SimulationCraft for Warlocks (4.0.3a numbers at 80 and 85) - Page 193

dedmonwakeen · 10/24/10, 8:08 PM

We haven't yet put out a binary release of the 4.0.1 client.
If you're brave enough you can download the source from the Cataclysm branch with SVN (Source Checkout - simulationcraft - Project Hosting on Google Code for details on how to check out the main trunk. Substitute where necessary for the Cata branch) and build it yourself with VC++ or gcc.

While it's possible we might put out an initial 4.0.1 release soonish we'd really like to get all classes supported and some general infrastructure issues sorted.

faidwen · 10/24/10, 8:49 PM

I've been using the 4.0.1 cata branch and compiling on my Fedora box, although it is missing a few classes, it is looking good so far!!

VoidStar · 10/25/10, 3:44 AM

Originally Posted by Xayide

I'm not sure if this line of thinking is correct. You shouldn't compare the DPS of each ability but the percentage of the overall damage they represent.
The results show that Immolate is responsible for 12% of overall damage while Shadow Bolt + Incinerate are responsible for 27%.

27/12=2.25 < 2.5, so you might still have a point...

I doubt the OP didn't consider this though... I looks like the math behind this choice is not that simple or some other factor that we are not seeing must be taken into account.

You could be right, but I'm pretty sure that the pDPS is the amount of the total dps that the ability in question represents, not the dps of the ability if that was all you happened to be casting. You'll notice that Incinerate, say, has a pDPS of 1086 for the Demo profile, but that the dps of spamming Incinerate would be several times higher.

The total damage percentage numbers are rounded to whole numbers which was why I didn't use them, as this is significant when we're talking about division of two numbers of similar magnitude.

When looking at this, I actually used the average damage values from the hits, crits and DoT tick hits and crits to double check, but I didn't post it because it's just a longer way of saying the same thing. I did try and think of a reason (other than simply the damage they do) why it might be beneficial to increase Shadow Bolt and Immolate damage, but I couldn't think of one!

All in all, I'm still fairly sure that Improved Immolate is a better use of talent points than Shadow and Flame for Demonology. Looking at logs of my own performance: it certainly is for me.

Innulock · 10/25/10, 5:56 AM

OP demonology profile:

immolate,time_to_die>=4,if=(dot.immolate.remains<cast_time)&!ticking

shouldn't it be |?

Xayide · 10/25/10, 7:46 AM

Originally Posted by VoidStar

You could be right, but I'm pretty sure that the pDPS is the amount of the total dps that the ability in question represents, not the dps of the ability if that was all you happened to be casting. You'll notice that Incinerate, say, has a pDPS of 1086 for the Demo profile, but that the dps of spamming Incinerate would be several times higher.

The total damage percentage numbers are rounded to whole numbers which was why I didn't use them, as this is significant when we're talking about division of two numbers of similar magnitude.

When looking at this, I actually used the average damage values from the hits, crits and DoT tick hits and crits to double check, but I didn't post it because it's just a longer way of saying the same thing. I did try and think of a reason (other than simply the damage they do) why it might be beneficial to increase Shadow Bolt and Immolate damage, but I couldn't think of one!

All in all, I'm still fairly sure that Improved Immolate is a better use of talent points than Shadow and Flame for Demonology. Looking at logs of my own performance: it certainly is for me.

Indeed, you seem to be correct. I ran the simulation with 2/2 Improved Immolate instead of 2/3 Shadow & Flame and got a 10-20 DPS increase, hehe. Still, I wouldn't recommend making that change if you can't be sure if there will be someone else bringing the 5% crit debuff 100% of the time.

shouldn't it be |?

Probably yes. I ran the sim with that change and got a very small DPS increase. It was so small though, that it can probably be disregarded. Nevertheless, | seems to make more sense. It is | in the Destruction profile.

VoidStar · 10/25/10, 8:10 AM

I assume that the reason that the demonology warlock doesn't always want to recast Immolate when it has less duration than the cast time is that there could be a Hand of Gul'dan currently "in the air" that will refresh the duration when it strikes. I certainly run into this situation when I'm actually playing.

It is odd though: if ANDing is the intention then it would seem that the DoT duration check probably isn't required.

