Tuesday, January 31, 2012

Critical Error

There's a lot of misunderstanding about the value of critical strike around the healing community that I see. It isn't that critical strike deserves to be the stat that you stack or even prioritize, but it gets an unreasonably bad rap for all the wrong reasons...just like you or I did as a teenager. We can't do anything about our teenage years (unless you're still in them), but we can ease the social awkwardness of poor critical strike.


Too Random At Any Speed


This is, by far, the most common reason/argument that I see healers use in advising other healers to avoid critical strike. The syllogism goes:

  1. Critical strike is RNG
  2. RNG is bad
  3. Therefore, critical strike is bad

For those unaware of the term "RNG," it stands for "random number generator" and it's used as a proxy for anything in the game system that has an element of chance to it.

Obviously, the conclusion in (3) is correct if we accept the premises of (1) and (2). So let's take a look at each of these premises and see if they hold up to focused scrutiny. I'm going to examine them in reverse order.

RNG is bad

A very common belief is that healers depend on reliability. DPS can exist in the chaotic maelstrom that is RNG, but no player tasked with keeping the raid alive can afford the luxury of rolling dice.

Is that true, though? I was talking about this with a friend of mine who tanks (Schriko of Ebon Plaguebringer) and he provided the perfect comparison point: tank avoidance stats. Tanks are no less tasked with keeping the raid alive (they are, in many ways, the choke point of the raid's jug of health...lose that stopper--the tank dies--and the rest of the raid's health is going to come pouring out soon after) and should presumably also be concerned with "reliability" over all else, if reliability was the be-all and end-all for those who's failures mean the raid dies.

However, no tanking guide suggests ignoring dodge and parry, the two "RNG" avoidance stats for tanks. Acquiring a fairly significant amount of each is an important thing for tanks to do (well, not parry so much for bear tanks, for obvious reasons!), even long before they can reach full combat table coverage.

The reason is that raid encounters are fairly long combat encounters. A tank will be exposed to a huge number of boss or add swings over the course of the encounter. The larger the sample size (that is, the number of opportunities to dodge or parry) the lower the volatility of the randomness. That is to say, if you have a 20% chance to dodge and only one chance to dodge, you're looking at an 80% chance to get exactly zero benefit from that dodge chance. However, if you have a million chances to dodge, you have so many chances to dodge that you'll almost certainly see right around 20% of those million chances dodged which means you reduced your damage over those million swings by a fifth, or 20% (assuming, for this very basic example, that all the swings were equal in damage done).

The same holds true for healers or DPS. The more spells you cast, the more chances for any type of RNG (critical strike, weapon or trinket procs, etc) to stabilize around its numerical value and play out like a static value (as the 20% dodge plays out like a 20% damage decrease over a long combat).

Critical strike is RNG

At its base, critical strike is clearly RNG...the value denotes the chance for a heal (or damage, but I'm mostly focusing on healers here) of double the value.

That said, there are some things that you should keep in mind when considering critical strike as a healer.

The first thing is that some healing class-specs have a much, much greater number of opportunities to generate critical strikes. For example, let's look at holy priests. The majority of their healing comes from Prayer of Healing, Circle of Healing and (if the raid can spend a fairly large amount of time stacked up) Holy Word: Sanctuary.

Prayer of Healing can hit five players at a time. Circle of Healing can hit six, glyphed. Holy Word: Sanctuary can hit as many people as stand in it, so up to twenty five people in a raid setting.

That means that on each Prayer of Healing cast, you have five chances to critically heal, six chances with Circle of Healing and an indeterminate but high number of chances with Holy Word: Sanctuary (assuming you're making smart decisions about when to use it).

This stands starkly in contrast to a healer (or DPS) who casts lots of single-target spells. Those players can still generally count on critical strike largely smoothing out over an entire combat, but are more volatile from cast to cast. A holy priest really isn't, because their model is to cast, in essence, a ton of small heals...which greatly expands their sample size even on a per-cast basis. For a holy priest, it's not all-or-nothing even on a per-cast basis.

Returning to the tank avoidance stat analogy for a moment, if there were a type of tank that had the ability to break each boss swing into seven smaller pieces of damage, each of which could be independently dodged or parried, then dodge or parry would be even less volatile for the tank. That's the tanking equivalent of healers who cast a lot of area of effect (AoE) heals (even if such a tank doesn't currently exist...but maybe it should!).

