Friday, July 23, 2010

Can We Prioritize Shield of Righteousness?

I've been claiming that forcing us to hit Crusader Strike (CS) every third GCD will make our other, signature tankadin spells feel like filler that we hit when we have a spare second rather than cool strikes of awesomeness that we want to hit on CD. To back this up, suppose (as is plausible) that we will prioritize Shield of Righteousness (ShoR) over all other non-holy power related spells. Then there are two possible places we can fit ShoR into a CS rotation: before a cast of CS, or after (there are only two GCDs between CS). In the following diagrams, an * denotes a CD conflict between CS and ShoR.

ShoR before CS:
ShoR | CS | ?? | ?? | CS*

ShoR after CS:
CS | ShoR | ?? | CS | ?? | ShoR | CS | ?? | ?? | CS*

However, there is a a tiny amount of flexibility in our CS rotation, we don't absolutely need to hit it every third GCD to keep holy shield up. We can occasionally resolve CD-clashes in favor of ShoR while still keeping HS up. If we push back CS in the first example, we get:

ShoR | CS | ?? | ?? | ShoR* | CS | ?? | ?? | CS*

Basically, any time we resolve a ShoR/CS CD-clash in favor of ShoR we set ourselves up for another CD-clash as soon as ShoR comes of CD. If we always resolve in favor of ShoR, we fail at generating enough HP to keep Holy Shield up and maximized.

As things stand in the current build, there is no rotation that

(1) Maximizes HS uptime & mitigation;
(2) Always prioritizes ShoR over CS.

Thursday, July 22, 2010

Protection Paladin Rotation Revisited

Today's new beta build proves that the developers are taking the protection paladin rotation in a very different direction than I imagined. First, the design of Holy Shield has changed again. It's now got a fixed 15s duration and it's possible to refresh immediately, for 100% uptime. But holy power now affects the magnitude of the block chance effect; you get 5% block chance per point of holy power. Moreover, Holy Shield is now exclusive with Inquisition; you can't have both buffs at once.

This means that, like the 969 rotation of WotLK fame, our Cataclysm rotation will be designed around maximizing Holy Shield uptime and effect. While there may be situations in which we prefer to spend holy power on Inquisition, those situations will not be raid tanking situations. If we need Inquisition to be competitive in generating threat compared to other tanks, this will be a reason to use those other tanks, who don't need to make the same survival for threat trade-off.

This means that we need to generate 3 holy power every 15s. Essentially, this forces us to cast Crusader Strike on cooldown. Well, not literally on cooldown if the cooldown remains 4s, but every third global cooldown, i.e., every 4.5s.

The protection paladin rotation will then look like this, I fear:

1. Holy Shield as soon as holy power = 3, and then every 15s after that.
2. Crusader Strike every third global cooldown.
3. Roll face.*

I find this disheartening. I was looking forward to a priority queue in which we needed to generate holy power, but this was largely taken care of by procs and other random events that would change up a rotation that otherwise consisted of the abilities that protection paladins know and love. Instead, it seems we will need to prioritize Crusader Strike in order to ensure maximum Holy Shield effect and uptime, which means de-prioritizing all of protections cool, role-defining spells like Shield of Righteousness, Hammer of the Righteous, and Avenger's Shield, making them into "filler" spells that you cast when you have a spare second, rather than "awesome" spells that you want to cast on cooldown. This is not exciting, fun, or dynamic in my opinion.

As I conclude in a post I made over on the official forums:

Tacking holy power onto CS makes it mandatory, but it doesn't make it fun for prot. This is the fundamental mistake I feel that the developers are making with the current prot rotation design.

*For those readers who may not be familiar with this idiom; the retribution paladin spell rotation in WotLK was often described as "faceroll," the idea being that, since it made only a small difference which spell you prioritized casting if two spells ended their cooldown at the same time, you might as well just roll your face on the keyboard to get a perfectly adequate rotation, rather than actively making choices.

