The current state of Affliction theorycrafting and guidelines for gearing

[Hepes]

Thank You for posting this I found it super helpfull

[Spektroman]

Then try achieve even levels of both stats through your gems.

I'm not sure I get this. Are you saying achieve even level of stat ratings? Like paladin tanks during Cataclysm tried to keep dodge/parry ratings even?

[Scathbais]

The best argument for getting full hit cap might come from checking what the top Warlocks have done since the release of MoP. Even back in MSV most of the Warlocks in the hardcore progression guilds went for full hit cap. I think we can assume they are all at the skill level of being able to react to misses. Still they didn't think it was worth it and even if it might not be backed up theoretically we should probably assume they have good reasons for sticking to the hit cap.

Finally, it's been pointed out that as we get closer to full T14 BiS gear we are not losing more DPS from aquiring the hit cap. In other words it's likely that the value of hit gets higher with better gear and it's safe to assume that the gain of not reaching the cap will be very low in future tiers.

I have been debating hit cap with another Warlock and I no longer agree with what you wrote here.
If I recall correctly, affliction locks were below hit cap in MSV/HoF/ToES UNTIL their gear allowed them to hit cap while also hitting relevant haste breakpoints.

My guess (can't simcraft it at work) is that once we get into 5.2 and beyond, affliction locks will once again drop below hit cap if it will allow them to hit the 9211 (extra corr tick)/9778 (extra agony tick) breakpoints.

[ati87]

My question is, assuming we don't try to go for hitcap but we don't complety ignore it.

At which point do we hit a sweet spot where you don't miss allot of spells but you aren't sacrificing to much dps

[Shauni]

12% is correct. Especially if you have some expertise racial it becomes just sweet.

[Deleted]

Hi,

i've been trying to maximise stats on my warlock and came to some strange results with simulationcraft. It tells me to ignore 4717 haste point and swap some haste with mastery (like 500). I've run the simulation many times with the same result.



Should i ignore that and keep the stats as they are (4717 haste point, and rest into mastery until i can afford 6,7k ish haste without hurting mastery) or follow simulationcraft suggestions?

My character armoury link

[Deleted]

well i would prolly pick an amount on your scaling high enough to include the next haste breakpoint and see from there, sense we're pretty close in terms of ilvl, chances are you'll see a minor dps increase at 6637 haste, atleast it does for me, altho its only 5-600 dps, so my suggestion is to make the reforging plots to 2000 and then see if there is a jump in dps at that mark, and see how big it is and then i would see what i need to reach that point and if it is worth it but you also need to keep in mind that it is "just" a simulation, it isnt based of actual encounters, and the only way to be 100% sure if it is worth it, is to do 2 runs, 1 with 4717 haste and 1 with 6637 haste and compare the two, however the general idea, i think, is that you shouldnt go after the 6637 breakpoint until you reach ilvl 495ish and you can reach it without compromising your mastery but personally, i would always pick haste over mastery bcoz i like it a lot better but that is just an opinion, and simcraft says that im wrong.

also, the essence of terror trinket isnt really a trinket you should use as afflic(im assuming you're using it as afflic, altho you're logged out demo) due to its low uptime and light of the cosmos is afaik a better trinket, the only time where EoT is worth using over normal or hc versions of LotC is when it is the EoT hc, otherwise use LotC, atleast thats what ive read.

[Whydrood]

I've had some debates about haste and hit caps. The best explanation I can come up with for telling people to aim for hit cap is this:

Min-Maxing is important to world/realm first players. These players were progressing into heroic modes in 463 blues while the majority of us were still wiping on stone guards or trying to figure out how to start the quests in Dread Wastes at 89. Facing that content at such a low gear level they needed to use every advantage they could to beat the enrage timers or deal with these mechanics. It is not unheard of for these players to switch specs/gear sets/reforges or even hopping on another toon to down a boss as fast as possible. While hit rating is a DPS loss when trying to maximize your DPS, for the majority of players it is actually more beneficial to be hit capped while you learn the spec and encounters, and the majority of warlocks on these forums would agree. I wouldn't suggest to a world/realm first player to aim for the hit cap as I would expect that player to know his class/spec well enough to be able to quickly deal with a miss. At this point in the expansion players are generally not going into MSV with a full raid of under geared players, so more and more the enrage timers that the 500 or so dps gain you get from not being hit capped becomes less and less beneficial to newer level 90 warlocks. Crafted gear (476 and 496) as well as 476 and 483 LFR gear are easily obtainable now so there is no excuse to tell someone not to hit cap just to reach a certain threshold. A world/realm first player isn't going to be coming in here reading the guides posted and say "hey I guess since someone on the forums says I shouldn't hit cap, I shouldn't hit cap." I however do expect this behavior from new warlocks or people who haven't played their locks in a while.

