Partial ticks

I don't find haste thresholds very fun. Hit capping is already a little bit annoying; whenever I get new gear on a caster class I have to reforge everything to deal with the cap. At least I can manage the hit cap, though, without needing to refer to external sites. I just reforge to get enough hit and then change the rest to something else.

Haste thresholds are like hit caps but far more complicated. There are multiple haste thresholds that apply under different buff scenarios, and you have to decide with haste threshold you want to aim for. Then, again, every time you get new gear, you have to reforge like crazy to get yourself right back at the correct haste threshold. In practice, I'm hardly bothering. As a result, Ohken has a significant amount of haste right now that I should really get around to reforging off.

It doesn't help that the mechanic doesn't exactly feel polished. There have been a lot of questions posted over where the exact thresholds are. Many people are grabbing 10 or more extra haste just to be on the safe side. It also doesn't mesh well with the warlock spell Dark Intent. If you cast it on someone who is just above a haste threshold, they get minimal benefit from it. However, if they want to reforge so as to take advantage of it, they are severely gimped whenever they don't end up with Dark Intent. Overall, it feels like I'm doing a lot of work around haste thresholds for something that, from Blizzard's point of view, is more of an accident than something they prepared for a gamer to experience.

Let me describe a way haste thresholds could be removed from the game. First, consider the background. Up through most of Wrath, haste didn't help HOTs and DOTs at all, and thus it tended to be weak for classes that rely on HOTs and DOTs. In Cataclysm, all HOTs and DOTs now scale with haste. The way it works is that haste improves the ticking speed but does not change the duration of the spell. If you have a spell that, at baseline, does 5 ticks in 10 seconds, and you get 100% haste, then the spell will do 10 ticks in 10 seconds.

The only problem is that now we have to deal with haste thresholds. If you have anywhere between 80% haste and 99% haste in the above example, the spell will do 9 ticks in 10 seconds. Thus, the haste from 80% to 99% isn't improving your character in the slightest. That's a large gap for a stat to be doing absolutely nothing!

To eliminate these gaps, all it would take is to add a partial tick whenever a HOT or DOT is about to fall off. Continuing the example, if a player has 90% haste, then the spell would do 9 ticks at full effectiveness and then a partial tick at half effectiveness. With 81% haste, there would still be a partial tick at the end, but it would be a tiny one. With 99% haste, the partial tick would be almost as strong as a full tick.

One rule of game design, one I think is widely held, is that you want to control the kinds of decisions players are faced with. You want those choices to be fun. As an example, deciding whether to cast a big heal or a small heal is a fun decision. Each decision takes about 100 ms to work out during a single battle, it's not punishing if you make a mistake on some of these decisions, and in a gaming session you make a long string of these simple decisions.

I think haste thresholds fail the fun test. The only way to even know what the thresholds are is to do extensive theorycrafting and play testing. Hardly any player has time to do this themselves, so what they do is consult sites like Elitist Jerks that have a complicated chart of thresholds and a guide on which ones to aim for under what circumstances. It reminds me of the cookie cutter talent builds that we all currently use. Like with cookie cutter builds, reading up on haste thresholds are something you have to do and then, like a monkey, replicate in the game. There's no real choice in it, and it's tedious going through the exercise.

I think haste thresholds should go, and I think I see a way to do it. WTB partial ticks.

Valuing proc-based throughput

There's an interesting discussion on the Elitist Jerks resto-druid forum on the question of how much value should be placed on proc-based throughput. This is highly relevant because the Mandala of Stirring Patterns is one of the best pre-raid trinkets available from the point of view of sustainable healing per second. However, a lot of its throughput comes from an int proc that only has 20% uptime.

Note I am writing here only about the throughput improvements. Everyone agrees that the regeneration benefits of a proc are at least as good as a static improvement that has the same average benefit. As long as you finish the fight with at least one mana left, you don't care whether your mana descends at a steady rate or that it goes up and down in spurts as your procs come and go.

