7.2 Spriest dps and rankings

[andresity]

Traditional dps ranking by [Spec Class] is fundamentally flawed because it doesn't take into account reasonable option value. Legendaries/AP are still substantial barriers to spec switching, but for the most part players have the option to invest playtime towards what is perceived to be their best spec and we see this bias in number of parses from certain specs. I think a more relevant ranking that reflects real world raiding would be by [Class]. For example,

Class A has 3 damage specs

• Spec 1 - Very good
• Spec 2 - Average
• Spec 3 - Very Bad

Class B has 1 damage spec

• Spec 1 - Average

Traditional [Spec Class] ranking will give the false impression that Class B is middle of the pack in the wild.

1. A1
2. A2 / B1
3. A3

Ranking by [Class] however will reveal the effect of players min/maxing such that we expect to see

1. A
2. B

Since blizzard servers are having a huge outage today, I decided to do exactly this. I play spriest, which is why this question interests me. Personal experience and traditional [Spec Class] rankings do not seem to match up for many people. How bad is Priest really?


Terms
Weighted
Class dps weighted among number of parses.

• Spec 1 - 70 parses (100dps)
• Spec 2 - 20 parses (80dps)
• Spec 3 - 10 parses (60dps)

Weighted dps of 92 (0.7*100+0.2*80+0.1*60)

Min/Max
Max(spec1,spec2,spec3). Using above example gives 100 dps.

Avg Rank
Dps consistency measured as the average ranking per boss from 1-12 using a weighted dps figure. Lower is better. If a class smashes one boss in particular but performs badly on all others, it will show as a high average ranking. A crude proxy indicator for relative dps volatility.


Results
All bosses, without Skorpyron (meme boss), without Skorpyron and Botanist (for spriests). Won't let me post images or links so you're gonna have to open them in a new tab. Even though mythic Skorpy isn't meme, included those numbers for comparison with heroic if you want to do that.

Heroic
75th Percentile
imgur.com/fibqE3X.jpg

95th Percentile
imgur.com/rhsNb02.jpg

Mythic
75th Percentile
imgur.com/1j7IjkL.jpg

95th Percentile
imgur.com/wAlMMom.jpg

Best Spec per Boss
95th Percentile Mythic NH
imgur.com/UPIBdrR.jpg

Some observations

• Monk and Spriest consistently score very low across the board.
• Rogue is not doing very good either but we should expect their weighted dps to increase over the long term as players gain better ability to switch between specs.
• High Priest dps volatility between bosses goes down the harder the content (more hp adds).
• A large part of the performance range imbalance of 10-13% is coming from warlocks(affliction) outperforming by a large margin.
• When we exclude warlocks, the performance range between classes decreases to 5-8%.
• Shaman, Hunter, Mage see 50-60k average dps increase over all encounters by switching specs per boss.
• Mage is very dps volatile between bosses. A lot of Mages are sticking with fire (up to 40% of total parses) even though it is the lowest performing mage spec by 10-25%.
• Mage players change specs the most depending on boss.
• Shamans and Druids are the most "stubborn". They like to stick with their favourite spec. Consistent 60/40 split between ele/enh and 85/15 between balance/feral respectively irregardless of which spec performs better on a boss.

Class Tiers
From 75 percentile Mythic NH (All Bosses, Spec Min/Max), classes seem to be divided into distinct performance groupings

God Tier +17.2%

• Warlock

Good Tier +8.7%

• Warrior
• Hunter
• Death Knight

Middle Tier +7.5%

• Paladin
• Shaman
• Mage
• Druid

Meh Tier +4.2%

• Demon Hunter
• Rogue
• Monk

Garbage Tier +0%

• Priest

TL;DR
Priest is not doing so hot in 7.2

Sheets
Data pulled from warcraftlogs.com

dropbox.com/sh/wwefkuwugsgc30w/AABVIZkiLFDYSSEQGKC0Tj9Ma?dl=0

[Relapses]

Finally, a mathematical reasoning to bench all your SPriests!

[niil945]

Not that this is a shocker to anyone except those very few vocal spriests who dance around all the threads where people are voicing these exact same concerns and they do nothing but shoot them down and talk about how they're gods among men in their own raid groups. Still, thanks much it was an interesting read.

