The most important consideration when optimizing dps is often overlooked: the goal of maximizing dps is to increase the likelihood of a raid’s success.
Simply put, we deliver damage so we may kill a raid boss, but not all damage or means of delivering damage are equally effective at ensuring raid success on a given encounter.
There are two basic ways to focus damage delivery: in short bursts, or sustained over time. The number of targets important to attack also varies from encounter to encounter. Each of these factors is important to consider when optimizing dps for raid success, as different abilities are better suited for different situations, with different ways to optimize their use.
The key here is that not all encounter situations are equally important to maximize for overcoming to ensure raid success. Some aspects of an encounter may be easily overcome, while others may require tweaking and optimization to overcome.
Consider a basic encounter with a non-threatening enrage timer, but an add randomly spawns that will wipe the raid if not killed in a short time. In such an encounter, optimizing your damage delivery for burst has a larger effect on the success of your raid than optimizing for sustained. Conversely, if killing the add is easily accomplished, optimizing for sustained would be more beneficial since it would help finish the encounter more quickly, reducing the risk of a wipe.
When working on an encounter that your raid has never defeated, success should be your primary concern, while rapidity may be if your raid has defeated the encounter many times.
I’m sure most of this is common sense to many of you, but I imagine the majority of raiders don’t fully consider which aspects of an encounter are critical to optimize for, or at least all the options available for optimization.
I’d like to focus the discussion in this thread on the varying ways we can optimize dps for the different types of situations we face during a raid encounter. It seems the most common methodologies for ability use and talent and item selection focus on maximizing sustained dps on a single target, but very rarely is that the sole concern, so it would be helpful to have clearly defined ways of dealing with the many other situations.
For example, it may be more effective to use wound poison on both weapons with a combat spec if your primary concern on an encounter is killing an add as quickly as possible, while deadly poison may provide more sustained damage given sufficient time to stack.
Another example would be the current primary raiding concern of my and many other guilds: killing heroic Anub. This encounter has fairly atypical damage delivery priorities – frequent periods of burst AoE with a reasonably high sustained dps requirement. In this situation, it is likely suboptimal to use rupture, so a talent/item build that maximizes rupture damage over abilities such as fan of knives or eviscerate would be less effective.
If you have any different types of examples, or theories on what is optimal for a given type of situation, please feel free to contribute.
I believe consolidating discussion of this nature in one place will help enable raiders to identify what decisions they should make to most effectively benefit their raids.
tl/dr: It is important to consider optimizing all methods of damage delivery, so let’s identify the best ways to optimize damage delivery of each type: burst, sustained, single target, and AoE.
I like where this is going but I think for clarification purposes it would be helpful to define what burst really is.
For instance, in your example you mention an add that comes out and must be killed or it will wipe the raid. At face value this is a burst situation but only if the add can be killed within the span of a cooldown or some other short measure. In a situation like Heroic Vezax, the Saronite Animus is certainly such an add as you describe but because his health is probably somewhere around 1/3 that of Vezax, the idea of "bursting" it down doesn't work. In such a situation, sustained DPS is still the way to go.
I'd argue that Heroic Anub'arak and his adds works in a very similar way. While each add themselves may not have a huge health pool, the period of time where AoE is required is very much a significant portion of time. I'd be hesitant to call this a burst period myself unless your raid can burst high enough to put significant time period separation between groups.
I really think the best example of what defines the need for periods of burst where a "secondary" target takes priority over the "primary" targets mechanics is a fight like Heroic Jaraxxus.
Yes, I left that specific example purposely vague, as there are extremes and grey areas for different time frames depending on each fight.
Some relevant time frames that I can think of:
1 to 30 second duration
30 to 120 second duration
120+ seconds
I also recall Aldriana doing some calculations on the crossover point for choosing between wound and deadly poison for the off hand and that being somewhere around 150 seconds.
If a fight is under 30 seconds, you can use all your cool downs exactly once.
If it is over 30 but under one 120, you can use kspree twice if glyphed, otherwise only once, and so on.
