2  Wisdom of Chi-ji

This section captures more subjective opinions about the game. In short, rather than trying to be completely impartial, this section veers off in a direction where I’m injected some of my own idea about the game, and the data that’s presented here.

2.1 On FOTM Specializations

We need to state an important piece of information:

All specializations are able to run a +20 key level.

Picking one specialization over the other solely based on power is probably a mistake. When evaluating, say, the Mythic Plus Score of a player, we have several factors in play:

  1. How skilled is the player?
  2. Who are the other team members? What are their skill at the game?
  3. How strong is the specialization?
  4. What dungeons were run?
  5. What item level does the player have?
  6. What talents did the player choose?
  7. What affixes are in play that given week?
  8. What region were the dungeon run in?
  9. What are the rest of the spec composition?

And so on. By far the most important aspect is player skill. Every other factor is much less important. The overarching trend in the analysis is that specializations are quite close in power.

In fact, once you look at the player population as a whole via sampling, some of the specs that does well in the very top, doesn’t do as well in lower keys. Some specializations require a lot of ramp before they can deal their damage. Some require very large pulls, which are less likely to occur at lower key levels when players aren’t as organized, talented, or experienced. And some have a very high skill ceiling so they are less likely to work well in the lower key levels. From the other side, some specializations are strong at lower key levels but taper off as the key level increases.

Indeed, the data suggests Matt Villers is right when he stated that there is a community perception problem which occur when people transplant the top-level group compositions down to lower levels and think things are working the same way down there.

2.1.1 Fluency

We can treat a specialization as a language where the words of the language are talents. To obtain fluency in a spec means you have the core language committed to muscle memory and you don’t think when playing. Rather, you have a flow-state in which you can focus on other aspects of the encounter than the pure mechanical parts.

Switching between specs will have you start over because the language is slightly different. It takes time to learn the details of each talent and how it interacts with the kit as a whole. Hence, a large part of the power of a specialization comes not from its tuning but the players fluency. When you switch specialization, your fluency drops for a while, until you can retrain yourself to play it well.

Maintaining a high level of fluency on a wide set of classes/roles takes a lot of time. So one might want to keep the number of specializations lower, and then sink more hours into mastering those. Comfort is a large part of playing well1.

1 You might want to gravitate to languages/specializations where you like the playstyle. If something feels clunky, it won’t be fun to play in the long run, even if it’s tuned to be powerful. Unless tuning is majorly broken, the differences are small enough they don’t matter unless you are trying to push at the top 0.1% or something in that ballpark.

2.2 PTR

PTR, the public test realm, is not a good marker for specialization strength. Testing is limited and changes happen all the time. Gear levels are not fully realized and people are locked to a certain gear level, which makes it harder to figure out exactly how specializations will scale. You should generally be cautious with data assembled from PTR.

  • You don’t know how data were collected in many cases. This means you don’t know the encounter profile either. Simple example: some specializations have an “execute” style skill where they deal extra damage in the terminal phase of a fight. If you never record terminal phases of fights, this won’t ever show up.
  • PTR is played by a limited set of people, who are often quite skilled. This may skew results.
  • Skills might not work.
  • You might not have access to legendary weapons, dropping from the final boss.

Generally, you should probably hold back on judgement until a couple of weeks into the season where we have better data. Usually, there are some larger tuning passes early in the patch, when the Raid race-to-worlds-first is over.

2.3 HPS as a metric

This section also applies in Raid, if you squint your eyes enough. At lower difficulty tiers, the healing required eventually hits a ceiling when people figure out the mechanics. This makes HPS into a zero-sum game where one healers zealotry provides another healers misery.

The amount of HPS a healing specialization can do is often equated with its direct power. However, there are reasons to be a bit cautious when using this metric. First, the metric isn’t useless. There’s a merit in being able to dish out large amounts of healing, especially to meet burst requirements. So overall, there’s value in using HPS to gauge power of players or specializations.

However, the power isn’t entirely proportional, and will begin to taper off at a point in time. You might want to consider some of the following thoughts.

When you run a dungeon, the damage intake can be divided in the portion that is unavoidable and the portion that is avoidable. The former scales with key level. As you run higher and higher keys, the amount of damage that’s present will go up and require more effort on the healers part. On top of that, you add the avoidable damage, which is present due to small (or large) mistakes made as the party plays the dungeon. It’s entirely possible that the lower key levels require far more healing than the higher key levels. People are simply better at avoiding damage, because they have to, or they die.

This creates a conundrum where healing at a higher level is more predictable because a larger portion of it comes from the unavoidable damage which is often orchestrated via a mechanical “dance” with the boss or trash.

In a way, it is perhaps better to flip the damage intake on its head: there’s an adequate amount of healing at which the healing check of the dungeon is being met. Everything above that is waste insofar the time could be better spent on doing damage, helping out with mechanics, crowd control, and so on.

Second, the amount of incoming damage can be mitigated. Every specialization has access to some kind of defensive measure to reduce or immune incoming damage. And using those at the right point in time can severely limit the amount of incoming damage that has to be healed through.

Third, not all healers are alike. Some healers have access to very potent damage reduction tools. Those will not typically show up on an HPS-meter, but they prevent a lot of damage to the party2.

2 Context also matter. A player who is getting a buff from an Augmentation evoker in a raid will be able to heal more compared to people who aren’t getting the same buff. In many cases, the reason someone pops off on the HPS meters is because the context allows them to reach a given level of healing.

Fourth, preventing damage is far better than healing through it. A correctly placed “kick” or “stop” from a healer can easily mitigate a massive healing requirement in the longer run. This healing requirement often spans several GCDs, which increases the GCD-pressure considerably.

