[5.4] Affliction Warlock and Encounter Specific Guides

[EvreliaGaming]

Your common sense fails then. It does not matter if you refresh after cast ends or midcast after tick, its exact same. You have exact same wasted time there. Either you look cast bar and try to press next spell just after bar ends, or you watch cast bar and try to press after tick marker.

Lets say for arguments sake that Malefic Grasp deals 1000 damage per tick. It ticks 4 times, so that's 4000 damage total for a full cast.

Now let's also say for arguments sake that when you interrupt your MG after a tick - wherever it is - you lose around 0.3 seconds due to natural human reaction time/latency.

Scenario A)

You need to refresh Corruption. You can refresh it any time during its final 8 seconds (random number picked for this example) at level 90 and suffer no penalty.

You are channeling Malefic Grasp and notice that Corr can now be refreshed, and UA can be refeshed in 4 seconds. You channel it for two seconds, and interrupt to recast Corr. You then channel for another two seconds, and interrupt to recast Unstable Affliction.

In total, you've cast 4 ticks of MG and you've refreshed Corr and Unstable Affliction. You've done 4k damage total and refreshed your Dots, and suffered a total of 0.6 seconds wasted to MG interrupts.

-------
Scenario B

You are channeling Malefic Grasp and notice that Corr can now be refreshed, and UA can be refeshed in 4 seconds. You channel it for its entire cast and then refresh Corruption and UA together. You have channeled 4 ticks of MG for 4000 damage, refreshed both of your DoTs, but have wasted only 0.3 seconds.

Also, this exact behaviour is modelled in Simcraft. So no, my common sense does not fail.

[Deleted]

Your example has nothing to do about clipping we talk about. Why you make scenario A) play artifically stupidly and not chain cast dot recasts like example B) does? If A) plays like he should, he should recasts dots just like B) does, so end results is exact same. Except A) has thanks to clipping more room to select exact time when he wants to do this refresh, he does not need to wait for channel to end. And this comes even more important with Haunt and other spells which should be cast as fast as possible after they come ready.
Clipping caster casts Haunt when it comes out of CD right after next MG tick. Non clipping waits for end of channel, and ends with overall less Haunt casts over fight.

[EvreliaGaming]

Your example has nothing to do about clipping we talk about. Why you make scenario A) play artifically stupidly and not chain cast dot recasts like example B) does? If A) plays like he should, he should recasts dots just like B) does, so end results is exact same. Except A) has thanks to clipping more room to select exact time when he wants to do this refresh, he does not need to wait for channel to end. And this comes even more important with Haunt and other spells which should be cast as fast as possible after they come ready.
Clipping caster casts Haunt when it comes out of CD right after next MG tick. Non clipping waits for end of channel, and ends with overall less Haunt casts over fight.

... do you even know how Haunt works? It doesn't have a cooldown. It's limited by Shard procs and nothing else.

I'm sorry that you don't understand, but I can't really explain it any more plainly than I have done.

[Thirdegree]

... do you even know how Haunt works? It doesn't have a cooldown. It's limited by Shard procs and nothing else.

I'm sorry that you don't understand, but I can't really explain it any more plainly than I have done.

I don't know which of you is right, but I do think I can better explain what sahtila is trying to say. In your scenario a) you have the player cast 2 ticks of MG, clip that to cast corruption, cast another 2 ticks of MG, the clip that to cast UA. He's asking why you don't have the player cast 2 ticks of MG, clip that to cast both UA AND corruption, then cast the next MG ending with the same result. (The answer being you're missing 2 seconds of UA)

[Tresdallah]

I don't know which of you is right, but I do think I can better explain what sahtila is trying to say. In your scenario a) you have the player cast 2 ticks of MG, clip that to cast corruption, cast another 2 ticks of MG, the clip that to cast UA. He's asking why you don't have the player cast 2 ticks of MG, clip that to cast both UA AND corruption, then cast the next MG ending with the same result. (The answer being you're missing 2 seconds of UA)

Exactly, which is a direct DPS loss vs. just time loss (yes I know time=dps essentially, but for the sake of the argument.)

