Dots and you: The Affliction Warlock Thread - Page 10

marano · 12/27/08, 11:28 AM

Well, I'm about to respec because I have 14% hit and am soft capped in raids. I can leave cataclysm out, but im just wondering what would provide more dps, 2 points in demonic power and 3 in ISB or no demonic power and full isb? Of course the pet of choice would be the imp.

Ideas and calculations much appreciated.

DiamondTear · 12/27/08, 11:54 AM

Originally Posted by Blacksen

Anyways, I generally try to cast one more spell whenever I see Haunt come off cooldown. Aiming to cast Haunt once ever 10 seconds is highly advisable, since Shadowbolt technically has a higher DPCT than Haunt.

I've found that often when I cast one more SB after Haunt comes off CD, one of my dots ends within 1.5 seconds of my SB cast finishing. Meaning I will either have Haunt drop for a while or I'm late for refreshing the dot. Since the DPCT difference between Haunt and SB is quite small, I'm wondering if it really is worth casting another SB before refreshing Haunt.

Originally Posted by marano

From what I've tested it is; MC provided better DPS than ISB, the Imp is superior in damage output than the FH and 1 point in Eradication should be more than enough for a consistent uptime.

How could you possibly test that reliably? SB benefits much more from raid crit buffs and debuffs which you won't have on dummies. If you test it in a raid, there will be too much RNG to get reliable results.

Originally Posted by marano

Hm, I guess so. Allthough DS being broken by splash damage is even worse.

It's not. You can refresh it after a tick if you take splash damage. The cast time is so long that you have plenty of time to react.

tkoreaper · 12/27/08, 3:12 PM

Originally Posted by Warlocomotif

For the main part you shouldn't usually be casting CoE, try convince your guild that a Boomkin/UnholyDK would really be worth it. Past that, it's not amazing, but it's better than the alternatives (assuming you're CoAing).

We do have some DK's in the raid and I'm sure at least one of them has EP, but shouldn't my CoE be overwritten cause I never see it fall off the boss unless the time runs out.

Akj · 12/27/08, 3:39 PM

Originally Posted by tkoreaper

We do have some DK's in the raid and I'm sure at least one of them has EP, but shouldn't my CoE be overwritten cause I never see it fall off the boss unless the time runs out.

They do not overwrite each other because CoE has a secondary effect (- resist) which unfortunately is worthless on most raid bosses.

♦ Blacksen · 12/27/08, 5:19 PM

Originally Posted by DiamondTear

I've found that often when I cast one more SB after Haunt comes off CD, one of my dots ends within 1.5 seconds of my SB cast finishing. Meaning I will either have Haunt drop for a while or I'm late for refreshing the dot. Since the DPCT difference between Haunt and SB is quite small, I'm wondering if it really is worth casting another SB before refreshing Haunt.

The DPCT between Haunt and SB is not small, especially as you get more haste. Haste increases the DPCT of longer-cast spells a lot more than of short-cast spells.

For instance, let's imagine this scenario:
- Haunt does 1500 dmg, Shadowbolt does 2500 dmg. They both have 1000 DPCT.
- You get 25% haste. Haunt's now doing 1500 dmg over 1.25 seconds, Shadowbolt is now doing 2500 dmg over 2 seconds. Shadowbolt's DPCT is 1250 (2500/2), while Haunt's DPCT is only 1200 (1500/1.25).

These numbers are completely made up, but they still prove the simple point that haste benefits your Shadowbolt more than any other spell.

The other big factor here is that Haunt crits don't benefit from Ruin, and Shadowbolt crits also have a chance to cast Shadow Mastery. If you have the 2-piece t7 bonus, even better.

It's okay to let Haunt fall off as long as another haunt is already in flight. 10-seconds really is the ideal rotation time.

In response to Curse of Elements, your guild should have a boomkin or unholy death knight. If you don't have either of those, you really need to pick them up. They keep their things up without any DPS loss. Unfortunately, neither Earth and Moon nor Ebon Plague will override CoE because CoE has spell penetration (which is useless in PvE).

