SimulationCraft for Warlocks (Cataclysm Edition) - Page 41

botafuco · 08/10/11, 5:53 PM

Thanks for looking into it Zakalwe. Maybe my mistake was simming it using my profile and not the T12H profile.

I just run the T12H profile and then I applied to it my BoD modification, and you are indeed right, there is a rough 130 dps loss.

Interestingly enough, I run a sim for your profile (I just pulled it off the armory now) and I got 30281 dps.
I then made my BoD modification to your profile and I saw a similar increase on dps, 30614 dps. Roughly 330 dps.

As I read the current action list (and I might be wrong) the doomguard is being summoned before the target gets a BoD. So my thinking is that the doomguard is idle until BoD is applied (and possible some small time after).

I didn't mean to try to cover any bugs by recasting BoD. My apologies if I expressed myself incorrectly there.

• Zakalwe · 08/10/11, 6:06 PM

No, what I'm saying is the simulator does not take any bane into account when simulating the doomguard - it simply assumes the doomguard will always assist you right away. The "bug" I was referring to was that he doesn't automatically do so in-game, but looking at the tooltip I guess it's actually intended behavior. Whether we should simulate that behavior I'm not sure.

Either way, you're right, there is a DPS boost by changing the action list in the way you are. It's not because of the doomguard, however, it's because the default action list syncs up BoD with DS, but if you're not syncing Meta with DS it's better to sync BoD with Meta instead.

The real issue here is that the default action list is optimized for the BiS profile, which uses the moonwell trinket and syncs Meta with that (and thus also with DS). I may have to make it check for the trinket when generating the action list.

EDIT: This is now fixed in SVN, and as a bonus you now no longer have to manually modify your action list if you're demo and have the moonwell chalice (unless you want to simulate *not* syncing meta with the trinket, of course).

Diivil · 08/11/11, 5:54 AM

If you import a profile with Synapse Springs and Moonwell Chalice the sim will put the springs above chalice in the default action list. Just switching them around was a 550 dps gain with my current gear.

• Zakalwe · 08/11/11, 6:27 AM

Originally Posted by Diivil

If you import a profile with Synapse Springs and Moonwell Chalice the sim will put the springs above chalice in the default action list. Just switching them around was a 550 dps gain with my current gear.

This has been fixed in SVN - it used to search through items for on-use effects in an ascending order, but I reversed it, meaning it should always put trinkets first, on the assumption that trinkets tend to be more powerful than the springs.

It's worth noting that before this fix it also wouldn't automatically sync up metamorphosis with the chalice, even for non-engineers, but it does now (in SVN).

netsach · 08/11/11, 9:27 AM

I didn't find the info on these forums, so i tried to simulate it using the svn version, here are the results i got :

Affliction : int 1 spell 0.8627 hit 0.727 crit 0.3219 haste ~0 mastery ~0
Destruction int 1 spell 0.8052 hit 0.7372 crit 0.3421 haste ~0 mastery ~0
Demonology (mastery factor = 3) : int 1 spell 0.7775 hit 0.6246 crit 0.3629 haste ~0 mastery 0.8073
Demonology (mastery factor = 1) : int 1 spell 0.8176 hit 0.6437 crit 0.3714 haste ~0 mastery 0.342

This was simmed using no equipement but pure stats (4500 int, 2400 spell, 1650 hit, 1000 crit, 1000 mastery, 2350 haste) and 10000 iterations, with fight duration fixed to effective dooomguard duration, and actionlist containing only buffs, no pet, no spells except doomguard on pull with a pre-pot. No demon soul nor meta were used, thoudh i tested it just to be sure but that changed nothing. Edit : I just realize i didn't though of adding hand of gul'dan in the sim for the additionnal demon crit, but i don't even know it the doomguard would benefit from it).

Side note : simming my own demo warlock afterwards and changing the mastery factor from 3 to 1 resulted in total doomguard damage going down from ~800k to ~550k (ie approx 31% loss) so it's a hint at what will happen when Blizzard fixes the triple mastery bug (similarily to what happened when they fixed the imp/racial bug)

angaroth · 08/11/11, 3:39 PM

Originally Posted by netsach

the triple mastery bug

Which bug is this?

netsach · 08/12/11, 4:31 AM

> Which bug is this?

The doomguard currently gets a damage multiplier in which the contribution from mastery is tree times what it normaly would be (ie once). More info here, here and here. I haven't found any "reliable" source to say where or what is wrong, nor why it's there. My personnal opinion is that the devs could use this mechanic to tune the demonology output without having any influence on other specs.

This bug is implemented via the following (in green color) fix in simc code (and simc doomguard numbers line up with live data)

// FIXME: Somehow he (the doomguard) gains three times as much from mastery as expected?
m *= 1.0 + ( o -> mastery_spells.master_demonologist -> ok() * snapshot_mastery
           * o -> mastery_spells.master_demonologist -> effect_base_value( 3 ) / 10000.0 ) * 3;

Scaron · 08/14/11, 2:05 PM

I notice that when simming myself as Demonology and changing to the "actions+=/metamorphosis,if=buff.moonwell_chalice.up" in order to simulate stacking metamorphisis with Demon Soul and Moonwell Chalice the DPS distribution of the 10k iterations I ran was slightly bi-modal such that a noticeable minority of iterations would net me a significant amount less average DPS than the majority of iterations.