Jaraxxus was probably the fight that got me thinking a lot about this. When they buffed him, my raid wiped a bit because the portals weren't dying fast enough to prevent two spawns, but we were killing jaraxxus fast enough to beat the enrage timer, so I started thinking of ways to maximize portal damage even if they sacrificed damage on Jaraxxus.
And yes, Anub is a good example of a fight where dps strategy can vary with your guild's strat due to the complexity of the fight. Overall it is many periods of aoe that add up to a fairly sustained amount of aoe throughout the fight, but if your raid is wiping because the adds don't die before another wave spawns, focusing on one wave of adds at a time may help you get over the hump.
Rogue DPS Simulation Spreadsheet can be used as a first estimate of burst dps, especially if you do not go below 30 seconds. If you reduce the estimation time in Inputs tab to 30 or 120 seconds, it will measure the dps of your spec in the first 30 or 120 seconds, which can give you an indication of your burst dps. I am saying an indication because in a real fight situation, you typically go into the burst phase with SND already running, whereas in the simulation, the first few combo points are used to start up snd. All the cooldowns are used around 10th second, with heroism, AR, BF, on-use trinkets, speed pot timed together. Killing spree is used around 7th second of the fight, just before all other cooldowns are popped. Also the proc trinkets will proc as expected based on their triggers and proc rates, in the first few seconds of the fight, and their effect on your burst dps will be properly measured.
You can change your specs, glyphs, use of abilities and choice of poisons to determine which spec comes out ahead in terms of burst.
The other aspect of this discussion related to the number of targets is more intriguing since I have not seen a good working model of it yet. It's when you model or simulate the use of FOK, blade flurry, and regular attacks while maximizing your AOE damage. I suppose the simsheet can be modified to use FOK instead of regular combo point builder to estimate AOE dps. I already have an option to estimate DPS assuming there is an extra target for blade flurry to hit, but FOK is not currently an option to use.
Best burst options with no cool downs available -- combat evisc v mutilate with pooling, etc
Balancing burst and sustained -- say you have a fight with a tight enrage *and* dangerous adds, you can change priority of your actions on the fly (pooling before fok versus continuing on the boss, weapon swapping, etc)
On fights like Jaraxxus or Twins, burst phases may come up before your cool downs do, so staggering is an option to balance the dmg output amongst them, but is otherwise inefficient use of each cool down over time.
I'm not sure I see the purpose of this thread; there are just too many variables to consider.
From a theorycrafting perspective you'd need to break this down to time chunks, time spent moving, whether SnD/HfB/rupture/EA/sunder/debuffs are already up, time in between burst phases, number of other players involved in the switch and their need for debuffs/buffs you provide, number of targets... the list includes too many variables.
Trying to account for each situation essentially amounts to an impossibility, and trying to discuss it will essentially turn into an encounter-by-encounter assessment of a rogue's job.
Furthermore, there are situations where your burst on a high-priority target is actually unnecessary; if your guild would kill a Jaxx portal before the second mistress without your help; you might be better served not switching target to it at all (instead dpsing the mistress to start an EA, or kicking jaxx, or god knows what else). Likewise for a Twins shield phase, perhaps (especially if you have empowered buff)...
What it fundamentally boils down to is "not all damage done contributes equally to the success of the raid"; important to recognize, but almost impossible to model.
I can see the use of answering some FAQs such as "how long do you need to stay on a target for DP to be more useful than WP for combat", "how many targets do I need for FoK to be better DPS than single-target rotations", and "how many targets do I need for FoK to be better DPS than anything, including keeping up SnD" (if this happens?); "how long do we need to DPS a target before sunder from DPS warrior becomes more efficient than EA" "Is it ever worth talenting/glyphing EA; if so, under what circumstances" but most other situations are just too complex or variable to model.
Trying to account for each situation essentially amounts to an impossibility, and trying to discuss it will essentially turn into an encounter-by-encounter assessment of a rogue's job.
