Thursday, December 30, 2010

What Slow Means - Or Why You Should Take Heal Off Your Bar

Correction: Once again foiled by bad information on the net. While I think the essential point of my post is still completely valid, the casting time of Healing wave is indeed 2.5 seconds with the proper spec, not 2.

As Cataclysm was approaching, the developers talked about how they wanted to change the way healing worked. They said they wanted to "slow it down." What exactly it means to slow things down, however, might not have met the expectations of everyone.

Healing in the Wrath environment could certainly not be accused of being slow. Commonly people talked about how healing was all about twitch reflexes. You twitch wrong, someone dies. You twitch right, everyone is back at full health and everything is fine.

I wondered as Cataclysm approached how slowing things down could keep things challenging. After all, successful Wrath-era healers were used to making a life-or-death decision every second. If we only have a make a life-or-death decision every two seconds, how could that be anything but easier? It turns out that slower did not mean exactly what I expected. In the sense that nearly every GCD was life or death things were very fast in wrath. In the sense, however, of the amount of damage your raid took compared to the amount of healing you were capable of doing, things were actually incredible slow.

In Wrath healing, druids and paladins would often have overheal in the 60%-70% range, or even a little higher. Discipline priests had fairly low overheal, at 30%, but when you looked closely, because shields could not overheal, their actual overheal on spells that healed people could get just as high or sometimes even higher than paladins and druids. If you are overhealing 70% and no one is dying, that means you are outputting three and a third times as much healing as is needed. In a 10-player raid, that means that one healer is able to output enough healing to keep the entire raid alive. The reason you brought more than one healer was a question of distribution, not of throughput.

In a Cataclysm raid overhealing 70% would be next to impossible. You would pretty much have to go out of your way to target people at full health with your heals to achieve a number like that. Distribution of heals is no longer reason you bring three healers to a fight, you bring three healers because you need that many to actually heal the amount of incoming damage.

This was easiest to see in early experiences healing heroics. When the entire group had just reached item level 3291 healing through boss damage could be very difficult. The boss could actually output more damage per second than some healers could heal. The reason it was possible to win was that the damage coming from the boss only exceeded the ability to heal by a small amount. If the fight lasts three minutes, and your group has around 550,000 health between then you can actually win if the boss does 3000 more dps than your healer can heal.

This is a new kind of calculation, one that we would never have done in Wrath of the Lich King. This is what it means to "slow down" healing. Instead of needing to land a heal every second to make sure the boss doesn't two-shot the tank, you need to maximize the amount of healing you can get over a long stretch in the hopes that it will be enough total healing to give the dps time to win the fight.

Many of us, when we heard that they were going to make mana matter, rightly reasoned that if the total amount of damage that will be done before the fight ends is close to or slightly exceeds the maximum amount of healing that your healer can do over that period of time then your healer cannot afford to use low efficiency heals. The flip side of this – which is entirely necessary for the new healing model but which I didn’t see coming – is that if the total amount of incoming damage per second is close to or slightly exceeds the maximum amount of healing that your healer can do per second, then your healer cannot afford to use low throughput heal.

A lot of priests have taken the wrong path here. For a shaman, the real question was how to do enough healing over the length of the fight, and Healing Wave was an important part of the answer. But let’s take a look at Healing Wave. Its casting time is only 2 as opposed to 2.5 for Heal, but its scaling is similar to Heal and while its base healing is lower it is higher on a per second basis. Tidal Waves hastes it by 30% twice every six seconds. It can trigger Earthliving Weapon. Ancestral Awakening matches Divine Aegis in the extra critical department and Nature’s Blessing probably slightly exceeds Grace in a heroic dungeon. If you are having trouble keeping up with incoming damage than Deep Healing increases its effect by around 1.25% per point of mastery while the discipline priest mastery likely increases Heal by around 0.1% per point. The tank also has Earth Shield. All together, Healing Wave is slightly more powerful than Heal, and has an average cast time of around 1.7 seconds to Heals’ 2.2 seconds.

For a discipline priest, the question was never how can I do enough healing over the course of the fight, it was always how can I do enough healing to keep up with the incoming damage on a per-second basis. With 15% extra intellect, Hymn of Hope, Rapture and Shadowfiend, discipline priests have enough mana to get through the fight, what they need is enough throughput to get through any given 10 second period.

So the solution is to just not cast Heal. If you are trying to heal damage to a single target, cast Greater Heal. If you are looking to fill time casting something because you don’t think you can afford to waste it, cast Prayer of Healing - you can think of it as a single-target heal with benefits rather than a group heal if you need to. If everyone is near full use Renew. If none of these is a good option then just don’t do anything at all. This may sound like a strange suggestion given that I’ve been talking about not having enough throughput to keep up, but casting Heal can be worse than nothing. Given that Greater Heal has about 2.8 times the throughput of Heal and an at-par mana efficiency Prayer of Healing has around 2.9 times the throughput of Heal, starting a heal now instead of waiting one to one and a half seconds to cast a better spell is a loss of throughput.

Strangely enough I don’t even think this is a complaint. For Heal to be a bad spell for discipline priests – and I think a questionable one for holy priests – is not that bad a place for the game to be. Priests actually have a lot more spells than the other healing classes, and expecting them to use them all is a little unrealistic. We should be much more concerned about the relative weakness of Power Word: Shield for discipline and Renew for Holy than about Heal which I feel completely comfortable leaving in my spellbook to be pulled out for Chimaeron.


1. And many of the members of your group may have faked their way to item level 329. This is especially easy for most tanks since, as plate wearers, they can easily buy a whole bunch of gear that is absolutely useless to them from faction vendors. Cloth with spirit counts towards a prot warrior's item level.

Wednesday, December 29, 2010

Mind Sear

A recent blue post indicated that Shadow Priest AoE feels too low. I was pretty sure that Mind Sear was not strong enough, but I hadn't taken the time to really look into it.

We downed Maloriak on Monday, and for those unfamiliar with this fight, it involves an AoE phase where you need to burn down some high health adds with a +100% damage taken debuff before the debuff expires. Naturally, I turn towards them and hit Mind Sear.

My damage to the adds was very bad, around 12% of the damage done to them compared t0 16.7% for the shaman and 18.7% for the hunter.1 Mind Sear ticks did around 13% less damage than earthquake ticks and, of course, hit all but one add instead of all.

But part of my low Mind Sear damage was most certainly bad play and incorrect gearing. Right now I have only one gear set, and I think of myself as healer rather than as a dps. As a consequence, almost everything I have has spirit on it or is reforged to spirit. I also reforge away as much haste as possible to get more crit and mastery - for single target and AoE healing respectively - since mana is still a larger consideration than throughput. This is not the ideal configuration for shadow dps, but, combined with play mistakes, it is especially bad for AoE.

Shadow Priest mastery is shadow orbs, a complex way of adding a percentage to virtually all damage we do. We have an 18% chance to get an orb every time we deal damage with Shadow Word: Pain or Mind Flay. Upon casting Mind Blast or Mind Spike the orbs are consumed, causing the spell to deal 1.25% more damage per point of mastery per orb and leaving behind a 15 second buff - Empowered Shadow - that gives 1.25% more damage per point of mastery (but not per orb) for all periodic effects.

With my relatively high mastery rating, Empowered Shadow is a 17% buff to periodic damage, including Mind Flay and Mind Sear. Unfortunately, I had Empowered Shadow up very little of the time when I was AoEing. What I should have been doing - and what I will do next time - is ensure that the last thing I do just before the add burn-down starts is Mind Blast the boss to refresh Empowered Shadow so I'll have it the whole time.

Having 100% uptime on Empowered Shadow during Mind Sear phases would have increased my Mind Sear damage by something like 14-15%. That would still leave me at the bottom of the AoE damage meters, but much less solidly at the bottom. It would also have made our AoE phases a little bit less tight.

Obviously Mind Sear is just not doing enough damage, but balancing it is actually really tricky. Given that there is no way to generate that buff while you are actively AoEing the question of where to balance Mind Sear compared to other AoEs is very tricky. If the switch from boss to adds is predictable and you have a 15 second or smaller burn phase to get rid of them then good play will give good uptime. If the adds come out an unpredictable times then uptime will be lower. If the burn phase is 20 seconds then uptime will be lower. If the burn phase is 10 seconds than good play becomes less important to getting good uptime and the predictability is less of a factor. Increase gear a couple of tiers and the difference between the buffed and unbuffed versions of Mind Sear are going to get larger and larger.

I think that right now Empowered Shadow might be a little too mind-bending just by itself, though improvements to my UI should do a lot to fix that. Despite the fact that it is complex compared to other masteries, it is probably not badly balanced, so I expect it will likely stay. Now that I know the developers think Mind Sear has some problems, I'll be really interested to see how they rebalance it given the complexity of the interaction with Empowered Shadow. Ideally I think the answer is to find some way to let Shadow Priests trigger or refresh the Empowered Shadow buff while AoEing, and changing Empowered Shadow to simply not affect Mind Sear is another, slightly less palatable option. Of course ignoring this problem is another option, since it can hardly be seen as the most pressing issue in the game, but as mastery levels increase, if this interaction is ignored it may be increasingly the case that Shadow Priests are either perpetually good or perpetually bad AoEers, or that their performance in AoE varies tremendously based on some very esoteric elements of the fight.


