Healing Theory, Part 10: Single-Target Rotations

All posts in this series can be found here.

As promised at the end of my last post on Warlords heals, the next step would be to analyze more complex rotations involving multiple spells. This post will go into single-target healing rotations for each class, building on the overview of individual spells I did previously. Single-target rotations are only one slice of the healing picture, but one of the more readily quantifiable ones, making them a good place to start. Also, the need to directly heal tanks is expected to be a much bigger part of Warlords than it was in Mists.

On Modeling Healing Rotations

One difficulty in expressing numerical results of combined healing spells, especially cross-class, is that healers don’t use “rotations” in the same way DPS do. They’re constantly reacting to the demands of the encounter and modulating output based on that. A model of a pure max-output rotation akin to DPS is slightly informative, but rather limited because that’s an unusual mode of casting. Since you spend nearly all your time somewhere in the middle of the sliding scale of mana usage vs. healing output, it’s hard to nail down exactly what numbers to measure or model. We can choose cases to model that are informative, but it takes more thought to motivate the decision of what those are. The assumptions also have to be made clear so people understand what the numbers represent.

The usual way I approach this is anticipated by earlier posts in this series, and should also be familiar to anyone who used my TreeCalcs sheet in Wrath/Cata/Mists (for this post, I actually put the heal chart from the prior post into the WrathCalcs/TreeCalcs shell). I assume a certain subset of spells are used whenever possible: on cooldown, whenever a HoT expires, or whenever procs/resources allow. All available cast time that’s left over is for “filler” or no-cooldown/spammable spells. In the case of the single-target model, the filler time is divided between a) the cheap direct heal, b) the expensive direct heal, and c) casting nothing. This flexibility in how to allocate the filler time is needed for two important purposes:

  • One class filling all time with its cheap heal (for example) may not be parallel to another class filling with its cheap heal. The first class might do less healing but also use less mana in that comparison, but then equal the other class in both healing and mana consumption if it mixed in the expensive heal some. Every class has a different mana vs. healing profile, and being able to adjust spell usage allows more sensible comparisons.
  • It provides a basis for evaluating mana. As discussed extensively in the post on mana, mana usage affects your use of non-cooldown heals much more than it affects your use of cooldown-bearing heals. In a more concrete framework like a rotation model, you can evaluate exactly what mana lets you do by varying the spell usage correspondingly. The conceptual chart in the final section of that post, describing healing done as a function of mana, will be revisited in this post.

This is of course idealized, as is unavoidable when theorycrafting healing, but it’s useful in a lot of ways. In particular, while in various low-healing situations you might use spells more in isolation and not nonstop like in a rotation model, we already have HPCT and HPM info on individual spells. Those can be looked up anytime on the spell chart form the last post. Models of rotations let us find out more about what happens when cooldowns and cast time become a limiting factor in how much healing you can do.

I’ve not yet modeled every talent and glyph combination for every class, but I’ve tried to include the ones that were especially relevant to the current analysis.

Spreadsheet Updates

Download the current version I used for this post here.

As before, this post can be read on its own, or with the sheet open to look at things in more detail. The spreadsheet has had a lot of additions since last time, to both add a rotation model and to provide a lot of analysis of the effects of spells and talents. It’s had some user-friendliness updates but also is a lot more complicated. I’m going to refrain from giving a full tutorial now—I’m going to be continually building more into it and if I ever get it ready for totally public use I’ll do so then. For now, most of what’s going on is evident from playing with the “Main” tab, where you can set up a character and see the output of the current rotation. The actual workings of the rotations are a bit opaque for now, but they generally work as discussed for each class below.

One of the handy parts of the sheet is that, next to every talent and stat, it shows how much HPS and MP5 they’re adding to the current rotation.

I’ve also replaced the abstract framework using raw spellpower coefficients with one using more traditional stats. HPS and HPM work as you’d expect. Throughout this post, I assume 4000 spellpower from gear, 1000 Spirit, and 600 of each other secondary stat. This is a generous amount of Spirit (more than you’ll have in low gear), but I expect gearing in Warlords to heavily favor it where possible, and it may well be a popular food/flask as well. The only impact of Spirit in the discussion is factoring into the “net MP5 usage” described for each rotation—essentially setting what the zero point is. For reference, spending 5000 net MP5 will drain a full mana bar in 160 seconds; spending 2000 MP5 will do so in 400 seconds.

Druid (and some further explanation of the methods)

With all filler time on Healing Touch, this rotation clocks in at 44.7k HPS while burning 4353 MP5 (net mana drain including regen). This is actually a slightly more aggressive approach than it looks. Druid burns a lot of mana on its rotational spells due to Rejuv and Swiftmend, and does a significant amount of baseline healing (Swiftmend is actually a poor addition to the rotation on its own, but it enables SotF). So we’re going to use less filler heals to compensate. If we cut to 50% filler time on HT (and the rest unused), we’re at 34.9k HPS for only 1610 MP5.