[EvreliaGaming]

I don't know which of you is right, but I do think I can better explain what sahtila is trying to say. In your scenario a) you have the player cast 2 ticks of MG, clip that to cast corruption, cast another 2 ticks of MG, the clip that to cast UA. He's asking why you don't have the player cast 2 ticks of MG, clip that to cast both UA AND corruption, then cast the next MG ending with the same result. (The answer being you're missing 2 seconds of UA)

You're not missing anything off UA because the refresh window at level 90 is so large, and as I specified in my example, UA was 2 seconds away from entering its refresh window, not 2 seconds away from expiry. You get a period equal to 50% of its base duration in which you can refresh it without penalty. If your dots are about to expire before your channel finishes (due to tunnel vision) then you would obviously clip to refresh them - nobody is contesting that. My point was and still is, that it is a tiny loss in DPS to clip a MG at 2 ticks versus clipping it at 4 ticks, on its own merit. You can see this inherently by its own DPET time.

If we go by the numbers I gave in the previous example of a tick amount of 1000 and a human error latency of 0.3 seconds, then the DPET of a 4 tick MG would be:
4000 / 4.3 = 930

Whereas the DPET of a 2 tick MG would be:
2000 / 2.3 = 869

Obviously these numbers are totally made up, but you get the point.

So once again, I will say - it is better to get a 4 tick MG off rather than a 2 tick one if you can. At level 90, with proper planning, you should always be able to do this. At level 85, that might not always be the case, since your refresh window is so small.

[Thirdegree]

You're not missing anything off UA because the refresh window at level 90 is so large, and as I specified in my example, UA was 2 seconds away from entering its refresh window, not 2 seconds away from expiry. You get a period equal to 50% of its base duration in which you can refresh it without penalty. If your dots are about to expire before your channel finishes (due to tunnel vision) then you would obviously clip to refresh them - nobody is contesting that. My point was and still is, that it is a tiny loss in DPS to clip a MG at 2 ticks versus clipping it at 4 ticks, on its own merit. You can see this inherently by its own DPET time.

If we go by the numbers I gave in the previous example of a tick amount of 1000 and a human error latency of 0.3 seconds, then the DPET of a 4 tick MG would be:
4000 / 4.3 = 930

Whereas the DPET of a 2 tick MG would be:
2000 / 2.3 = 869

Obviously these numbers are totally made up, but you get the point.

So once again, I will say - it is better to get a 4 tick MG off rather than a 2 tick one if you can. At level 90, with proper planning, you should always be able to do this. At level 85, that might not always be the case, since your refresh window is so small.

Bolded: That's what I ment when I said missing, poor wording on my part.
Italics: I agree.

[Deleted]

You're not missing anything off UA because the refresh window at level 90 is so large, and as I specified in my example, UA was 2 seconds away from entering its refresh window, not 2 seconds away from expiry. You get a period equal to 50% of its base duration in which you can refresh it without penalty. If your dots are about to expire before your channel finishes (due to tunnel vision) then you would obviously clip to refresh them - nobody is contesting that. My point was and still is, that it is a tiny loss in DPS to clip a MG at 2 ticks versus clipping it at 4 ticks, on its own merit. You can see this inherently by its own DPET time.

If we go by the numbers I gave in the previous example of a tick amount of 1000 and a human error latency of 0.3 seconds, then the DPET of a 4 tick MG would be:
4000 / 4.3 = 930

Whereas the DPET of a 2 tick MG would be:
2000 / 2.3 = 869

Obviously these numbers are totally made up, but you get the point.

So once again, I will say - it is better to get a 4 tick MG off rather than a 2 tick one if you can. At level 90, with proper planning, you should always be able to do this. At level 85, that might not always be the case, since your refresh window is so small.

You are still using that silly example where you artificially made clipping player to play worse than non-clipping example so that you got results you wanted. So far there has been no reason shown why clipping would be bad outside playing purposefully badly. And I dont thing you can show it anyway as shadow priests have been using clipping like this whole end of Cata expansion.
It would seem this is minor thing, but its very usefull to know that its no dps loss to clip MG casts early if one wants for example move, cast new spell or target switch. Makes boss fights feel more fluidy.
Also gotta remind, if MG cast is followed by another MG cast, it should be ALWAYS clipped. Trying to cast full MG and then another MG after it is automatic dps loss.

[EvreliaGaming]

Bolded: That's what I ment when I said missing, poor wording on my part.
Italics: I agree.

Glad you understand.