It may turn out that ORing it would be better than ANDing it, or end up about the same as your test indicates Xayide, but I don't think that it's automatically wrong, and not unreasonable that it might be different for demo and destro.

TL;DR: I think it's a more interesting question than first appears, and a matter for consideration, rather than simple correction.

Xayide · 10/25/10, 12:10 PM

Originally Posted by VoidStar

I assume that the reason that the demonology warlock doesn't always want to recast Immolate when it has less duration than the cast time is that there could be a Hand of Gul'dan currently "in the air" that will refresh the duration when it strikes. I certainly run into this situation when I'm actually playing.

It is odd though: if ANDing is the intention then it would seem that the DoT duration check probably isn't required.

It may turn out that ORing it would be better than ANDing it, or end up about the same as your test indicates Xayide, but I don't think that it's automatically wrong, and not unreasonable that it might be different for demo and destro.

TL;DR: I think it's a more interesting question than first appears, and a matter for consideration, rather than simple correction.

What I noticed after posting though is that Demonology (at least) oscillates a bit between runs, even using 10k iterations (usually less than 50 DPS). Even using 20k iterations it still oscillated. I guess Demo might rely heavily on RNG...
So my assertion about the "|" is probably worthless. It would be nice to hear from the OP about if it was intended or not.
As you said, it probably wasn't since the DoT duration check becomes useless by ANDing it with "!ticking". Maybe some interaction with the remainder of HoG's CD and its cast time should be woven into the conditions, although this shouldn't be a problem with the BiS profile's haste level.

I agree it might be reasonable for it to be different for Destro.

The Imp Immolate vs S&F issue has proven to be consistently higher despite the oscillations though.

• Zakalwe · 10/25/10, 1:20 PM

The & is a typo, it's supposed to be |. I probably won't bother regenerating the OP or the wiki output for the fix though, since it has virtually no effect.

Caltiom · 10/25/10, 1:47 PM

The AND operator was certainly a mistake.

The default action priority list offered for armory/wowhead imported characters should now be again up to date. Keep in mind that this is really just a bonus that is offered to run simulations directly with only a name and a server, but it is actually part of the "user input" and not the simulation process of simcraft. These default lists get out of date even faster than the simulation mechanics (which aren't always up to date either, with blizzard still messing up everything on a regularly basis).

I pimped the drain affliction profile too, it seems there is still a bit of hidden dps in every profile. Regarding the dot refresh conditions, the way to go is to separate different priorities of a single spell. For example for Unstable Affliction there is a refresh window [ cast_time, cast_time + tick_time ]. And tick_time in this case is 2*GCD. So you could create 3 different priorities:

actions+=/unstable_affliction,if=dot.unstable_affliction.remains<cast_time
actions+=/unstable_affliction,if=dot.unstable_affliction.remains<(cast_time+gcd)
actions+=/unstable_affliction,if=dot.unstable_affliction.remains<(cast_time+tick_time)

The most desirable one is certainly the first one to maximize DPET (because we have a finite fight duration, so a little bit of time is lost if you refresh earlier than the last_tick, even with the new dot mechanic). But it might be still better to use the second or even third condition before a drain_life, which might otherwise reduce the Uptime of UA just a little bit.

And it also helps reducing uptime loss when both UA and BoA are in "sync" and would have to be refreshed at the same time. In this case, the new DoT refresh system really helps so that you can refresh both within their windows between the next to last and last tick.

The profiles oscillate quite a bit, yes. There is an error margin given in the report, which should give mathematical boundaries of it. To analyze small changes - which does not only include ones made by the user, but also for the scale factor calculation - iterations really need to be set to a high value.
The main reason why the error values are not that small for the warlock is the Tier 10 4p buff, and to some other extent other procs/mechanics. Keep in mind that this does not have to be a bad thing for the simulation per se, on the contrary the "range" in simcraft can show you how broad your dps range is in game, and why every wol report should be taken with a grain of salt.

Zakalwe, what is your position on pre-combat potions? I think it would be good to activate them for the 401 profile ( with use_pre_potion=1 ).