I disagree with this. The rogue class has only so many things it can do and it takes a full group with a combination of classes/specs to fight through the various things a boss throws at you. As such, there is a finite number of situations that can be explored where rogues are useful and I don't believe they're different enough for every boss that you'd have to describe every single encounter separately.
For instance, in your example you mention an add that comes out and must be killed or it will wipe the raid. At face value this is a burst situation but only if the add can be killed within the span of a cooldown or some other short measure. In a situation like Heroic Vezax, the Saronite Animus is certainly such an add as you describe but because his health is probably somewhere around 1/3 that of Vezax, the idea of "bursting" it down doesn't work. In such a situation, sustained DPS is still the way to go.
As Leto referenced, Aldriana has done calculations based on fight length and which offhand poison to use. A factor that would make things like the Animus in General hard a "burst" type of situation, at least in regards to wp\wp, is that heroism is generally up for a significant portion of your Animus dps. This is an assumption (a fairly safe one I think), but I think something like 50% heroism uptime would absolutely favor double wound over using DP in the OH. I know some of you dislike weapon swapping because of a conviction that it will be nerfed, but weapon swapping is another great way to up your burst on a fight like Animus, that lasts more than a few seconds and has a high heroism uptime.
I'm glad to see a more detailed discussion on this - it will be valuable for new bosses in icecrown as well.
Combat's popularity as a spec can likely be attributed to the option for burst. As mutilate I used to get railed on for low Steelbreaker damage, even though overall DPS was very close to combat rogues I had no cooldowns to blow on the most important part of the fight. Combat continues to have that utility on fights like twins (burst down shields or use with empowered buff), Icehowl (stun phase perfect for cooldowns), Jaraxxus (Volcanoes and portals), and to a lesser extent Anub'arak (burning down certain add waves that line up poorly with shadow strike).
Perhaps it would be a good starting point to review previous and existing boss encounters and strategies in order to distill common themes and situations wherein a rogue would need to differentiate, hypothetically, his gear, talents, glyphs, and play style. From there, we could try to input the parameters for these situations into existing models and determine if what, if any, differences actually exist.
Furthermore, I think that a holistic, raid-centric discussion of rogue performance is an area that has been somewhat lacking due, in part, to changing raid dynamics over relatively short time scales (1.5-2 years). This sort of discussion will bring to light many important issues with regard to damage output beyond the typical average DPS, such as damage acceleration (rather than damage rate), the roles of expertise and hit rating in reducing damage variance, and so on. I do think that some of these questions are going to be quite difficult to accurately address using Excel spreadsheet-style models and would demand something much more sophisticated.
I've started putting together a spreadsheet to model combat FoK dmg with a variable number of targets and various states (slice and dice active, autoattacking or not, sundered, faerie fired, etc), and so far the results are interesting.
I haven't modeled overall dmg while using FoK yet, only the individual dmg and dps contribution of FoK itself, but it is looking like it may not be worthwhile to maintain slice and dice for sustained aoe when using FoK on 3 or more targets. Note this is for the aoe contribution... obviously not using slice and dice would hurt your dps on your current target.
It isn't fully modeled, but I'll walk you through my logic using best case examples so someone else may chime in if they see any flaws. On 3 targets with roughly my stats/spec using a 2.6 speed 196dps main hand fist and 1.8 speed 216dps dagger off hand, I found average FoK dmg (including poisons -- double wound) to be ~15600. That puts the dpe of that FoK around 312. The minimum energy required to maintain snd would be 40 energy (1 ss) per 22 seconds, assuming a relentless, ruthlessness, and ss glyph proc to refund snd cost, and add 2 combo points. With a dpe of 312, that 40 energy could have yielded 12480 FoK damage. 40 Energy used over 22 seconds is 1.82 energy/second. With my haste, a 1.8 speed off hand, and raid buffs, snd yielded an additional .88 energy/second gain through combat potency, so a net loss of roughly 1 energy per second by maintaining snd.
