Death from Above Tuning

Another cross post from the official forums.

Some thoughts on rogue level 100 talents and Death from Above in particular.

One important caveat for the rest of the results in this post.  SimulationCraft implements the auto attack loss from Death from Above as an 800ms swing timer pushback.  However testing in game suggests that the auto attack pushback caused by Death from Above is closer to 1.1-1.4 seconds.

Pathal’s Results:  http://pastebin.com/3hCTcPAb

My Results:  http://www.warcraftlogs.com/reports/cqpYxDrNF3BXmQAT#type=damage-done&source=1&view=events

For this test I tracked the swing timer with a weak aura and attempted to use DfA as close to the end of the swing timer as possible to show the maximum possible loss from DfA and then determined the auto attack loss as the time difference between melee swings minus the 2.47 second swing timer.

With this potentially modeling inaccuracy in mind which may overstate the value of DfA some results.

Combat:  https://www.dropbox.com/s/782cnmxis4tut96/combat_dfa_new.html?dl=0

This shows 4 combat configurations based on the T17M profile.  T17M uses Shadow Reflection by default, DfA_no_AR is a modified version of the DfA profile that does not use DfA during adrenaline rush due to energy capping issues.

This shows pretty clearly that DfA is a non-trivial dps loss for combat compared to no talent(~3.5%) due to a combination of factors, most importantly the high energy cost of DfA compared to evis (25 energy vs. 10 energy after relentless strikes is taken into account) results in a reduction of AR uptime (20.69% vs. 22.5%), deep insight uptime (28.28% vs. 33.68%) and killing spree uses (9.5 vs. 10.3).  It is not clear how important the auto attack loss since the changed AR uptime confounds energy regen however intuitively is will be most detrimental to combat with its high passive damage compared to the other specs and energy regen that relies on auto attacks.

Assassination:  https://www.dropbox.com/s/vfe56xuwaawh452/assassination_dfa.html?dl=0

Assassination has DfA as a minor dps increase over no talent but still not competitive with shadow reflection in the default profile.  The underlying issue with DfA for assassination is similar to the issue for combat, both specs get substantial portion of the value of their finishers from side effects rather than direct damage.  Combat loses cooldown and BG cycling speed and assassination loses envenom uptime (75.31% vs. 81.04%) due to the energy cost of DfA compared to a regular envenom.  Auto attack loss does not seem to be a substantial issue for assassination with ~0.3 auto attacks lost per hand per DfA.

Subtlety:  https://www.dropbox.com/s/qjr96kcmo872tsp/sub_dfa.html?dl=0

Unlike combat and assassination subtlety has no major side effects on eviscerate.  This makes subtlety the best possible spec for DfA and the data bears this out.  While DfA is a useful dps increase for subtlety it is outdone by shadow reflection (the default sub profile uses venom rush).  Even if DfA showed a slight dps increase for subtlety it would be a hard sell since DfA is by far the hardest of the three talents to use requiring cp and energy pooling and raising all the loss of control and landing in a bad spot issues combat rogues have been complaining about with killing spree for 3 expansions now.  Compare this to SR which is effectively a passive talent that is macroed into primary cooldowns.

Reducing the energy cost of DfA to 35 energy like a standard envenom/eviscerate would help DfA be much more competitive for assassination and combat since it would reduce the envenom uptime loss and reduced cycling speed respectively.  It would also probably push DfA above SR for subtlety, matching damage with degree of difficulty.

Another possibility would be buffing the AoE pulse portion of DfA to be a useful form of AoE damage.  Currently it hits for very little damage and would help give the talent a niche on the tier.  One problem with the rogue 100 tier in general is both SR and DfA are primarily single target dps talents without any other major niche, this means that the weaker talent between the two will get left by the wayside.  Increasing the AoE pulse damage of DfA would give DfA a niche beyond single target which may keep it useful even if the single target damage is lower than that of SR.

In a similar vein Pathal suggested in #ravenholdt that the AoE pulse could proc poisons, this would grant a larger benefit to combat and assassination than subtlety and would increase the effective damage of the AoE pulse without having to buff its damage.

Combat Mastery: Tuning and Weapon Speed

Another cross post from the official forums.

There are some issues with combat mastery tuning that are causing a number of problematic tuning issues with the spec.