• dragon12 · 12/27/08, 5:54 PM

Originally Posted by Blacksen

The DPCT between Haunt and SB is not small, especially as you get more haste. Haste increases the DPCT of longer-cast spells a lot more than of short-cast spells.
[numbers]
These numbers are completely made up, but they still prove the simple point that haste benefits your Shadowbolt more than any other spell.

This is incorrect. Haste benefits all spell cast times equally (relatively).
The formula for haste is:
New Casting Time = (Base Casting Time)/(1 + (% Spell Haste / 100))

Regardless of the base casting time, it is always divided by (in this case) 1.25, hence preserving the relationship between the DPCTs for the two spells.

Arnath · 12/29/08, 7:02 PM

Just a minor edit: Cuffs of Winged Levitation drop in Heroic Occulus, not Heroic Malygos.

adolchristin · 12/30/08, 9:16 PM

Got a lock I'm leveling as affliction who's nearing 80 and I had a quick question about the merits of improved shadowbolt. The first post says that it's not quite as useful as other talents; the spec i thought about taking at 80 is:

55/0/16

Basically forgoing Dark Pact and any points in imp shadow bolt for maxing out Molten Core. It seemed to me that the increased fire damage would be up more often than the imp shadow bolt proc. I don't have any real mathematics to back this up; I thought I'd defer to the experts.

I saw the discussion about 2 pages back which involved dropping points from cataclysm to grab molten core but it seems that it'd be better to keep both of these at the starting levels of gear I'll be rolling with when I hit 80.

couchtyp · 12/31/08, 6:42 AM

Hi adolchristin.

The build is fine IMO.

You could remove 1 point from Molten Core and Eradication and put them into ISB, because you are throwing enough shadow spells on bosses to keep Molten Core up and you would not need a 100% uptime anyway. The third point in Eradication gives the least haste gain, so maybe it is better to spend the point elsewhere.
But moving this point would not be a notable change in DPS - at least at early gear levels.

When you are breaking the hit-cap (hit-rating from gear only, no hit-gems, no hit-enchants) you can move talent-points from Cataclysm to ISB to max it; or to Destructive Reach and Grim Reach (from Suppression) to get a bit more movability at the cost of mana-efficiency.
But I don't see that happen before the next content-patch.

Cigaras · 12/31/08, 6:46 AM

Done some testing with SimulationCraft and there are my results:

Eradication uptime:
1/3 = 13%

2/3 = 21%

3/3 = 25%

Molten Core uptime:
1/3 = 67%

2/3 = 89%

3/3 = 96%

so pretty obvious - 1/3 is a must in both, 2/3 is quite a reasonable increase but 3/3 is a waste. After multiple tests swapping glyphs and so on i got 2 winners spec1 and spec2 with difference like 1 or 2dps

using CoA, SL and Corr or Immolate glyphs (Immolate was slightly better). Having 3/3 in both or mixing 1/3 2/3 in Err/MC and 3/5 ISB was bellow for about 10-20dps. I guess in next content patch with better gear there will be a bigger difference and it will be easier to decide.

Ebi · 12/31/08, 6:56 AM

Preface
I've seen a whole bunch of different numbers thrown around in different threads regarding how effective talent points placed in Eradication are. I tried running the numbers myself and came up with numbers that I hadn't seen in those posts. I rechecked my math and wrote an Eradication proc simulation that I think shows my numbers are right.

TL;DR Version
1 point in Eradication gives an uptime percentage of 11.76%, or about 2.35% effective spell haste
2 points in Eradication gives an uptime percentage of 17.18%, or about 3.44% effective spell haste
3 points in Eradication gives an uptime percentage of 21.05%, or about 4.21% effective spell haste

The Math
Call the probability that an eligible Corruption tic results in an Eradication proc p.
Call q the probability that an eligible Corruption tic does not result in an Eradication proc, i.e., q = 1 - p.
Assume 100% uptime of Corruption. Assume cows are spherical.