Wednesday, July 21, 2010

Assessing Blizzard's intentions for the Protection Paladin Rotation in Cataclysm

The Cataclysm beta is, naturally, still in a state of flux, but I think the current build has enough information in it that we can infer what Blizzard's design intentions are for the protection paladin's single-target tanking rotation in Cataclysm.

The first things to notice are the importance of two medium-duration buffs: Inquisition and Holy Shield. Holy Shield has been changed from it's current, 100% uptime implementation to have a 1 minute cooldown with a variable duration depending on holy power: 10s base, +10s for each stack of holy power consumed, for a maximum possible duration of 40s. Inquisition is a percentage increase to holy damage, which means a big buff to protection's threat output, with a variable duration of 10s per holy power and no cooldown. These will likely be our top priorities. We will want to get as close to 100% uptime on Inquisition as possible, and we will likely also want to hit Holy Shield every time its cooldown finishes, but only when we have 3 holy power.

This has significant implications for our rotation, because it means that we need to generate holy power at a rate of 9 per minute. That comes to one every 6⅔s, or every 4-5 global cooldowns (GCDs). Since Crusader Strike (CS) is our only attack that generates holy power, that means it'll have to fit into our rotation, despite being physical damage and so suboptimal for threat (since it powers Inquisition, it does indirectly provide holy damage, in addition to seal procs). The new Grand Crusader talent gives Shield of Righteousness (ShoR) a chance to end the cooldown on CS and give it a significant buff to damage, which means we will want to hit CS in conjunction with Grand Crusader procs whenever possible to generate holy power. Since it's only a 20% chance and ShoR has a 6s cooldown, we will have to use some non-buffed Crusader Strikes to generate sufficient holy power.

ShoR, in addition to providing a buff to CS, will also likely be our hardest-hitting spell once more in Cataclysm, and so will probably be our highest priority spell once holy power is taken care of. The new Shields of the Templar talent makes our ShoR crits cause our next Judgement to automatically crit, Judgement has always hit relatively hard in our rotation, and of course Judgement is important for keeping our attack speed debuff up on our target, so Judgement will probably be next in priority.  It's possible the numbers will be such that Judgement will be higher priority than ShoR, but I imagine that this is not their design intent. (Evidence: we have a talent that makes ShoR crits buff Judgement, and if Judgement is too good, getting the cooldown reduction talent in the retribution tree may seem mandatory, which I don't think they want as it's not easy to get.)

Hammer of the Righteous (HotR) has had a 6s cooldown ever since being introduced, but for now it has no cooldown in the Cataclysm beta. This suggests that it is meant to be a low priority spell, used to fill GCDs when ShoR and Judgement are on cooldown and holy power is taken care of. It will probably be better for threat than CS if we don't need holy power, being holy damage.

Based on the above, it looks like protection paladins will want to rely on the following priority list for single-target threat in Cataclysm (with the caveat that things can certainly still change dramatically):

1. Make sure Inquisition is up
2. Holy Shield if holy power = 3
3. Shield of Righteousness
4. Judgement
5. CS if you have a Grand Crusader proc or if the duration of Inquisition or cooldown of Holy Shield is almost up and holy power < 3
6. Hammer of the Righteous

Of course there is some vagueness in this list, because of the "almost" in #5. You may want to start prioritizing CS over HotR with even about 12s left before you need holy power, if you have 0, because of the 4s cooldown on CS and possible cooldown conflicts with ShoR and Judgement.  But you also won't want to prioritize it immediately upon reaching 0 holy power, because ideally you'd like to use CS only when Grand Crusader procs.  This might be the "too much complexity" that Ghostcrawler is worried about.

A simple way to fix this to make protection's rotation less complex would be to make HotR generate holy power.  Since protection uses holy power only to maximize buff duration, this wouldn't be overpowered, and it would simplify things a lot.  They'd need to then give us another reason to hit CS occasionally, presumably through a proc mechanic (Grand Crusader might not be enough on its own).