As far as haste thresholds, I still maintain that they are important as long as you can reach them without sacrificing too much hit or mastery. It is not so much the amount of ticks each cast gives that is important, as it is the frequency of those ticks (Thanks to Somewhat in the official WoW forums and Zagam from Icy-Veins that explained this to me). Here is an example I used on the offical WoW forums that pretty much explains the haste cap/pandemic debate:

Think of it this way, a boss has an ability, lets call it Rain of Blades. This ability does 45,000 damage to all raid members every 0.5 seconds, and lasts 6 seconds (hitting 12 times), this is a total of 540,000 damage over 6 seconds. Lets assume this boss just stands there and casts nothing but Rain of Blades for 5 minutes and has Pandemic meaning this dot never falls off. Over the course of 5 minutes, this ability will tick 600 times, meaning the boss does 27 million damage to every raid member with this ability.

Now lets give the boss some haste, enough haste to add 3 ticks onto this ability. With this haste, the tool tip for Rain of Blades now reads "Inflicts 45,000 damage to all raid members every 0.4 seconds, and lasts 6 seconds", (hitting 15 times). Now the boss is doing 675,000 damage over 6 seconds. Over the course of 5 minutes, this ability now hits 750 times, meaning the boss does 33.75 million damage to every raid member with this ability.

In gaining that haste, even though the dot never drops off the boss is doing 6.75 million more damage, to each raid member. In a 10 man raid, that's 67.5 million more damage. In a 25 man raid that's 168.75 million more damage.

Even with pandemic and 100% up time, the frequency of your ticks matter, this is not debatable.

Now looking at this example, it really doesn't explain how pandemic works in, so let me try and explain that as best as I understand it. The haste buffed Rain of Blades ticks every 0.4 seconds and lasts for 6 seconds. If the first cast of Rain of Blades is refreshed with 3 seconds remaining, the dot will have a new timer of 9 seconds, but it will still be ticking every 0.4 seconds. You aren't interrupting the ticks, you are just extending the duration of the dot. Now when you are just below a haste threshold where getting a bit more would add an additional tick, you should also note the duration of the dot, as this becomes important when multi-dotting or on fights like feng where timing refreshes becomes important. Take for example Corruption. At 2302 haste rating (non-goblin without the +5% spell haste), corruption only lasts 17.07 seconds. At 2373 (closest I could get to the 2367 threshold) it lasts 18.94 seconds. So not only did I gain an extra tick of corruption by adding 71 haste, I also increased the time corruption would be up with pandemic by 2.8 seconds, which is substantial. The same thing happens at 4700 and 4717 haste rating with a 5% spell haste buff. You gain nearly 3 seconds on corruption just by gaining 17 haste. Also haste decreases the channel time of MG, but increases the frequency of those ticks, meaning more damage overall from MG/DS. I don't have exact figures, I have barely begun clearing HoF normal, and keep in mind this doesn't explain exactly how heroism/DS:M/BL effects dots, but it gives a general idea of why I personally think haste thresholds are important enough to go after.

[Ridcully]

At 2302 haste rating (non-goblin without the +5% spell haste), corruption only lasts 17.07 seconds. At 2373 (closest I could get to the 2367 threshold) it lasts 18.94 seconds. So not only did I gain an extra tick of corruption by adding 71 haste, I also increased the time corruption would be up with pandemic by 2.8 seconds, which is substantial.

I think this is the main reason to go for haste breakpoints. It also means that you shoudn't go too high above them, it will decrease duration.

I done some tests, Pandemic maximum duration formula seems to be floor(number_of_ticks_without_pandemic*1.5)*tick_duration + time_until_next_tick
So technically even number of ticks provide a longer maximum possible duration, up to 1 second to be exact.

[Chrispotter]

Just go for full hit cap. Want to wipe on elegon heroic because your haunt missed the spark? Simulations does not = real raiding. Yes that Sim can go 20000 Iterations. Are you going to pull the boss 20000 times? No?, Try 50 Iterations ( closer to a real pull number for most people), its very possible your theoretical dps is like 75% of what it is on the 20000 Iterations, because of RNG misses. Just go for full hit cap or get fustrated and raged at. Missing haunts with Cd's up is a big deal.