[Deleted]

If you haven't already you should post this on the official forums (kinda goes without saying but still).

[Archaea3]

Mythic botanist isn't padding for spriests.

[Xraychicken]

Mythic botanist isn't padding for spriests.

What he said. WarcraftLogs only counts damage done to the boss you kill in each phase towards your overall dps. Tichondrius is the cheese fight for spriests since a majority of guilds tank 3-6 bloods on the boss for a majority of the fight and all that damage is counted in logs. (Obviously dotting the bloods technically isnt cheesing since its a ST increase but it inflates our numbers in comparison to other classes)

[andresity]

What he said. WarcraftLogs only counts damage done to the boss you kill in each phase towards your overall dps. Tichondrius is the cheese fight for spriests since a majority of guilds tank 3-6 bloods on the boss for a majority of the fight and all that damage is counted in logs. (Obviously dotting the bloods technically isnt cheesing since its a ST increase but it inflates our numbers in comparison to other classes)

I should have typed more and made it clearer so it's on me. As stated in the OP, [Without Skorp] and [Without Skorp/Botanist] wer included in the mythic figures for completeness for anyone who wants to compare them with heroic numbers. I assumed people would have only really looked at the [All Boss] numbers for mythic and should have been clearer on that.

• Botanist is excluded specifically because of S2M cheese on the heroic version on that fight.
• While Heroic Tichondrius can be a padding fight, the padding benefits all classes who can take advantage of it more evenly. Tichondrius resembles Skorpyron in many ways.

[andresity]

What he said. WarcraftLogs only counts damage done to the boss you kill in each phase towards your overall dps. Tichondrius is the cheese fight for spriests since a majority of guilds tank 3-6 bloods on the boss for a majority of the fight and all that damage is counted in logs. (Obviously dotting the bloods technically isnt cheesing since its a ST increase but it inflates our numbers in comparison to other classes)

What he said. WarcraftLogs only counts damage done to the boss you kill in each phase towards your overall dps. Tichondrius is the cheese fight for spriests since a majority of guilds tank 3-6 bloods on the boss for a majority of the fight and all that damage is counted in logs. (Obviously dotting the bloods technically isnt cheesing since its a ST increase but it inflates our numbers in comparison to other classes)

I should have typed more and made it clearer so it's on me. As stated in the OP, [Without Skorp] and [Without Skorp/Botanist] were included in the mythic figures for completeness for anyone who wanted to compare them with heroic numbers for whatever reason. I assumed people would have only really have looked at the [All Boss] numbers for mythic and I should have been clearer on that.

• Botanist is excluded specifically because of S2M cheese on the heroic version on that fight. Expect Spriest dps to keep rising up towards 2mil as more people start learning how to cheese it.
• While Heroic Tichondrius can be a padding fight, the padding benefits all classes who can take advantage of it more evenly. Tichondrius padding resembles Skorpyron add dynamics in many ways but is much more consistent for classes having different ramp up/burst timings due to the adds living longer.

[andresity]

What he said. WarcraftLogs only counts damage done to the boss you kill in each phase towards your overall dps. Tichondrius is the cheese fight for spriests since a majority of guilds tank 3-6 bloods on the boss for a majority of the fight and all that damage is counted in logs. (Obviously dotting the bloods technically isnt cheesing since its a ST increase but it inflates our numbers in comparison to other classes)

I should have typed more and made it clearer so it's on me. As stated in the OP, [Without Skorp] and [Without Skorp/Botanist] were included in the mythic figures for completeness for anyone who wanted to compare them with heroic numbers for whatever reason. I assumed people would have only really have looked at the [All Boss] numbers for mythic and I should have been clearer on that.

• Botanist is excluded specifically because of S2M cheese on the heroic version on that fight. Expect Spriest dps to keep rising up towards 2mil as more people start learning how to cheese it.
• While Heroic Tichondrius can be a padding fight, the padding benefits all classes who can take advantage of it more evenly. Tichondrius padding resembles Skorpyron add dynamics in many ways but is much more consistent for classes having different ramp up/burst timings due to the adds living longer.