1. The mage did 22% with Arcane Explosion spam, but there is no reason to compare yourself to the high outlier when you are the low outlier.

Thursday, December 23, 2010

Giving Atonement a Chance

First of all, I'd like to draw attention to the following blue post about bug hotfixes:

The range of the heal on Atonement has been increased to 15 yards. It has been verified to be working correctly for the priest and party/raid members.

It is nice that not only did the fix the Atonement bug, but they actually checked that it was fixed.  I don't think I've seen a post like this before, but I'm glad to know that they are actually checking to see if they fix the priest bugs that they say they are fixing.  If I recall correctly, and it is obvious that I don't, when Surge of Light first came out they indicated that they fixed it in approximately one thousand consecutive patches without ever changing anything about it1. It was one or two patches of claiming to have fixed early Rapture bugs before they were finally resolved.

The changes to Atonement go beyond that, however. Atonement is now affected by healing bonuses. This means that Atonement will always heal for 6% more than the Smite thanks to "double dipping" on Twin Disciplines. It also means Atonement can benefit from Archangel and Grace. I'm not quite sure whether these all stack and whether they stack additively or multiplicatively yet, but I do know that I have had heals well over 20k on occasion, which seems like a lot for a cheap 2 second cast spell.

This significantly increased throughput and mana efficiency added up to making Atonement viable. I gave it a test run in heroics before raiding with it last night and I was happy with it in both cases, though it seemed more useful in heroics than in raids. The main advantage of this build at the moment is that you don't have to give up much to get it.

This is the spec that I tried out. Comparing this with my previous healing spec you can see I took the five points out of Inner Sanctum and Soul Warding which were both of questionable value to begin with. I also swapped Inspiration out for Surge of Light, because Smite procs Surge of Light, Atonement does not proc Inspiration, and Inspiration doesn't have a very good up time with current crit levels. This choice may be a mistake, but I haven't had time to do a good quantification of either talent so far.

How many Smites I actually use varies a great deal from fight to fight. So far it has been important to keep in mind that the Atonement spec is not an all-out Smite spec. Smite is one heal among many, and the rest of the toolbox needs to be there. In particular, for our Chimaeron attempts I was not using Smite at all because it didn't provide the sort of healing the fight needed. Smite may actually be very useful for that fight, but it will need some more experimentation to figure out how to approach healing overall. For Magmaw, on the other hand, Smite is great for healing the tank.

I picked this spec out as overpowered because of unusually high scaling, and I was waiting for it to come into its own in future tiers. The most recent changes have buffed it enough to make it good at the current tier, and presumably it will need to be dramatically merfed in future tiers.



1. I think the actual number was somewhere in the four to six range. Anyway, for months on end every time they patched anything they would say the talent was now working correctly despite the fact that they hadn't fixed a thing.

Wednesday, December 22, 2010

Archaeology

So I actually love archaeology. I really wanted to have some kind of profession that involved more than just clicking on something, and archaeology goes that for me. I think the items they added to the game for it are worthy of the time you put into getting them. It is not necessary for raiding but it provides you with cool or useful things.

I did have some constructive criticisms of archaeology, however. One of my first complaints was simply the predictability of dig sites - there were always three finds for three relic pieces each. They already fixed that one by putting in a random element, so I figure I'd better get on complaining before it is too late.

Another thing I'd like to see from Archaeology is some way of controlling what you are working towards.  Basically right now you are completely at the mercy of the random numbers.  If you want to work on a particular race's artifacts you can't choose to dig for them as opposed to digging for other things.  This is very disempowering.  If I am interested in the ancient history of the nerubians, why would I not be spending my time digging at their sites rather than at troll sites?  There are a number of ways this could be accomplished:

  • An ability on a medium cooldown (one hour is probably about right) that allows you to abandon a current dig site to spawn a new one
  • An ability on a long cooldown (seems like one day) that allows you to abandon one of your current research projects to get the new one
  • A way of trading in fragments from one race for fragments of another or of trading fragments with other players - I can understand if they don't want fragments to be a commodity on the AH, but there are ways to avoid that
  • An ability or daily quest that allows players to pick a particular site or even a particular zone they want to dig an either activate it immediately or increase the chance they will get it
Of course that list is not exhaustive, nor are the things on it mutually exclusive.

I don't suggest any of these things as ways of speeding up the process of getting meaningful archaeological finds.  Regardless of the amount of customization given the players, drop rates and frequencies of rares can be adjusted to give whatever progression rate the developers feel comfortable with.  I would very much, though, like to see a little bit more interactivity with archaeology - something that goes beyond the process of digging for things and puts a little bit of the process of deciding where to dig or what to make in the hands of the player.

Tuesday, December 21, 2010

Sliver of Pure Ice

My item level is up to 348, but I am the first to admit that value is somewhat spurious. Item level takes the highest level item that you still possess that you have equipped in a slot at some point. As a result, you get credit towards both finger or trinket slots based on your highest level ring or trinket, even if you only have one of them and thus couldn't possibly use it in both slots.

In fact, my armory says that my item level is 348 but conveniently notes in brackets that I have item level 337 equipped. Why the big difference? It's Sliver of Pure Ice from Icecrown Citadel.1

It's not that I've been extremely unlucky with drops and have never acquired a trinket from a dungeon. But I've never acquired a better trinket from a dungeon.

Case in point: I have a Blood of Isiset in my inventory. It's 69 levels higher, so it seems like it should be clearly better.

Blood has a 105 second internal cooldown and a 10% proc chance. Right now I'm casting a spell about every 2.5 seconds, so it should average around 25 seconds to proc again, giving it an overall cooldown of 130 seconds. So it gives 1512 spirit 15.4% of the time. That's equivalent to 232.6 spirit all the time. Spirit provides 0.1225 mana per second with my current intellect, so this is about 28.57 mana per second overall. Intellect provides 0.0733 mana per second with my current levels of intellect and spirit, so the 152 intellect is worth 11.13 mana per second. The use ability on the Sliver is pretty easy to work out, 1830 every 120 seconds is 15.25 mana per second. Looks like blood wins on mana by 2.49 mana per second, the equivalent of about 20 spirit.

Blood provides 252 mastery instead of 152 intellect. I haven't parsed any of my logs yet to get a good casting pattern, but we can easily compare these things for two of my most frequently cast spells: Greater Heal and Prayer of Healing. For Greater Heal, Intellect is around 6.5 times as good as mastery for increasing throughput. Furthermore, Intellect is consistent and mastery is random as it is dependent on criticals. For Prayer of Healing, aegis triggers every time, so Intellect is only 2.9 times as good as mastery. In the best case scenario, then, the Sliver is around 75% better in terms of throughput. If we converted the throughput from Intellect into master for a direct comparison, the Sliver would be around 189 master rating better. Bear in mind that this is being extremely generous towards mastery rating by assuming that I cast virtually nothing but the spell where is shines the most.

I don't want to directly compare throughput to mana, but if you asked me whether I wanted 189 mastery rating or 20 spirit, I think the choice is fairly clear.

The developers recently mentioned that they know that "secondary stats" are worse than primary ones. So spirit, mastery, crit, and haste are worse than Intellect. I think they may be a bit delusional, however, about exactly how much worse they are. This is spec dependent - I think Blood would clearly win out over sliver for a Holy priest - but 69 items levels for a trinket itemized for the same spec should guarantee a stronger item. Instead, the most important thing to look for on a trinket is whether it gives a passive bonus to Intellect. Item level comes second.


1. It's also because I have my Runed Signet of the Kirin Tor equipped from a recent teleport to Dalaran.

Monday, December 20, 2010

Tol Barad

I had my first experience of trying to attack Tol Barad last night and it was very frustrating. It seemed just about as impossible to win on offense as it seemed to lose on defense. Obviously the zone can change hands, but it feels like the defenders have far too great an advantage in the battle.

Although it ignores some of the complexity of the zone, the analysis to see why this is is prety simple. In order to win, the attackers need to have more people than the defenders do in one location while at the same time having at least as many people as the defenders do in the other two. Given that the number of people on each side is supposed to be equal, this is a pretty difficult task.

The graveyards help the attackers in a way - the distance is always larger for the defenders, but this is greatly overwhelmed by the fact that the defenders have a shorter distance to attack a *different* building than the one they died at. Since the attackers already have to manage a noticeably higher unit count in an equal battle in order to have a chance of winning, they cannot have lots of extra people at uncontested bases. Whenever the horde reinforces the base where a fight is happening, it means the attackers move more people to that base. This means that the defenders can simply go for one of the other bases after they die, knowing the attackers couldn't have kept it at full strength.

But what's more, the Eye of the Strom-like capturing mechanic makes it impossible to put up a valiant defense. If there are five of us and ten of them then it doesn't matter how well we play, we are going to lose the base. Even if five eventually manages to beat ten, the other team has still bought themselves a lot of time and made sure the attackers can't get to three bases in the mean time.