To evaluate this question, one premise I assume is that we are talking about progression content. It is a separate question to consider content that is on farm, where your raid reliably beats it and you want to beat it more reliably and more quickly. For such content, it doesn't matter that much what you do, much like it doesn't matter how you level, so I find it much more interesting to talk about how to address progression content.

For progression fights, it's highly unrealistic that players will be at max health throughout the fight. What's more likely to happen is that players hover around 50-75%, and you race to get them above 75% before the next wave of incoming damage happens. The cycle is that health bars drop after big damage, they float slowly up as healers heal like mad, and then they go down again at the next wave of damage.

If your healing is sufficient for the content and for the skill of the raid members at dodging fire, then the peaks and valleys of health in the cycle will be about the same. When I see my guild lose in such raid encounters, it's due to either unhealably severe spike damage, or it's due to players standing in fire. The former I can't do anything about. For the latter, I theoretically can help with, but when I fail to do so, what I see is the health bars going down fairly gradually.

For the content I'm familiar with, which is about half the Cataclysm raid bosses, the length of the cycle of hurt is on the order of thirty seconds. For example, on Argolath, a meteor slash comes in around 10-15 seconds, but it hits a different half of the raid each time. Thus you have 20-30 seconds after a player is hit by one meteor slash to heal them back up for the next one. Furthermore, any one wave of hurt tends to take off something like 50% of everyone's health bars. It's pretty rough, but it's not like many Wrath fights where everyone would suddenly lose 75-90% of their health.

How does proc-based versus static throughput compare in this environment?

It seems to me they are actually pretty close. First of all no proc is going to be wasted. Since players aren't at full health in the steady state, there will always be plenty of healing to use the proc on. Second, if you don't get a proc at the ideal time, it should just mean that players hover at a lower health for a little while. You will be able to pick them up just fine when the proc eventually comes around.

This is a minority view on the thread I linked. The majority view is that it's better to have static int instead of an int proc even if the average int goes down by 10-20%. The main arguments are just what I addressed above: they argue that many of the procs will be wasted, and that the procs might not be there when you need healing the most. In short, the proc won't save lives. I just don't see it. Saving a life only requires an emergency heal of 10-20% of a player's health, and for healing on that magnitude it doesn't matter what your immediate throughput is. What's more important is whether you can push the raid back up high enough to survive the next spike, and for that procs are fine.

Perhaps things are different in heroic mode, and there is more a need for emergency, unpredictable heals. I wouldn't know. At the rate my guild is going, I'm not sure we'll be doing much heroic content before the next tier comes out.

As well, I'll certainly agree things are different in a lot of ways once you master a tier and have everything on farm. You'd rather avoid proc-based gear for on-farm content. Additionally, sustainable throughput is much less interesting, and in particular haste is a lot more valuable.

Sustainable HPS theorycrafting

Tangedyn writes:

I was not satisfied with TreeCalcs having 2 different metrics. Although sound in theory, in practice hardly anyone knows how to use the 2 different metrics which is why most people seem to simply ignore MPS and look at HPS... which severely overinflates the value of Haste Rating. My model combines the two into one measure called "Sustainable HPS", basically the maximum sustainable HPS given available mana.

I'm delighted to see this kind of theorycrafting. Here is a link to the spreadsheet.

Tagedyn calculates over several healing models. To give an idea of the results, here are the relative stat values for the model where you roll lifebloom and rejuvenation on a tank and then focus on the raid:

• int: 1.0000
  
• sp: 0.3907
  
• spi: 0.3707
  
• mastery: 0.2099
  
• crit: 0.1694
  
• haste: 0.0756
  

A few things to note:

• Intellect is the best, followed distantly by spirit, followed distantly by mastery and crit.
  
• Haste is rock bottom of the barrel, both for raid healing and tank healing. This is counterintuitive, because it is king for max throughput. Max throughput just isn't that important.
  