First SimC results:

https://www.dropbox.com/s/rhxwuppixnal30u/combat_weapon_final.html?dl=0

This shows three different combat weapon configurations, MH dagger, OH sword, double swords and MH sword, OH dagger.  All use the ilvl 660 pvp set with mastery enchants except slow_fast_haste_stack which uses haste enchants.  Fast_slow_RvS_maintain uses RvS only for debuff maintenance, all others use RvS as primary cp builder during deep insight.

As additional confirmation that this is not an artifact of SimC Vigilate tested Slow/Slow and Fast/Slow with 660 pvp gear and 680 tier gear.  This shows Fast/Slow ~5.6% ahead of Slow/Slow which is consistent with the SimC results.

Slow/Slow
Fast/Slow

There are a few problems with this:

  1. Fast/Slow is a very counterintuitive, I don’t believe that fast/slow has ever been the optimal weapon configuration for combat, and dagger main hands haven’t been a thing since combat daggers was killed in early wrath.  Beyond the historical rationale it just doesn’t make a ton of sense for the weapon that is hitting with spamable abilities like SS and RvS to be weaker than the one that is just there for auto attacks and mastery procs and has a 50% damage penalty attached.
  2. Even in the Slow/Slow setup the OH_dps is more valuable than MH_dps, this is problematic for the same reasons as above, the OH weapon being more valuable doesn’t make much sense.
  3. Combat passive damage (defined as auto attack, poison and MG damage) is very high.  In the fast/slow case it is 71.4% and even in the slow/fast case which should move more damage into active sources still has 63.4% passive damage.  I don’t think we have ever been told the target passive damage percentage however using the other two specs for comparison sub has 38.5% and mut has 34.6% in the ilvl 660 pvp gear.
  4. Despite SS being higher DPE and CPPE than RvS, RvS still appears optimal to use during deep insight (T17 2pc may change this).  Personally I don’t mind that however given that changes were made to prevent this a few build’s ago this is likely a problem.

The root of all of these issues is combat’s mastery is overtuned.  At 225% OH weapon damage an MG proc out damages a sinister strike (100% MH weapon damage) given equal weapons.  While this isn’t strictly tuning related having our main CP builder deal the same damage as a MH auto attack and RvS deal less damage than a MH auto attack seems very problematic.

Shadowboy did a useful analysis of this issue with slow/slow vs. slow/fast (before fast/slow was discovered to be comparable dps) and proposed a revised tuning that moves damage from MG into SS and RvS that appears to solve all of the issues above.  He proposes nerf MG to 140% weapon damage, buff SS to 160% weapon damage and buff RvS to 120% weapon damage.

Full post on his proposal with SimC results demonstrating its effect:

http://venomousthoughts.wordpress.com/2014/09/09/combat-off-hand-in-warlords-and-the-problem-with-main-gauche/

 

Late Beta Feedback Part 1

Another cross post from the official forum beta feedback thread.

Comments on three issues that don’t require new mechanics to solve.

1) The new Sinister Calling mechanic is much better than the old one because it eliminates the wonky snapshot and stack behavior or hemo however there are a few problems with the mechanic. The mechanic may break down at high gear levels, at around 60% MS, something that seems readily achievable with weapon enchants by the middle to end of the second tier the effective duration of rupture will be ~12-13 seconds. Given sub’s high cp gen the rotation should be reasonably sustainable but I’m not sure the managing a very short duration dot will be particularly enjoyable. The mechanic also makes it impossible at almost all gear levels to avoid refreshing rupture during FW, a generally important optimization for sub. At 12-13 seconds I believe rupture would be shortest duration active maintenance dot in the game.

Complicating the duration issues is SV, such a large portion of subtlety’s damage is tied up in maintaining an active bleed on the target that making those bleeds advance faster makes the rotation feel unforgiving and MS procs punitive despite being a net damage increase. It isn’t clear to me why SV as a mechanic needs to exist, rupture maintenance is already an important part of sub dps, elevating it with SV seems redundant.

With the change to SC its not clear why hemo snapshotting and dot rolling was removed since the mechanic wasn’t particularly problematic on its own. It seemed like a good solution to narrow the gap between positional uptime and lack of positional uptime. The mechanic was hugely problematic in the context of the old SC extra tick mechanic but with that mechanic changed it didn’t cause any problems.