Some other interesting things I found:
FoK appears to out dmg single target dps with 4+ targets... 3 is close, but I wouldn't feel safe hinting at that until the model is refined more.
Using a 2.7 speed Mh likely yields more damage than a 2.6 speed main hand multiple tiers of itemization ahead (I checked from 188 dps to 216 for example).
A 1.8 speed dagger in the off hand is likely the best balance of combat potency regen + fok and poison damage (2.6 or 2.7 speed weapons hit harder, but you can FoK less).
Mace spec is looking to be the best for aoe, followed by cqc, but I still have to split up the armor penetration by hand as well as the crit chance.
Using a 2.7 speed Mh likely yields more damage than a 2.6 speed main hand multiple tiers of itemization ahead (I checked from 188 dps to 216 for example).
Care to speculate on why this would be the case? For instance, I'm looking at using The Facebreaker vs. Golden Saronite Dragon which have average damages of 511.5 and 508, respectively. In addition Facebreaker has what I would guess are better FoK stats (more AP, crit, and ArP). However, I suppose GSD would have a higher chance to proc Wound. Is it possible in your sheet to break out physical vs. poison damage generated by the FoKs?
The 1.8 vs. 1.4 makes sense to me with respect to the Anub fight in particular, since each wave will be killed much faster than the amount of time it would take for you to regenerate enough extra energy to produce an extra FoK
Care to speculate on why this would be the case? For instance, I'm looking at using The Facebreaker vs. Golden Saronite Dragon which have average damages of 511.5 and 508, respectively. In addition Facebreaker has what I would guess are better FoK stats (more AP, crit, and ArP). However, I suppose GSD would have a higher chance to proc Wound. Is it possible in your sheet to break out physical vs. poison damage generated by the FoKs?
The 1.8 vs. 1.4 makes sense to me with respect to the Anub fight in particular, since each wave will be killed much faster than the amount of time it would take for you to regenerate enough extra energy to produce an extra FoK
Ah yes, thanks for noticing... I didn't adjust the stats when switching, just the speed/dps. It was a ~150 dps upgrade on 5 targets, but now it is showing as a ~40 dps upgrade, so it would likely be a downgrade overall since you'd be doing less dmg to your target with GSD. The main stat disadvantage was the loss of the armor pen. From spinebreaker to GSD dropped me from 728 to 677.
The reason GSD hits harder is the weapon speed -- it gets more damage from attack power, and a higher chance to proc poisons. The poison dmg per hit went up ~30, while the physical dmg per hit went up ~50.
Some things to keep in mind when modeling FOK damage:
1. When determining whether to snd or not snd, what's the optimal number of combo points to put into an SND. Result will obviously vary by fight length and number of FOK targets
2. How to calculate FOK damage itself. It depends on your level of gear. The value differs greatly based on your buffs and debuffs. Things to consider: 4% physical from wound debuff (can assume it's up), 3% crit (may or may not be there depnding if you gain it from paladin or ele shaman or rogue), sunders - aoe tanking, slow rate of sundering or no sundering, faerie fire - 1 target if lucky, 5% spell crit - would not count on it, 13% spell damage - only if you have an unholy dk. With full buffs (i.e. full sunders on each target etc), I find that FOK with double wound in BIS gear and 18/51/2 spec will do 5.5k per target (of that 1k is wound poison damage). However, that number can drop by about 1k if there are no debuffs (save the 4% physical vulnerability, you apply on your own).
3. What energy regeneration to assume. Part of this is whteher autoattacking is on or not, another is whether FOK procs focused attacks and combat potency. There were some 3.2 changes that were related to combat potency and all offhand attacks, but I can't recall whether it applied to FOK. Also only MH attacks procing focused attacks, is that still true?
Either way, I still intend to do an independent model of FOK damage and we can compare the results.
Ah, can't believe I forgot savage combat. I got the other buffs/debuffs though, but I'll double check tomorrow.