[swilksy11]

I guess new shadowfiend will have to do about 150k dps overall.. Its up about 1/6 of the time so 900k dps when its out. Totally balanced.

[Ailylia]

I think for me, the most disappointing part about all this is that we have to work really hard just to put up those mediocre numbers. Whereas there are other classes that are pretty much "hit whatever is on cooldown" and they do better dps. Heck, my BM hunter I barely have put any effort on puts up better numbers than my spriest for many fights.

[Spicymemer]

I said spriest would dead before this tier ended.

H2P discord told me to stop spreading negative information.

I was right.

From a mythic standpoint there is only 2 fights where spriest is actively valued, that is botanist and star augur and mostly for the same reason, execute range DPS and self healing.

[Fleckens]

I said spriest would dead before this tier ended.

H2P discord told me to stop spreading negative information.

I was right.

From a mythic standpoint there is only 2 fights where spriest is actively valued, that is botanist and star augur and mostly for the same reason, execute range DPS and self healing.

Well, that's not really true, we are not in the best state and I hope we do get buffs but we can still sort of pull our weight on progression on the harder fights. Star augur not so much anymore but it's much easier now anyway, but elisande and gul'dan last phase we are still very useful and I assume those are the 2 bosses most guilds are progressing on now.

[Spicymemer]

Well, that's not really true, we are not in the best state and I hope we do get buffs but we can still sort of pull our weight on progression on the harder fights. Star augur not so much anymore but it's much easier now anyway, but elisande and gul'dan last phase we are still very useful and I assume those are the 2 bosses most guilds are progressing on now.

Gul'dan is a melee stacked fight, you take 5-6 Ranged for progress.

Why bring a Spriest when a Boomkin or Warlock are clearly better choices that accomplish the same playstyle and you need Hunters/Warlocks for their AoE and Utility.

[Waynebrady]

It's ok, they say we're just balanced around the next tier of content! That's all! They tell us Shadowfiend will be king next tier, we'll just keep waiting while other classes stomp all over our DPS this tier with little to no effort.

[Archaea3]

Gul'dan is a melee stacked fight, you take 5-6 Ranged for progress.

Why bring a Spriest when a Boomkin or Warlock are clearly better choices that accomplish the same playstyle and you need Hunters/Warlocks for their AoE and Utility.

So assuming 2 tanks, 4 healers, 6 ranged, that's 8 melee. If you switch 1 melee to ranged it would be 7:7. That's a difference of 1 player, hardly sounds like stacking melee.

- - - Updated - - -

My two biggest problems with this analysis are 1. using the mean the calculate the average. And 2. Assuming dps specs can just switch between spec easily.

For 1. Outliers drastically affect the mean. Affliction warlocks are a clear outlier on many fights causing the data to skew towards high dps. Using the median is more reliable and meaningful qualifier here. An ideal balance would mean all classes are performing better than the median 50% of the time and worse 50% of the time. Using that metric shadow is 3/10 over the median - in the same tier as Marks, Outlaw, and Windwalker and doing better than Fire, Unholy, Destro, and Enh.

For 2. Changing specs is more than pressing a button. Most specs require different gearing, different legendaries, and AP. Most players aren't going to be able to keep up the AP and legendary grind for multiple specs. I don't agree that viewing by class is an accurate way to judge the health of a spec.

http://imgur.com/a/1aKx2

So what do I think this means? Yes, shadow priests are underperforming. So are Arcane, Ele, Marks, Outlaw, Windwalker, Fire, unholy, Destro, and Enh. But shadow is not last place, that's arguably fire mages. If you look at progression fights, which I'm arbitrarily calling the back half of NH, Shadow comes out right around the median. So shadow is easily relevant on Mythic progress.

[Assumptions and data: all fights used for numbers unless otherwise stated. Mythic 75% numbers used. First column shows # times a spec beats the median on all bosses. Second column is number of times a spec parses last. Third column is beating the median on the last half of NH.]

[Fleckens]

So assuming 2 tanks, 4 healers, 6 ranged, that's 8 melee. If you switch 1 melee to ranged it would be 7:7. That's a difference of 1 player, hardly sounds like stacking melee.