This, of course, is not limited to holy priests. Restoration druids also have a lower volatility per cast, due in part to their AoE heal (Wild Growth) but also due to the fact that the ticks of nearly all heals-over-time (HoTs) can critically strike. HoTs, like AoE heals, also subscribe to the model of smaller but more numerous heals, which substantially reduces the volatility per cast.

Restoration shamans, with Chain Heal and Healing Rain, fall into this boat if they're primarily raid healing. Holy paladins have been pushed into this boat by the redesign of Holy Radiance (and its resultant synergy with Light of Dawn).

To be perfectly honest, discipline priests can end up here, too, if they find themselves spamming Prayer of Healing as raid healers, which is a fairly common style for discipline priests.


Critical Conclusion


Let me clarify what I'm saying here. I am not remotely saying that critical strike is the best stat for each of these class-specs. What I am saying is that critical strike is not a poor choice due to its "unreliable RNG aspect." Critical strike tends to stabilize over raid encounters that aren't extremely short and spec or playstyle can even stabilize critical strike on a per-cast basis.

Looking at critical strike as an unreliable stat that makes healing raid encounters as chancy as rolling dice is simply the wrong way to approach analyzing our secondary stats and gearing.

There are, certainly, other factors to consider* in critical strike but unreliability is really not one of them.

The idea of this post is not to make a full review of critical strike as a stat, but rather to strenuously object to one of the most common philosophical arguments put forth within the healing community about critical strike.


*Factors To Consider Not Expanded On In This Post
  • The cost, in rating, to get a percentage of the various secondary stat
  • The talents that are affected by critical strike
  • The spells you use that are affected by critical strike
  • The effect on mana that the various secondary stats exert
  • The value other secondary stats have to your spell usage (opportunity cost of critical strike)

Friday, January 20, 2012

Should Raid Nerfs Be Disappointing?

For those of us who enjoy raiding as a significant part of our World of Warcraft experience, "nerfs" (the weakening of encounters) to raids that are still current content tend to be a hot topic. Most of the (visible) response tends to be negative, with an even more vocal minority of that group moving toward the wailing and rending of garments.

Even the more moderate, though, generally are a bit disappointed when the raid encounters they're working on are made easier. Should we be disappointed, though?


"It's Too Easy Now"


What makes an encounter too easy? I can guarantee that if you weren't finding yourself a few percent short of getting the last tendon down on heroic Spine of Deathwing once January 31st rolls around, you're not suddenly going to find yourself blasting through the remaining content on February 1st. Challenge still abounds.

The fact of the matter is that the "correct" difficulty level of an encounter is entirely arbitrary. Had the raid encounters been released back in December with 5% less health and 5% less damage done, would that have been the "right" difficulty level? While a few more people may have been agitating from the start that the raids were too easy, the vast majority of those who are upset now would never have been upset at all, because they'd never have had the higher figures set in their minds as real and proper.

Increasing the difficulty of a game with a computer opponent where the rules are set by the game itself is trivially easy. You can always increase the power or mechanics of the computer opponent  until it is impossible to complete. So clearly the answer to what should be the correct level of difficulty is not "As high as possible." As high as possible is always unwinnable.

So where do difficulty levels come from, in gaming in general? The simplest answer goes to accessibility. Games manufacturers are doing what they do in order to make money. That means selling a game to the widest possible audience. There are two competing dynamics there.

At the most basic level, the largest possible audience is simply anyone that can fire the game up...it includes basically every computer or console user.

However, a game that only involves starting up obviously wouldn't be compelling to anyone. Therefore, it needs to have enough challenge that those who play it find it fun. If it is too challenging, such that it is impossible for any human to complete, then, again, almost no one will find it compelling.

So game designers have to create difficulty levels that are hard enough to be compelling, to make players feel like they're acquiring a skill and being forced to improve it, while being easy enough that a sufficient number of players can be successful.

The operative term and variable, there is "sufficient." For every designer, that's going to be different. No designer can aim for 0% of the game-playing population and 100% is probably untenable as well. Where you draw the line, in terms of percentage of population being able to complete the game, is arbitrary from a game design standpoint. Ultimately, it's probably going to be based on some complicated linear algebraic model determining what difficulty level sells the most games.

Raids in World of Warcraft are no different. Blizzard has a figure in mind for how many raid teams should be able to complete the raid, on normal and on heroic (different figures for each, obviously). The release difficulty is their best guess as to what will hit those figures.