On second thought, I think a more elegant solution would be to make Grand Crusader increase the holy power generation of CS from 1 to 3, in addition to or instead of buffing its damage done.  A 20% proc rate on ShoR, if ShoR is used on cooldown, would translate into an average of 2 Grand Crusader procs per minute, or an average rate of 6 holy power per minute.  This is not enough to both keep Inquisition up and use maximized Holy Shield on cooldown, but perhaps that would be a good thing from a design standpoint.  It would force protection paladins to make choices during battle: do you change up your rotation to work in extra, unbuffed Crusader Strikes, or do you forego the some of the threat bonus from Inquisition or some of the mitigation from increased Holy Shield duration?

Holy Power in Cataclysm

Last Friday's developer chat contained, big, big news about the paladin class.  Ghostcrawler had recently said
If we do one thing for class design in Cataclysm, it will be changing the paladin (all 3 specs) rotation up a bit!
He wasn't kidding! As a long-time player of the paladin class, I have often felt that the developers haven't had any great ideas about what to do with the class, and have thrown a bunch of band-aid fixes at it and second-hand abilities from other classes. We've never had an interesting core mechanic; for a long time protection and retribution have been just about watching cooldowns and that's about it.  While it was nice in Wrath of the Lich King that we finally got to be powerful, I can't say that we were the most interesting class to play.  I have stuck with it because for some reason it's what I chose on launch day, and I've enjoyed being along for the ride, trying to make the best of whatever they give me.  But it seems like the days of feeling like the afterthought class are over.  Cataclysm is giving us some interesting mechanics to call our own.
 Q: What is the goal when re-designing the paladin class? How do you plan to change rotations, talents, etc? 

A. All of the paladin specializations will make use of a new resource called Holy Power. Holy Power accumulates from using Crusader Strike, Holy Shock, and some other talents. Holy Power can be consumed to augment a variety of abilities, including: 

An instant mana-free heal: Word of Glory 

A buff to increase holy damage done: Inquisition 

A massive physical melee attack for Retribution paladins: Templar’s Verdict 

Holy Shield’s duration is now extended by Holy Power 

Divine Storm’s damage is now increased by Holy Power 

We also introduced several new heals for Holy Paladins including Healing Hands (an AoE heal-over-time that is applied to all players standing near the paladin), Light of Dawn (a cone heal with a 30-yard range), as well as a new heal called Divine Light, which is similar to a priest's Greater Heal, and the new instant heal mentioned above, Word of Glory. 


Q: Can you give us a sneak peek at one or two of the new Retribution abilities or talents? 

A. Templar’s Verdict: An instant weapon attack that causes a percentage of weapon damage. Consumes all applications of Holy Power to increase damage dealt:
1 Holy Power: 55% Weapon Damage
2 Holy Power: 125% Weapon Damage
3 Holy Power: 225% Weapon Damage

Word of Glory: Consumes all Holy Power to heal a friendly target for a specific amount per application of Holy Power (0 mana cost, 0 cooldown, instant cast).
As I said, this is big, big news.  People are already calling it a knockoff of the death knight's runic power and the rogue's combo points, but while there are similarities,  holy power is not the same as either of those mechanics.  More importantly, it promises to make combat for retribution and protection paladins much more dynamic, as we shall see in future posts exploring the likely rotations that will result from the new design plan for retribution and protection paladins. (I haven't given as much thought to how holy power will affect paladin healing, mostly because I haven't played a holy paladin in a long time.)

One thing that the developer's haven't said explicitly, but which seems like a safe inference from the information we have, is that we can have a maximum of 3 holy power at any time.  It's not clear how it decays if not used, but I'm guessing that it doesn't decay in combat, and it's possible it won't even decay out of combat.

Friday, May 7, 2010

How Good is the Tiny Abomination in a Jar?