[Deleted]

well im thinking simcraft is bugged or something, sure im only in ilvl 491 gear, but depending the skilllevel i put(good or elite) in haste becomes more valuable compared to mastery for me, at elite its like 2,52 for haste and 2,32 for mastery, at good, its about 2,32 for haste and 2,22 for mastery, so in theory i should be stacking mastery, but when i was at ilvl 490 it stated that mastery was the better stat, so im a bit confused, ofc this could be bcoz i had a difference between mastery and haste at around 2300(4717 haste and 7k mastery) that might have cause it, now im trying a more haste oriented build, reaching the next haste breakpoint by gemming(yes i know) to see if it is a dps increase.

[nightfalls]

I wouldn't suggest to a world/realm first player to aim for the hit cap as I would expect that player to know his class/spec well enough to be able to quickly deal with a miss.

It's the other way around from what I hear most, you definitely want to hit cap. More consistent DPS and less having to focus on misses versus focus on other rotation aspects (and on beating the encounter) is more important than some theoretical, RNG-based simulated gain.

[Deleted]

I cant imagine playing without a hitcap. Missing a dot when one of ur buffs is about to expire is bad enough but missing drain soul on some quickly dying add is a loss of up to 4 shards, which is horrible.

I know thats why some locks aim for safe 12%, but even then u sacrifice the perfect execution for some potential more dps with rng that can mess u up.

[Bonkura]

The main thing about hit cap is that we're not gaining much more from not grabbing it as gear gets better. It will constantly be about 1000 DPS all the way from 470 ilvl to T14H BiS. In other words, it could be argued for to ignore the cap in order to obtain certian haste values. When we're sitting comfortably with at least 6637 haste and mastery above that and still can go for full hit it will not be worth the RNG and QoL issues we get since it's going to be theoretically 1% or even less that we benefit from it in the long run. Also please keep in mind that SimCraft doesn't refresh dots on procs, which should increase the value of hit rating.


Before you ask... I'm taking a break from WoW until 5.2 so I'm not very motivated to keep on theorycrafting. If I decide to come back for hardcore raiding I'll try keep this thread going for next tier. Meanwhile I'm just going to recommend you people to follow the suggestions of the following posters: SirFlipper, Micke, Nagassh, Mugajak, voidspark, Ridcully, evralia and Rustjive (don't forget to check out his blog). They provide some good objective and constructive posts. I'm really bad with names so I'm sorry if I forgot someone.

[Deleted]

I cant imagine playing without a hitcap. Missing a dot when one of ur buffs is about to expire is bad enough but missing drain soul on some quickly dying add is a loss of up to 4 shards, which is horrible.

I know thats why some locks aim for safe 12%, but even then u sacrifice the perfect execution for some potential more dps with rng that can mess u up.

By not playing on such a knifes edge. A good player wouldn't leave it up to chance, they'd refresh dots with enough time to get another attempt in if the 1% chance that they miss comes through and that wouldn't drain soul a second before a mob dies. If missing one thing is causing you to wipe or your dps to drop drastically, then you've got bigger problems than your hit being 1-2% below the cap.

That said, I stopped avoiding the hit cap once I got out of MV gear, I found it useful in poor gear where there wasn't enough stats on it to comfortably hit the hit cap and still have plenty to invest elsewhere, misses were never a huge problem, I'd turn the air blue every time a haunt missed, but this rambling on about "oh I missed a dot and it fell off" etc are just silly, if you're under the hit cap then play accordingly - leave yourself time to react, don't play poorly and then blame the playstyle.

I do prefer going for the 14-15% mark once you're in better gear though, I'm capable of hitting 14-15k haste + mastery while being just under the hit cap, feels better than being 3% or so below, but obviously personal preference.

@Bonkura - Hope you've regained your motivation by next patch, never enough quality posters and theorycrafters in these turbulent times for warlocks.

[Deleted]

Just go for full hit cap. Want to wipe on elegon heroic because your haunt missed the spark? Simulations does not = real raiding. Yes that Sim can go 20000 Iterations. Are you going to pull the boss 20000 times? No?, Try 50 Iterations ( closer to a real pull number for most people), its very possible your theoretical dps is like 75% of what it is on the 20000 Iterations, because of RNG misses. Just go for full hit cap or get fustrated and raged at. Missing haunts with Cd's up is a big deal.

Want to choose your boss specific reforging by being intimidated by fear and peer pressure? What is so hard at simulating 6 seconds of spells on one spark?