Take the first Corruption tic after an Eradication proc that is eligible to proc another Eradication, i.e., 30 or more seconds since the previeous proc. There is probability p that 12 of the next 30 seconds will be hasted, after which the next Corruption tic will be eligible to proc another Eradication. There is probabiiity p * q that 12 of the next 33 seconds will be hasted, following which, again, the next Corruption tic will be eligible. The expected duration until the next time a Corruption tic will be eligible following an Eradication proc is the infinite sum:

30 * p + 33 * p * q + 36 * p * q^2 + ... + (30 + 3 * k) * p * q^k + ...

Which can be separated into two infinite sums:

30 * p + 30 * p * q + 30 * p * q^2 + ... + 0 * p + 3 * p * q + 6 * p * q^2 + ...

The first infinite sum is a simple geometric series multiplied by a constant:

30 * p * Sum(q^k, k = 0, 1, 2, ...)

Which is therefore equal to 30 * p * 1/(1 - q). But since, from above, we've said that 1 - p = q, then 1 - q = p, so that gives us 30 * p / p = 30 for p not equal to 0.

The second infinite sum is trickier. To solve it we rely on some calculus. Since Sum(q^k, k = 0, 1, 2, ...) = 1/(1 - q), take the derivative of both sides:

d/dq Sum(q^k, k = 0, 1, 2, ...) = Sum( d/dq q^k, k = 0, 1, 2, ...) = Sum(k * q^(k-1), k = 0, 1, 2)
d/dq 1/(1-q) = d/dq (1-q)^-1 = -1 * (1 - q)^-2 * d/dq (1 - q) = -1 * (1 - q)^-2 * -1 = (1 - q)^-2

The second infinite series, expressed in summation form is

3 * p * Sum(k * q^k, k = 0, 1, 2, ...) = 3 * p * q * Sum(k * q^(k-1), k = 0, 1, 2, ... )

Which, from the calculus equations makes it equal to 3 * p * q / (1 - q)^2, which is simplifies to 3 * p * (1 - p) / p^2, which further simplifies to 3 * (1 - p) / p.

Which gives the total duration equal to:

30 + 3 * (1 - p) / p

Therefore, for p = 0.04 (i.e., 1 point in the talent) the duration equals 102 seconds, for p = 0.07 the duration is sightly less than 70 seconds, and for p = 0.1 the duration equals 57 seconds. Therefore, for one talent point the uptime ratio is 12 / 102, for two points it's about 12 / 70, and for three points it's 12 / 57. These fractions are approximately 0.1176, 0.1718, and 0.2105, respectively.

The Simulation
Written in Perl because I'm lazy, impatient and overconfident. It models the latency between corruption completion and recast as an exponential distribution. As the time that a single simulation is performed over is increased, the number appear to converge to the fractions given above.

#!/usr/bin/perl -w

use strict;

# globals

# change these for different simulations
my $procProbability = 0.04;
my $simTime = 600;
my $iterations = 10000;
my $uptime = .9999;

# these shouldn't change
my $duration = 18;
my $tics = 6;
my $ticTime = $duration / $tics;
my $eradDuration = 12;
my $downtimeSecs = $duration / $uptime - $duration;
my $eradLockout = 30;

# Modeling the lag time as an exponential distribution.
sub randlag()
{
  return $downtimeSecs * -log(1 - rand())
}

# Run the simulation once and return the eradication uptime in seconds
sub runSim()
{
  my $lastErad = -$eradLockout; # Low enough so that any initial procs will not be suppressed
  my $curTime = 0;
  my $ticsLeft = 0;
  my $eradTime = 0;
  my $clip = 0;
  my $clipTime = 0;

  while ($curTime < $simTime)
  {
    if ($ticsLeft == 0)
    {
      # Model lag, and recast
      $curTime += randlag() + $ticTime;
      $ticsLeft = $tics;
    }