I'm going to take a short break from the basic, introductory posts I've been writing to investigate a more advanced question that has been on my mind lately.  The Tiny Abomination in a Jar trinket is often said to be best-in-slot (BiS) for retribution paladins in the current endgame (patch 3.3.3).  In particular, a bit of googling will turn up this oft-quoted trinket ranking for retribution paladins:

Tiny Abomination in a Jar (heroic) > Tiny Abomination in a Jar Death's Verdict (heroic) > Whispering Fanged Skull (heroic) Death's Verdict > Deathbringer's Will (heroic) > Whispering Fanged Skull Herkuml War Token > Deathbringer's Will > Darkmoon Card: Greatness > Comet's Trail

Usually unmentioned is the fact that the ultimate source for this ranking is the Elitist Jerks forum post "Retribution BiS lists for 3.3."  The lists in that thread are derived entirely from Rawr.  Now, I use Rawr, and normally I trust it (although there are some issues with its calculations for the value of Haste Rating).  But in this case, there is not much reason to put a lot of faith in it.  Rawr does not accurately model the special effect of the Tiny Abomination in a Jar.  The author of the BiS lists posts has implemented a workaround for his purposes, but it is a very rough approximation and still not accurate enough to put a lot of faith in.  Still, no one really doubts that Tiny Abomination in a Jar is BiS for retribution paladins.  But I'd like to dig a little more into the justification of that claim.

I recently acquired a Tiny Abomination in a Jar, and got to use it in our last Icecrown Citadel 25-player (ICC25) raid.  Here is a link to the World of Logs combat log parse from our Festergut kill.  Now, my total effective dps for the fight is reported to be 12698.9.  However, that number seems to be inflated by damage from the boss ability Blighted Spores, which for some reason shows up in the breakdown of damage-by-spell for me.  That added 169870 to my overall damage done, so subtracting that from the total and dividing by the length of the fight (214s), we can see that my effective dps from my own attacks was 11964.4.

How much of that dps can be attributed to the special effect of Tiny Abomination in a Jar?  Fortunately for my purposes, I was already at the 8% hit cap before I got the trinket, and so the Hit Rating from the trinket was entirely wasted (I've since upgraded my gear so that I make use of some of it, but at the time of this parse I hadn't yet).  So we can neglect the Hit Rating contribution of the trinket entirely, which will simplify things.  The attack triggered by 8 Motes of Anger shows up in the combat log as Manifest Anger.  Over the course of the fight, I had 18 Manifest Anger hits and 19 critical hits (appropriate, given that my raid-buffed critical strike chance is ~50%), for a total damage of 101016.

However, that's only a small part of the damage attributable to the trinket; one of the primary reasons the Tiny Abomination in a Jar is so good for retribution paladins is that it also procs our active seal.  So some portion of the 703917 damage done by my Seal of Vengeance was caused by my trinket.  In particular, an average Seal of Vengeance hit during the course of that fight did 3519.6 damage.  So we can estimate that the 37 total Manifest Anger procs caused an additional 130224 damage, for a total damage of 231240 attributable to my Tiny Abomination in a Jar.  Notice that the damage from seal procs is actually more than the damage from Manifest Anger itself; thus, the damage from the trinket is more than doubled, and that explains why this trinket is so good for retribution paladins, but not best-in-slot for other melee dps classes.  Tiny Abomination in a Jar contributed an effective dps of about 1080.6 in the Festergut parse linked above.  Impressive, for a single piece of gear!

Now, to compare that to other trinkets, I intend to use Rawr, since Rawr should model the other trinkets in the list much more accurately.  However, since Rawr's estimate of my total dps is likely to be higher than it actually would be in practice, we won't want to compare dps attributable to the trinkets (even though, on such a comparison, Tiny Abomination in a Jar already comes out ahead).  Instead, I will compare the ratio of dps attributable to the trinket to overall dps.  So, in the Festergut log parse above, Tiny Abomination in a Jar accounted for 1080.6 / 11965.4 = .090310395, or roughly 9% of my overall dps.  Still quite impressive for a single piece of gear!

Now, let's compare the heroic version of Death's Verdict, our other BiS trinket.  According to Rawr, with no trinket in my second slot, I ought to be able to do 12121 dps on an undead target in a stand-and-deliver type fight like Festergut (with the Strength of Wrynn buff, which I had in the parse linked above).  We shouldn't put a lot of stock in that number, but we need it as a reference point.  Now, Rawr thinks that adding Death's Verdict (heroic) should add 732.63 dps. So then I'd be doing 12853 dps, and the portion of that overall dps attributable to my trinket would be about 5.7%.