With regards to snd length, I didn't model it, but I looked at Aldriana's snd cost modeling and did best case analysis and it still didn't seem worth it. Unless I'm mistaken, one ss for 3 combo points would be more energy efficient than 2 ss's for 5 combo points, and both those situations would be rare.
For energy regeneration I took the formula from Aldriana's sheet with regards to probability of a hit, etc. It was showing around 15.5 energy per second with vitality and combat potency with a 1.8 speed off hand.
I haven't messed around with mutilate FoK modeling yet, mainly because FA still doesn't proc off the offhand attacks to the best of my knowledge.
Jaraxxus was probably the fight that got me thinking a lot about this. When they buffed him, my raid wiped a bit because the portals weren't dying fast enough to prevent two spawns, but we were killing jaraxxus fast enough to beat the enrage timer, so I started thinking of ways to maximize portal damage even if they sacrificed damage on Jaraxxus.
Might not be the place to ask this question, but... do the portals have an armor value? I've always thrown up a 1CP Expose Armor on them to help melee down them quicker. Is it a wasted global or is this helping?
Also, on an unrelated note....
This probably crosses into the territory of anecdotal evidence, but I'm gonna offer it anyway because we're just talking here, right? As far as Glyphs go I've always personally felt that, as counterintuitive as this might seem, Glyph of Slice and Dice was better for my overall DPS than Glyph of Killing Spree for a number of reasons:
1) I often feel like Killing Spree is wasted if there is more than one target because I can't control what it hits. I tend only to use it when there is only one possible target it can hit, or when I can pop Blade Flurry at the same time so it's more likely to hit what I want it to hit if there's another or multiple targets. Taking this into account, the fact that it has the same cooldown as Blade Flurry is kinda convenient.
2) Any fight in which I have to switch targets brings with it an increased chance that I'll have to burn or waste combo points on my current target when the new one spawns. Since these combo points often turn into less-than-maximum duration Slices and Dices, the extra couple seconds from that glyph becomes really valuable. I feel like this glyph is ALWAYS incredibly useful without having any inconvenient conditional requirements -- if nothing else, the benefit it provides to my quality of life as a rogue outweighs the raw added DPS of another glyph like KS or TotT on whatever fights those glyphs shine.
Below is a perfect example of the kind of question that points back to my thoughts on the general usefulness of the Glyph of Slice and Dice. I'm not saying that Glyph of SnD trumps Glyph of FoK if you had to pick only one for an AoE spec/fight; however, assuming it's worth keeping SnD up even while fanning, if you only raid with one spec, Glyph of SnD is at least still useful here too.
Originally Posted by Mavanas
Some things to keep in mind when modeling FOK damage:
1. When determining whether to snd or not snd, what's the optimal number of combo points to put into an SND. Result will obviously vary by fight length and number of FOK targets.
Well, the benefits of Glyph of Killing Spree are somewhat fight-specific, to be sure; and, quite honestly, ToC as an instance doesn't necessarily lend itself to the glyph - it's good on beasts, but it's significantly less impressive on every other fight in the instance. It can be good depending on your strat for Twins, but is largely wasted on Faction Champs and (I suspect) Jaraxxus - though I admit I haven't totally verified the timing on that last one. Anub is certainly not a KSp glyph fight, but then, you want to be getting the FoK glyph for that fight anyway, and (depending on strat) possibly Tricks as well, so you'd likely be swapping out everything buy Glyph of SS anyway.
Now, that said, what to get instead doesn't strike me as at all an obvious decision. I suspect the spreadsheet will tell you that the answer involves getting glyph of rupture and going low-rupture instead of evis-only, but I'm not convinced that proves to be optimal either. It's something worth thinking about, though.
Well, the benefits of Glyph of Killing Spree are somewhat fight-specific, to be sure; and, quite honestly, ToC as an instance doesn't necessarily lend itself to the glyph - it's good on beasts, but it's significantly less impressive on every other fight in the instance. It can be good depending on your strat for Twins, but is largely wasted on Faction Champs and (I suspect) Jaraxxus - though I admit I haven't totally verified the timing on that last one.