Rather than typing out more individual cases, we can plot the whole variation on a graph:

Screen Shot 2014-07-28 at 12.51.41 AM

The leftmost part is where we slide from spending 0% of our open time on HT up to 100%. Then after the kink, we start replacing HT with RG. Because HT is a very good spell, the efficiency of added mana is strong earlier on while we have time to add in more HT, but then drops quite a bit when we need even more healing and the only way to get it is to replace them with RG.

(I will not show this individual graph for each class; I used this one to illustrate the concept. However, at the end of the post is a combined one showing all healers, so if you want to refer to that as you look through the classes, it might help put the numbers you read in perspective.)

Note that we have built the graph from the end of part 6 in a concrete case with actual numbers. Neat! The only important difference is that the current graph is measuring net mana consumption including Spirit regen, so “0 mana usage” here refers to what you can do while holding steady on mana, not to what you can do without spending mana at all. It’s only a visual difference, with the y-axis being further to the right in this case. I took Spirit regen into account here to help relate these numbers to your mana usage in practice, and to better account for talents that affect regen.

Back to Druids. You’ll see when we get to other classes that these numbers are good. Most classes can build a reasonable rotation of around 30k HPS with little net mana use, but getting up to 45k HPS without resorting to excessive mana use is beyond every other healer besides Monk. While it’s true that Druid is strong, there are a few important components to note here:

  • Living Seed. Around 12% of the healing in the all-HT rotation is from Living Seed. Whether you get full value from this depends on how well the Seed is utilized when spamming HT nonstop on a tank.
  • 2 talents, Germination and especially SotF, are very strong in this context. In a very efficient rotation, they can add over 20% output by increasing Rejuv healing so much.

Any way to look at it, Druid is very strong on a single target. Being able to reach 25k HPS without spending any mana on direct heals, due to the strength of Lifebloom and Rejuv and the talents that boost Rejuv, gives it a very high baseline.

Disc

The first issue with Disc single-target is, as hinted in the previous post, its relative helplessness without Clarity of Will. Filling with Heal can get to a decent mana-neutral rotation for 22.5k HPS, but mixing in Flash Heal increases the mana burn rate rapidly for a modest gain, topping out at a maximum of 30.7k HPS for a full burn of about 8900 MP5 (switching to Surge of Light for this purpose).

There’s actually very little Disc can do to improve on pure CoW spam (35k HPS), and those improvements come in terms of efficiency. Technically adding PWS is a small throughput increase if you account for Borrowed Time into a 2.5s cast, but at half the efficiency of CoW it’s not attractive to add in (the prior post discussed some problems with the costing of instants). Penance, inversely, is a small throughput loss compared to CoW, but very cheap, so it’s some free mana savings. Mixing in Holy Fire and Smite (some weak casts) to allow Archangel is also an HPS wash that saves significant mana. It saves even more mana by allowing full value from the PW:Solace talent.

So we can actually build a very efficient rotation, but it tops out at 32.8k HPS (spending only 1264 MP5). This is actually similar to what Druid could do for a similar low mana cost. This is fine so far, but the Druid can extend much higher, while Disc spells just don’t have the single-target throughput needed to provide that option. Due to the mana return from Solace and the efficiency of Penance, Disc can even do reasonable healing at zero net mana use (at the 1000 Spirit used for this post), reaching about 26.5k HPS.

Holy

At an efficient mix between the two filler options, this rotation produces 35k HPS for 3196 MP5. That’s a pretty nice place to be, though not as efficient as the Druid for similar healing. Surge of Light is very strong in a context with heavy Heal and Flash Heal use.

There’s not too much to say about this one. There’s no one overly strong spell, but Serendipity and Surge of Light together allow for good mana savings while mixing in nontrivial amounts of Flash Heal. This lets throughput reach pretty standard levels without high mana use.

I’ve left out Saving Grace, even though it can improve the average rotation, since I’m not sure that’s a realistic usage model. The talent’s a little odd right now because the stacking debuff encourages you to use it very rigidly, casting 3-4 times and then letting the debuff fall off (you never want to restack the debuff anytime other than immediately after the previous SG). I’m not sure this is the desired outcome for the talent. I’ve similarly left off the Level 90 talents, which can technically be slight throughput increases on a single-target, since they’re inefficient for this purpose and are usually used for other things.

Paladin

Because the EF direct heal is comparatively stronger than before, and there’s Empowered Beacon of Light, it’s better to recast on the tank every time you have 3HP than to blanket around the raid (for this purpose). One EF rolling on yourself adds a very small amount overall. Sacred Shield is almost as good on a single target, and both talents have a lot more value when you consider multiple targets, so this will be an interesting comparison in group healing models. Selfless Healer is generally useless to Holy in its current form.

I added Stay of Execution even though I’m not usually including things with medium or long cooldowns, since it’s a talent whose explicit purpose is single-target.