But their best guess may be wrong.

When that happens, when their internally-tracked figures are showing that not enough raid teams are meeting the challenge, it tells them that the raid is actually too hard. Not that it was the correct difficulty and now they need to provide some charity to all of these fail raiding guilds.

"Too hard?" "Correct difficulty?" Based on what, again? Based on the same calculus of game difficulty that every game company works from: that enough people can eventually meet the challenge without it being so easy that it's not even challenging to do.

What does that mean? It means that the nerfed raid difficulty levels arriving on January 31st are the "correct" difficulty...by definition. That definition being the only way game difficulties are ever chosen from the arbitrary spectrum that ranges from "success just by loading the game" to "unwinnable by any human."

By choosing to play World of Warcraft, by choosing to raid in World of Warcraft, by choosing to play any game, you are entering into an implicit agreement that the correct difficulty level is chosen by the game company producing it. It can't be any other way. There is no "platonic game difficulty level" that a game company could use.

So if you've defined the correct level as whatever Blizzard defines it to be, and Blizzard (like all other game designers) has defined it to be "whatever allows X% of players to win the game" then Blizzard nerfing (or buffing) raid encounters is not reducing it from its "correct level." It's Blizzard tuning it to its correct level, because they got it wrong at release.


This Isn't New, It's The History Of Video Games


Video games largely find their genesis in the cabinet-style arcade games of the 1970s and 1980s. Pong is probably the best-known arcade pioneer, an incredibly simple video game that was meant to simulate table tennis. Through the 1970s and 1980s, and into the 1990s, arcade games exploded in popularity (while consoles and video game computers that were birthed from that success had a more up-and-down development curve), with such iconic hits as Missile Command, Pac-Man and Space Invaders. Double Dragon. Street Fighter II. (I forgot how much I missed arcade games. Or maybe I miss youth. Wow, this got depressing fast.)

What is probably less well-known about arcade games is that many of them had a feedback loop to dynamically regulate difficulty levels.

The game would spend a certain amount of time evaluating how well the player was doing, how quickly the player was progressing and then amp up or amp down the difficulty level to better match the player. After all, if a game was under-tuned or over-tuned, players wouldn't play it and you couldn't "hotfix" ten thousand or a hundred thousand cabinet games scattered around the country by releasing a patch over the ARPANET.

So what was the correct difficulty level? The correct difficulty level was just slightly above the player's skill level, which forced him or her to continually improve or face dying over and over at a certain point. That's perfect from both sides...the player has something to continually strive for without facing a brick wall that he or she can't get past, and the game producer has a steady stream of quarters it's earning as the player keeps dying in the process of improving. Everyone's got what they wanted.

Blizzard is doing essentially the same thing with these nerfs, just writ large. The period of polling or evaluating player skill is two or three months (instead of, say, 30 seconds). Instead of evaluating a single player, it's evaluating a playerbase. The game is then adjusted (sometimes with hotfixes, sometimes with a patch) and another stable evaluation period occurs.

In retrospect, are you disappointed that perhaps some of your Pac-Man scores were illegitimized by nerfs? (Of course, I'm dating myself a bit by implying that I have scores on Pac-Man cabinet games...many of you may not be old enough for that. That said, I'm sure you understand my point here. Also, this parenthetical comment has gotten really long.)

Dynamic retuning of a game that is in progress (and current content raids can be considered "in progress" for as long as you're doing them this tier) is not at all new. In fact, it's quite standard practice. Going hand-in-hand with the fact that are no platonic difficulty levels for games is the belief that everyone is best off by making sure that the designer's vision of where accessibility and challenge intersect is continually being enforced.

So instead of being concerned or annoyed that Blizzard has decided you need a booster seat to get to the creamy desserts in the middle of the table, it's probably more appropriate to view it as Blizzard deciding they made the table far too large and that only the people with freakishly long arms are getting enough dessert and that they didn't intend the desserts to only be for freaks.

And remember that the "desserts are meant for more than just the long-armed freaks" principle is solidly entrenched in video game history...gamers are happier and, therefore, game companies are wealthier when games are continually redesigned to better match the original vision.

Monday, January 16, 2012

Dr. Strangebubble (Or How I Learned To Stop Worrying And Spam The Shields)

"Shield spam" has become a term of ill-repute in the discipline priest community. It's spoken of either in whispers about the "bad old days" or else in the sharp, mocking tones of what not to do. Only the most degenerate discipline priests would encourage shield spam. It speaks to loose healing morals and, yes, even a bit of healing perversion. Shields are to be enjoyed in small quantities, not to gluttonous excess.