Xayide · 10/25/10, 2:23 PM

I did 3 runs with 40k iterations each and the final DPS were 17105, 17104 and 17105, so it's safe to say that's enough iterations for Demo (might be overkill but whatever). 10k might be too few considering its uneven DPS distribution.
Having said that, I ran twice with 2/2 Imp Immo and had 17111 and 17113... So we can safely say, DPS-wise, it's the same. But considering S&F also provides an important debuff that you can keep up in case that Fire Mage dies, go with S&F.

I didn't change anything on the action list though. There's probably nothing to change anyway...

Obs.: I ran that with the "|" correction, although that might not matter =)

Zakalwe, what is your position on pre-combat potions? I think it would be good to activate them for the 401 profile ( with use_pre_potion=1 ).

Isn't that covered by the (!in_combat) condition? What is the difference between that and use_pre_potion=1?

VoidStar · 10/26/10, 3:56 AM

Well, Madlax gave us a method of gaining one dps by pre-combat Soulburn/Soul Harvest, so a six or seven boost seems well worth taking

I realised that the Simcraft benefit of switching Shadow and Flame to Improved Immolate was likely to be small, but I find the difference to be greater when I look at logs of my actual performance.

Anyway, I'm not going to keep banging on about this and fortunately it's very easy for people to check whether the change would be beneficial for them anyway: just take the damage done numbers from your damage meter of choice and if (Shadow Bolt + Incinerate) < (2.5 x Immolate) then consider swapping, if not or if your raid needs the crit debuff from you then don't

Neisfur · 10/26/10, 4:08 PM

Originally Posted by Slashburn

Just a quick question, what is exactly happening at this haste sweet spot of 1406?

Originally Posted by Dragones

A extra tick is added to immolate.

Are we sure its the extra tick to immolate that is causing demonology's haste scale factors to plummet? Reason i ask is demonology refreshes immolate, thus as long as demonology doesn't let immolate fall off, the gains from extra haste (w.r.t immolate) are purely linear - the damage contribution of the extra tick becomes a nonfactor.

Haste's plummeting value could be due to a cutoff for squeezing in an extra metamophosis into the fight, because we're casting shadow bolts faster. Or perhaps its not the etrax tick's damage, but rather the extra tick coincides with the longest possible run-time for immolate making refreshing of Hand of Gul'dan easier. Note the simcrafted warlock let immolate fall off 4x. Perhaps its the greater success at refreshing immolate, rather than the extra tick of damage that is the cause of for haste numbers falling off? Can someone post the number of times the lower haste sim-warlock let immolate fall off?

It also might be interesting to see if there's another local maximum around the previous immolate tick, and if its dps is lower/higher than the current one.

something2 · 10/26/10, 4:09 PM

I noticed in the Destro profile (didn't check others) that the thread used on pants is still the spirit version, this should probably be changed to stamina.

• Zakalwe · 10/27/10, 1:29 AM

Originally Posted by Neisfur

Or perhaps its not the etrax tick's damage, but rather the extra tick coincides with the longest possible run-time for immolate making refreshing of Hand of Gul'dan easier. Note the simcrafted warlock let immolate fall off 4x. Perhaps its the greater success at refreshing immolate, rather than the extra tick of damage that is the cause of for haste numbers falling off?

This is exactly it. Extra ticks to dots are not a huge DPS jump in themselves. It just so happens that two of our specs have very significant gains from secondary mechanics (Conflag damage and HoG refresh) when immolate gains a tick.

Tòki · 10/27/10, 9:23 PM

When Referencing to the Destro Locks DPS scaling factors.

The Current Simcraft listed has the “soft haste cap” of 1406. (Which the sim has used for the iterations)

If you are unable to hit the soft cap of 1406, using the listed scaling of

Warlock_401_Destruction Sta=0.0350 Int=4.6378 SP=3.1804 Hit=6.3289 Crit=2.7720 Haste=1.4481 Mastery=3.0874

And the thresholds (in a raid situation) of 157, 781, 1406, 2030

Does this correctly state that you are better to role all the haste out of a gear set down to the next soft cap of 781 and place all the Haste and Crit you can into Mastery?

Or by moving under the 1406 haste cap, does that in some way change the scaling factors of Haste?