Something needs to chance in the zone. Either make the buildings click based like AB flags instead of EOtS buildings is in possibility. Another is to get the attackers a real advantage from taking the towers. Or maybe just give attackers a general buff, or make it so that attackers count for more than omet defendesr while taking a building. Any of these might work, but the zone right now is just not working well.

Edit: I just defended Tol Barad last night in a real battle for the first time. I've defended with 10-12 people before, but this time it was around 30. I explained the strategy to my teammates: If the horde sends a large force to take a node, don't reinforce it when you die, simply go and retake another node after rezzing. The horde never managed to control two bases, for more than half a minute. I felt they were playing better than us in battles, and generally winning with equal sized forces, but it was hopeless for them. We even had a few people mining and picking herbs instead of helping.

Friday, December 17, 2010

Train of Thought

Train of Thought is a very interesting talent, and fairly hard to evaluate. It's a talent that saves you mana, but only if you spend mana on an expensive spell. The question of how much mana it actually saves you is a difficult one to answer.

If you don't cast a good number of Greater Heals, Train of Thought is a very bad talent. Suppose you mostly use Heal as your single target heal, and use Greater Heal only when Inner Focus is making it free. Because Inner Focus must be available at the start of the cast, but it not used - triggering the cooldown - until the end of the cast, the real time between Greater Heals would be 45 seconds plus the casting time of Greater Heal. We'll use 2.28 seconds for the casting time of greater heal1.

So in this case Train of Thought would reduce the time between free Greater Heals from 47.28 seconds to 42.28 seconds. These free heals also benefit from +25% crit, so they are more powerful than typical Greater Heals. In order to figure out how much mana this saves us, we can't look at the cost of Greater Heal, because we wouldn't have been casting Greater Heal if it weren't free. Instead, we should look at how much mana we would have spent healing the same damage with our spammable spell of choice, Heal. In my current gear, a free Greater Heal is worth around 3.21 Heals of healing. Since heal costs 2118, that means we save 6799 mana per Inner Focus boosted Greater Heal. This is actually more than the cost of Greater Heal, thanks to the 25% crit bonus. If we save 6799 mana every 47.28 seconds, that gives us 143.8 mana per second. If instead we save 6799 mana every 42.28 second, that gives us 160.8 mana per second, only 17 mana per second for two talent points. Definitely very bad points.

The question is, what if we cast Greater Heal instead of Heal to heal the damage between free casts, reducing the time between free heals more. Things get a little more complicated in this case. We can figure out how many of our Greater Heals will actually be free to cast - and +25% critical strike chance - using the following:


In order to calculate the time between casts, we would need to know how much our average heal heals for, and how much damage we are trying to heal per second:



To get the average heal, we'd need to know the ratio of free heals, as well as the average amount a normal Greater Heal heals for (G) and the average amount an Inner Focus Greater Heal heals for (G'):



Fortunately, three equations is exactly the right number for three unknowns. The simple formula for the ratio of free Greater Heals is:



So, given a damage amount we can get the average free Greater Heal ratio and the time between casts, which allows use to easily calculate the amount of mana spent per second. The question is, is it cheaper to cast Greater Heals than Heals? If it is not, then Train of Thought saves us mana only when the damage exceeds the amount we can heal with Heal, and we should cast Heal as often as possible. If it is cheaper to heal the same damage with Greater Heal, then we should ignore Heal, even if the incoming damage is low.

There are two results from this simple analysis. The first is that it is cheaper to cast Heal, the second is that the amount of incoming damage you can actually heal with Heal is so low that Heal is basically never a viable option so it is a bit of a moot point.

Of course these results come because the model is too simple. In reality we have three spells - Power Word: Shield, Penance and Borrowed Time hasted Renew - that are all strictly better than Heal but that have real or effective cooldowns. We would never cast Heal if casting one of these were an option instead. In additional, while Power Word: Shield isn't strictly better than Greater Heal, it would have to be a very bizarre situation for us not to use it, and both Penance and the hasted Renew are better than Greater Heal in both efficiency and throughput.

So, what happens to the calculations when we cast Penance, Power Word: Shield, and hasted Renew on the "cooldown"?2 Fortunately we can use the same formula above for the ratio of free Greater Heals. What changes is that the damage we are trying to heal with them is much lower, and our mana expenditure depends not only on how many Heals or Greater Heals we cast, but also has a baseline set by the cost of our superior spells. When we factor this both of our conclusions from the simple case are reversed. While Heal becomes a viable option for some reasonable levels of incoming damage, Greater Heal becomes the more mana efficient way to heal at all levels of damage.

This may seem counter-intuitive at first, so I'll talk a little bit about how this happens. I can understand if your intuition tells you that if one thing is more mana efficient, then it should still be more mana efficient when used as a gap filler. The reason this doesn't end up being true is that Train of Thought is doing more work for us when we are casting other spells. While we still have the same number of real seconds between Greater Heal casts, many of those seconds are used up casting things we are going to cast regardless of what else is going on. At most we are going to fill around 30.86 of our 47.28 seconds with Heal or Greater Heal. The rest of the time we will be using more power heals with better efficiency. This allows us to increase the time between Greater Heals to Heal the same amount of damage. More time between heals means a higher ratio of free heals.

At around 8400 incoming dps - the highest incoming dps that Heal can take care of in my current gear - using Heal and the more efficient spells requires 1010 mana per second. At the same damage level, using Greater Heal instead requires only 900 mana per second. So Train of Thought is saving us 110 mana per second for two points, which is a lot more respectable than 17. Of course it's also important to consider that healing with Greater Heal instead of Heal saves us time, so we have more leeway to fit more heals in if we need them. While at this level of healing casting Heal means casting pretty much 100% of the time, using Greater Heal means only casting 70% of the time. Train of Thought is buying us a lot of time to move around and do other things, in addition to mana savings.

I haven't had Heal on my bar this entire expansion and this analysis just helps to reinforce that decision. While there would still be substantially more modelling to do to quantify the benefit of Train of Thought against the benefit of another talent, such as Divine Aegis, it is pretty safe to say that Train of Thought should be regarded as a "must have" talent rather than an optional one and that Greater Heal should be the staple spammable single target heal for Discipline priests.


1. I chose this number because it is the casting time of greater heal given that you have just enough haste rating to make sure your Renews get a fifth tick when hasted by Borrowed Time. Until throughput becomes more of an issue than mana, I will regard this as the "correct" amount of haste to have.

2. I'm assuming shielding a single target and casting one Renew per shield.

Thursday, December 16, 2010

So Much Mana

They just hotfixed in a huge buff to priest mana. Rapture was increased from 2.5% of mana to 6% and holy concentration from 20% more spirit regen to 40%.

A discipline priest will get about 4 rapture ticks a minute. Over a five minute fight, at 3.5% additional maximum mana per tick, that means about 70% of their mana bar restored. In my current gear, that would be around 67k more mana to spend over the fight, or about 16.3% more mana to spend.

I'm not sure why this change was made at this point in the expansion. It could be that the developers were getting a lot of feedback from priests that suggested that they weren't powerful enough or that they were constantly running out of mana.

This is one of those times where my experience doesn't seem to line up with the experience other people are having. I've certainly run out of mana while healing as discipline, but if mana is supposed to be something then running out of mana has to be a possibility. I was able to heal heroics as soon as I hit the item level 329 requirement, even if my mana management strategy was occasionally to just hope for more crits. On our Argaloth kill I was the top healer.

I think what's really going on is that the majority of discipline priests, and the development team, are stuck in a Wrath of the Lich King style of healing. Though I still find myself doing it occasionally, I have almost entirely trained myself out of the reflex of throwing a Power Word: Shield on anyone who takes damage. I never use Flash Heal for any reason. I only cast Prayer of Mending when the damage distribution on the fight makes it sensible. I spend the majority of my time casting Greater Heal and Prayer of Healing. Basically, aside from Penance, there is extremely little overlap in how discipline priests should heal now and how they did heal before.

Most of the time buffs are accepted with open arms, but I'm a actually pretty skeptical of this one. If they want me to cast Power Word: Shield, then giving me more mana for my one cast every 15 seconds is not what they need to do. They need to either make Power Word: Shield large enough to warrant its cost and to provide a real life-saving benefit, or they need to make it cheap enough to justify throwing out regularly. This change reinforces the idea that you should cast Power Word: Shield on the tank every time Weakened Soul wears out and never cast it otherwise. Power Word: Shield without a Rapture proc is just as expensive and ineffective. Power Word: Shield with a rapture proc is now a significant net mana gain. It is a rotational ability that keeps your mana up, and more closely resembles Life Tap than Wrath of the Lich King Power Word: Shield. When a buff like this happens, it makes me worry about what upcoming nerf is going to compensate.

Wednesday, December 15, 2010

Preliminary Results

The results are in. We've killed Argoloth so I have some preliminary examples of real raid healing. Did I meet the developer expectation of 50% healing from shields?