Another class with some difficulty keeping up. Like Disc, Paladin can do a solid amount (25k HPS) for free, in this case because it has some very cheap spells (and Glyph of Illumination, which now is almost always a net benefit). But similarly, Flash of Light can’t push you higher with any decent efficiency. Much of the strength here is the Paladin’s baseline cheap Holy Power generation from Holy Shock and Holy Light, and cashing that out into Word of Glory on the tank. Upgrading the Holy Lights to Flash of Lights at significant cost is not a huge gain, since you spend a lot of extra mana but aren’t getting much more Holy Power. Even going up to 6000 MP5 spent or higher will only get to around 30k HPS.

Unlike Disc, Paladin is not quite as bad as it looks here, because by focusing on one target we ignore the benefit of Beacon (other than through the perk). In real situations, nonstop Beacon flow from all your group healing will cut into the needed healing on the tank, making the lower throughput more acceptable when you do heal them directly. But if you’re behind on one target and need to pour heals into them aggressively, it will be harder.

One caveat is that this does not account for cooldowns, and Avenging Wrath is a far stronger single-target throughput cooldown than anyone else has. If situations where heavy burst throughput is required only occasionally, a Paladin can actually cover them quite easily. This might ameliorate the disadvantage of having trouble maintaining heavy throughput on a single target. Not to mention there is also Lay on Hands for serious problems, which is a much bigger deal than it used to be, now that HP bars are so large relative to heal size.

Shaman

Shaman is a little tricker than most to set up in a “pure single-target” model, because on only one target, you have no reason to Riptide more than once every 18 seconds, or to Chain Heal ever. This means that your filler heals don’t all have Tidal Waves, whereas in most realistic situations you’ll at least be using enough Riptide to have Tidal Waves whenever you cast on the tank. I experimented a bit with including Riptides on off-targets in the model (most reasonable way to do this seemed to be counting their healing done and mana cost to both be 0), just to produce Tidal Waves. This is mostly a wash in the end so I didn’t include it.

Earth Shield is also tricky to model. I chose a somewhat arbitrary refresh time (18 seconds) and assumed 100% uptime. It turns out that the per-cast HPCT and HPM for the Earth Shield cast itself are reasonably close to neutral for the whole rotation (specifically, pretty similar to Healing Wave), so things don’t change much if you have to refresh more or less often. And it’s a big healing increase simply to have it on there. If it gets eaten rapidly, the constant GCDs to refresh could be awkward, but the numbers come out pretty even in any case. If gets eaten so fast that you can’t reasonably keep it up, you’ll lose some output.

Finally, I didn’t include Healing Rain. It can add a a little throughput on a single target due to the Perk, but is quite expensive for this. I also don’t want to dash Shamans’ hopes for Warlords quite yet, so I’m maintaining the possibility that there can be situations where you don’t cast Healing Rain.

With all that explanation, Shaman come out average or slightly below. A Healing Wave-only filler rotation (except for Unleashed Fury, which is always used on HS) nets 29.2k HPS for 2847 MP5. It is worth pointing out that, similar to the Paladin case, you’ll mix in some non-tank healing in the real world, and some of that will come essentially free (because, as mentioned above, Riptides on other targets cancel out their own time and mana cost through Tidal Waves, leaving the tank healing unchanged). So only counting healing on one target ignores some amount of potential output. Moreover, the Healing Rains and Healing Stream Totems that the Shaman is using in any group situation (not included here), will contribute to tank throughput a little as well.

Like Druid and Paladin, mixing in the expensive heal comes at a steep cost, because the basic heal is reasonably good and the expensive heal is only a marginal improvement. Shaman can output near 40,000 HPS with a heavy mana burn but still won’t reach Druid and Monk levels.

Monk

The most unusual single-target rotation. The backbone of the huge efficiency of the Monk setup is nonstop Soothing Mist, which continues even while using more expensive filler heals. Note that this model benefits from one very favorable assumption, which is that your Statue is also Soothing the tank. It quite reasonable that this will often be the case when healing on a single target matters, but it is not controllable. Because so much of the strength of this rotation comes from using Soothing as often as possible, removing the Statue bonus would lose on the order of 20% of the healing. One other generous assumption here is that mastery spheres from the single-target heals are fully counted. That wasn’t worth trying to dig into further since it only accounts for 3-4% of the healing (single-target heals don’t produce as many spheres as AoE heals).

Finally, it’s rather hard to account for the GCDs used restarting Soothing Mist, since you have to do so every time you interrupt to cast certain spells, and you get an inconsistent amount of Soothing time per GCD used. I actually ignored this problem for now (as well as the GCDs used on Mana Tea), meaning that if you crank filler usage up to 100%, you get a nonphysical rotation that uses too many GCDs. This is fine, since Surging is such an expensive filler that 100% use isn’t a very important case anyway. Monk (with the assumptions mentioned above) surpasses the other healers in throughput even at much lower filler usage.

Monk starts out at a strong baseline. Even with no Surging Mist use, Monk can manage 29.7k HPS and remain mana-neutral or slightly positive. This is with nonstop Soothing Mist (stopping only to cast Expel Harm and Chi Wave), and Enveloping Mist only using the Chi from Expel Harm and Chi Brew.