That is why I'm here to tell you to reject those stereotypes. Even good-hearted people can blast out bubbles. It's not the evil that you've all heard it is. It can--nay, should--be done. Don't listen to your neighborhood priest telling you that it's sinful.

That said, we should probably define what we mean by "shield spam" because multiple behaviors have been lumped into that one term and they are not all the same.

Let's start by examining where the term really sprang into prominence and trace our story from there. This, it turns out, may run a bit long but I promise there's a point at the end. And a pizza party. Pertinent point and pizza. And cold beer.

Come with me, then, back to those heady days of Wrath...


Spammy Shield Spam


Welcome to Dalaran! Do you have any emblems of frost to turn in? And how about those vrykul, eh?

I hope I've adequately set the mood. Back in this era of World of Warcraft, discipline priests were slowly but surely rising in the world of healing. In prior eras, discipline was really more of the PvP spec and holy was the healing spec for priests. However, the tree was revamped in Wrath and PvE discipline priests were becoming more and more populous.

Oh, sure, there was still residual persecution that they faced. They were regarded with suspicion, not being holy. Absorbs were a bit of a new mechanic to much of the playerbase and no one really paid attention to absorb values, which made discipline priests appear to be rather ineffectual, what with Power Word: Shield and Divine Aegis being completely ignored by almost all healing meters.

However, as raiders were graduating from Trial of the Crusader to the final raiding tier of the expansion, Icecrown Citadel, people were understanding the power of the absorb. High-level players had understood it long before, but it gained popular acceptance. Now, discipline priests were not just being accepted but often actively recruited as they carved out their own niche. Discipline priests weren't healers, they were shielders! You could almost think of them as a huge damage debuff on each and every boss.

This was a double-edged sword. It made discipline priests a rather necessary part of any raiding team capable of picking and choosing their raid compositions, as much so as getting a shaman in for Bloodlust or having a druid who could Rebirth. Having such a valuable niche in raid teams is obviously a nice thing...damned are the poor souls who play class-specs marginalized by current raid design.

On the other hand, it was a terribly boring playstyle, especially in 25 man raids. In 25 man raids, very literally a discipline priest's job during an encounter was to do nothing but blanket the raid in shields. And then do it again. And again. Until the boss was dead. Other healers were there to do the actual healing, after all. In 10 man raids, it was a little better, purely because there weren't enough people to shield before the Weakened Soul debuff prevented you from casting more shields. At that point, it was reasonable to cast an actual healing spell.

So discipline priests were in a bit of a bind...the niche that gave them ready access to raid spots also trapped them into a rather unenjoyable style of playing. (This wasn't purely confined to discipline priests, as I know there are restoration druids out there who have nightmare stories of being Rejuvenation bots. But I think it was most pronounced for discipline priests.)


Our Story Runs Into A Twist


Blizzard agreed that it was a trap. Their design philosophy had been slowly attempting to weed out repetitive, or One-Button, play. Discipline priests sometimes using nothing but one spell in raids didn't escape their notice. So they changed it. So mighty was their change to the discipline priest spec that they named an entire expansion for it: The Cataclysm.

Perhaps you heard about it? The change resulted in many fires around the world.

What they did was make Power Word: Shield really weak. It was brilliant. Why would you spam a spell that wasn't very strong? You wouldn't. That Blizzard, always thinking.

Well, as it happened, trying to reduce a spec's dependence upon their best and most iconic spell by making it terrible didn't do wonders for balance, that nasty issue that boils down to "Do you get to raid or not?" Discipline priests were pariahs again. Go holy or go home. The chants were hurtful.

So Blizzard did what any wise design team would do. Turn the dial the other direction as far as it would go. Suddenly, Power Word: Shield was really strong. Discipline priests were great again...in fact, they were recording by far the best parses overnight. No other class could compete.

"Hmm," said Blizzard. "That was unexpected." Also, there was this issue of discipline priests spamming their newly god-mode spell on everyone again. Also unexpected.

So now Blizzard needed a solution that did not involve turning that dial at all. Turning the dial on Power Word: Shield, you see, revealed a problem inherent in WoW balance. A spell you can use whenever you want will either be strong enough that you will use it first, last and always or else so weak that you will never use it. For if it is the strongest spell right now, it will probably be your strongest spell a moment from now. And if it is not your strongest spell right now, it probably won't be your strongest spell a moment from now.