    $ticsLeft--;

    if ($curTime - $lastErad >= $eradLockout && rand() < $procProbability)
    {
      $lastErad = $curTime;
      $eradTime += $eradDuration;
    }

    if ($ticsLeft > 0)
    {
      $curTime += $ticTime;
    }
  }

  # If the final eradication proc was clipped by the duration of the sim,
  # chop off the unused portion
  if ($simTime - $lastErad < $eradDuration)
  {
    $eradTime = $eradTime - $eradDuration + $simTime - $lastErad;
  }

  return $eradTime;
}

# Main starts here

my $sumTimes = 0;

for(my $i = 0; $i < $iterations; $i++)
{
  $sumTimes += runSim();
}

my $averageRun = $sumTimes / $iterations;
my $avgPct = 100 * $averageRun / $simTime;

print "For the average run over $simTime, uptime is $averageRun ($avgPct%)\n";

exit 0;

dedmonwakeen · 12/31/08, 10:27 AM

Regarding simulationcraft uptimes:

Just a reminder that the "time" in "uptime" does not refer to seconds. The percentage is not a chunk of time relative to fight length.

Rather, it is the frequency with which a buff is USEFUL. It is the ratio of how often the buff is present when looked for during spell calculation. I know this is different/confusing, but it is actually a much more useful piece of information when plugging the simulation output into formulation calculations.

fallenman · 12/31/08, 1:20 PM

Something that I think a lot of people are missing, is that once you are decently geared, any 1 point into ISB is going to be more dps than even a single point in MC. I did some math based upon my last patchwerk, and MC's 1st point was worth an increase in roughly 3200 damage overall, while any 1 point (ISB is more linear since it's a flat 2% per point) in ISB was worth more than 4000 damage overall.

Shocktar · 12/31/08, 1:57 PM

I disagree for 10 mans.
I just ran SimCraft with three different builds:
Talent Calculator - World of Warcraft (55/0/16 with full Eradication)
Talent Calculator - World of Warcraft (54/0/17 with a point in Eradication moved to ISB)
Talent Calculator - World of Warcraft (54/0/17 with that point moved to MC instead)

Results indicate that full Eradication is best, if only by a small amount. Picture attached. Gear is armory. Please note that this is an unbuffed Warlock, meaning no mage, moonkin, elemental shaman, or shadow priest. I'll rerun now with a full raid, see if results change.

Edit: For a 25 man raid, it would seem that he is correct, by a small amount. I'd assume that this is due to significantly increased crit (which makes the imp better too). In this case the imp build is using the imp glyph, instead of the Agony glyph.

Ebi · 12/31/08, 2:15 PM

dedmon, looking at the warlock simulation code, there might be a bug in the eradication code. Line 1171 reads:

if( s -> sim -> roll( p -> talents.eradication * 0.04 ) )

shouldn't it read:

if( s -> sim -> roll( 0.01 + p -> talents.eradication * 0.03 ) )

dedmonwakeen · 12/31/08, 2:40 PM

Originally Posted by Ebi

dedmon, looking at the warlock simulation code, there might be a bug in the eradication code. Line 1171 reads:

if( s -> sim -> roll( p -> talents.eradication * 0.04 ) )

shouldn't it read:

if( s -> sim -> roll( 0.01 + p -> talents.eradication * 0.03 ) )

Bah.... I keep getting burned by those types of talent ranks. I'll have a new release out in a couple of days. (For folks building their own executable, r1227 has the fix.)

FYI..... Feel free to use Issues - simulationcraft - Google Code to keep me informed of bugs. I do my best to troll EJ for any simulationcraft references, but there is a ton to read..........

At least it helps understand the larger gaps between theory and sim for ranks 2 and 3. (The uptime dropped about 2 percentage points with the change.)

And thanks again for the oversight. A couple of guys have signed up to make Warlock tweaks, but more eyes are always better.