• Intellect is way ahead of other stats. Always choose higher-intellect gear when you have a choice. Notably, trinkets with int should be the better choice even across ilvls. Trinkets with both int and spirit are vastly better.
  

One thing I wish Tegedyn would add is support for raw mp5. This would be helpful in evaluating, for example, the on-use effect of the Jar of Remedies.

Always be casting?

At current gear levels, the best rotations for maximum sustainable throughput are a little counter-intuitive. The maxim "always be casting" does not hold for us right now!

To see why, start by looking at some heal-per-minute (HPM) estimates from the invaluable TreeCalcs spreadsheet on Elitist Jerks.

• Lifebloom: 18.94 (assuming a rolling 3-stack)
  
• Swiftmend: 9.94
  
• Wild growth: 9.67
  
• Rejuv: 5.53
  
• HT: 4.96
  
• Nourish: 4.92
  
• Regrowth: 2.11
  

The exact values will vary depending on gear, raid buffs, and boss debuffs, but these give a rough idea. Also, these values will change in patch 4.0.6. Rejuvenate will get much cheaper, thus having a much higher HPM. Wild growth will do 30% more healing, which means 30% higher HPM. The 4.0.6 changes only further emphasize what I describe below.

These values assume that there is no overhealing. For progression content, if you can afford to let people drop to 80-90% before healing them, then this is pretty realistic. The main exception is that lifebloom on the tank will probably do a lot of overhealing, but you really need to keep that lifebloom stack rolling no matter what. If you let it drop, you don't just lose the healing from it, but also the mana regeneration.

Other things equal, you want to cast high HPM spells. At the top of the list, keep lifebloom up and cast wild growth and swiftmend whenever they are off cooldown. At the bottom of the HPM list is regrowth. Regrowth is just abysmal for sustainable throughput, so save it for emergencies or for omen of clarity. Overall, the extremes of the HPM list give us the usual conventional wisdom for sustainable throughput.

The spells in the middle are more interesting. They all have roughly the same HPM. Nourish and healing touch are nearly identical, and rejuvenate is a little bit higher. This similarity in HPM leads to a counterintuitive result: for sustained throughput, it doesn't matter very much which of these three spells you use! The only thing to be aware of is that if you cast the mana-heavy ones, you need to cast them less frequently. Thus, the more you cast rejuvenate and healing touch, the more dead time you need to sit still and let your mana regenerate.

If you are tank healing, then the thing to do is keep rejuvenate up and then cast healing touch every once in a while. How often you should cast healing touch depends on how much mana regen you have. This is a counterintuitive approach, because most druid bloggers recommend constantly spamming nourish and then casting an occasional healing touch whenever you can. However, I don't see the advantage of doing so. You do just as much healing and use just as much mana, but by casting healing touch you are able to move around between casts.

If you are raid healing, then matters are simpler: just cast rejuvenate. Rejuvenate has better HPM than any other option, and it's an instant cast, so nourish and healing touch have literally no advantage. The only time you'd want to cast something other than rejuvenate on the raid is if you need to heal someone that already has rejuvenate on them. All these rejuvenates are expensive, but like with tank healing, simply insert enough dead time that your mana doesn't drain too fast.

Both rotations have a lot of dead time in them where you just sit there twiddling your thumbs and watching the pretty graphics. How much dead time is enough? It's a matter of theorycrafting, and I don't know. As a rough ball park, though, I believe it should be substantial, something like 40% of the time. Instead of nourish spam, which you can do indefinitely, you cast spells that are roughly twice as expensive, roughly half as often.

I know this dead time is counterintuitive. Dps casters must Always Be Casting to get their highest dps, but that's only true when they are in a position of never running out of mana. Also, all the blogging restoration druids are recommending never-ending nourish spam whenever your mana is tight. Perhaps their intuition is that nourish is "free", because while you cast it you regenerate enough mana to cover the cost. However, you still have to pay the cost of the spell. If you instead don't cast, for the same amount of time, you get to cast rejuvenate or healing touch sooner, and this makes up for the lost healing from nourish.