The answer resided in the other dial that Blizzard had ignored in Wrath (which caused that previous logic about "strongest spell now->strongest spell always" to be binding in nearly every case) but used with abandon in Cataclysm...the mana cost dial. True to the Cataclysm healing model, which held that you don't simply min/max based on power of spell but min/max such that you don't go out of mana in the first minute, they increased the mana cost of Power Word: Shield greatly.

That changed the game significantly for discipline priests. The shield was still really strong, and it gave mana back every 12 seconds, so it was still useable...but "shield spam was dead for good" (I place that in quotation marks because it was probably the most common quote of the past year, it even featured on the cover of a Time magazine).

This was driven home with a peculiar sternness on forums and guides for discipline priests. Perhaps the people saying it felt it was necessary to say it with passion to break the dirty habits people had built up, but a great deal of scorn and derision was heaped upon the concept of shield spam (in my experience, at least). Not that it really required stern insistence...any discipline priest that tried to spam shields during a raid encounter, Wrath-style, would find themselves as merely a decorative prop pretty quickly, completely out of mana (OOM) and therefore unable to help much. This was a lesson as subtle as a jackhammer to the head.

And lo...it is still impossible. Unlike the latter stages of Wrath, where no spell could really run you out of mana, casting shields non-stop on a raid group, even in 10 man, will solidly OOM you.


So What's All This Talk Of Shield Spam, Then?


The problem is, "shield spam," which was originally applied to casting shields on each raid member robotically, over and over, is now being applied to any behavior involving using lots of shields.

The fact of the matter is that absorption is still the healing space of discipline priests. That is where you can add the most value while still allowing all other healers to do what they do best. While holy paladins have entered the space to some extent (to say nothing of the abominations known as blood death knights), it's still an important niche that discipline priests largely dominate. Damage prevented or mitigated is more valuable than damage taken and then healed up (the easiest way to see this is to compare the value of absorbing a 30k hit to a player with 24k health remaining, versus attempting to heal it up after).

While Divine Aegis does yeoman's work in terms of absorption, there's no faster way to get on-demand absorption out on the raid than what you might call "spamming" shields.

The key here, of course, is that you must choose when it makes sense to spam those shields. Your mana won't support doing it all the time. And the mechanics of bosses (along with a duration lowering change that I neglected to mention earlier) are such that ill-timed shield spam will come at great mana cost for no benefit as the shields mostly drop off due to expired duration without absorbing any damage at all.

So you absolutely have to pick your spots intelligently. However, if you do, there's a world of benefit to be mined from strategic shield spamming. Getting ten shields out ahead of a big raid-wide damage spike can effectively increase your raid team's "clutch heals per second (HPS)" by which I mean that if you're using GCDs during a low-stress period to put up shields that will be eaten by a high-stress period, it frees you up to cast other spells during that high-stress period. You're borrowing time from a low-stress period to lessen the difficulty of a high-stress period, in a way other healers largely can't. Absorption shields are about the only "preventative healing" (though HoTs can be used in a somewhat similar way, though less efficiently for the same purpose).

You won't be able to shield to the exclusion of all else, even if you pick your spots. There are times, especially in 10 man raiding environments, where helping your fellow healers move life bars will be necessary. However, the more you bias your casting patterns in favor of Power Word: Shield, the more you can help your raid team even out the spikiness of the overall incoming damage of the encounter. It also has the benefit of clarifying your stat choices for gemming, enchanting and reforging. If you know you will be using as much of your mana as possible (over the entire encounter) on Power Word: Shield, then you know mastery is your best bet.

This is Smart Shield Spam(tm). This is good. This is a perfectly responsible way to play discipline priest. There are other perfectly responsible ways to play a discipline priest, but don't let others lazily say "Shield spam is bad, mmmkay?" Make sure they define what they mean by shield spam...if they're talking about Wrath-style bubble botting, nod at the obviousness of their observation. Yes, you can't do something that you don't have the mana for. However, if they're advocating shielding as little as possible (some even claim you should largely only be shielding the tanks or only shielding to proc Rapture) then reject that supposed truism. Intelligently spamming shields at the right times can be an extremely powerful way to play (I would argue the most powerful way to play)...you should experiment and see how far you can stretch Power Word: Shield, especially as you push into progression fights that stretch you to your very limits.