Evyle · 12/31/08, 4:47 PM

I currently raid at or really dang close to hit capped 288 personal + 3% from suppression/cataclysm and +3% from shadow priest (we usually have 1-2 shadow priest and 1 boomkin in raid). I am trying to weigh the benifits of that +3% from talents vs getting more hit gear and taking dps talents. Build is 53/0/18 to 55/0/16

For the early affliction tree there is not a lot of good choices to put the must 15 points into. So I went with 3/3 suppression - the other choice would of been to get 3/3 fel concentration (not used all the time.. but you do take spalsh damage every now and then). So affliction is no "real" dps gain by picking another talent.
For distruction I feel I am making a very real dps sacrifice/choice to get the 3/3 cataclysm. Soo after the 5/5 isb, 5/5 bane, 5/5 ruin - there are 3 points for wiggle room.
1 point in molten core will have a high up time for the 10% help to my immolate
going from 1 to 2 eradication will help also (possibly even 3/3 eradication with that 1/3 molten core).

So the 3/3 cataclysm (in my case) could equal the 1/3 molten core and the 3/3 eradication. The problem is I then end up gearing for the other 76 or so hit which is approx 92 spell damage. I have gone from the 14% personal + hit (with 5 yellow gems being used) down to the 11% personal hit and a lot more red gems (no yellow now).

(edit added below data from Leulier spreadsheet)

With my current gear and raid buffs the 53/0/18 build would net me 5080 dps. I could swap out 92 worth of + spell damage for the 76 hit .... that would free up 3 talent points in destro to get 1/3 molten core and 3/3 eradication.
DPS (accorrding to the spreadsheet) would go from 5080 to 5013 (quite a loss).

Some say "the hit is free with the better gear"... your still making a choice to use that neck with + hit instead of + crit/haste. So from what I can tell its better to take your + hit from talents (3/3 suppression & 3/3 cataclysm) instead of swapping to + hit gear.

Bandoer · 12/31/08, 8:48 PM

Assuming that the optimal time to cast Haunt to cast it so the debuff refreshes exactly as the old one falls off, is it better to cast Haunt slightly early or slightly late?

The simplest tradeoff example I think is that you are a 10 sec into the last Haunt debuff and you don't have any dot refreshes to do, and your Haunt cast time is 1.3 sec. Do you:

1 - Haunt
2 - Lifetap then Haunt
3 - Shadowbolt then Haunt

If you cast Haunt early, it is basically like refreshing a dot early. You chop off a slight bit of damage from your previous Haunt cast. Modeling-wise, I think the theoretical loss from early cast would count as a loss from your filler spell.

If you cast Haunt late, you have a loss from undebuffed dots. I would think it is random whether a tick falls during the undebuffed time, so the distribution of ticks falling in the non-debuffed time would be uniform. Only thing that might not be randomly distributed is Corruption depending on any wonkiness with Haunt refreshing corruption. The loss I would count this as is the value of all ticks currently on the target multiplied by a factor representing the proportion of ticks falling in the undebuffed time. But this makes my head hurt.

Pretty sure the differences are miniscule but would just like to know. The difference between casting Haunt at 8sec vs 16 sec might be interesting too.

DiamondTear · 01/01/09, 8:33 PM

You all know that is the Immolate Buggy and sometimes not goes on the Boss - so: WHAT to to? Reset Immolate & Destroy my Rotation, loose DMG? loose MANA?

Don't cast immolate before the previous one fades.

Sayessa · 01/02/09, 7:06 AM

I believe there is an error here:
# [Punctilious Bindings] (Naxx25) 129 dps
# [Unsullied Cuffs] (Sartharion10) 126 dps

First, [Unsullied Cuffs] are from Sartharion25, and second, I don't belive that they are worse than [Punctilious Bindings]. It looks like the red socket wasn't taken into account. The socket alone with a 19 spellpower gem is worth 26,6 dps, which puts the cuffs at 153 dps and makes them the best wrists currently in game.

fallenman · 01/02/09, 10:57 AM

Deleted.