The right answer is not even close. World of Logs shows me getting 10% of my healing from PW:S and 25% of it from Divine Aegis. Their Divine Aegis numbers seems a little sketchy, however, since if you add up all the critical heals and Prayer of Healing heals I did in the entire fight and take 37.5% of that - my current return on Aegis thanks to mastery - it doesn't reach the amount of aegis absorbs that World of Logs calculated. In fact, I applied only about 80% of the shields that they say I used, and I know I didn't use them all. It wasn't until halfway through the fight that I clued in to the correct healing strategy to avoid wasting Aegis.1

So my real healing from shields was around 31%. On that fight I can get 100% shield usage if I play correctly, so we can use that number as an accurate representation of the amount of shields I would put out on a fight.

The reason that this matters, as I've noted before, is that mastery rating is tuned according to the amount of shields that discipline priests do. Mastery rating increases shields, so how powerful it is is dependent on what fraction of your total healing comes from shields. However, another change that has been made since I last did analysis on this is that Prayer of Healing always triggers Divine Aegis. That means that mastery is significantly more powerful than it would have been while critical rating is significantly weaker.

If I had 179.28 more mastery rating, I would have increased by mastery bonus from 25.12% to 27.62%, thus increasing my shields by just about 2%. That would increase my total healing and shielding by about 0.6%.

179.28 crit rating would have increased by critical strike chance from 19.06% to 20.06%. That means that 1.24% of my non-critical spells would instead have been criticals. Since my non-Prayer of Healing heals totaled around 166.6k, that would mean critting with an additional 1399.5 worth of heals. My effective critical multiplier is 1.5 plus divine aegis. Aegis is 30% times plus the mastery bonus for 37.5%. This gives a total critical multiplier of 1.5 * 1.375 = 206.25%. That means I would get an additional 1399.5 * 106.25% healing by critting with those heals for 2186.

Now for my Prayers of Healing, I get the Divine Aegis on the base even if I don't crit. That means that I only rack up 50% * 1.375 extra healing on a crit, or 68.75% of the base spell. While I did overheal with Prayer of Healing, this is largely due to the fact that the rogue was standing behind rather than in front of the boss, and so prayer of healing hits on him were largely totally overheal. I did 546940 healing with non-critical Prayers of Healing. If 1.24% of those were crits and those crits added 68.75% more to the total healing done then that would contribute 4644 extra healing.

The total extra healing from a 1% gain in critical strike chance, then, would be 6830. With just over 2M healing for the total fight, this is an increase of 0.33%.

These calculations show Mastery as a far better stat than critical strike rating. Of course this is contingent upon casting a lot of Prayer of Healing. What would these numbers look like if instead of making Prayer of Healing always trigger Aegis, they had simply decided to give Discipline Priests a talent to increase its effect of Prayer of Healing?

The buff for Prayer of Healing is contingent upon not getting criticals. If you get a critical heal then the buff does nothing. Let's say I averaged my critical strike and mastery ratings to eliminate my current reforging bias. Then I would be critting 17% of the time and my mastery would be 12, giving me a 30% increase to shields. At that point, the buff to Aegis to always apply to prayer of healing would be boosting the effect of 83% of my Prayers of Healing by 39%. The overall buff to Prayer of Healing, then, would be about 32%. What if they had just put in a talent that increases the effect of Prayer of Healing by 32%?

Then my shields would be only 19% of my heals, so 1 Mastery would be a 0.36% increase in healing. 1% crit, in this case, would be a 0.56% increase in healing, nearly three times as much as Mastery.

It would seem that Discipline priests require two different gear sets. One that is geared towards critical strike rating for single target healing and one that is focused on mastery for AoE healing. For now I might try to even out the two stats since I don't know what raids will have in store for me yet. Ultimately, these quibbles probably make very little different since both critical strike rating and mastery are fairly weak. In the best case scenario, if all my new secondary stat points went to mastery, an entire tier of new gear would provide less than a 2% healing increase from these ratings.


1. Meteor slash happens every 17 seconds, and Divine Aegis has a 12 second duration. Since meteor slashes alternate between two groups, a group will take one only every 34 seconds. Thus, if you spam three or four Prayers of Healing on the group when the slash hits, the Aegis will wear off before the next slash. If you space your Prayers of Healing out over the 34 seconds then you have the full Aegis from all of the prayers roll through to the next slash. You just have to battle against those Wrath-tuned instincts that tell you that if anyone is below max health you need to spam until they aren't. This should mean that a discipline priest can take care of meteor slash damage by themselves.

Tuesday, December 14, 2010

Gear Cubed

Cataclysm has succeeded in reintroducing mana as a resource for healers. How much mana your healer has is a major factor in whether or not your can win a fight. Because of this, dps paying attention to encounter mechanics has become important, because if they take more damage than they have to and the healer has to keep them alive, then the healer is more likely to run out of mana.

Our first few heroic runs were good demonstrations of this. I was drinking every two trash packs and our ability to beat the boss was very tied to my mana pool.

Three days later, I was never drinking and the idea of running out of mana on a boss was laughable unless we really, really blew it.

The difference was not that our dps got so much better at handling the encounters. My guild members are pretty competent, so when it became clear that we had to actually do things right we did things right. The difference is explained by how encounter survivability works.

In Wrath, healers didn't really think about mana and their heals were powerful enough to heal the tank to full very quickly, so tanks didn't worry about avoidance. All anyone worried about was whether the tank could live through a large burst. Survivability was pretty much all about the tank's health pool. Because of this, it scaled linearly with gear.

In cataclysm, the tank is not in danger of being bursted out because a heal can land unless they make a mistake with encounter mechanics. That means that avoidance is quite relevant since it decreases long term damage. Since it takes multiple heals to move a tank from low health to full health, the power of the heals is very relevant. Since it is plausible to run out of mana before the end of the fight, and more mana means the ability to cast more heals.

Instead of survivability scaling roughly with the stamina of a single character, it scales with avoidance * healing throughput * mana. In other words, it now goes up with gear cubed instead of linearly.

Adding 13 item levels was enough to turn heroics from a significant challenge into something pretty close to a cakewalk. Another tier of gear and heroics will probably be a daily chore to get Valor Points.

This is the formula that explains pre-Wrath gear check bosses. A gear check boss was one that was not that difficult provided that you had sufficient gear, but that became completely impossible below a certain gear level because your tank would simply die or your healers would run out of mana. Wrath had gear check bosses as well, but they were bosses that required a certain amount of dps gear to beat an enrage timer, not ones that relied on healer and tank gear.

These two types of gear checks are not mutually exclusive. Last night we downed our first Cataclysm raid boss and he had a tanking/healing gear check though the damage he did as well as a dps gear check by way of an enrage timer. Fights like this are extremely easy to overgear and trivialize. Of course it was a very simple fight, and the second boss we attempted seemed significantly less simple.

I hope that as we go forward with Cataclysm bosses we find fewer gear sensitive bosses and more skill sensitive bosses. So far I've seen an even split.

Monday, December 6, 2010

My Cataclysm Spec

First of all, The Spec, and then, the complaints.

I've written a lot about Atonement and its ups and downs. But all the number crunching in the world doesn't change the fact that it will only heal you if you are within 8 yards of the target of the Smite. For most raid bosses, that would mean standing well inside of them. For the most part, Atonement will simply do nothing against raid bosses, so I am opting out of the Smite spec completely.

Given that I won't be taking the Smite spec, there is actually very little choice about what I *will* be taking. There are 42 points in the three and I need to select at least 31 of them. Since I am not taking the five points dedicated to Smite, I have to choose 31 out of 37. Reflective Shield essentially doesn't do anything, so I'll rule that out right away, leaving me with a choice of four talent points I don't want.

Many of the talents are simply not candidates for cutting, so I'll focus on the ones that I think I could choose to do without:

Improved Power Word: Shield - Power Word: Shield without a Rapture tick is an exorbitantly expensive and underpowered spell. It is considerably worse than Renew, Penance and Greater Heal for both throughput and efficiency. While it is almost 10% more mana efficient than a Flash Heal, it only about 65% of the throughput. I know as well as anyone that instant casts matter, and that shields are better than heals some of the time, but ideally I want to avoid casting this spell more than once every 14-15 seconds. A talent point for a 5% increase to a spell I cast less than 10% of the time is not a very exciting point.

Mental Agility - I don't actually cast a whole lot of instant spells. If I am casting Power Word: Shield, Renew and Prayer of Mending each once every 15 seconds then it's about 91 mana per second, which is likely a decent investment of talent points, but it's certainly not obviously something you must take.

Inner Sanctum - I think I will mostly be using Inner Fire, not Inner Will, so this would be a 2% spell damage reduction per point. I think this is probably very bad. If I have occasion to need to run faster for a fight and switch to Inner Will then this talent is probably pretty good.

Soul Warding - If I don't want to cast Power Word: Shield more than once every 14-15 seconds, then I probably don't need to cast it more than once every three seconds.

Power Infusion - I have never seen Power Infusion as a terribly useful healing cooldown. I see it more as an external dps cooldown for the raid's casters. For that it is a great talent point. Since I am a healer, I would consider dropping this point if I am not using it for healing.

Rapture - It should be offensive that I am even listing this talent, but, as I say, Power Word: Shield is a bad spell that you don't want to be casting. If you are using it, then Rapture is over 160 mana per second for three points, which seems like it's definitely a good use. But when if I didn't take it and didn't cast Power Word: Shield at all? Right now I'm a fairly convinced that this would make me worse off, but it's hard to be sure without a lot of analysis. Of course, I probably won't do that analysis for reasons I'll explain soon.