[Rustjive]

I actually agree w/ Sahtila here. If you consider MG an infinitely long channel there is no difference between casting after the 2nd tick as opposed to the 4th. The key is that you cannot queue spells after the 2nd or the 4th tick, so effectively it is the same.

The difference lies only in that in any given fight you want to extend the length of your channels so that you minimize downtime due to reaction/latency. However, knowing that 2 ticks of MG > cast > 2 ticks of MG > cast is worse than 4 ticks of MG > cast > cast is not the same as saying that 2 ticks of MG is ALWAYS worse than 4 ticks. The DPS time loss from 2 ticks > cast is the same as 4 ticks > cast, and 2 ticks > cast > 4 ticks is the same as 4 ticks > cast > 2 ticks. This is an important distinction.

[EvreliaGaming]

I actually agree w/ Sahtila here. If you consider MG an infinitely long channel there is no difference between casting after the 2nd tick as opposed to the 4th. The key is that you cannot queue spells after the 2nd or the 4th tick, so effectively it is the same.

The difference lies only in that in any given fight you want to extend the length of your channels so that you minimize downtime due to reaction/latency. However, knowing that 2 ticks of MG > cast > 2 ticks of MG > cast is worse than 4 ticks of MG > cast > cast is not the same as saying that 2 ticks of MG is ALWAYS worse than 4 ticks. The DPS time loss from 2 ticks > cast is the same as 4 ticks > cast, and 2 ticks > cast > 4 ticks is the same as 4 ticks > cast > 2 ticks. This is an important distinction.

Except, it wouldn't be - 4 ticks > cast > 2 ticks. It would be 4 ticks > cast > 4 ticks. Someone who only casts 4 ticks MGs will have less instances of having to interrupt over the course of the fight.

Load up Simcraft, find this line:
actions+=/malefic_grasp,chain=1

and add

"interrupt=1"

Then tell me what happens.

[Huevos]

I think it's hard to argue with SimulationCraft, but I don't see why there would be more human error trying to interrupt a channel with a non-channeled spell than trying to follow a channel with a non-channeled spell. It seems like you're just as likely to either wait too long after the last tick or try to time it too closely and cut off the last tick.

It seems like the goal should be to maximize MG chaining so you don't have to interrupt. Does waiting until the end of the cast somehow result in this? I understand why interrupting consecutive MGs should be a no-no, since we have huge DoT refresh windows now, but I need to see some mathematical proof of why waiting until the end of the cast is better than clipping.

It's worth mentioning that SimulationCraft can remove the human error, but it seems to me that it'd be able to clip without error just as well as follow a finished MG without error, and there is nothing inherent to MG that suggests it does increased damage the longer it ticks.

[Rustjive]

Except, it wouldn't be - 4 ticks > cast > 2 ticks. It would be 4 ticks > cast > 4 ticks. Someone who only casts 4 ticks MGs will have less instances of having to interrupt over the course of the fight.

You are working with dissimilar timeframes. If you have time for 8 ticks of MG it would look like this:
2 ticks > cast > 6 ticks and 4 ticks > cast > 4 ticks. Do you see why it is identical? The aim is not to maximize 4 tick channels - the aim is to end MG as few times as possible. Those are different things.

Edit: not sure why I didn't read Huevos' post but he has it exactly right. MG ending is a 'natural' interrupt, which is the same as a player interrupting it. In both cases you're trying to reduce those.

Edit 2: And the reason that interrupt=1 sims lower? It's because, as far as I can tell, that sims interrupting the channel as soon as a higher priority item comes up, without respecting the upcoming tick.

[EvreliaGaming]

You are working with dissimilar timeframes. If you have time for 8 ticks of MG it would look like this:
2 ticks > cast > 6 ticks and 4 ticks > cast > 4 ticks. Do you see why it is identical? The aim is not to maximize 4 tick channels - the aim is to end MG as few times as possible. Those are different things.

Edit: not sure why I didn't read Huevos' post but he has it exactly right. MG ending is a 'natural' interrupt, which is the same as a player interrupting it. In both cases you're trying to reduce those.

Edit 2: And the reason that interrupt=1 sims lower? It's because, as far as I can tell, that sims interrupting the channel as soon as a higher priority item comes up, without respecting the upcoming tick.