Finally I mentioned in the AoE issues in an earlier post on sub AoE but the problem remains, even with the new version of SC. The MS interaction is sub’s big new mechanic and presumably the reason MS is sub’s fancy attuned stat however its value is diminished on AoE or in solo play. While stat weight variance between different situations is expected the lack of MS interaction on solo play and AoE seems problematic. This may also pose problems for subtlety on fights that do not allow high positional uptime, something that may be a problem even with the new, larger, backstab arc (this could be its own item but I’ve mostly given up on it).

2) I’m a huge fan of the new Bandit’s Guile mechanic, I think it has the potential to add a lot of dynamism to the combat rotation but the direction indicated by these two tweets (https://twitter.com/Celestalon/status/502167046247809024 and https://twitter.com/Celestalon/status/502222409139630080) seems worrying. If SS is strictly superior to RvS in every way, as suggested by the last tweet, then delaying BG doesn’t make a lot of sense. Not only are you reducing the uptime of the now 50% deep insight buff but you are also losing out damage and combo points versus normally advancing the rotation. This seems to remove the possibility of optimizations related to synchronizing cooldowns with BG. Shadowboy’s computations (http://venomousthoughts.wordpress.com/2014/04/21/bandits-guile-mechanic-in-warlords-of-draenor/) suggest on sustained single target a 5 second BG delay per cycle in moderate insight works out as dps neutral at best. This seems to relegate the new BG delaying mechanic to skipping enforced downtime, which seems like a waste of a potentially cool mechanic.

Even in the case of enforced downtime its not clear if BG delaying makes sense given the short duration of BG. I mentioned this in one of the first posts of the earlier thread, BG’s 15 second duration doesn’t make a ton of sense given BG’s rotational role as a pseudo-resource much like a monks TeB. Since RvS does not refresh BG using RvS to delay deep insight across even a short period of enforced downtime risks BG dropping completely. At the very least RvS should refresh the duration of the BG buff to make delaying less risky however even better would be increasing the baseline duration of non-deep insight BG. Combat already suffers more than other rogue specs on enforced downtime fights like Lei Shen due to BG falling off and the increased strength of deep insight (50% up from 30%) makes enforced downtime even more problematic. Given the substantial resource investment required to get into deep insight (600 energy, ~30-45 seconds) it seems odd that the buff should fall off so easily.

3) The combo point UI remains inadequate. We’ve talked a lot about the lack of change to combo point UI in this thread but it hasn’t been changed yet so lets talk about it again. The current combo point UI does a terrible job communicating the actual behavior of combo points, its no wonder many rogues believe combo points aren’t really on the rogue and are just being automatically redirected, that is what the UI communicates. Yes we can fix it with addons (kinda, more on that in a moment) but we shouldn’t have to use addons for the fundamental resource system of our class.

The funny part is addons can’t actually solve the issue because the GetComboPoints() function still returns 0 when we do not have an attackable target selected. Its still technically trackable by addons but it requires fully implementing the out of combat combo point decay logic which shouldn’t be necessary. We’ve heard the justification of “consistency” before but I don’t buy it, the functionality is no longer consistent so why should the UI have to be?

Finally on the topic of combo points there seem to be a lot of mechanics that just ignore anticipation. The assassination pvp 4pc bonus sets your combo points to 5 rather than generating 5 combo points. The sub FoK perk only generates extra combo points per target hit but not extra anticipation stacks. With combo points now on the rogue anticipation should work like the Monk talent Ascension and simply grant extra combo point bubbles on the UI. Combo points in the current system feel like a mess, caught half implemented and not totally functional. Three months ago this wasn’t a problem because beta is beta, things are expected to be half implemented and only partially functional but as we’re now in the polishing stage its very worrying.

Fixing Subtlety AoE

Another cross post from the official forums.

Talked with Vigilate in #ravenholdt about this earlier so its no surprise we agree but seeing sub AoE in action tonight on a fight like Beastlord where there are almost always 5+ targets made me rethink what I said previously about the sub AoE rotation.  The sub AoE rotation doesn’t feel bad but looking at the damage meters (obviously acknowledging that the tuning pass hasn’t happened yet), it feels really weak.