Borrowed Time - Going back to not casting shield very often, Borrowed Time might not be doing much. Hasting one spell by 14% every 15 seconds is an average of 1% haste for two talent points, which isn't spectacular. That's not really how Borrowed Time works, though. The 14% haste can be applied to multiple instants - where Penance counts as an instant here - and then to a spell with a cast time. Also, the 14% haste on a Renew is likely a 25% or 20% increase in the effect as well as a 14% reduction in the cast time. Some simple simulations are showing me that this is more than 1% haste per talent point. Still not great.

Strength of Soul - Reducing the Weakened Soul debuff duration seemed like a pretty good idea when Rapture had been reduced to a 6 second cooldown. At a 12 second cooldown, this talent is near useless even if you are casting Heal on the tank.

Train of Thought - It's far from obvious that this talent is a good one to take. How much mana it actually saves you depends on how often you cast Greater Heal. If you are using Inner Focus for Greater Heals on the cooldown then Inner Focus is worth 141 mana per second, which is a really good deal for one talent point. Train of Thought increases this to 159 mana per second, a gain of 18 mana per second, or 9 per talent point. If, however, you are casting three Greater Heals per 30 seconds, then this would increase the mana gain from greater heal to 211. So the two talent points were worth 35 mana per second each. If you are casting five Greater Heals every 20 seconds then the mana gain is 318 per second, and the Train of Thought talent points are worth 177 mana per second, or 88.5 each. The trick is, that they are only worth that additional mana if you were going to cast the Greater Heals anyway. Of course it does make casting those Greater Heals more attractive, so even if you substituting them only because of the talent it will still give some advantages.

Focused Will - 10% of my total health is presumably going to be more in Cataclysm than it was in Wrath. For example, I don't expect bosses with damage auras that tick every 2 seconds to do 10% of my total health. I do, however, expect this to trigger from occasional raidwide AoEs or from having a fire spawn under me. However, it only helps if another hit comes within 8 seconds. Ultimately I find it hard to believe this is going to be good for raiding.

So why did I make the decisions I made? As I indicated above, I think that Strength of Soul is probably going to do nothing useful at all. In fact, Weakened Soul might be a handy timer for me to know when to recast my shield, and reducing it may mean having to pay attention to more information without getting any benefit. There are several talents here that require analysis to figure out how good they are, but my intuition tells me Focused Will is the worst of them. It will depend on the structure of boss fights, and it might easily turn out to be a better investment than Inner Sanctum. For now, I'm going to have to go with my gut and I'll do the calculations when I actually have fight data.

I just trash talked a huge amount of talents and then elected not to pick ones that I thought would have no or virtually no effect on raiding. More importantly, I trash talked every talent that affect Power Word: Shield. Basically, I think Power Word: Shield is a terrible spell that only becomes castable because of Rapture. Assuming Rapture still interacts the same way it did in Wrath with large AoEs that hit the whole raid at once, Soul Warding will probably edge out Focused Will for utility. If they have fixed that bug then I'll probably switch those.

The real reason I take Rapture, Improved Power Word: Shield and Borrowed Time is that I simply can't avoid it. If Atonement isn't going to work on bosses, and I don't want Power Word: Shield talents, then there are only 24 talent points left in the tree. I have to pick something.

So why am I even discipline in the first place? First of all, it's because Penance, Divine Aegis and Grace are all insanely good and Holy just doesn't have anything with that kind of oomph. Secondly, it's because I feel like external tanking cooldowns are too useful to pass up, and Discipline gets two of them.

I'm not going to be writing for probably at least a week, maybe two. When I do come back I hope to have a little bit of data so I can answer a few of the following questions:
  • How much better off am I using Rapture and casting shields than I would be if I simply didn't cast shields at all? (Even though this point is moot because you can't avoid taking PW:S talents)
  • What is the real benefit of Train of Thought? Should you cast Greater Heals instead of Heals when you have it?
  • How good is Mental Agility?
  • How good is Focused Will?
  • How many of my spells are hasted by Borrowed Time and how often is it boosting a Renew?
  • How much damage is Inner Sanctum preventing?

Friday, December 3, 2010

The Archangel Roller Coaster

It's been a pretty wild ride for Smite priests since the first look at the Cataclysm talent trees. The smite sub-spec has gone from a nice little bonus, to extremely overpowered, to probably better than using Heal, to totally unplayable and now I don't really know what to make of it.

The recently Archangel from returning 3% mana per stack of Evangelism to only 1% mana per stack of Evangelism. Since we are probably going to have a little over 100k mana, that a difference between 15k mana returned every 30 seconds and 5k mana returned every 30 seconds. Of course the difference is even more dramatic than that in a sense. Smite costs over 3.5k mana and a full Evangelism stack reduces that by 30%, of just over 1k mana. If you need to rebuild your stack then you need to cast one smite and full cost, one 94% cost, one at 88% cost, and so on until you get back to 70% cost. That means you are paying the cost of 30% + 24% + 18% + 12% + 6% of an extra smite, or 90% of a smite. That's around 3.2k mana. So in reality, using Archangel only nets you around 1.8k mana. It also costs you about half a smite worth of damage and healing. It also gives you a 15% healing bonus for 18 seconds, but then again it also costs you a GCD. On the balance it is a net increase in healing, but I'm not sure it's worth taking.

This, apparently, made the Smite spec unusable. The solution they came up with is as follows:

Smite's scaling coefficient has been increased. The damage of Smite is now very similar to Heal in value. This change has been made to ensure the Archangel Discipline sub-specialization remains viable.

There is a really glaring problem with this. Smite already had a far higher coefficient than Heal. If you make Smite hit for about as much as Heal in tier 1 raid gear, then by tier 4 heroic raid gear it is going to hit for quite a lot more than Heal. At that gear level, Smite has throughput much closer to Flash Heal than it does to Heal, and it has better mana efficiency than Heal. This is a very problematic place for it to be.

With the current numbers, tier 1 Smite priests are going to be a little sub-par and probably will have too many mana concerns, while for end-of-expansion content they will be very noticeably overpowered - providing around the same healing as another healer would by spamming their fast high-powered heal and doing about half the damage of a real dps character.

If mana was the problem for discipline priests, perhaps they should have addressed the problem by lowering the cost of Smite, rather than by increasing it's effect. The disparate coefficients of Smite and Heal were already heading for trouble. Pushing them further apart is going to be a bit of a catastrophe.

Tuesday, November 30, 2010

Berserk Timers

Berserk timers are now as brutally overused as untauntability was in Burning Crusade. There are virtually no bosses in Wrath that do not have a berserk timer. Berserk timers are an awkward kludge, and ideally the game shouldn't have them at all.

What's wrong with berserk timers?

First of all, they are "immersion breaking." No matter how much you play the game as a game I think this matters. Boss abilities fit a theme of the boss and demonstrate the power of the creature you are fighting. Calling meteors from the sky, freezing us in ice, summoning minions and good old fashioned breathing fire are all things that powerful creatures do. Suddenly becoming unbeatable if you fight them for exactly six minutes, on the other hand, is really out of the blue.

Second, berserk timers are often put in just for the sake of it. My guild tried to take on 25-player Beasts of Northrend with 14 people, both for gear and to see if we could do it. After a couple hours of trying we finally managed to get to the end of the fight and lost to the 15-minute berserk timer. It is far from clear what purpose that timer is serving other than to stop us from having fun. Fighting 25-player encounters with 14 people is generally not the best way to get loot. Similarly, a Death Knight managed to solo Sartharion, and the greatest difficulty to overcome was that he only had 15 minutes to do it. Sartherion's berserk timer didn't stop us from beating him with 4 people in appropriate gear, so it is serving no purpose but to limit the fun of people doing strange things.

Third, when berserk timers are actually setting the difficulty of the fight, they limit the raid's ability to play through mistakes. This is especially true in 10-player raids. If a fight requires 2 tanks and 3 healers, then losing 1 dps means losing 15-20% of your damage. If the berserk timer is the limiter that is making the fight challenging, then this is an unrecoverable loss, and the moment someone dies you might as well just wipe and start over. Even if someone dies 75% of the way through the fight, if you were only on pace to make the berserk timer by a few seconds, this can be an unrecoverable position. Playing through mistakes and making the best of bad situations is something the game should encourage, not discourage.

Once upon a time, Berserk timers were a good solution to a problem. The problem was that if you brought a tank and 39 healers to an encounter, then you would win for sure1. A time limit on how long you have to win gives dps a place in competitive content. Without them, you eventually die.

Berserk timers do solve that problem, but there are lots of other ways to solve that problem, and very few modern bosses need berserk timers to make dps matter. If the boss summons adds then either you need enough dps to kill the adds before the next ones come, or you need enough dps to kill the boss before there are too many adds to tank. Stacking buffs or debuffs that make the boss more and more dangerous as the fight goes on serve the same purpose as enrage timers, but don't break immersion, and give a little more leeway to play exceptionally and scrape some extra seconds out if you need them. Rotface summons oozes faster and faster as the fight goes on. On Blood Queen you will eventually run out of bite targets and get mind controlled. Heroic Northrend Beasts releases extra bosses on a timer. Heroic Anub'arak only gives you six frost patches to work with, so the third burrow is basically automatically fatal.