Incorrect. Simcraft has a built in 0.250 second latency period when interrupting channels. If you lower this less than 0.180, there is no DPS loss at all (and if it were not respecting an upcoming tick, it would be a loss regardless.) If you lower it to zero, it ends up being a gain (because it can clip a tick with zero latency/error and therefore it can clip after a tick with no penalty).

And yes, you're right - the aim is to end MG as few times as possible, but you're looking at too small a time frame to see the follow-on effect. Someone who interrupts their MG after two ticks because they want to refresh a DoT will, quite obviously, over the course of a fight, interrupt their MG more times than someone who only casts 4 tick channels.

[Deleted]

Incorrect. Simcraft has a built in 0.250 second latency period when interrupting channels. If you lower this less than 0.180, there is no DPS loss at all (and if it were not respecting an upcoming tick, it would be a loss regardless.) If you lower it to zero, it ends up being a gain (because it can clip a tick with zero latency/error and therefore it can clip after a tick with no penalty).

According to authors below it starts to be dps increase already under 0.18sec latency.
So simcraft shows also that not dps loss to interrupt MG cast. Noticed EJ had discussion about this too, from reading EJ posts and simcraft action list it seems that simcraft does not chaincast multiple dot refreshes intelligently which makes interrupting MG's look bit worse than it should compared to fully casted MG's. In other words, it seems that with simcraft situation can happen where it refreshes one dot, then does 2 tick MG, then refreshes another? Where proper way would probably be to refresh dots same time. Especially lvl90 char should prolly chain refresh dots after each other if possible instead of casting MG's between. Wonder if simcraft action list can be adjusted for that, could be small dps increase for affliction profile.

EJ:
------- Gohinad -------
2.
actions+=/malefic_grasp,chain=1
why is dps loss if you add "interrupt=1" to this line
-------- Zakalwe --------
2. I imagine this is because we don't let the sim interrupt channels perfectly - we add a delay at an average of 250 milliseconds after the most recent tick, and it's better to pay that price for 4 ticks than for less.

EDIT: The 250 millisecond average delay after channeled spells is tunable using the option "channel_lag". If you set this to anything lower than 180 milliseconds or so, interrupting malefic grasp is no longer a loss, but a gain.
-------------------------

And yes, you're right - the aim is to end MG as few times as possible, but you're looking at too small a time frame to see the follow-on effect. Someone who interrupts their MG after two ticks because they want to refresh a DoT will, quite obviously, over the course of a fight, interrupt their MG more times than someone who only casts 4 tick channels.

Exactly opposite. Clipping allows to cast longer continous chained MG's and refresh dots closer to optimal times, or selecting refresh times more freely according to situation which in turn allows less GCD's used for dot refreshes which means more dps.

[EvreliaGaming]

According to authors below it starts to be dps increase already under 0.18sec latency.
So simcraft shows also that not dps loss to interrupt MG cast.

No, it shows that if you can react to it within 0.18 seconds (which unless you are a robot, you can't), then, and only then, is it not a DPS loss.

The very fact that you are arguing that you should refresh DoTs all at once (although correct) is contradicting your entire point. You're saying its not a dps loss to interrupt MG, so in your argument it should not matter.

[Rustjive]

And yes, you're right - the aim is to end MG as few times as possible, but you're looking at too small a time frame to see the follow-on effect. Someone who interrupts their MG after two ticks because they want to refresh a DoT will, quite obviously, over the course of a fight, interrupt their MG more times than someone who only casts 4 tick channels.

There are a variety of reasons why you would 2 tick channel at least once, even in Patchwerk (procs, for example, where extra 2 ticks of proc boosted DoTs > a small amount of human reaction DPS downtime).

Finally, the issue was mainly with your wording:

It's actually a tiny DPS loss to interrupt MG with a different spell before its finished casting.

Every time MG ends you lose DPS time, that's all. It's not because you're interrupting it. You can imagine where there's an optimal playstyle that SimCraft can't simulate where it selectively interrupts MG channels, right? There is a way (which isn't being done now) to optimize number of MG interrupts vs. DoT uptime (which is fairly abysmal in Sim right now) - 94% UA, 96.8% Corruption in the last 5.0.4 T13H level 85 report.

[Deleted]

No, it shows that if you can react to it within 0.18 seconds (which unless you are a robot, you can't), then, and only then, is it not a DPS loss.