20000 iterations could represent 400 players each with their 50 pulls. When you have 400 people with gear x having more "win" outcomes compared to "looses, when they are paired against 400 people with gear y, you would often conclude that gear x is better. It is also likely that this is only possible if and when gear x has higher average dps value when measured with 20k iterations in simc. Or then either 400 people/50 pulls or the 20k iterations could be considered a too small samples.

Anyhow 75 % of average dps means perhaps the lowest 50 results, though seems it is only the min. dps that can be 75 % of the avg., and Chrispotter does not know what the second lowest dps is, or what is the average dps of lowest 50 results.

Let each player pick 50 results from the 20k iterations with probability of 1/20k each, and do not reduce the pool so same results can be picked twice. In this approach the 50 lowest results are under 0.25 percentile and probability for getting all results from under 0.25 percentile is 0.0025^50. You need zentillions of players before you could have it "very possible", for one of them, to get 50 lowest results forming their average dps for 50 pulls.

Every gear has 50 lowest results when simulated with 20k iterations, and even if you are hit capped your min. dps can be something close to a 80 % of avg.? Just because you hit cap does not mean that you can't become that worst person among zentillions, in the eyes of the RNG generator, and escape your punishment.

75 and 80 % of avg dps, ie. min. dps, numbers above are arbitrary, and would change if you ran several 20k simulations or one 100k simulation or anything else.

Missing haunts with Cd's up is a big deal.

Getting more mastery is a bigger deal. Period. Sealed. Double tagged. You're it.

Avg. dps wise it is so for everyone. But some people who choose to stay at 3 % miss may not have enough mastery to cope with the haunt misses, so they are simply trusting the average RNG to not give them a bad day, letting the exposure to some failing to be there. When mastery is incresed to make more dps than one Haunt, loosing them is more probable, but first ones are also welcomed as necessary trade-offs. Difference between avg. dpses is yet about same for 3 to 6 % as it is from 0 to 3 %.

I wouldn't suggest to a world/realm first player to aim for the hit cap as I would expect that player to know his class/spec well enough to be able to quickly deal with a miss.

It's the other way around from what I hear most, you definitely want to hit cap. More consistent DPS and less having to focus on misses versus focus on other rotation aspects (and on beating the encounter) is more important than some theoretical, RNG-based simulated gain.

Fact that the first players are not the ones who take suggestions, or ones that are knowable and quick, would not be an excuse for everyone to want to hit cap.

Also not hit capping may not be consistent compared to capping, if the dps distribution is becoming wide. There is nothing inconsistent however whether it is more dps. Not capping is not about gambling or coin toss, in the same way as in two profiles that would have same avg. dps, where wider distribution is only gaining dps on the right side of the avg.

Instead when you draw two distributions over each other they look like this basicly. There is a dps gain more than 95 % of the time while amount of what you gain can vary.



(First cross point is not the <5th percentile where table turns, but only close to it. Area below the curve may not be the same on the left side of the cross point for both distributions that is. Actual distributions are likely to have smaller dps gains on left side of avg. and larger ones on the right side.)

More consistent DPS and less having to focus on misses versus focus on other rotation aspects (and on beating the encounter) is more important than some theoretical, RNG-based simulated gain.

All rotation aspects are theoretical (unintelligent included), to most they are not even simulated, and they take your focus away from something, and many of them can be giving less dps than non cap, without you knowing it. Without the sims there is very little to know about whether your aspects have a failure chance to produce dps, like in the picture. Also it is not right to call not capping as a "RNG-based gain" because it is based on constant boost into hastery that affects the whole fight. If you reduce miss chance from 15 to 14 % (for free), that's a RNG-based gain there.

If you think your aspects are true and useful because you can go to a dummy and cast a dot with a proc on, and it does more damage, you are still not proving it more important (which has a number describing expectable dps) over non capping. Things you do with procs are RNG based in a long encounter, always different, and you have no real idea of how much dps you are gaining by giving your focus to them. There is probably someone that has made a mistake while dealing with procs so it could be a dps decrease in the game as well.

Focus itself is RNG based: it could only run out if you need to do something real with your rotation at the same time as you are missing. If that is how you can describe the actual playing.

I cant imagine playing without a hitcap. Missing a dot when one of ur buffs is about to expire is bad enough but missing drain soul on some quickly dying add is a loss of up to 4 shards, which is horrible.

I know thats why some locks aim for safe 12%, but even then u sacrifice the perfect execution for some potential more dps with rng that can mess u up.

