7.1.5 PTR Notes Changes for DPS DKs

[Schizoide]

Hm, you should try simming a couple of different specs and compare to how the stat scaling compares to Frost. Frost does have a pretty serious scaling issue, it's far worse than Unholy in that regard.

Nope. Data from actual raids says Frost scales much better than Unholy.

Some of that is, of course, sample bias. The "top spec" for any class gets the best players, with the best gear, playing it. But that only applies to the gap between Frost and Unholy. Frost scales well compared to other DPS specs also.

Again, that doesn't mean Frost performs better than other specs-- it's middle of the pack, because its base numbers start lower. But based on this scaling data you would expect Frost to continue to be competitive in future tiers. Unholy, not so much.

[Deleted]

At least this guy understands.

[DarkBlade6]

meh /10char

[Schizoide]

Those are from actual raids, not sims.

No offense meant, but I suggest you read and think before posting next time

[wooootles]

Huh whatever dude, you can think what you want.
By the way, Real life scenario =/= Sims

You can say whatever the fuck you want, the statistic (from real performance , not computer generated sims) don't lie.
We start high and end up at the bottom, its simply due to bad scaling. (Click 75th, 80th, 90th, 99th)

It's from warcraftlogs, not sims.

[Deleted]

Nope. Data from actual raids says Frost scales much better than Unholy.

Some of that is, of course, sample bias. The "top spec" for any class gets the best players, with the best gear, playing it. But that only applies to the gap between Frost and Unholy. Frost scales well compared to other DPS specs also.

Again, that doesn't mean Frost performs better than other specs-- it's middle of the pack, because its base numbers start lower. But based on this scaling data you would expect Frost to continue to be competitive in future tiers. Unholy, not so much.

I'm not sure how accurate it is to measure scaling by ilvl through Warcraftlogs, since that varies heavily on how well players can compete with each other. Furthermore, the data I get when simming differs massively from the data you showed, again, this could be due to different factors such as bad APLs, but the difference is pretty massive.

Still, I believe Frost isn't scaling as bad as it was early on in the expansion, mainly because the base numbers were buffed so hard that it made the scaling less worse simply because of that. I don't think it is on par with other classes yet, though. Of course, whatever metric we pick to evaluate that might be skewed for one or another reason, and the only real way we'll ever reach a conclusion is when Nighthold launches and we get actual numbers, until then, we could fling shit all night long without reaching a conclusion.

I do absolutely agree Unholy is a mess in pretty much everything. Bad scaling being one of the worse offenders. Really hope they fix that and the spec RNG problems, I enjoyed playing it a lot, but it'll never be competitive in it's current state. I simply don't believe they'll ever properly fix the scaling issue, and the RNG will always be a killer for it.

[Schizoide]

Of course it relies on player competition-- in the best way, because the data comes from real players in real raids!

You know I spend a ton of time in sims too, but whenever you can source actionable data from the actual game and Simcraft disagrees, that simply means Simcraft is wrong.

There is no reason to be concerned about Frost scaling right now, and complaining about Frost does Unholy a disservice. Unholy really is poor, in both raw performance and scaling.

[Eruntien]

Are you saying in a Frost vs UH or just in general? If in general wrong as at best we were in the middle.

[Schizoide]

Again, you are equating scaling with performance. They aren't the same thing.

(I feel like I'm being very patient here. It's become very clear why the devs ignore anyone complaining about scaling.)

[Casterbridge]

meh /10char

You do understand that when you change the percentile in warcraft logs you are not looking at scaling right?

If you want to look at scaling in warcraft logs change the ilevel, and keep the percentile level within say 75% to 85% at most, if you start looking at higher than that your run into the danger of looking at people who are cheesing mechanics to get higher level parses.

Least that's how I would do it.

[DarkBlade6]

Those are from actual raids, not sims.

No offense meant, but I suggest you read and think before posting next time

Well then explain to me, why do we end up at the bottom of the chart when you compare highly skilled players with the very best gear. Where is our superior scaling when it matters?

- - - Updated - - -

You do understand that when you change the percentile in warcraft logs you are not looking at scaling right?

If you want to look at scaling in warcraft logs change the ilevel, and keep the percentile level within say 75% to 85% at most, if you start looking at higher than that your run into the danger of looking at people who are cheesing mechanics to get higher level parses.

Least that's how I would do it.

Yes I understand that, but still you can say that the person parsing in 99th probably have some of the best gear/exp/skills ETC, compared to a 75th who it was probably his first kill with mediocre gear.

[Schizoide]

Frost is not at the bottom of the "chart". It performs generally in the middle of the pack. Again, I suggest you read up on what scaling actually means, because you still don't get it.

You can actually look at itemlvl scaling directly in WCL, but the data is much more difficult to consume and compare to other specs than the site I linked to earlier, which digests it for you.

https://www.warcraftlogs.com/statist...gregate=amount

Basically you want to look at the slope of the line, and compare that to other specs in the same view. But that's impossible to accurately eyeball. Thus, the other site I linked.

[DarkBlade6]

I don't think we're looking at the same ''data''


How's that good scaling ?