If Blizzard wanted to change this situation, they could make nourish do much more healing and thus have better HPM than healing touch and rejuvenate. Then what we'd cast nourish as much as possible and only use the more expensive spells because we become GCD locked. However, honestly, I kinda like the idea of having some free time to think and move around. It's a nice change of pace. It makes healing more deliberate and less spammy.

Sustainable throughput

When people talk about druid gearing, they usually talk about throughput versus efficiency, and they talk about regen rates. Throughput is how much healing per second can be done if you don't worry about mana. Efficiency is how much healing you do per mana. Regen is how fast your mana comes back. All three of these are good, and it ends up being a three-way tradeoff.

In principle, the latter two can be combined, so as to make it a two-way tradeoff. The two goals would be maximum throughput versus sustainable throughput. Sustainable throughput is the amount of healing that can be put out without going out of mana.

Once we get there, I suspect most people wouldn't pay much attention to maximum throughput. You can always use cooldowns and apply pre-hots to get through the hard parts of fights, so maximum throughput should rarely be a problem. The real reason druids want better throughput, I think, is that it indirectly helps sustainable throughput. When lifebloom heals for more, you can spend less time casting healing tuch and more time casting nourish.

The level of sustainable throughput depends on how long the fight is. That's because you start the fight with a full bar of mana. With a very short fight, maximum throughput is the same as sustainable throughput, because even at maximum burn rate you won't run out of mana. As fights get longer, sustainable throughput decreases toward an asymptote. For absurdly long fights, the initial mana bar becomes irrelevant.

I haven't tried to theorycraft the computation of sustainable throughput. I would love to see any work that others have done on it. It is likely helpful to divide it into two components: one for the infinite-length fight that ignores the initial mana bar, and one that figures out the advantages of burning through the initial mana bar over the course of the fight. The full sustainable throughput is the sum of the other two.

Maximum sustainable throughput might not only affect gearing, but spell selection. In principle, it could even be helpful to pause for a few seconds and, let mana regen, and then cast a more effective spell after a few seconds. I don't know whether that's really true or not, but in principle it could help. Sustainable throughput gives a way to analyze decisions about spell selection.

Sustainable throughput is more valuable when damage is less spiky. With spiky damage, you just have to drop big heals on the tank so that they don't get gibbed. With slower damage, you can afford to wait a few seconds and do things more efficiently.

All in all, I'd like to see more emphasis on sustainable throughput. It looks clarifying for gearing and spell selection, especially with Cataclysms smoother damage patterns.

RestoDude on gearing and professions

RestoDude has an excellent summary up on gearing up a restoration druid in Cataclysm. It assumes you nothing, and it includes stats,

I hope it gets updated over time. I have a few comments on the details:

• Spirit is not restoration-only. For restoration it gives mana regen, and for moonkin, it gives hit. Blizzard is always trying to make spirit valuable for all casters, and this is their latest effort.
  
• It's true that mana regen is especially valuable at lower gear levels, but spirit isn't the best way to get it. In early Wrath, it made sense to gem for spirit if you were low on regen. In Cataclysm, you should gem intellect.
  
• Herbalism is better than RestoDude says. It gives an extra cooldown besides Tree of Life for improved healing. You can either use it separately or macro it together with Tree, but either way it's a very helpful cooldown. Haste is great.
  

These are just nits. It's a great one-stop article for gearing up restoration druids in Cataclysm.

Regen stats

While mana is still plentiful at level 80 -- you get enough regen just by the mandatory intellect and spirit on our gear -- it looks like things will be different in Cataclysm. We won't be able to just stack all throughput stats. We need to watch out for mana as well.

Rank 4 Healing Touch did some math that compares intellect versus spirit for mana regeneration. In short, at level 85, one point of spirit gives about 0.5 mp/5 and one point of intellect gives 0.90--almost twice as much!