They didn't mention shards to me, bastards. What dps is 4 shards on that encounter worth? Is it possible to walk to another add and take 4 shards in a time shorter than what is required to take full gain from 4 shards? If you for example only do Haunts then that is atleast 4*8 seconds when you MG over the Haunts, but something needs to be refreshed.

You can't get so much benefit from constantly refilling your shards, unless you think that some of them are waisted by missing Haunts etc.. And the more benefit you gain from Haunt comes from the fact that you have more mastery, and the dps with debuff minus dps without debuff is higher. If you cast your last Drain Soul during the adds last gcd of life you are either all knowing about how they are going to die, with one gcd accuracy, or you just didn't prepare. Someone who doesn't prepare is not entiteled to get more shards and more expected dps, just by his hit rating. There are situations where he has taken more hit but does not take the shards anyway and is doing lowest possible (avg.) dps.

---------- Post added 2013-02-02 at 05:38 PM ----------

The main thing about hit cap is that we're not gaining much more from not grabbing it as gear gets better. It will constantly be about 1000 DPS all the way from 470 ilvl to T14H BiS. In other words, it could be argued for to ignore the cap in order to obtain certian haste values. When we're sitting comfortably with at least 6637 haste and mastery above that and still can go for full hit it will not be worth the RNG and QoL issues we get since it's going to be theoretically 1% or even less that we benefit from it in the long run.

1k dps in BiS is almost 500 % more than what people tend to think. Everything in this thread is also less than 1 % of total dps. So stop doing everything else instead of using hit.

Bad RNG and QoL, or errors, are not scaling at some tremendous rate of their own. Or they won't stay as the same ratio of you dps. When missing with 3 % rate, you might think you also loose 3 % of your time, and that this is 3 % of dps. Dps goes up with gear, but also dot dps does while very little of the dot dps is usually suffering from the misses in comparisson to fillers.

Also please keep in mind that SimCraft doesn't refresh dots on procs, which should increase the value of hit rating.

Thank you for the almost obvious. If you would ever simulate dots with procs you would tell whether mastery increases its value also, and which stat increases its value more, and does hit rating also reach the value of mastery, ever, if that was the direction.

[aztr0]

I always thought the hit cap thing was self explanatory.

[rezoacken]

Also not hit capping may not be consistent compared to capping, if the dps distribution is becoming wide. There is nothing inconsistent however whether it is more dps. Not capping is not about gambling or coin toss, in the same way as in two profiles that would have same avg. dps, where wider distribution is only gaining dps on the right side of the avg.

Instead when you draw two distributions over each other they look like this basicly. There is a dps gain more than 95 % of the time while amount of what you gain can vary.



(First cross point is not the <5th percentile where table turns, but only close to it. Area below the curve may not be the same on the left side of the cross point for both distributions that is. Actual distributions are likely to have smaller dps gains on left side of avg. and larger ones on the right side.)

The bold part is not true at all and I feel obligated to rectify it since it is very misleading for someone reading your post who doesnt have basics in probability and would be lead to believe that his dps "would be better 95% of the time" (almost always). It depends on how far the average are from one another (hit cap average vs not hit cap average) and also the variance of the distribution compared to such difference. What you affirm is really not true at all for the graphic you have provided nor for the numbers I usually see for cap vs not-cap sims.

Math stuff:
If you have two Normal distributions X and Y so that X=N(Mx,Sx) and Y=N(My,Sy) then the probability P(Y>X)=P(Y-X>0)
Y-X is a Normal distribution of average My-Mx and variance Sx^2+Sy^2
Therefore:
P(Y-X>0) = P( Z > -(My-Mx)/sqrt(Sx^2+Sy^2)) where Z is a standard normal.

With the graphic provided (yeah I know its handdrawn) where the variance are close to each other, and numerically many times the difference in means, this number would be a little over 50% which is very far from a great 95%.... Especially when you consider that 50% would mean that both method are equal.

Example (numbers out of my ass, I don't have the time to sim noncap vs cap to get accurate numbers):
Hit cap: average 100 000 standard deviation 5000
Non Hit cap: average 102 000 standard deviation 5200
The probability of Non hit cap being better than hit cap would be 61%.

[nightfalls]

On a side note, a claim that one method "yields higher numbers 95% of the time" means absolutely nothing, even if the claim itself is true.

[rezoacken]

That's not true. It means something, even if that is not enough information to chose a method. You could also argue whether its true or not too.