If we compare the normal version of Death's Verdict (I skip over Whispering Fanged Skull (heroic) in the priority list because that's the trinket in my other slot), the latter would amount to 648.4 dps, or about 5.1% of my total dps, according to Rawr.

Now that's only one data point, and it would be nice to compare two (or ideally more) combat log parses in otherwise equivalent gear rather than a combat log parse and a Rawr estimate.  I'm not going to strongly defend the methodology of this post.  But, despite potential methodological difficulties, the difference between even the non-heroic Tiny Abomination in a Jar and the next-best trinket is large enough that I feel confident that it really is significantly better, even if we can't confidently arrive at a precise quantification of how much better.  And remember, that's just the special effect we're talking about; if you can utilize some portion of the Hit Rating on the trinket, it just gets better.

Final verdict: Tiny Abomination in a Jar really is best-in-slot for retribution paladins, or at least for me, by a wide margin, even if the Hit Rating on the trinket is entirely wasted.

Friday, April 30, 2010

White & Yellow

A terminological distinction that I will often employ on this blog is the distinction between "white" and "yellow" attacks.  These terms come from the color of the damage number that appears above the head of your target in Blizzard's default interface.  Generally, by "white" attack I mean an weapon auto-attack (melee or ranged), rather than a spell or ability.  "Yellow" attacks are everything else (the distinction is mostly irrelevant for casters, who typically don't attack with their weapons).  The difference is important, because auto-attacks tend to follow very different rules from spells and abilities, as we will see when we discuss the combat table.