Adding in Surging Mist adds healing quickly and not too inefficiently, due to the powerful Enveloping Mists they enable you to cast more of. Enveloping Mist has great value when you’re already Soothing most of the time, which in this model there’s no reason not to be. It will become a little less effective in real-world situations where you have to stop Soothing to cast other things more often, but only marginally. By adding Surgings, Monk can pass 50k HPS, higher than anyone besides a Regrowth-using Druid, at a mana cost that that’s not at all outlandish, 6882 MP5.

So Monk is clearly very strong. It’s hard to say exactly how strong, because they are benefiting from some helpful assumptions (most importantly that the Statue is always assisting you). The message is still clear that Soothing and Enveloping are two of the best tank heals, a fact which will remain even when you have to spend some of your time healing other targets. And even when the Statue is healing someone else, or the tank’s not getting mastery spheres, they are still providing value by doing effective healing on other targets. Monks exemplify a theme in these single-target rotations that applies to almost all classes: leaning as much you can on your efficient baseline heals (e.g. using them often even when the tank’s almost full) will go a long way towards overall efficiency in the long run.

Synthesis

To start with, here is a graph summarizing the HPS/MP5 numbers for all the rotations discussed above.

HT10 Graph

I put this at the end because it’s best looked at with a full understanding of what it means for each class: what spells/talents are included, and what assumptions are needed for a single-target model that work for or against the class. The relative position of the classes is generally anticipated by the comments above, but this helps to see it all at once. To restate the most notable patterns:

  • Druid and Monk can push higher HPS than anyone else, in both cases at reasonable cost. This has to do with HoTs primarily.
  • Paladin struggles to do much on a single target; they really need to be using Beacon to get value out of their heals. Even that won’t be a massive increase (especially since Holy Light and Flash of Light are cast on the Beacon anyway).
  • Disc can get to reasonable point somewhat efficiently through CoW spam, but has almost no other relevant options.
  • Holy seems like a very solid example of what a class should look like. There’s an efficient and mana-neutral option that gets you pretty far using Heal alone, and mixing in Flash Heal provides meaningful added returns for the extra mana. Note how the slope of the right half of the Holy line is steeper than Druid, Paladin, and Shaman, showing that Holy gets better value for added mana spent.
  • Shaman is weaker than Holy but stronger than Paladin (although not quite as efficient), subject to the difficulty of modeling Tidal Waves. Of the two healers in the middle, I like Holy’s mana vs. HPS tradeoff slightly better (in terms of good choices for gameplay).

Finally, note that the classes performing best (Druid and Monk) are there largely because of a more Mists-like rotation, with high baseline HPS and then more optional filler heals. The classes with no or almost no strong rotational heals to build on (Disc) are the ones languishing.

All of the above should suggest some spell tweaks to help with class balance. For this post I’ve spent a lot of time (and words) simply doing the analysis of 6 classes, so I don’t have detailed tuning proposals, but I think the things you’d focus on most are:

  • Possibly too strong: Soothing Mist, Enveloping Mist, Healing Touch, Soul of the Forest, Surge of Light
  • Possibly too weak: Penance, Power Word: Shield, Holy Fire, Flash Heal, Flash of Light

Conclusion

As you can see, there is a lot to discuss even on a seemingly narrow part of the healing world—only healing one target. Starting here was good, because the mana dynamics are complex, and we’ll need the robust framework to go into the even more openended world of analyzing multi-target healing. I’m looking forward to it, and hope this has been an informative look at the healing classes of Warlords so far.

Theorycraft 101: The Statistics of WoW Spells

I’ve been doing a lot of in-game testing of spells lately, as a part of making spreadsheets and other projects. In particular, with the new beta, I’m more inclined to vet the info for any spell I look at by measuring in-game, rather than simply putting the coefficient from wod.wowhead into a spreadsheet, because:

  • The designers are changing spells a lot, and tooltips are out of date much more often than on live.
  • The passives, talents, and Draenor Perks aren’t all familiar, and you have make sure you know what all needs to be multiplied in between the coefficient in the data and the final damage amount.
  • There are frequent bugs on beta, and actually testing means you can help catch/report them.

There are a lot of various techniques and tricks you get used to for doing this stuff quickly, but I wanted to dash off a quick post on one that both saves work and is mathematically interesting.  It looks like Theck is starting a series on general concepts of theorycrafting, and while I don’t expect to do anything that elaborate, I do want to write down ideas that are familiar to me but might be helpful to people who are just getting into it.

Background

The focus of this post is how to measure a spell damage value in-game, but I should give at least an outline of what to do with that information once you get it.

Take your measured value and divide by your spellpower or attackpower, and you have your coefficient (including any modifiers). If it matches the Wowhead data, or the Wowhead data plus the modifers you know about (20% from a Draenor Perk is common), then everything checks out and you’re done. Otherwise, you’re looking at a modifier you don’t know about, a bug, an incorrect tooltip, or something else.