Totally false. Its very easy with cast bar which shows ticks accurately. Gnosis is popular and there are other options too, just gotta make sure castbar knows how to handle situations where channel has extra tick because of chaining, not all castbars can do it. You dont need to react in 0.18 seconds, you just look cast bar going towards tick marker and time button press properly. Its exact same situation as looking full MG cast and timing next spell after MG ends. If your next spell takes more than 0.18 sec to come after channeled spell ends, you are not good player or you have hardware/network problems.

The very fact that you are arguing that you should refresh DoTs all at once (although correct) is contradicting your entire point. You're saying its not a dps loss to interrupt MG, so in your argument it should not matter.

Read again, carefully, what I wrote in end of my previous post about.

Btw, this whole simcraft sidestep in this discussion is pointless anyway. Clipping simcraft results doing less dps than non-clipping results means model or action list of simcraft is not giving accurate results for this case.

[Badahs]

This is a stupid argument. As already stated, with Pandemic at level 90, your window to refresh dots is so large that you will never need to clip MG. Without any haste, MG has a channel time of 4 seconds. Our shortest dot, Unstable Affliction, has an uptime of 14 seconds and a cast time of 1.5 seconds. That means you can cast UA at any point where it has 8.5 seconds left and there would be no dps loss (14/2 + 1.5). 8.5 seconds. Obviously more than the 4 second MG cast. Thus you have ample time to line up all your dots to recast between full MG channels because the longer dots give you even more leeway.

As for mechanics at level 85 (for those of us that like to bang our head against the wall, and i know I'm one of those people), I find that clipping MG is a pretty substantial DPS increase, even if that may not be the case at default simcraft settings. The reason why it is an increase for me is that it allows me to concentrate on the single goal of refreshing my dots at the optimal time rather than refreshing my dots and managing a full MG channel. This helps because currently our dots tick so fast that at end level gear you have about 1.3 seconds between ticks in order to refresh. I would rather concentrate on catching that 1.3 second window rather than letting a dot fall while i'm channeling MG (which would result in a dps loss of not just the dot falling but also the lack of the ghost dot) or refresh a dot one or two ticks too early which would also be a dps loss.

I actually think that the new level 85 mechanics are much more unforgiving than the previous mechanics but a lot of players are not noticing because we do so much awesomeness in dps now and we have so much burst and so much sustained. so so so much awesomeness that I'm afraid we are probably going to get nerfed.

[Huevos]

This is a stupid argument. As already stated, with Pandemic at level 90, your window to refresh dots is so large that you will never need to clip MG. Without any haste, MG has a channel time of 4 seconds. Our shortest dot, Unstable Affliction, has an uptime of 14 seconds and a cast time of 1.5 seconds. That means you can cast UA at any point where it has 8.5 seconds left and there would be no dps loss (14/2 + 1.5). 8.5 seconds. Obviously more than the 4 second MG cast. Thus you have ample time to line up all your dots to recast between full MG channels because the longer dots give you even more leeway.

First, it's not a stupid argument, as there was a discrepancy in dps that needed explaining. Second, you seemed to have missed the point as it was not about whether you should clip MG in order to keep DoTs from falling off. The argument was that you stand to gain dps by waiting until the end (4th tick) of MG to refresh you DoTs instead of interrupting it after any other tick. What causes a dps loss is ending MG (following MG with anything other than another MG). The idea is that you would somehow end less MG casts by always waiting until the 4th tick to cast something else, which you could presumably do on a Patchwerk fight because of our huge refresh windows. I still fail to see how this is true and think the dps loss that Evralia is citing in SimulationCraft by interrupting earlier is probably due to SimCraft not modeling this exactly the way it needs to be executed.

The important thing to take away is that we want to end as few MGs as possible. This means waiting until the refresh windows for as many of your DoTs as possible overlap and chaining your refreshes once you hit this window. To make it simple, if UA and Corruption are both in their refresh windows and will be for the next 6 seconds and you are casting MG (for the sake of simplification lets say it's just hitting it's last tick), you could UA -> MG -> Corr and maintain perfect DoT uptime, or you could UA -> Corr -> MG, and still maintain perfect DoT uptime. The second provides more dps because you're interrupting one less MG.