The key problem is none of our AoE abilities hit hard enough.  FoK and CT do not do enough damage, quick dummy numbers as sub put FoK at around 800 damage with CT at around 1.8K (mastery stacking gear set up for assassination),  To put these numbers in perspective, FoK is dealing less damage than an OH white swing while CT is comparable to an OH white swing.  Things aren’t as problematic for assassination due to the power of poisons but with subtlety, which has to rely solely on FoK and CT shows very clearly how weak they are.

In addition to raw number tuning the subtlety AoE and ST target rotations have basically nothing to do with each other.  You keep up SnD and rupture because of massive cp overflow but otherwise you are just spamming FoK->CT->FoK->CT ad infinitum.  All of the mechanics that define subtlety on single target, vanish, shadow dance, find weakness, fit into the AoE rotation about as easily as an elephant in a phone booth.  A simple solution to this would be a new AoE opener which applies FW to all targets it hits or even simpler FoK applies FW to all targets if cast from stealth or during shadow dance.  This doesn’t have the complexity of a single target FW but it allows sub’s AoE rotation to integrate more smoothly with its primary mechanics.

In addition to the FW issue sub suffers from the “s word”, stat scaling on AoE is messy.  Given the low number of FoKs per CT while AoEing subtlety has substantial energy overflow degrading the value of haste.  More problematically, the new sub MS mechanic is entirely non-functional on AoE when backstab is replaced wholesale by FoK.  I understand that stat weights will change in different situations but to have the mechanic that makes MS interesting as subtlety’s attuned stat be entirely non-functional on AoE seems wrong.  A simple fix would be to allow FoK MSs to proc additional bleed ticks just like backstab. Since FoK and backstab have the same energy cost this is unlikely to cause odd side effects in the single target rotation.  Combine this with the previous idea of AoE FW and potentially a damage buff to FoK and CT and subtlety could have a solid AoE burst niche, in line with its single target burst niche.

Beta Rogue Rotation APM

Another cross post from the official feedback forum.

A few days ago I posted my thoughts on rotations for each spec on beta, a couple days Vigilate posted his thoughts on combat on Sentry Totem which mostly lined up with my own.  One thing sorely lacking from both our analysis was numerical backing.  Today in #ravenholdt someone asked what the actions per minute for combat would be in WoD compared to MoP so I decided to compute this for all three specs.

A couple notes about these APM computations.  These are simple steady state models, they do not take into account time varying stats or temporary buffs unless otherwise noted.  Vanish is not taken into account which would increase the APM of each spec somewhat.  The stat amounts for each spec are based on the ilvl 660 premade gear which I believe is equivalent to WoD Normal/MoP Flex level gear, additionally to see if “haste will fix it” computations are done for two other haste points.  The upper haste point of 1500 haste rating is something I believe will be reachable during the first tier since the premade gear set is somewhat light on haste and not enchanted which can add an additional 540 haste.  The MoP models use WoD stats but otherwise uses the MoP launch mechanics, so for example 40 energy sinister strike and no ruthlessness for combat.

Mastery for combat and crit for assassination are held constant between the haste points.  All models assume Lemon Zest as the level 100 talent, the combat model adds 3 target LZ with blade flurry.  3 target LZ is not considered for other specs because AoE rotations are not set in stone yet and number of targets adds more complexity than it adds in clarity.

Some top line numbers (full results linked below)

Assassination_APM_WoD_No_Execute: 31.77
Assassination_APM_MoP_No_Execute: 37.44

Assassination_APM_WoD_Execute: 39.75
Assassination_APM_MoP_Execute: 34.32

Combat APM_WoD: 27.69
Combat APM_MoP: 28.85

Combat_APM_WoD_3rd_Haste_Point: 30.9 
Combat_APM_MoP_3rd_Haste_Point: 30.7
Combat APM_WoD_AR: 50.23
Combat_APM_MoP_AP+SB: 58.17

Subtlety_APM_WoD: 38.11
Subtlety_APM_MoP: 37.02

These numbers confirm a number of the observations made about rogue dps specs.