What this means is that, as things stand, spirit is mediocre Always stack intellect in preference to spirit. You'll get better regen, and you'll also get spell power to improve throughput. Further, when comparing same-ilvl gear, always prefer the gear that has no spirit, no matter what else it has on it. Even if you need the regen, you can make up the lost regen by gemming half as much intellect as the spirit you gave up.

Personally, I find this a little disappointing, and I hope Blizzard rebalances the regen stats. I find it fun to balance regen versus throughput, and even to balance different kinds of throughput (crit, haste, mastery, spell power). On the regen side, it makes no sense that the one stat that is purely for regen is underpowered. It becomes one of those things where players who read the tool tips and follow basic logic end up playing worse than players who go snoop around on the Internet for obscure calculations.

It seems like all Blizzard would have to do is double the regen from spirit and halve the regen from replenishment. Then, intellect would be a so so regen stat but also give throughput, while spirit would be the place to go if you are really hurting for regen. Intellect would still win out for most people, but at least spirit wouldn't be so bad that people avoid gear with it.

Type H for Heals has done some great analysis on how haste affects heal-over-time spells.

One punch line is that, from the perspective of healing per cast time (HPCT) and healing per mana (HPM), haste doesn't improve a spell until you get an extra tick on the spell. For restoration druids, the only relevant extra-tick number is 37.5%. If you get to 37.5% haste -- from all sources including raid buffs -- then your rejuvenation has an extra tick at the end.

You also get a rejuv tick at 12.5% and a regrowth tick at 16.7%, but these are such low numbers that practically all druids will be past those points. There are also several thresholds for lifebloom, but those don't matter because...

The analysis is completely different for tank healing! When tank healing, even if you don't get an extra hot tick, you benefit from haste by being able to cast the hot a little sooner. Thus, for tank healing, you get higher HPS at all levels of haste. HPS will take an extra large bump when you get past 37.5%, but otherwise haste is all good. Assuming you have a 5% haste buff from a raid, you need 1016 haste rating to get there. There seem to be some rounding issues, so it's better to get a little more.

All this said, I'm not sure what the practical impact is. Haste is still great both before and after that number. First, you are probably at least rolling HOTs on the tanks even if you are raid healing, and for rolling HOTs, haste benefits them at all levels of haste. Second, every little bit faster that regrowth gets, the better.

Haste in 4.0

It looks like haste is an even bigger stat than ever for 4.0. We are casting a lot of direct heals, now, and more haste means they land faster. In particular:

1. Regrowth is now our top-up spell, and haste speeds it up until a cap of 50% where it has a 1.0-second cast. Faster regrowths on a tank mean it has a better chance of landing before the next melee swing. Faster regrowths around the raid mean we can top up more people after a decimate before they die.

2. Healing touch hits like a truck now. I’m seeing it land at twice a regrowth and three times a nourish. However, the cast speed is just sluggish enough that it’s difficult to use well. Getting it’s 2-second cast time lower makes it much easier to use. Haste is valuable for this purpose until healing touch is at 1.0 seconds, somewhere around 150%.

3. Rejuv and regrowth have better throughput when hasted, which is especially relevant for tank healing. This is useful up until stupid levels of haste. (Instant rejuv? Yes please.)

As much as it makes sense to talk about a soft cap, it's around 50%. However, 50% is already nearly impossible to attain at level 80, and it should be even harder at level 85. For practical purposes, resto druids have no haste cap in 4.0.

So how much haste should you get? Right now, mana regen seems to be fine for most players. Many are even spamming regrowth and not running out. So don't worry too much about regen.

What about throughput? To the extent you really want to worry about raw throughput, haste should be balanced against intellect, crit, and -- if you tank heal -- mastery. Right now, though, throughput is pretty high, and for many healers it is probably as high as matters for the raids they are doing. In that case, it's more important to get heals landed than to make them bigger, and landing heals faster is all about haste.

TL;DR: Haste is great. Get all you can without going OOM.