I'd like to see them do away with enrage timers entirely. It's totally fine for Firefighter - after all, we hit the big red button - and maybe for certain other circumstances where there is an in game reason that the fight would be a limited time only. But making bosses do 500% more damage after a certain number of minutes is a crutch for encounter designers, is an awkward kludge solution to an old problem, makes no sense within the game world, and rarely makes the game better.


1. I realize this is a bit of an overstatement, many fights would have required more than one tank, and beating fights with nothing but tanks and healers would have taken forever. There was a real concern, though, that encounters could be trivialized by overloading on healers.

Monday, November 29, 2010

Zul'Gurub

I've done the Zul'Gurub quests, and I want my raid back.

First of all, the quests that involve Zul'Gurub would have worked just as well in an instance as they did on the world map. You never walk in through the front door, you get teleported in and do strange little tasks.

Second, they took maybe 10 minutes in total, all of them put together.

I can't entirely rule out the possibility that we will be sent back there at higher level, but if you are so far Wowhead knows nothing about it.

I realize that no one even set foot in Zul'Gurub aside from a few dedicated mount collectors, but what they put in to justify removing it doesn't have a whole lot of value either. My position is clear skewed by the fact that I was one of those people farming for a mount, but comparing what was given up to what was put in its place - and the development lost plus the development that was put into the new stuff - it seems like a bad trade.

Tuesday, November 23, 2010

Too Hard and Too Easy

One of the problems with PUG raids is that raid content is generally accepted to be too easy but in reality much of it is too hard for new and casual players. Some time ago, Ardol at World of Warcraft Philosophized posted about why the leveling process doesn't prepare you for raiding. He makes some good points about the differences between solo and group play that contribute to why people manage to be profoundly bad at raiding with their class despite having to go through the leveling process just like everyone else.

Much more importantly to the issue of PUG raiding is that the leveling process, in its current form, doesn't give you the most important skill in raiding: dying, running back, asking yourself what you can do differently next time, and trying again.

In raiding, you have to be ready and willing to lose and to learn from your mistakes. Without this, you really don't have any opportunity to improve your play. But when raids break up after a single wipe, you never have an opportunity to get better even if you are ready and willing to analyze your play and practice.

When a fight isn't explained and a wipe results in everyone just giving up instead of discussing what went wrong, a newcomer to the fight won't even know why they wiped or whether anything could have been done to avoid it. Sometimes to an experienced raider the causes of wipes can be pretty obvious even on a fight they've never seen before. For a new raider, however, how do they know the difference between - to use Toravon as an example - the raid losing because too many people stood too close to a pair of frozen orbs and the healers weren't on the ball; versus the raid wiping because the fight went too long and he got off a lethal white out? Their experience of either will be similar: everything was going fine until suddenly it wasn't and nearly everyone died.

The problem this creates is that not only does leveling not teach you how to raid, raiding doesn't always teach you how to raid. Unless you are raiding with a group of people who actually share their experiences and who are willing to try again, you are just spinning the wheel every time and hoping you come out a winner.

Dying in WoW is of little consequence. Usually it won't waste more than 10 minutes of your time. Using 10 minutes to die is a great use of time if you actually manage to learn something from it.

People often say that WoW has become too easy, but I want to be totally clear about something: WoW is shockingly hard, but the number of things to do in WoW that are actually really that hard are fairly limited. Firefighter, Yogg +0 and heroic Lich King are all very hard. Algalon, heroic Anub'arak, and heroic Sindragosa were pretty challenging as well - even if the former two seemed harder than they really were thanks to limited attempt mechanisms.

But there is a distinct lack of build-up to the encounters of that difficulty. There are parts of the game that don't seem that tough and then the difficulty curve suddenly gets very steep. PUG raids usually want to take on content up to the point where the difficulty ratchets up, but don't want to try to go further because they can't reasonably expect a PUG will win. Because they only take on the content that is "easy" they expect they should win without a great deal of difficulty. This adds to the frustration of losing, increases the chances the group breaks up after a small number of wipes, and reduces the learning opportunities for PUG raiders.

Reversing this trend means having to start early. Players should get used to the idea of dying and trying again before level 10, and should be reminded that that is part of the game as they go through the leveling process. In his inaugural developer blog post, Ghostcrawler points out that the fun of playing on God Mode wears off pretty quickly. That philosophy needs to be applied consistently through the game, not just to heroic raid encounters. The player base should expect that they will need to try more than once to accomplish important things, and they should get used to feeling that good feeling you get when you succeed after failing, learning and practicing.

Wednesday, November 17, 2010

A Leader's Responsibility

Wipe aversion is actually very understandable in PUGs. One of the key differences between a PUG raid and a guild raid is that a guild raid is actually pretty certain that they are going to win. Maybe they won't win tonight, or even this month, but raiding guilds don't let that dampen their enthusiasm. When the members of a guild lose faith that they will ever beat an encounter the guild often starts to fall apart. A single boss can not only wipe raids, but it can outright destroy guilds, if it is capable of making people feel that no matter how hard they try and no matter how much they practice they will never win.

In a PUG, winning is far from inevitable. You really might just plain lose and walk away with nothing. It's a potentially disheartening affair, and at some point continuing to fight the boss is throwing good money after bad. If you no longer think the raid is going to succeed, then you should bail and do something more productive with your time.

Much worse than this from a raid cohesion perspective is that once one person leaves, the threshold for another person to leave plummets drastically. After all, the first person has to be unwilling to sit through one more attempt. The second person has to be unwilling to sit through the time it takes to replace the first person plus the time for another attempt, and so on. One person leaving - especially from a 10-player raid - can be enough to set off a cascade.

Obviously no one in a raid wants this to be the outcome, but people running the raids have an additional investment. The majority of people, even people who seem like jerks most of the time, actually have an internal sense that when they start something and it fails, they deserve the blame. That feeling of shame and blame may be warring with other feelings - such as rage against the group's lousy dps - but people mostly don't like to fail and as a raid leader you can't really leave the blame with other people. If the group failed, it's because you put together a bad group.

Unless, that is, you put together the group based on an objective measure that is supposed to tell you whether people are capable of succeeding or not. If you go into 10-player ICC with a group wearing ilvl 264 gear, then you "should" win. If the group can't beat Saurfang and degenerates into name calling then it is pretty easy to say to yourself that the members of the group must have just been bad players.

Because the tool is objective1 the raid leader doesn't have to shoulder the blame for its outcome. The tool really becomes the raid leader; and given that PUG raid leaders rarely explain fights, go over what happened when there is a wipe, give motivational speeches, or talk to individual raid members about how to improve their play, the tool might as well be the raid leader for all other purposes as well.

None of this is to say that PUG raid leaders are wusses or to put the blame for anything on their shoulders. Very few people decide lead PUG raids precisely because being a leader carries with it a burden. Moreover, people who are really great raid leaders are mostly raid leaders in guilds. The skills it takes to be a great raid leader aren't really that different than the skills it takes to be a great leader of people in general. Those skills are not common.

Blizzard thinks that implementing an Raid Finder system similar to the Dungeon Finder system is a problematic idea for a lot of good reasons. GearScore, however, serves as a makeshift Raid Finder that players have created themselves. Using an objective entrance criterion that should theoretically be tied to the group's chance of success takes the raid leader and their responsibility out of the equation.

In an environment where a single wipe can lead to a complete failure, you'll get very few volunteers to shoulder the responsibility for the success of the raid. The number of PUG raid leaders who want to use GearScore to determine raid inclusion is just a consequence of that fact.


1. It isn't really that objective because you have to decide what GearScore you are using as a minimum. My feeling is that most raid leaders choose values that are well above what is really needed to do the dungeons and rely on overgearing rather than appropriate gearing to beat content. Of course they do this for the same reason they use GearScore in the first place - to try to minimize wipes.

Tuesday, November 16, 2010

Reframing the Argument

The fastest way for a discussion to degenerate into uselessness is for each side to convince themselves that the disagreement stems from the other side being stupid. If you assume that you understand someone else's point of view, and that the disagreement is occurring only because they do not understand your, clearly superior, point of view, then you aren't going to get very far1.

Of course, one of the challenges is that people rarely actually articulate the reasons why they are doing something, and often aren't completely aware of the reasons themselves. People rationalize and then argue against the words used to express the other party's rationalizations. It's some pretty fantastically unproductive stuff. It certainly doesn't get any easier when one side never even presents a point of view at all.

People who start raids in trade chat requesting that everyone who comes have a certain GearScore rarely get drawn into the debate about whether GearScore is a useful tool or not. Other people have that debate for them. The arguments against are generally that skill is far more important than gear. The arguments for are generally that since there is no readily available way to measure skill, gear is the best we can do2.

But those arguments actually don't have much to do with the problem or with either side's motivation. Whether or not GearScore is objectively a useful tool for forming successful raids doesn't make a huge difference. The group that doesn't like GearScore doesn't like what relying on GearScore does to the game and to the community. The group that uses GearScore doesn't think that the GearScore haters understand the reality of forming PUG raids.