2 brief points just in case they help people:

  • Some passives don’t show up in the spellbook any longer (typically, ones that do nothing but give passive bonuses), but they usually show up in the “Specialization” tab on Wowhead (a submenu under “Spells”).
  • Checking the tooltip in-game can help track down discrepancies (unless it turns out to be a case where it’s entirely wrong). Since the Wowhead data is itself from tooltips, if you know all the modifiers that should be included, they should always match. If the game tooltip is what you expect after taking bonuses into account, but your observed damage/heals are different, then the client tooltip data is wrong and you can’t rely on it. If the game tooltip differs what you’d expect based on the Wowhead tooltip, the game one factoring in a bonus that you don’t know about.

Weapon-based attacks are a little more complicated, but I won’t run through that all here.

Spell Ranges

For spells with constant damage/healing (such as HoTs and DoTs), taking the in-game measurement is easy; you only have to look at one tick. Also, many non-DoT spells have constant damage/healing right now, since they no longer have base values like they used to (it’s just coefficient*spellpower, both constants). It looks like there’s a system for artificially re-inserting variance like there used to be, but it’s not done or not used everywhere yet.

But how do you measure the average damage/healing when it’s not constant? The average is what you want, because 1) when you need to compute a spellpower coefficient, that’s what you want to start from, and 2) when you’re actually making a model, you typically only care about average damage values.  The instinct is to take a large data set with a lot of casts, and take the mean.

An Abnormal Distribution

Why is that the instinct? Probably because it would be correct in nearly any real-world context. If you’re scientifically-trained, or have done statistics in basically any other context, it’s probably second nature to process a large set of measurements by taking their mean (and possibly standard deviation) and going from there. It may never even have occurred to you to do anything else.

However, the reason we do that rests on an assumption–one that’s so universal in the real world that it’s rarely worth thinking about: the assumption that measurements of uncertain phenomena are normally distributed. The “normal distribution” is a specific statistical distribution that earned that name because of its ubiquity. What makes the normal distribution special is the Central Limit Theorem, which states that when you combine a large number of identical probability distributions, a normal distribution results (I’ll defer to the wikipedia article for a less abbreviated summary of how that works).

normal_distribution_500x263

WoW spells, however, do not follow a normal distribution. Nor are they (like many real-world measurements) an aggregate of microscopic phenomena that, regardless of the behavior of each individual molecule, aggregates to a normal distribution due to the Central Limit Theorem. They follow a predictable, known, artificial distribution: a perfectly uniform spread between the min and max values.

Random_Uniform_Distribution

What this means is that when playing the role of the scientist, looking at data and trying to figure out what the underlying behavior is, we have a huge advantage. We know the exact form of the phenomenon we’re observing, and all we’re missing is two numbers: the min and the max. Get those, and we know the complete underlying distribution with full precision.

The Upshot

Which brings to the thesis of this post. When you attack a target dummy 10 or even 100 times, there are only two trials you need to record: the min and the max observed values. The rest is quite irrelevant (which is logical: if you’ve seen a hit for 100 and a hit for 120, then a hit for 119 adds literally zero information to what you know about the spell*). In both cases, the mean value of the spell that you should put in your spreadsheet is 110.

It’s interesting to think about why this is different if you measuring something in a real-world scientific experiment. There, if you’d measured 100 and then 120, your best estimate of the true mean value would be 110. And then when you measured 119 on the next trial, your updated estimate would be the mean of 100, 120, and 119, which is 113. If those were the only three trials you did (bad procedure, I know), your best information going forward would be that the true value is 113.

So the difference is quite material. In the WoW setting, taking that extra step of averaging in the 119 data point is not only unnecessary work (possibly substantially so, depending on how much data you’re taking), but it’s also incorrect.

*Don’t fall into the trap of thinking that, maybe it at least allows a secondary inference of how close to the true min and max you’ve gotten so far. For example if you get 100, 120, and a whole bunch of 118-119’s, that’s not evidence that the 120 is less likely to be maximal. That’s your intuition for normal distributions talking.

Conclusion

There are two potential audiences for this post. The first are people who already did this intuitively when measuring WoW spells. I was in this category for while before thinking about the statistic rationale. For them, I hope it was interesting to stop and focus on this odd method of taking data that feels like a shortcut (glancing through numbers for the min and max and ignoring the rest). Specifically, how it’s a product of the way that games generate random events and how it differs from the way that nature generally does The other are people who are making WoW spreadsheets and can hopefully save some time and effort with this explanation.

In any event, I’m glad all the talk about theorycraft on beta prompted me to write this down. I do want to be better about writing down things I think about when working on WoW projects, especially when the topic has a little math intrigue. Hopefully there will be more to come.

Healing Theory, Part 9: First In-Depth Look at Warlords Heals

All posts in this series can be found here.

