant cast with a long GCD (mage on target dummy, I assume his GCD>1.0s).
Manually chain casting IS, with a 1.37s GCD. Combat log showed
49 Not Yet Ready messages, running from 0.1s to 1.0s after the previous cast.
7 Early successful casts (reported at 1.2 to 1.3s after previous success)
20 On-time casts (reported at 1.4s after previous success)
22 Late casts (reported at 1.5 to 1.8s after previous success)
Note that the combat log didn't have any "Not Yet Ready" or "Another Action is in Progress" message between 1.0 and 1.37 seconds. Considering the distribution of the other timings, it seems likely that keys were pressed in that time zone. So, they were either successfully transmitted (effectively queued via some combination of client, network, and server), or failed silently.
Do you have a link for "testing of pure instant cast spam" tests? How big is "slightly over the gcd" compared to reported numbers for SF spam?
2, and ensuing posts) I'll have to look into this more. It would change some things (mainly, it would make Wrath stronger generally).
3) I'll just change the ">" in my brief stat list to an "=", I already say something similar in the gems section.
4) Yeah, will add.
5) I will, I just don't know much about it.
EDIT: I just realized I did not answer your "how big" question: The average "gap" was 150ms.
Come to think of it, I believe Althor did a similar test with his Priest: Cast 100 times and compare actual number of casts vs the maximum number possible in that time period, dividing for the average penalty.
Unlike Zak/Althor, Tiga was actually able to get MORE casts than expected, more than would be possible even with zero gap between all casts. He posted additional parses (and his analysis) to our dev forum recently enough I should be able to find that for you.
I have come to the point where I am likely to get the tier 9 setbonus. I have got enough badges to get the I lvl 232 t9 Balance gloves. I would be replacing Gloves of the Frozen Glade (Thorim Hardmode). This would mean a small decrease in dps looking at my current stats. However, my question is: Does the 2 piece balance tier 9 bonus trigger nature's grace?
A note to add for Alchemy is that you also get double duration on elixirs and flasks.
[EDIT: I now see that you have this included.]
I really like the article, but the haste cap theorycraft doesn't seem to make much sense to me. According to your spell mechanics section, you can't spell queue below 1s, therefore under optimal conditions the best you can chain cast wraths is 1s + latency time/reaction time. Therefore, wouldn't that imply that the haste soft cap is significantly lower than 401 quoted above?
If your latency + reaction time reaches 0.13s, which is quite plausible, wouldn't that essentially remove the whole concept of the haste soft cap, since it would be achieved at 0 haste.
Is this the correct link for WiseEclipse?
4shared.com - online file sharing and storage - download WiseEclipsev1.0.1.rar
Also, I gave the code a quick scan--looks like it's just my macro, but triggering on UNIT_SPELLCAST_SUCCEEDED. Which is good--I think that's exactly what we want. Will add it in.
Some updates in.
|All times are GMT -4. The time now is 11:27 AM.|
Forum Infrastructure by vBulletin 3.6.12 ©2000-2007, Jelsoft Enterprises Ltd.