Thursday, April 29, 2010

Combat Results

When you attack something with an attack or spell in WoW, there are several possible results.  Most of these results are relatively clearly communicated to the player, but not all of them.  Before discussing the combat table itself in a later post, first I want to enumerate and describe the different results that appear on the table, as well as other results that can occur when you attack a target.  Future posts will discuss how to increase or decrease the chances of each of these types of result.
  1. Hit:  The "default" or "normal" result of an attack is a hit.  The results of the hit depend on the attack or spell; a normal melee attack does weapon damage, reduced by armor and other damage reduction.
  2. Critical:  A critical hit with a melee attack normally does twice the damage a hit would have done.  A critical hit with a spell does 1½ times normal damage by default, though many talents and class abilities modify this.
  3. Miss:  An attack may also miss.  There are many ways for an attack to fail to result in damage, of which the miss result is only one.  Misses are to be distinguished, for example, from the following two types of result.
  4. Dodge:  The target may dodge the attack, resulting in no effect.  This is a different type of result than a miss, and is reported differently.  Only melee attacks can be dodged (what exactly counts as a melee attack as opposed to a ranged attack or spell is not always obvious, however).
  5. Parry:  The target may parry the attack, resulting in no damage.  Blizzard has announced plans to change this result in Cataclysm; in that expansion, a parry will result in 50% damage reduction for the parried attack and the next attack (it's not clear at this time if "next attack" means the next attack on the parrying target from any source or just the next attack on the parrying target from the same source as the attack that was parried). * Only melee weapon attacks can generally be parried, but see below.
  6. Deflect:  Occasionally, but rarely, you may see a "deflect" result reported in your combat log.  A deflect is the same as a parry, but is for some reason reported differently.  You will never see this result for an auto-attack, but only for spells and abilities (a paladin's Hammer of the Righteous attack, for example, can be deflected, by not parried).  Most spells cannot be deflected, except by a hunter under the effect of Deterrence.
  7. Block:  The target may block the attack, resulting in partial damage.  Presently, the amount of damage reduction from a block is a static number depending on the target's Block Value statistic, which is increased directly by gear and talents and indirectly by Strength.  This can result in no damage done if the target's Block Value is greater than the damage done by the attack (after applying armor and other damage reduction).  Such full blocks are treated as fully avoided attacks.  In Cataclysm, full blocks will no longer be possible.  Blizzard is planning to remove the Block Value statistic entirely and make all block results reduce damage from the attack by a percentage.
  8. Resist:  Like blocks, resists can be full or partial.  Weapon attacks generally cannot be resisted; this type of result applies only to spells.  A full resist registers in the combat log as a "resist," while a partial resist will specify what portion of the damage was mitigated by resistance.  Exactly how much damage is resisted depends on factors such as the level of the target relative to that of the attacker and whether or not the target has any resistance (Fire Resistance, Arcane Resistance, etc.) to the relevant spell school; there is always also a random component.  Some spells, particularly those that do not deal damage, cannot be partially resisted.  If they are resisted at all, they are fully resisted for no effect.
  9. Absorb:  If the target of an attack is under the effect of a damage shield spell like a priest's Power Word: Shield or an ice mage's Ice Barrier, the attack may be absorbed.  Damage shields typically absorb a fixed amount of damage, and every point of potential damage subtracts from that amount.  An attack's result will be reported as fully absorbed if the remaining potential of the shield is greater than the amount of damage the attack would do.  Unlike a full block or full resist, a fully absorbed attack may still apply a secondary effect.  For example, if the damage from a mage's Frostbolt is absorbed by a damage shield, the target may still be affected by the slowing effect of the spell.  Partial absorption is also possible when the amount of damage done by the attack is greater than the remaining damage absorption potential of the damage shields on the target.
  10. Glancing Blow:  This result is only possible when the target of the attack is a non-player character of a higher level than the attacker.**  A glancing blow does reduced damage based on the difference between the attacker's Weapon Skill and the target's Defense Skill, which is a function of the target's level (all non-player characters of the same level have the same Defense Skill).  The case we mostly care about is a player with maximum Weapon Skill attacking a Boss level target; in that case, the damage reduction is 30%.
  11. Crushing Bow:  This result is the counterpart of the Glancing Blow when a higher-level target is attacking a lower-level target.  In the current state of the game, it is not possible for a maximum-level player to receive a Crushing Blow, and only non-player characters may deliver Crushing Blows, so the only time you will ever see this result is when you are leveling and attempting to fight a mob at least four levels higher than you.
  12. Immune:  Targets are sometimes immune to the damage and effects of spells and attacks.  For example, a paladin under the effect of Divine Shield is immune to all damage and the effects of all player spells (with the sole exception of a priest's Mass Dispel) and most non-player spells, though there are several non-player spells in the game that ignore player immunities.  Sometimes a target may suffer damage from a spell even when you get an "immune" result for that same spell; in such cases, it is the secondary effect of the spell that the target is immune to.  For example, many bosses are immune to the slowing effects of spells like Frostbolt, but may still take damage from them.
  13. Evade:  This result occurs only when the target of an attack is a non-player character.  Any target that evades one attack will evade any attack until whatever condition is causing it to evade is removed; such a target is said to be "evading" and is essentially immune to all attacks and effects and regenerates health as though it had exited combat.  This state is triggered when a mob is pulled far enough away from its spawn location that it drops aggro and quickly speeds back to where it came from, when it is trapped by or stuck in the terrain of the game world, or when it is impossible for it to reach its attacker with any attacks.  Essentially, this result indicates that something has gone wrong, and the mob needs to "reset" itself; it's a kind of error state rather than a true combat result.
Those are all the possible results of an attack in World of Warcraft.  Not all of them can occur for every type of attack or every combination of attacker and target.  The chances for several of these results can be influenced by the attacker's or the target's stats; several others can be altered by situational factors like the way the attacker and the target are facing.  The real interest in distinguishing all these attack types is so that we can be clear about how to influence the chance for each result, so that we can increase the chance of good results and decrease the chance for bad results, as I will spell out in detail in future posts.

*Blizzard has officially decided against changing parry; it is now a 100% damage reduction against the parried attack, just like a dodge or miss.
**This is not strictly true; a glancing blow can occur against a target of the same level if the attacker's Weapon Skill is not at its maximum possible value.  I neglect this complication in part because Weapon Skill is going away in Cataclysm.