I’ve been waiting since alpha started to sink my teeth into the numbers of six new healing toolkits. Rather, I’ve been doing so for a while, but trying to get the information to the point where I can write about it in an organized way. I’ve been assembling a large spreadsheet of 6.0 heals since even before alpha started (since we had spell data). It’s mostly for my own purposes, to have a good reference for the properties of all the spells. I’ll post it here for people who want the full detailed background, while using the post to discuss various interesting points. The goal for this, and ensuing posts, is to work both for people who don’t peruse the actual sheet and just read the post (so I copy various numbers and such into the post), and also for people who want to look further into the sheet after I use a post to highlight some interesting points.

This is the sheet (download).

HT sheet image

It’s more utilitarian than user-friendly; as I said, it’s mostly an easy place for me to store/compare numbers (it’s not a character setup tool). The main focus for each heal is to compute an index for their HPM and HPCT (healing per unit cast time). In some cases, for non-spammable heals, HPCD (healing per cooldown time) is used to measure how much healing is added if the spell is used as often as possible. Some notes:

  • All of these are in unusual units, which is fine since they’re only meant to be compared against each other. For people who want details, HPCT and HPCD are in units of (spellpower coefficient)/seconds, and HPM is in units of (spellpower coefficient)/(% base mana cost).
  • All heals use the haste, crit, mastery, and multistrike values from the top of the sheet (Multistrike affects all spells equally so is not that interesting, except that Holy Priest heals get 25% more benefit).
  • Mastery is also in unusual units. “0.2” means, the amount of mastery stat that’s equivalent to 20% crit, or what was once called “20 points of mastery.”
  • The main computation in each row is to combine a spell’s spellpower coefficient from wod.wowhead.com with haste/crit/mastery/MS, any class passives or other auras, and any Draenor perks or other bonuses, to compute an overall effective spellpower coefficient.
  • I make assumptions as needed about talents, glyphs, and other variables. Usually the guideline is, a spell’s row in the sheet represents whatever case I was most interested in when thinking about how it would be used in practice.
  • Similarly, sometimes a spell has multiple entries if I want to see e.g. both glyphed and unglyphed, or if I want to see a combination of spells in a single row.
  • Blanket disclaimer: the beta build is changing often, changes aren’t always documented, some spells’ behaviors don’t match their tooltip data, and so on. I got some help from Dayani of Healiocentric, and we vetted all of the rows against in-game behavior in beta builds 18505 or 18522 (often with the techniques described here), and are continually updating them.

On to some of the interesting patterns.

Basic Direct Heals

Each class’s pair of single-target direct heals is meant to be a bigger part of gameplay in 6.0.  First is the “efficient” heal. Here’s how they look with the balanced stat array used in the sheet:

When you look only at basic, efficient, single target spam, Druid has the nicest spell (although its advantage is mostly due to Living Seed), and Disc and Paladin have the weakest. These numbers are of course out of any context of the remainder of a class’s toolkit which may affect tank healing, but they’re a starting point. The “fast” heals are worth adding before going further:

  • Regrowth (glyphed): 46 HPM, 8.51 HPCT (unglyphed RG is slightly better per-cast, but can’t be spammed effectively).
  • Regrowth (glyphed) without counting Living Seed: 31 HPM, 5.67 HPCT
  • Flash Heal (Holy): 49 HPM, 6.86 HPCT (Assumes every 3rd cast is a Heal due to Serendipity)
  • Flash Heal (Disc): 34 HPM, 6.32 HPCT. But Disc has a better spammable tank heal:
  • Clarity of Will: 80 HPM, 6.79 HPCT
  • 3xFlash of Light + Word of Glory: 37 HPM, 5.12 HPCT (HPM goes up to 46 if you replace one of the FoLs with a Holy Shock).
  • Healing Surge: 44 HPM, 7.60 HPCT (with Tidal Waves)
  • 3xSurging Mist + Enveloping Mist: 85 HPM, 10.14 HPCT (While Soothing)

As before, comparisons aren’t exactly apples-to-apples, since different classes use these spells in different ways and have other tools that supplement them (some discussed more below). But the basic numbers are helpful to look at. Glyphed Regrowth holds its own in the most optimistic assumption that Living Seed is always fully utilized, but is very inefficient otherwise.

Disc has a lot of difficulty healing a tank efficiently except for the Clarity of Will talent, which is currently very effective. Worth noting that while most classes have other tools such as HoTs that will supplement the heals shown here, Disc has nothing that adds HPS to a single target when worked in (Power Word: Shield and Penance are both lower HPCT than Flash Heal or CoW spam). PWS can add slight net HPCT if you Borrowed Time into a 2.5s spell, but is very inefficient.

The winner in this category is Monk, who can spit out solid healing in 4 GCDs with 3 Surgings and an Enveloping (the HPM value does take into account the later recoupment of Mana Tea). It’s not quite that simple, since you made need a GCD to set up by starting a Soothing on the target, and around half the heal is in the form of a HoT that you cast last so the heal’s somewhat spread out (but you may well have Chi coming in). But in the Warlords world, efficiency and the total output of a cast is more important than small shifts in heal timing. The HPCT is slightly misleading since it includes 30% of 6 seconds’ worth of Soothing Mist, which requires staying on the target for longer than the 4-5 GCDs to cast all this (for reference, it’s at a solid 8.57 HPCT without that).