I completely agree with the GearScore haters and the majority of the WoW blogging community who say that using GearScore to form raids have a negative impact on the game. For me, though, this has nothing to do with the fact that GearScore can't measure skill, and it has a lot more to do with the fact that GearScore can't measure willingness to learn, ability to take direction, social skills, dedication, and any of the other traits that would make someone who has neither gear nor experience a good addition to a raid. The problem with GearScore isn't that it shuts out people who somehow deserve to be on raids, it's that it shuts out people who just want a chance to play and learn.

It's also worth noting that GearScore can't measure whether or not you are just a complete asshole. When it serves as the only entrance criterion it basically gives people a pass on destructive behaviours and attitudes.

For a trade chat raid leader, though, giving people a chance doesn't seem like a viable option. The reality of PUG raids is very harsh. We all know that a PUG raid can break up after a single wipe at worst, and that for something like Vault of Archavon the lifespan of the raid is half an hour to forty-five minutes maximum even if people are willing to stick it out and try again. A leader under these circumstances doesn't feel like they have any room to give new people a chance.

It's also very questionable whether giving new people a chance is very beneficial in this kind of scenario. No one explains the encounter before it begins. There is no discussion of what went wrong if things do go wrong. The closest thing you are likely to get to constructive advice is being called a "n00b."

The functioning of PUG raids in the current environment seems antithetical to the idea of giving new people a chance to play, to learn and to prove themselves. Despite the attention that GearScore gets, then, it is only a symptom, not the cause, of the problem. The real problem is one that people on both sides of the debate are caught in.

While ultimately if we can solve this problem GearScore may lose its place as a tool of choice for selecting raid membership. But banning or breaking the addon doesn't do anything to the underlying problem that PUG raids are the most obvious avenue for new and casual players to experience raiding and are simultaneously the worst possible environmental for new and casual players to experience raiding.


1. Unless you are in a situation that are are capable of outright winning. When the entire problem has to do with opinions and social interactions it is very rare that this is an option.

2. Of course the arguments are often a lot closer to, "You asshole, you are ruining not only WoW, but life as well!" versus "Noob!" with one side or the other quickly resorting to, "You are a virgin who lives in your mom's basement." This is trade chat afterall.

Monday, November 15, 2010

I Can't Tank

Come on, make Discipline a tanking spec. What have we got to lose?

Friday, November 12, 2010

GearScore

If you read WoW blogs, or trade chat, it will have been impossible for you to avoid noticing that there are a lot of people out there who really, really don't like GearScore. It was only in the last month or so, when I read a post on the Pink Pigtail Inn that I really got to thinking about why GearScore had become a problem for the community.

So when I saw people complaining that GearScore was meaningless and that "Skill > Gear" I tended to side with the GearScore users. This was a visceral reaction to bad logic. Just because skill is more important that gear doesn't mean that gear doesn't do anything. Since some random guy starting a raid in trade chat has no idea whether some other random guy looking for a raid in trade chat is any good at all at playing the game, there is really no point in comparing the two. Both of the following things can be true at once:

I'd rather have a player in 251s than a player in 232s
I'd rather have a skilled player in 232s than a bad player in 251s

And for all of those "Skill > Gear" people, I would also rather have a bad player in 251s then a skilled player in 187 greens. Skill can't make up that deficit - for reasonable values of bad, at least.

What I was ignoring for a long time is that it's really not my problem. When I raid, I raid with my guild. If I for some reason want to do a PUG raid and they want me to have a certain GearScore to qualify, I have it. More generally, if they want me to have a certain achievement, or meet some other qualification then I have whatever they are looking for. No one has turned me down for a raid and I have little reason to think anyone ever will.

The fact that is isn't my problem, though, suggests that I should ask exactly whose problem it is. When I really got to thinking about this, I decided I should respond to Larisa's post that I linked above. I noticed my response was getting to long so I started writing my own blog post. Unfortunately that got a little too long as well - after some editing, I still have 25 pages on the subject1.

Because that's a little bit more than I would expect anyone to be interested in reading, I wanted to try to express some of the basic ideas I have on the subject more succinctly through a few blog posts. Naturally, I am aware of the fact that very few people actually read my blog, so I don't hit a much larger audience by moving from 25-page-internet-essay format to blog format, but still.

As I said, the key to understanding the problem with GearScore is really understanding who it is a problem for and why we should care. GearScore - and other related problems, like requesting people link an achievement for killing the boss before they are allowed to attend a raid to kill that boss - is a problem for new and casual raiders. It is a barrier to entry for people who want to check raiding out or who like raiding but who don't have a large number of hours to devote to raiding each week.

Just as importantly, I think we should realize that GearScore and achievements are not the problems in themselves. The problem stems from the current climate for PUG raids, and GearScore and achievement requests are band-aids for that problem.

I want to talk about this because I think it is one of the most important issues facing the game. People can say all sorts of negative things about WoW and specifically about the direction the game is heading in, but I think the fundamental answer to most of the complaints is the let people who don't like the game move on to other things and continue to make the game better. The experiment of trying to let everyone - or at least a lot of people - get involved in raiding is an interesting one, and one that was bound to create unforeseen problems. Finding solutions to these problems has a huge potential to improve the game for players of all levels of commitment and interest.


1. If you'd like to read it, it's here, though I wouldn't say it's really finished or polished.

Tuesday, November 9, 2010

Paying Attention

If you read some recent developer comments about the game, you find out some strange things. It seems that it is impossible to heal through bosses with priests and shamans since paladin and druids AoE heals are so much more powerful. It seems that until recently most tanks were getting two-shotted because bears were so ridiculously tough that when the bosses hit hard enough to threaten bears, they hit way too hard for the other tanks.

What's interesting to me about this is that they balanced the bosses this way. They balanced the bosses around the broken tank and the broken healers. How could this possibly happen without simultaneously noticing that the tank and healers in question were dramatically overpowered?

Obviously at this point they have noticed the problems and they are figuring out how they are going to fix them. Despite that, I worry that the developers do too much of their balancing by tweaking numbers up and down based on experience and not enough of it based on math. There is definitely room for both things - no matter how good you are at creating models of situations, boss fights are extremely complex. Models and math can only get you so far and you have to dive in at some point and see how it works and whether it can be beaten with real raids. But some rudimentary math well before you got to the point would give you things like maximum theoretical healer AoE throughput and how long they can afford to keep that up in level 359 gear. It can give you things like the health of a bear tank versus a warrior tank, the size of Savage Defense shields and the frequency that the bear puts them up.

Monday, November 8, 2010

Two Streams

There are two streams of dungeons and raids in WoW - normal and heroic. Unfortunately there aren't really two streams at all. There is one stream, and it cuts through both of the difficulties. The expected progression in cataclysm is to level to 85, then do normal dungeons to gear up for heroics, heroics to gear up for normal raids, normal raids to gear up for heroic raids and then heroic raids.

Being a novice RTS player, when I got Starcraft 2 I decided to play through the campaign on Normal difficulty. It turned out Normal was pretty easy and I bumped it up to Hard fairly quickly. Now that I'm better at playing the game I've done most of the missions on Brutal difficulty - mostly because I want the achievements. I didn't, however, play Normal to practice for Hard, or Hard to practice for Brutal. I played Normal because I thought it was the right difficulty level for me. Then I played Hard because I thought it was the right difficulty for me, and I had more fun doing it.

The idea of having more than one difficulty level is a good one because it lets different players play at their own paces. If you want to play on Casual, you can do that. If you want to play on Brutal, you can do that. If you want to use cheat codes to be invincible, those are conveniently built into the game. You play how you want to play. Imagine if in order to play Normal you had to play through Casual to unlock it. For experienced RTS players, especially those with a background in the original Starcraft, this would make the game needlessly tedious.

Unfortunately this is the path of progression in WoW. Bosses on normal difficulty don't offer the challenge that top-end players are looking for. Forcing top end players to play through on this difficulty creates lots of negative outcomes.

First, it increases the shared sense in the community that WoW is just easy. After all, you make the best players do things that are in fact easy for them.

Second, it puts people will greatly different expectations in terms of skill, gear and efficiency in groups together through the dungeon finder. Even top end players have to run heroics to maximize their ability to get top-tier gear. These players create a lot of friction in the dungeon finder by having expectations of speed and efficiency that the other players on their run don't necessarily want to live up to.

Third, and I think most importantly, it drastically reduces the excitement of defeating bosses. Part of the excitement of beating a boss is getting to move on to the next new thing. When you've already been forced to run the instance on an easier difficulty, there is no next new thing - you are just repeating what you've done before. When we beat Al'ar it was exciting to go on to Void Reaver. When we beat Opera, it was exciting to unlock the back door, climb the stairs and face Curator. When we beat heroic Marrowgar, we got to go fight Lady Deathwhisper, who we had already killed a dozen times.