Melbuframa · 01/02/09, 6:28 PM

Because this is the first time I’ve looked at the gear listing on the first post, Can you explain how you came up with those rankings as some of them don’t make any sense. It looks like you are massively overvaluing Spirit and Hit.

• dragon12 · 01/02/09, 8:14 PM

Originally Posted by Melbuframa

Because this is the first time I’ve looked at the gear listing on the first post, Can you explain how you came up with those rankings as some of them don’t make any sense. It looks like you are massively overvaluing Spirit and Hit.

The minimum Spirit could be valued at would be 30% of spell power; I would guess that the extra few % are from lifetap scaling.

RoyalBoss · 01/02/09, 8:37 PM

Everywhere i read it says you should use CoA over CoD. I'm a little confused why. I'm a destro spec for pvp right now so i can't really test it with haunt. But here are some numbers (Yes I have the CoA glyph)

(This is also done with just a spellstone & fel armor up)

In 2 GCDs you get 56 seconds worth of CoA. So every minute you have to use 2 GCDs. Working on target dummys my CoA after the full 28 seconds (or 14 ticks) my CoA does on average 5081 damage. Add in the 10% CoA damage you get from Imp CoA and you get roughly 5589 damage in 28 seconds.

So in 1 minute you will get 11,178 damage from CoA in 56 seconds.

My Curse of Doom after 1 minute on average ticks off 11,660 damage.

The only thing I can't really test right now is if CoD is actually affected by Haunt. Because if both CoA & CoD are affected by haunt, then technically they will both be doing about the same damage in the same amount of time. Just using CoD you get that extra GCD, which in turn wouldn't that give you that 1 extra GCD for a cast, refresh a dot, or lifetap?

Everything I read says that CoA > CoD. Can anyone please tell me exactly why, or what I am doing wrong?

Thanks,

(PS.)
I also think switching between them would be smart. If by the time your first doom goes off and the boss is at less than 50% throw up CoA. Or if you about know how long your fights on average last, judge how many CoDs you think you'll be able to get off successfully before casting CoA.

I know it's a pretty big DPS loss if your CoD isn't allowed to go off, but it seems you should at least always start with it up?

Mystearica · 01/02/09, 8:56 PM

Originally Posted by RoyalBoss

Everywhere i read it says you should use CoA over CoD. I'm a little confused why. I'm a destro spec for pvp right now so i can't really test it with haunt. But here are some numbers (Yes I have the CoA glyph)

(This is also done with just a spellstone & fel armor up)

In 2 GCDs you get 56 seconds worth of CoA. So every minute you have to use 2 GCDs. Working on target dummys my CoA after the full 28 seconds (or 14 ticks) my CoA does on average 5081 damage. Add in the 10% CoA damage you get from Imp CoA and you get roughly 5589 damage in 28 seconds.

So in 1 minute you will get 11,178 damage from CoA in 56 seconds.

My Curse of Doom after 1 minute on average ticks off 11,660 damage.

The only thing I can't really test right now is if CoD is actually affected by Haunt. Because if both CoA & CoD are affected by haunt, then technically they will both be doing about the same damage in the same amount of time. Just using CoD you get that extra GCD, which in turn wouldn't that give you that 1 extra GCD for a cast, refresh a dot, or lifetap?

Everything I read says that CoA > CoD. Can anyone please tell me exactly why, or what I am doing wrong?

Thanks,

(PS.)
I also think switching between them would be smart. If by the time your first doom goes off and the boss is at less than 50% throw up CoA. Or if you about know how long your fights on average last, judge how many CoDs you think you'll be able to get off successfully before casting CoA.

I know it's a pretty big DPS loss if your CoD isn't allowed to go off, but it seems you should at least always start with it up?

Out of boredom I went to the Target dummy. Completely unbuffed(even w/o fel armor) my CoA did roughly 20k dmg in 1min w/o the glyph, with 1 missed haunt & SE falling off cuz I'm being lazy for this test. CoD did roughly 12k dmg both with & w/o Haunt up. My assumption is that CoD isn't affected by haunt & CoA > CoD.