Paladin is the trickiest case for looking at spam heals without any context. Talents will typically add a constant HoT on the tank from either Eternal Flame or Sacred Shield. If you assume those have 100% uptime, the value of Holy Power in a tank healing context is the value of the Word of Glory. An interesting fact that can be verified in the sheet, which is a change from 5.0, is that 50% of the total healing from an EF (i.e. an EF transferred from Beacon) is less than the full upfront healing from the Word of Glory portion only. The HoT is less than half the total value of an EF now, and in addition, a direct cast on tank benefits from mastery and gets 10% from Empowered Beacon of Light. So if your sole goal is to the heal the tank, it is better to cast each WoG directly on them than to blanket EFs around the raid. So a reasonable way to look at a tank-spam filler heal is to check the 3HL-WoG and 3FoL-WoG units, as I did above, and they both come in somewhat unimpressive on the HPCT front. They will gain a lot of efficiency in real raiding situations, when Beacon can be leveraged probably (which it can’t in these examples), but their ability to pull a tank’s HP up rapidly is limited.

AoE Heals

AoE heals are a little bit simpler to look at, because they can often be evaluated as standalone casts. Their use does depend on the number of available targets and in some cases the positioning, but the best case is reasonably clear due to their target limits. A few of the most important high-throughput options that each affect 5 or 6 targets:

The obvious standout here is Wild Growth. In a recent beta patch (18297) its healing and mana cost were increased by factor of around 4, making it the most AoE healing from one cast by far (except for some less rotational spells like Light’s Hammer). Its mana cost is pretty typical for the healing done, but being able to do that much from one cast is still an advantage. It is a particular advantage for Druids who leverage powerful buffs that affect “one cast”, Clearcasting and Soul of the Forest (the Swiftmend-WG combo, with SotF, is 132 HPM and 41.8 HPCT). One way or another, Wild Growth should likely come down from its current position, if only because it’s too asymmetrical. That said, a larger than average heal is reasonable since the spell is limited both in being a HoT and having a cooldown (unlike most others here). The only other similar one, Healing Rain, also has a high heal per cast to show for it. When heavy AoE heals are in order, other classes can cast multiple spells consecutively if needed, but Wild Growth and Healing Rain will only be cast once.

At the opposite end of the spectrum is Prayer of Healing, which is quite mediocre even after having its heal and mana cost recently doubled. Prayer of Healing is a spell I’d always hoped to see redesigned due to its anachronistic group-based mechanic, but so far the only nod to that is Holy Priests’ ability to replace it with Clarity of Purpose for a position-based AoE heal. Nevertheless, Prayer, while spammable, rates poorly on the efficiency scale and has little to show for it. Holy Priests still fare quite well because of Circle of Healing always being the first cast before you start having to spend on PoH. Furthermore, Holy has Binding Heal which is far more mana-friendly than it ever was, allowing a BH-BH-PoH combo that adds a lot of efficiency (up to 86 HPM and 11.58 HPCT with Glyph of Binding Heal).

Disc currently languishes in AoE healing with its efficiency option being Holy Nova (clocking in at 55 HPM and 3.51 HPCT).  However, I suspect that it was oversight for Holy Nova to not have its healing doubled when PoH and CoH did in the last build. At 110 and 7.02 on a no-cooldown heal, it will be a solid option whenever the positioning requirements allow. Disc will lack a good higher-throughput cast. The overall strength of PoH is likely fine when Archangel is used well, but the mana cost might need some help.

Monks come in just slightly behind on burst AoE throughput, with Uplift itself accomplishing up to 9.48 HPCT when used on 8 targets, and that’s ignoring where you get the Chi from. Fortunately, Spinning Crane Kick is a reasonably efficient spammable AoE heal and Chi generator, allowing ongoing AoE similar to casting PoH. But the numbers are below average, and the flexibility is poor with the uncontrolled nature of Uplift. Thunder Focus Tea will help slightly but it still rather inflexible in how it adds healing.

Miscellaneous Heals

The above was far from every important heal, but a sampling of reasonably comparable heals from the various classes. What I want to do for the rest of this first post is survey the chart for anything that’s interesting or stands out.

Lifebloom‘s bloom is currently 6 times what the spell data lists. 3 times is expected since they recently condensed what used to be 3 LB casts into one cast. I’m not sure what the other 2 is, and at any rate, the bloom might be need to be smaller than thrice the previous bloom. It’s currently quite a lot of healing for one GCD and no cost, albeit with a delay. I’m not too concerned about the apparently huge HPM numbers–high HPM can be a little misleading on spells that are limited in use, such as cooldowns or Lifebloom’s target restriction (if that’s not apparent, recall that some spells are free but not overpowered). However, the particular case of glyphed Lifebloom is 2.091 HPCD for no appreciable mana cost. (I haven’t discussed HPCD much but it means that using it on cooldown is equivalent to someone doing 2.091 HPS nonstop on average). This is far higher than 0.873 ticking HPS of the HoT ticking on its own; adding that 1.2 rotational tank HPS is probably not the intended use of Glyph of Blooming.