[Grimdeeds]

Holding out hope here that Unholy will still get some mention soon. However, the Sloot interview with Celestalon where the dev once again reiterated their intent not to bring up underperforming specs too high for fear of causing hardship on people having to switch specs, makes me very leary of anything positive at this point.

I still don't understand how they can be okay with the artifact knowledge/power system when it handicaps their own ability to more closely balance classes/specs.

Unholy really needed some love here and it's so far unmentioned.

The link about the scaling and showing how bad off Unholy is not just at performance but at scaling too, is really disheartening.

By all means, let's get some more people in here to remind us the sky is not falling, the world is not ending, umm, what else do they say as a way to dismiss any constructive feedback asking for changes? Life will go on...?

[Deleted]

The most concerning thing is that they never mention us, nthey ever talk about us. Not. A. Single. Word

[wooootles]

I don't think we're looking at the same ''data''
How's that good scaling ?

In that case I think the sample size is too small to gather any meaningful data. 887-889 suddenly has Unholy in the top 1/3rd for example...

[Deleted]

snip

I feel you. It's going to take years until the bad scaling meme goes away, both for Obliterate and for Frost overall. Same goes for other misconceptions, like haste causing GCD-locking problems (it's the opposite now: less haste means more GCD-locking because certain mechanics do not scale with haste and use up a static amount of GCDs, like RP from Frost Fever, Remorseless Winter or ERW).

People still use logic from a time when we were locked from the get go with a friggin' 1s GCD making haste nearly useless, a time when Howling Blast/Rime was a relatively low AP% spell, a time when KM could proc on any autoattack and affected skills representing nearly 70% of overall damage, a time when numbers had to be tuned with PvP in mind, a time when... you get the idea.

Baseline 1.5s GCD reduced by haste has been one of the biggest changes for this class in a long time. Same goes for the Rime and KM changes. On top of that, we have many talents that interact with haste and crit now. And don't forget things like NbtB or Ambidexterity (Ob and FS weapon damage scaling is bonkers, and that's happening in an expansion where our weapon it's always many ilvls above the average of our gear). It's no wonder that live data indicates that Frost looks better scaling-wise, because on a theoretical level that was what math was pointing to.

Back in Kahorie's sim and Consider's blog times, maybe we didn't have the most sophisticated tools, but hell if the theorycrafting level wasn't leaps and bounds better (and I hate using the "back in my time" card, believe me).
At least people tried to back up their (flawed) simulations, (flawed) napkin math and (flawed) empirical data with some some sensible theories to discuss them instead of just parroting memes.

Yeah, I'm pretty salty and I don't know why. I play Blood 95% of the time. Or maybe THAT is why I'm salty. At least until 7.1.5. Heh.

I don't think we're looking at the same ''data''
How's that good scaling ?

Uneven/low number of parses relative to lower ilvl brackets. See how Ret is also down there, which is kinda wrong.
Check that website Schizoide linked before and read about their methodology.

[Slim1256]

This is an example of a scaling comparison:



EDIT: I should note that I would take the ends of these (both sides) with a grain of salt, due to small sample sizes.

You can do any comparison you want like this with WCL.

[Tiyr]

Yes, I did. Did you?

https://cheald.github.io/warcraftlogs-parses/

I suspect you're one of those people that doesn't understand what scaling means, and use it as a synonym for overall performance. But they are not remotely the same thing.

That page could potentially be misleading. He is using the top 1000 logs from each spec, which means that the percentiles of logs that are being used will vary depending on how much a spec is being played.

When you look at data where all logs are from the same percentile, the results look much different. For example, with his data Frost DK looks like one of the best scaling specs while Outlaw looks like the worst. However this is what it looks like when you compare only logs from the same percentile:



In the above example I am using 95th percentile in heroic (heroic for much more logs at different ilevels) but it doesn't really matter what percentile you use. When using only logs from the same percentile, Outlaw seems to scale better than Frost DK.

[Schizoide]

It's very difficult to eyeball line slopes. That's why it's very difficult to draw conclusions from WCL's scaling graphs. If you look at the bottom of these two pages, it plots scaling by itemlvl per boss.

https://cheald.github.io/warcraftlog...th_knight.html
https://cheald.github.io/warcraftlogs-parses/rogue.html

Cheald's graphs include heroic data too, as mythic has insufficient parses for many specs. They go by itemlvl, so that shouldn't make any difference.

Looking at these graphs, you can see that outlaw gets somewhere around 4200 DPS per itemlvl on Xavius (its best boss), compared to Frost getting 4500 DPS per itemlvl on Il'gynoth, its worst boss.

He also has charts by difficulty, if you want to narrow it down, and unholy does better in mythic (due to legendaries and artifact power investments, I assume), but the overall conclusions don't change.

https://cheald.github.io/warcraftlog...ifficulty.html

@Feelycraft: Great post, I hope you stick around. I miss Consider too! He was incredibly sharp. Nobody since can compare.

To all the people pointing out very low sample sizes at highest itemlvls, particularly in mythic, thank you! It's good to see more clueful people in the thread.

Focus on Unholy. Frost is OK.