The solution to these problems is simple. Segregate the difficulties and allow people to choose which one they want to play based on how they want to play the game. Normal instances gear you up for normal raids, heroic instances gear you up for heroic raids - as simple as that. Make the heroic raid precisely one tier more powerful than the normal raid gear so that people who run the heroic don't have to gear up for the next heroic tier in the next normal tier.

Having difficulty settings is a good way to let people choose how they play the game, but a bad way to build in progression. If your way of playing the game is my tedious stepping stone then instead of giving us different choices, the difficulties are forcing us to step on each other's toes.

Wednesday, November 3, 2010

Power Word: Barrier

We were promised Power Word: Barrier some time ago1 but it ended up being cut. I believe this was because they just weren't getting it to work the way they wanted to. At the time I remember thinking it was a very difficult spell to implement. Obviously one way to implement an AoE shield is just to dump an individual shield on each person in an area, but that doesn't really have the right feel to it.

The implementation they came up with fits the idea of standing under a big shield that you share with your friends. The shield has a certain amount of damage it can absorb in total, but also a limit per hit so that if one person takes a massive hit they don't hog the whole shield.

The problem with this implementation is what happens when you cast it just one person. I'll use level 80 examples because they are more in our familiar range of understanding at this point. If you have around 4000 spell power then the shield can take about 51.6k, with 10.3k being the maximum per hit. If you cast that in front of the boss where the tank is standing then the tank gets 10.3k off every hit for the next five hits before the shield collapses. If a boss is swinging for 25k damage, this is basically Pain Suppression. If a boss is swinging for 35k then it's only 28% damage reduction, but that's an awful lot. What's important, though, is that it is a 52k instant cast heal on a single target in 10.3k installments. Think of it as a stupidly overpowered Renew.

Compare this to its AoE use. Suppose you are fighting the 10 player Sindragosa encounter. Her frost aura does 3750 damage every three seconds - and will be reduced by almost exactly 20% by resists. If non-tanks have an average of three stacks of mystic buffet, and they are gathered up under a shield, that means the shield will be block two ticks of the aura. This is not really that impressive. In fact, considering the positional requirements, it is arguable that Prayer of Mending is equally powerful or more powerful for this application, and that has a 10 second cooldown instead of a 2 minute cooldown.

For single massive AoE blasts there is a good chance that Power Word: Barrier will be inordinately effective. This is because, most likely, each person under the barrier will benefit from the full reduction of the barrier from a single hit even if that total overwhelms the total absorb of the barrier. That is, if your entire 25-player raid is standing under the barrier when the Lich King casts infest, it probably blocks 10.3k from each person, despite the fact that this means it absorbed over four times its maximum amount. This would be consistent with the behaviour of Anti-Magic Field, the spell that Power Word: Barrier was clearly based on. If the code does not have this mistake then it would basically be a 2-minute cooldown double strength Prayer of Healing.

Power Word: Barrier is good for two things: 1) For use as an external damage reduction cooldown on an individual who is standing in an isolated location or who is not standing near anyone else who will be taking damage; and 2) To take the edge off a powerful AoE attack that hits the entire raid when the raid can afford to stand together in preparation for the attack assuming that the way the coding works allows it to absorb far more than is intended.

When they announced Power Word: Barrier I really wasn't sure what kind of mechanics they could use to make it work. Clearly, they want it to have the thematic idea of sharing the shield between the people under it rather than giving each person their own shield. That requires a very large shield to be useful, but they certainly don't want it to be a super single-target shield.

These are some very significant problems to overcome. Unfortunately the design they settled on does not overcome them. The most likely outcome of Power Word: Barrier is that discipline priests will be seen as the external tanking cooldown healer - and will quite possibly be overpowered for this reason.


1. I believe they first announced it for patch 3.1, but it might have been 3.2

Monday, November 1, 2010

Resilience

The 4.0.1 problems with PvP and the attempt to solve them really highlighted some of the problems with Resilience. When 4.0.1 came out there was a big problem with people getting one-shotted or nearly one-shotted in battlegrounds. I forgot to put on my PvP gear in one Arathi Basin battle and took an Arcane Blast critical for around 28k. Apparently some ret paladins were doing single hits of up to 20k.

Obviously people being able to strip off half your health with a single attack in PvP is a big problem. On the other hand, people being totally unable to do any damage is a problem as well. Their initial reaction to the problem was to increase the effect of Resilience by 50%, so my 40%-ish damage reduction increased to 60%-ish damage reduction. Now that 28k Arcane Blast crit would only be 11k. Without the critical it would be a measly 5.5k.

In fact, I was usually pretty much invincible. I would heal myself through three people attacking me, despite stuns, silences and fears. Of course Blizzard noticed that this resilience chance was too much, so they lowered the effect of Resilience it by 25%1.

But when I go into battlegrounds, I'm still ridiculously tough. Other people on my team still get one-shotted. Why is this happening? It's because PvP balance comes from a single stat that starts at zero and gets better the more points you have of it.

Blizzard has put a lot of formulae in the game - some more complex than others - to try to prevent stats from getting better the more you have. For armor, the damage reduction formula means that armor expands your effective health linearly. For dodge and parry the diminishing returns formula does not quite come up with a linear result, but it was put into place to avoid these stats from becoming too powerful when you stack too much of them. Resilience has no such diminishing returns, it's power is just left to expand unchecked.

The math behind this is really simple. If you get a certain amount of resilience you take 1% less damage from your enemies. So if you started with zero, then you would now be taking 99% damage instead of 100%. But if you started with enough for a 40% reduction, you'd now be taking 59% instead of 60%. The thing is that while in one sense 59% is 1% less than 60%, the actual amount of damage you are taking with the extra resilience is not 1% less than what you were taking before. 59 divided by 60 is 0.983, so you are taking about 1.7% less damage then you were before you added that resilience. Resilience is 70% better when you already have enough for a 40% reduction than it was when you had none at all in terms of helping you live longer.

Making matters worse is that, as shown above, they like to fix PvP problems by adjusting the amount of damage reduction resilience gives. When they increased the value of resilience by 50%, the moved me from 40% to 60% resilience. I now took 40% instead of 60% of the damage my opponent's did. I was taking two-thirds of the damage, and my time to die from the same pressure by 50%.

A very well geared PvPer can get closer to 50% resilience. Increased that by 50% makes it 75%. So they take 25% instead of 50% of the damage they would have without it. They take half damage, and their time to die is doubled. Buffing resilience may make PvP more survivable on the whole, but it also drastically increases the disparity in power levels between the best and worst geared PvPers.

On top of this, so much PvP balance being dependent on resilience really messes up early expansion PvP. There were a lot of complaints about the first arena season in Wrath because people died so fast there was no time to have a good fight. The response from Blizzard was that we should wait until people get some resilience. There was a little more to it than that - people had easy access to relatively high level PvE gear via Naxxramas and much more limited access to PvP gear - but the increasing returns nature of resilience will always mean that PvP works without a certain range of resilience values and breaks down outside that range.

The incredibly high coefficient on resilience in Cataclysm will probably end up keeping resilience damage reduction values very low - unless they decide to improve them, damage reduction from resilience shouldn't get to even 20%, let alone 50%. Because of this, the problems with resilience will be much less pronounced in Cataclysm than they were in Wrath. But bad math continues to be bad math even when numbers close to zero make it not matter much.

The resilience formula should look just like the armor formula so that effective health in PvP expands linearly - rather than asymptotically - with resilience.


1. That is, they lowered it by 25% of the new value, which was 150% of the old value, so it was now 112.5% of the old value. Despite the fact that they gave an example of what they meant, I had to read their post a couple of times to figure out what they had done. Changing current values by percentages is quite uninformative.

Thursday, October 28, 2010

Percentage Changes

There was a patch to the beta today where they adjusted the strength of a variety of spells, mostly modifying only the base value, not the scaling. I'm not exactly going to complain about that. After all, they understand quite well that base values and scaling are two very different things and what it means to adjust them.

On the other hand, some the specific wording of their notes was a little weird:

Divine Light base healing has been increased by 30%. From 8538-9512 to 11100-12366

This is actually an improvement from how it used to be. They used to just list the percentage damage and not the damage numbers.

The thing is, that a percentage change in base healing or damage of an ability conveys basically no information. If a base amount increases 20%, that could mean anything from a 0% to an 20% increase in the actual output of the spell.

I'm not as much worried about the information I get, because I can go look things up if I don't know all the facts. What I'm more worried about is what this says about how they think about these things. When paladins aren't healing enough, do they think to themselves, "We'd better increase their base healing numbers by 30%"? Given they increased the base of the other direct heals by 30% that certainly appears to be exactly what they think. But I really don't understand how this train of thought works.

Each paladin healing spell gets a different proportion of its healing from the base amount compared to the spell coefficient, so they increased the power of some of these spells relative to one another. They also changed the relative value of critical strike, haste and mastery rating relative to intellect.

Obviously the concern was they wanted paladins to do more healing without increasing their scaling with spell power. This makes sense in a way, given that apparently paladins didn't heal for enough despite the fact that at a glance it looks like they have the highest scaling with intellect on direct heals. It just strikes me as very odd that they would choose a certain percentage and increase all the base amounts by that much. It looks a lot like a stab in the dark rather than the results of an actual analysis of how much more healing paladins should be doing.