One point covering a few spells is that instant casts lose some of their value in the Warlords world. Compared to a 1.5s cast, an instant shifts the timing of the heal by a GCD (the beginning of a GCD vs. the start) but doesn’t change the HPCT. This only matters when the target dies during the GCD, which is more of a frequent concern in Mists than Warlords. Otherwise, there is little reason to spend extra mana or burden a spell with a cooldown to make that heal happen 1.5 seconds sooner. Movement is a possibility, but unlike DPS you don’t have to do something while moving; there’s much less compulsion spend mana on an inefficient spell. Swiftmend (41 HPM, 7.92 HPCT) and Power Word: Shield (43 HPM, 5.66), both come in about even with inefficient spammable heal options. This is a little odd for spells bearing cooldowns. It may be an attempt to shift away from having as many “rotational” heals, but their use is somewhat limited if they exist mainly to produce gameplay from the excitement of saving someone from near death.

Earth Shield‘s current behavior is that there’s no ICD on the use of charges, so it gets consumed quickly. This means that maintaining 100% uptime eats a lot of GCDs, and the spell’s stats as a cast in its own right (45 HPM, 5.01 HPCT) matter more than they used to. It’s a weak spell for the healing it does, so its only value is the 20% heal buff. This is good as always, but now that it falls off easily, it may well not be worth spending a GCD on recasting during busy moments.

Unleashed Fury combines with Healing Surge to make a huge tank heal (especially given HS’s 40% crit buff from Tidal Waves). UF’s 95% buff to one cast amounts to a free direct heal for no extra time, with only a 15 second cooldown. This can jump a tank’s HP upwards quickly when needed, or simply act as free rotational HPS.

Chain Heal is very weak (52 HPM, 3.72 HPCT with Riptide and Perk). I could have mentioned that in the AoE section, but I think it should to come up. There might be some concern about a spammable smart heal after the smart heal reductions in Warlords, but right now it’s not attractive to cast over even single-target heals. High Tide helps, but still leaves it as a mediocre Uplift–not what you want from a L100 talent (92 HPM, 6.56 HPCT on a Riptide target + 5 others).

I haven’t touched on damage-heals much. Smite is very weak (23 HPM, 1.41 HPCT), with its use having shifting almost entirely into building Evangelism stacks rather than doing healing of its own. This is good gameplay on its face–use Smite during lulls to set up Archangel, but don’t cast it when you actually need healing. The numbers on Archangel use will be better evaluated in the context of full rotational analysis later, but it’s a little concerning to have to cast 5 such inefficient spells to get it (even Holy Fire is 64 HPM, 2.55 HPCT). While I prefer this Archangel-oriented setup for the Atonement spells, because it’s a nice way to keep the mechanic while making the smart healing unimportant, we have to check that it’s not a trap with current numbers.

The one healing rotation that hasn’t been mentioned so far is the Crane Stance Mistweaver. Simply put, it does not do much healing right now. It’s not totally firm what the expectation is for this stance since it’s new, but in the past they’ve gone for around 50% healing and 50% DPS output from a meleeing Monk. The cheap filler rotation, JabTiger Palm is 29 HPM, 0.71 HPCT with all of Crane Stance’s buffs ramped up. Autoattacks add around 0.61 HPS to this, and free Surging Mists every 5 combos adds around that much again. So you can view Crane managing around 2 HPCT with the filler spells, which is approaching half of a normal healer’s “weak” filler heal. Even though this isn’t a full rotation analysis, a quick look at the HPCT of Crane’s bigger heal (1.51 for Rising Sun Kick) shows it’s not going to improve that average. Blackout Kick manages 3.08 HPCT for an AoE heal, well under half of Uplift for the same Chi. Exacerbating this is that Spinning Crane Kick now does very little healing (2.01 HPCT) and Crackling Jade Lightning is extremely expensive, so getting Chi in Crane punishes you heavily in terms of mana. This may in the end be worse than the very low healing output, since it’s a resource loss that stays with you when you shift back into Serpent Stance. It’s possible that regardless of throughput, the efficiency of melee heals can’t be that terrible or else using them will be too much of a tradeoff against even your normal healing ability.

Conclusion

This was my first effort at synthesizing the large amount of information in the sheet into broad comparisons across the healing classes. I plan to keep going with it, in particular to start adding more rotations to move beyond looking at single spells at a time. You can see some of that happening already in this post, but most of the work so far has merely gone into the spell table. I did not realize when I started how it would be (thanks again to Dayani for providing another set of eyes on nearly the entire thing by this point). But it’s gotten me to where I wanted to start for Warlords analysis, with the time put in up front to organize all this spell information in a way that facilitates future analysis. This is the first output from that process, but I’m looking forward to adding more as beta continues.