- - - Updated - - -

My two biggest problems with this analysis are 1. using the mean the calculate the average. And 2. Assuming dps specs can just switch between spec easily.

For 1. Outliers drastically affect the mean. Affliction warlocks are a clear outlier on many fights causing the data to skew towards high dps. Using the median is more reliable and meaningful qualifier here. An ideal balance would mean all classes are performing better than the median 50% of the time and worse 50% of the time. Using that metric shadow is 3/10 over the median - in the same tier as Marks, Outlaw, and Windwalker and doing better than Fire, Unholy, Destro, and Enh.

For 2. Changing specs is more than pressing a button. Most specs require different gearing, different legendaries, and AP. Most players aren't going to be able to keep up the AP and legendary grind for multiple specs. I don't agree that viewing by class is an accurate way to judge the health of a spec.

http://imgur.com/a/1aKx2

So what do I think this means? Yes, shadow priests are underperforming. So are Arcane, Ele, Marks, Outlaw, Windwalker, Fire, unholy, Destro, and Enh. But shadow is not last place, that's arguably fire mages. If you look at progression fights, which I'm arbitrarily calling the back half of NH, Shadow comes out right around the median. So shadow is easily relevant on Mythic progress.

[Assumptions and data: all fights used for numbers unless otherwise stated. Mythic 75% numbers used. First column shows # times a spec beats the median on all bosses. Second column is number of times a spec parses last. Third column is beating the median on the last half of NH.]

While I somewhat agree that you can't just say that the classes have other specs so it's fine you also can't completely ignore that shadow has no other option, switching specs for the other classes is still significantly easier than having to completely reroll to another class(not that I'm claiming shadow is in such a bad spot that I would need that to do that, all we really need is some quality of life changes as well as a bit of a single target buff).

[davesignal]

Most of those specs can swap to other specs, though (as per usual) shaman and monks aren't that lucky.

[Spicymemer]

So assuming 2 tanks, 4 healers, 6 ranged, that's 8 melee. If you switch 1 melee to ranged it would be 7:7. That's a difference of 1 player, hardly sounds like stacking melee.

- - - Updated - - -

lots of talk

1. 6 is the maximum, anything over 6 means guldan is clearly on farm status for this theoretical situation, 5 is the standard, if 4 is viable, then that is the optimal.

2. Spriest has been broken since 7.0, Spriest AoE has been horrible for longer than this expansion, but now super fun repeatable content is based around spamming AoE, its a real issue now. Remember which spec was an utter joke that would never be chosen for CM runs? Spriest.

[andresity]

For 1. Outliers drastically affect the mean. Affliction warlocks are a clear outlier on many fights causing the data to skew towards high dps. Using the median is more reliable and meaningful qualifier here.

As stated in the OP, dps is weighted by percentage of parses from different specs. This is to accurately reflect real world raiding experience by all players as a collective. Individual experience is not informative for analysis of a system. I don't care if the warlock in this one guild or raid plays destruction, demonology, or affliction. I care about the reasonable power potential of each class relative to other classes.

An ideal balance would mean all classes are performing better than the median 50% of the time and worse 50% of the time. Using that metric shadow is 3/10 over the median - in the same tier as Marks, Outlaw, and Windwalker and doing better than Fire, Unholy, Destro, and Enh.

What you are describing is a measure of and design for a smooth linear skill curve.

f(performance) = a * skill
where a is some constant number

While an interesting subject of it's own, this has little to do with class balance when we are taking 95th percentile numbers specifically in order to dampen variables related to skill.

For 2. Changing specs is more than pressing a button. Most specs require different gearing, different legendaries, and AP. Most players aren't going to be able to keep up the AP and legendary grind for multiple specs.

Spec switching is an additional question of interest and one that mainly concerns future trends. As stated in the OP, players have the option to invest playtime into what they perceive to be their best performing spec and we see exactly this from biases in number of parses per spec within any given class.

I don't agree that viewing by class is an accurate way to judge the health of a spec.

No such claim was made. Rather, the inverse claim was made. Viewing by Spec is not an accurate way to judge the health of a Class.