Can anyone help me explain this slight bi-modality? I'd just like to figure out why it's there and see if I can't avoid the situation that caused it when I'm actually in a raid situation.

Darkmantle · 08/15/11, 1:49 AM

Originally Posted by Scaron

I notice that when simming myself as Demonology and changing to the "actions+=/metamorphosis,if=buff.moonwell_chalice.up" in order to simulate stacking metamorphisis with Demon Soul and Moonwell Chalice the DPS distribution of the 10k iterations I ran was slightly bi-modal such that a noticeable minority of iterations would net me a significant amount less average DPS than the majority of iterations.

Can anyone help me explain this slight bi-modality? I'd just like to figure out why it's there and see if I can't avoid the situation that caused it when I'm actually in a raid situation.

This is presumably because of the random chance for your meta cooldown to be the full 3 minutes. That would produce a very long wait time for moonwell chalice and due to the huge effect meta + chalice has on our dps it would produce an overall low result. The major problem with mastery scalling meta is that the higher our mastery gets the more cricial meta becomes to our overall dps.

• Zakalwe · 08/15/11, 2:09 AM

The problem with that theory is that the BiS profiles don't seem to exhibit the same bi-modality.

Also, Scaron, simulating you I don't really see what you're describing. Could you link your report, or at least the DPS distribution graph?

• Zakalwe · 08/15/11, 8:14 AM

Couple of changes in the OP: Since we finally have fully optimized T12N profiles, I've included a chart (with a link to a report with scale factors) for normal mode gear as well. And I've changed the tier set and trinket proc evaluations to use a T12N baseline.

I also made the default demonology action list much more robust to the effects of movement, which had a dramatic effect on the movement sims.

Gilgrissom · 08/15/11, 9:28 AM

There is another problem with Darkmantles theory: If all you do is add the condition on using Metamorphosis "actions+=/metamorphosis,if=buff.moonwell_chalice.up", this will not change anything if Metamorphosis ends up with a CD higher than 2 minutes, IF you are already syncing up the Trinkets with Metamorphosis like "actions+=/use_item,name=moonwell_chalice,if=cooldown.metamorphosis.remains=0|cooldown.metamorph osis.remains>cooldown". I would assume that you did not change this part and are in fact using Moonwell Chalice on Cooldown, which will then mean that you are not syncing up Metamorphosis with Moonwell Chalice/Trinkets if Metamorphosis has a higher cooldown than 2 minutes.

However, without actually seing your action priority list and the report that shows this bi-modality, it is pretty hard to theorize what could cause this. Please upload the report to somewhere and link it here.

Scaron · 08/15/11, 9:43 AM

I'm embarrassed to admit that the previously mentioned bi-modality appeared only in release 3 of Simcraft. When I re-ran the release 5 sim, the bi-modality was gone. I'm assuming that there was a bug or other inaccuracy somewhere in v.3 that created these results.

Gil, in answer to your question, I was using the same priority action last as you did above. However, as you were using release v.4, it likely was an idiosyncrasy that worked its way out of the Sim after v.3.

I apologize for my oversight.

• Zakalwe · 08/15/11, 10:07 AM

Originally Posted by Gilgrissom

There is another problem with Darkmantles theory: If all you do is add the condition on using Metamorphosis "actions+=/metamorphosis,if=buff.moonwell_chalice.up", this will not change anything if Metamorphosis ends up with a CD higher than 2 minutes, IF you are already syncing up the Trinkets with Metamorphosis like "actions+=/use_item,name=moonwell_chalice,if=cooldown.metamorphosis.remains=0|cooldown.metamorph osis.remains>cooldown". I would assume that you did not change this part and are in fact using Moonwell Chalice on Cooldown, which will then mean that you are not syncing up Metamorphosis with Moonwell Chalice/Trinkets if Metamorphosis has a higher cooldown than 2 minutes.

You're reading that line wrong - it will always sync up MWC with meta. The second part of the expression will never evaluate to true, because the cooldown remaining on meta will never be higher than the full cooldown of the trinket except when the trinket is still on cooldown anyway. (In other words, meta's full cooldown duration will never be above 2 + 2 = 4 minutes. Obviously, since it starts at 3 minutes.)

The line you're looking at is generated in the same generic manner for all on-use trinkets. It's constructed that way to make sure trinkets with a lower cooldown are used in between metas.

Gilgrissom · 08/15/11, 10:37 AM

My idea was that he did not change the line using the MWC and was in fact just using the MWC when it was ready:

actions+=/use_item,name=moonwell_chalice

instead of

actions+=/use_item,name=moonwell_chalice,if=cooldown.metamorphosis.remains=0|cooldown.metamorph osis.remains>cooldown

This would have desynced MWC and Metamorphosis if Metamorphosis had a cooldown greater than 140 seconds. (On a sidenote, this is the case for the automatically generated use of Blood Fury, the orc racial, which is being used on cooldown, regardless of Metamorphosis.) This could have explained the couple of iterations that were at significantly less DPS than the average. Turns out, the bimodularity was a "bug" of an older simcraft version, so we can just ignore the last couple of posts I guess.