In Raid, healing is a zero-sum game where the healers in the raid compete against each other for the available healing. This means playing with a really good player means you’ll do less healing overall. This makes any kind of reliance on the amount of healing you do suspect and should probably be taken with a grain of salt.

2.4 DPS as a metric

The thoughts in this part largely inspires a lot of the work on the site. We generally don’t use DPS as a metric of specialization power at all, because it will mislead computation. Rather, we focus on what keys are timed instead, since this is a far better measure. Not only does this choice protect against certain specializations “padding” the damage meters; it also protects against support specializations such as Augmentation Evoker.

DPS is a better metric of spec power than HPS in most cases. The more overall DPS you can do throughout a dungeon, the quicker the run is. That speedup helps the survivability of the party. You simply kill the boss before you run out of mitigation tools.

However, the damage pattern of different specializations aren’t the same. Some specs are very bursty, delivering their damage in a spike, and others deliver their damage in a more sustained fashion. Furthermore, some specs can focus their damage at a single target far better, and other specs are good at AOE.

The burst-damage means that key level can have an effect on the power of a spec. Some specs require the mobs to live for a certain amount of time, so their sustained damage and rotation can get to work. If the mobs die before this happens, they are at a disadvantage.

The above means that it is important to funnel damage into the right enemy. Many mob packs do not have even health, so if a spec can shave the most healthy mob a little bit, it speeds up the dungeon far more than dealing more damage to the weaker enemies in the pack. You want the pack to die at the same time, so you can start working on the next one. The horror-situation is if a single mob lives, so you have to spend time killing off that mob.

Hence, DPS is important, but it isn’t a foolproof metric. Timing the dungeon can easily be done with lower DPS-numbers if that damage is dealt in a more meaningful way. It’s also important to stress that survivability might be more important than raw DPS in many cases. If people die, the loss of DPS is far greater, often to the point of ruining a key.

2.4.1 DPS on healers

At a certain key level, dealing damage as a healer will become quite important. Healers are tuned so they deal very little damage compared to the other roles. This ensures their healing output is more important than their damage output. But they still deal enough damage over the course of a dungeon run that it can matter. The total damage might be roughly equal to a dungeon boss or so. And that amount of damage is not negligible.

Different healers provide different damage patterns. As an example, some healers are good at AOE damage, whereas others damages with a single-target focus. Some healers can do damage and healing at the same time, while some flip between damage and healing.

In a typical composition, you’ll often have DPS which can focus a primary target while still dealing damage to secondary targets. This means the speed of the dungeon is often determined by the highest health targets in packs. Thus, efficient single-target damage is somewhat more valuable to a healer because it allows you to “shave off” health from either a high-health target or a boss.

2.5 Early season

Early in a new season, players generally don’t play for score. They play for loot. This means they’ll stop playing as soon as they hit their +20 dungeon keys, as there’s no real reason to push any higher. There’s going to a few players who like playing for score early on, but a lot of players will be waiting until they have a bit more gear before they push.

The consequence is that individual player skill expression is being dampened, and as a result, it is harder to figure out what the power of the specializations are.

Furthermore, when people are playing for loot in the vault, the dungeons which are easier become overrepresented. You run the easiest dungeon for the week, and then you exit. You may push an alt, or you may do something else in the game. It is likely more beneficial to wait until your character has accrued a bit more powerful gear, as it will make pushes that much easier. You also combat overtuned dungeons this way, because tuning is likely to happen in the first few weeks.

A statistical model will have a harder time separating players (and thus specializations) due to this. Instead, the difference in dungeon difficulty will dwarf the effect of players, since in a certain way, they aren’t trying their best to get higher scores.

Worse, dungeons are often in need of tuning early season, which puts a lot of the statistics on even more uncertain ground.

In short, take early weeks in a season with a grain of salt. As people start pushing, you will see the real flavor of the specializations, and you will begin to see which specs are generally stronger. There may be early indicators in which direction a specialization leans, but that’s the best we can get out of the data early on in a season.

Clearly, this observation applies not only to ZugZugCreation, but every site relying on early statistical data from dungeons.

You should be cautious when people talk about specialization power in the first few weeks. People will not have access to their 4-piece tier sets yet, and historically some specializations gets a far larger bump in power from the set than their core skills. General rule 2-3 weeks in, and you start having a good grasp from the data.

2.6 Popularity noise

Popular specializations generate a lot of noise in the data. M+ score has a grind factor: the more you play a given character, the higher M+ score you are likely to obtain on it. This means popular specializations, or “in-meta” specializations, will look stronger at lower difficulty-levels. People play them more, so the median M+ score of the specialization is far higher, and it looks like the spec is very strong.

However, once you reach the top players, things might shift around. The score is now controlled far more by player skill, and grinding yourself to a higher score by “failing upward” becomes much harder to do. This means score becomes a better indicator of skill, which in turn improves the statistical certainty.

It’s important to keep this in mind when you consider off-meta specializations. A specialization might be incredbly strong even if it has less popularity in the game. The shift is rather drastic. If something is a small amount more powerful on average, a lot of people tend to gravitate toward it, even if the spread is so large that the “weaker” spec is strong about 40% of the time due to the random nature of dungeons.

2.6.1 Sibling popularity

The pure dps classes such as Warlock and Rogue will always be in a state where the less popular specs seems pretty weak. The friction of switching to a sibling-spec is usually fairly low, and a lot of the knowledge of the class can be shared. Thus you will often see players flock toward the specialization which either is strongest, most fun to play, or is the least in trouble.

This doesn’t mean the sibling specs are weak however. The same rule applies, and popularity noise is introduced, making the specs seem weaker than they really are.