I would think it's wasted on Jaraxxus. You can't use it on the portals without first popping Blade Flurry and tricking to a tank because the first Mistress spawns too quickly. Same with volcanoes and infernals. So on its own cooldown, what can you do with it except burst Jaraxxus between portals/volcanoes? DPS on him isn't likely to be the deciding factor in that encounter.
I have to agree, KS is godly on portals, 3-4 rogues ToTT+KS on portal and it is down in seconds, way before portal even starts thinking of spawning another mistress. Without KS, we have had 2 spawn like 90% of the time - until ranged switch...
Do you Blade Flurry first, or no? Because there's a 50% chance (or 66.6~% chance if Jaraxxus is in range) that your Killing Spree doesn't even hit the portal every time if not...
So, if you move quickly and position correctly, you can get all the KSp hits on the portal with the possible exception of the last one, which you can make "safe" by Tricksing the add tanks before so doing. And doing so does absolutely ridiculous portal damage - which is a good thing. However, the value of KSp is not the question here - the value of the Killing Spree *glyph* is the question. And so far as I know the portals are exactly 2 minutes apart, in which case you can get every one with or without the glyph, making the glyph less than wholly useful.
Oh, I know Killing Spree itself is good for the portals. I just didn't think it was easy, or even possible, to guarantee that much of its damage would be dealt to them without also using Blade Flurry (which is still probably a good idea, because it allows you to target the Mistress for combo point generation even before the portal is killed). Of course, if you have four rogues who can KS the portal like that, it might be dead fast enough that it doesn't matter. I've never been lucky enough to be in that kind of raid (although I wouldn't phrase it that way when rogue loot drops and I had to roll or bid against three other people for it). = P
As far as I can tell, the Nether Portal has similar armor to Jaraxxus, since my average hits were similar checking a couple parses, but it may be different.
That is an interesting point about the slice and dice glyph with regards to aoe. If I'm not mistaken, that would drop the minimal energy consumption to maintain snd to 1.6/second, which would still lead me to believe it inefficient to maintain for sustained aoe (where aoe dmg is the priority... you might see more overall damage from autoattacks/sinister strike), but it might be useful on certain fights. On anub for example, if your raid kills the adds fast enough, you may be able to get more snd time to cover the next add phase from the cp generated while dpsing anub.
Another thing to consider that has been highlighted by my analysis is just how important autoattacking is while using FoK. With 5 targets for example, I saw the combat potency regen from autoattacking add around 1500 dps.
Make sure to setup your ui to easily indicate whether you are autoattacking or not -- I use a separate mik's pane that scrolls faster than the others.
Ald is completely right, glyph of KS is useless on Jara. We are using 2 to 3 rogues and that is enough to easily get the portals down with no additional spawn. The tactic we are using is SS the portal while getting to it, trick another rogue while positioning furthest away from the boss and activating KS right away. This way KS never jumps to the boss.
When portal spawns it looks like:
PORTAL - - - - - ROGUE BOSS
When I'm positioned before KS is:
ROGUE PORTAL - - - - - BOSS
I don't use BF on portal, because we have to trick add tank and lose quite much of dps on the portal itself, and funny enough KS started before first mistress spawn, does not hit it.
After portal is dead, we don't touch the add, but going straight to the boss, while tank drags the mistress to the boss too. Then all rogues activate BF and add is dead without losing even a hit from the boss' dps
Rinse and repeat.
Might not be the place to ask this question, but... do the portals have an armor value? I've always thrown up a 1CP Expose Armor on them to help melee down them quicker. Is it a wasted global or is this helping?
I went back to last weeks logs and had a look and I think the sample size is just too small to say. Maybe someone else might want to do some testing at some point but I believe they probably do and should have EA put up on them (and the volcanoes as well). Anecdotally, it seems we always get the portals down faster with EA than without. (And yes, I run double wound for these so the Savage Combat debuff goes on and off like crazy in the logs; I included that just in case.)
10/09/09, 4:03 PM
Similar Threads