  1. Assassination is much faster.  A number of people have commented that assassination on beta feels much smoother than it does on live and these numbers bear that out.  Assassination gains more than 4 APM over MoP which helps make the spec feel much smoother at low ilvls.
  2. Combat is slower.  While both assassination and subtlety both gained APM going from MoP to WoD due to true haste buff and lemon zest, combat loses 1 APM due to increased SS energy cost.  Despite this combat’s energy regen appears to scale faster with haste than it did in MoP, note that despite starting 1 APM slower in WoD at the 3rd haste point combat is as fast as its MoP equivalent and if you were to extend that more, for example to T18 levels of haste, it would more than pass the MoP APM.  This could cause issues for combat down the line, while the enhanced AR perk will probably keep combat from ever GCD capping during AR as Vigilate has pointed out previously very few people actually enjoyed the 0.5 second GCD during ToT.
  3. The gap between AR and not AR is lower than MoP but still problematic.  The loss of shadow blades slows combat down by about 14% during adrenaline rush, however the APM gap is still quite substantial.  The drastic difference in speed between the two states exacerbates the sluggish feeling of the spec without adrenaline rush up.  Additionally, while I do not attempt to quantify restless blades interaction here the slowdown during adrenaline rush leads to a longer effective adrenaline rush cooldown which makes the spec feel slower still.

There are probably other pieces of additional information to be gleaned from the full spreadsheet so it is provided below.
https://docs.google.com/spreadsheets/d/1HPVnbxQFi4e5d9UtfvBmuZOITHEkVZVbUSFl81LaQg8

A number of people have expressed an interest in how these are computed. The spreadsheet isn’t particularly well documented but it should be understandable. At a high level these computations are based on the resource budget modeling technique discussed previously in the context of ShadowCraft.  If you have questions about the spreadsheet leave them in the comments or come join the conversation in #ravenholdt.

Beta Rogue Rotation Feedback

Cross posted from official forum feedback thread because I doubt many people are following the beta feedback threads too carefully.

Testing done with a level 100 premade.  Level 100 talent was Lemon Zest.

Single Target:

Assassination- Overall feels pretty good, energy regeneration in level 100 premade gear (ilvl 660) seemed about right especially with lemon zest proving an addition 0.6 energy per second.  There are two problems with the spec, vendetta and complexity.  Vendetta remains a really boring button to press with absolutely zero rotational interaction.  Especially compared to the cooldowns of the other two specs vendetta feels very lacking.  I completely forgot about the vendetta crit perk till I was writing this just now as I didn’t notice it during testing.  The spec also feels like its missing a mechanic.  The new emphasis on envenom is good but it isn’t enough to make the spec challenging and interesting.  The combination of the new pandemic mechanic plus the rate that spec plays at means that you’d be hard pressed to generate combo points fast enough to clip envenom.

Combat- Combat feels very slow, not just compared to current live speed but compared to the other two specs.  While combat does have faster energy regeneration than the other two specs the much higher energy cost per combo point makes the spec feel very slow.  The energy cost of sinister strike probably should be returned to 40 energy and then rebalance adrenaline rush so the spec doesn’t feel too spammy.  I’m also somewhat worried about the UI requirements of the new bandit’s guile system, I’m a huge fan of the design but this mechanic is going to require addons to play correctly.  You can track the progress in your head reasonably well but it requires a fair amount of concentration, especially if you are also trying to mentally track bandit’s guile progress (As a reminder this really needs better UI support).

Subtlety- Subtlety hasn’t changed much from live where it’s a lot of fun and so it still plays very well on beta.  The pace seems good and find weakness optimization is still a fun mechanic, albeit a frustrating one with the default UI.  The problem is that is really all the spec has going for it right now, the resource generation is high enough (even without full HAT saturation) to keep SnD and rupture up trivially without having to refresh during dance which was one of the major sources of complexity in the spec.  If the design intent is for subtlety is for the spec to be intentionally more difficulty it certainly isn’t living up to that now.  The durations of SnD and rupture need to be reduced substantially, at least 20% to 30% to make subtlety rogues have resource constraints that need to be actively managed.

AoE (Two AoE cases examined, priority target with incidental cleave on two targets and sustained three target AoE):

Note: Crimson Tempest is currently proccing Relentless Strikes per target which probably isn’t intended, I’ve tried to consider the impact of that additional energy to evaluate these rotations without this behavior.  I’ll also note that even if this isn’t intended it does make the current rotations better.

Assassination- Lemon Zest speeds up the rotation very nicely when you have multiple targets to cleave.  I’d like to see the default UI provide a bit more information about Lemon Zest, specifically how long your shortest deadly poison has left but that may be enough of specialized piece of information for addons to handle.  For sustained AoE rupture multidotting and weaving in crimson tempests is a lot of fun, on three targets it feels like you have infinite energy and crimson tempest as AoE poison damage increase synergizes nicely.  Even discounting the possible CT, RS bug (weaving envenoms rather than CT) the energy regen feels quite high. I don’t know if this is the intended or optimal damage AoE rotation for assassination but it is a great rotation so I hope it is.

Combat- The combat single target rotation is slow and AoE is even slower.  The blade flurry energy penalty makes the spec feel even worse.  Blade flurry provides passive maintenance for lemon zest which helps counter the blade flurry energy penalty but still slows the rotation substantially.  For sustained AoE CT basically effectively replaces eviscerate with a time to die breakeven point depending on the number of targets.  CT does not however feel like an exciting button to press, at least not without the possibly unintended RS interaction.  The same issues that plague the single target rotation plague combat here, the rotation feels incredibly slow and when compared to the way the other two specs speed up on AoE it feels even worse.

Subtlety- Combo points per target on FoK feels exactly as overpowered as we all expected but given subtlety’s current AoE weaknesses on live I’m not sure that is a huge problem.  The combo point overflow exacerbates the lack of resource limitations and multiple target lemon zest maintenance is a bit more frustrating for subtlety compared to the other two specs but it isn’t a deal breaker.  On sustained AoE CT as a rolling dot can result in some pretty beefy dots on a lot of targets but doesn’t provide a lot of AoE burst which seems like a good tradeoff.  The one complaint I have with sub AoE is the general lack of interaction between its major single target mechanic, shadow dance, and AoE, it makes the two rotations feel somewhat disconnected.

To recap:
Assassination Single Target: Good pace and flow, vendetta is boring and spec doesn’t have enough skill rewarding mechanics.
Combat Single Target: Way too slow, BG needs better UI support and may still be too hard to manage without addons.
Subtlety Single Target: Good pace, dance is still a good mechanic, not enough resource limitations to challenge players.
Assassination AoE: Good pace and flow, rotation is really fun.  Awesome design, please don’t change.
Combat AoE: SLOWWW, not particularly interesting but basically function.
Subtlety AoE: Good pace, very high combo point generation is fun, may need more interaction with single target.

Welcome to #ravenholdt

After months of looking jealously at Death Knights and their awesome IRC community #acherus we rogues have decided to stop lazing around and blatantly rip them off. #ravenholdt is the new rogue IRC community on quakenet. The goal of #ravenholdt is to create a place for the rogue community to chat, sometimes about class theory, sometimes about raiding, sometimes about nothing in particular. If you are some kind of old school gamer who has been hanging out in IRCs for years you can skip the rest of this post, if not some IRC basics.

What is IRC?
Wikipedia helpfully defines IRC as “a system that facilitates transfer of messages in the form of text.” That really is about all there is to it, IRC is an old technology for transferring messages over a network. Owing to its age IRC is pretty barebones but unless you want to do fancy stuff it is also pretty easy to use. IRC servers are organized into channels, separate groups of users who can communicate directly with each other. Channel names always start with a hash sign (#).

How do I join?
To join an IRC channel you need an IRC client, there are a wide variety of IRC clients out there for just about every platform. Since I’m just trying to cover basics I’m just going to talk about the quakenet web client.
http://webchat.quakenet.org/
From this accessing IRC is very simple, put in a name and #ravenholdt and you are set.

I’ve joined the channel, now what?
Chat, there isn’t that much to it. A couple useful commands:
/msg user message- sends a private message to another user.
/nick name- changes your current nickname

What is this “Log in to Q” button for?
You can register your user name with quakenet by following these directions.
https://www.quakenet.org/help/q/how-to-register-an-account-with-q
Registering an account allows you to keep user settings between logins but it isn’t necessary to join or chat in #ravenholdt.

What does the “@” by some peoples names mean?
The @ symbol means that a person is a “channel operator,” basically a moderator for the channel. We don’t plan to need a lot of moderation but we’ve got some mods just in case.

There is a lot more to IRC than I’ve covered above but that should be enough to get started. Welcome to #ravenholdt.