Moonkin PvE Discussion, 3.3.3 - Page 7

Cdin · 03/23/10, 10:19 PM

Ok, I did several sets of simulations using SimulationCraft. Most of them were based off of my toon, but I did do a test using Lappe assuming he is one of the best gear moonkin currently in game.

For each set of gear I ran two simulations changing how DoTs are applied in each sim. The first past was using the default rotation code of:

actions+=/moonfire,if=!ticking
actions+=/insect_swarm,if=!ticking

With the second I try to mimic the logic outlined by Hamlet. The code for it is:

actions+=/moonfire,if=!ticking&buff.solar_eclipse.cooldown_remains&buff.lunar_eclipse.down
actions+=/insect_swarm,if=!ticking&buff.lunar_eclipse.cooldown_remains&buff.solar_eclipse.down

For the first gear set I used my current gear set. Using the default rotation My sim DPS was 10773 when I switched the DoT code my sim DPS dropped to 10737.

In the second set I switched out my Nibelung for the 264 versions of FBS and SSS. For that gear my default sim DPS was 10476 and the custom rotation was 10454.

In the third set I switched my IotDS for the normal version of Muradin's Spyglass to push myself well over the caps. The Default DPS was 10593 and the custom DPS was 10562.

Finally I wondered if I might be the problem, so I exported Lappe from Paragon. Using the same two rotations I got a sim DPS of 11712 for the default rotation, and 11668 DPS for the custom rotation.

In short, in all the situations I tested, refreshing DoTs on cooldown resulted in more DPS then only refressing the dots after Eclipse just finished.

• Hamlet · 03/23/10, 11:34 PM

SimC is giving a DPS increase from constantly refreshing an unglyphed IS? That seems odd.

-------

Our Resto Shaman said that Water Shield didn't proc on BQL today. I'm going to guess that Shroud of Sorrow was fixed and Owlkin Frenzy won't proc anymore either.

Erdluf · 03/23/10, 11:47 PM

Cdin,

I ran team robot with your character, from armory, except that I changed GoIS to GoStarfall.
I used a queue lag of 30ms and a gcd lag of 100 ms. Other settings were defaults (including 300s fight), except for number of runs (1000).

Default rotation (always maintain DoTs) gave 10057 DPS.
Change IS to only be cast when Solar is ready (off cooldown) and MF only when Lunar is ready gave 10187 DPS.

Both numbers have reported +/- of around 27 (1000 runs each).

Note, one difference between our 1-dot rotations is the start of the fight. My logic casts both MF and IS before the first nuke (or whenever both cooldowns happen to be up). Your logic wouldn't cast IS until after the first Lunar, and not cast MF until 30+ seconds into the fight. That means my idol was at full strength about 10s into the fight, your took something like 25s to reach full strength.

The differences in either case are small (your numbers favored continual DoTs by 0.3%, mine favored 1-dot by 1%). By comparison, I've seen 2% boosts by adding end-of-fight logic to the rotations (no IS with <14s left in fight, no MF with < 15s left in fight, Typhoon+MF when no time for a final nuke (Wrath has travel time in Robot)).

The Wrath travel time may account for part of the difference between simulators. Travel time means it takes longer for Robot to proc the first Lunar, on average, but Solar->Lunar gaps should be smaller on average (since wraths completed during Solar can still proc Lunar).

dedmonwakeen · 03/24/10, 12:01 AM

Originally Posted by Erdluf

The Wrath travel time may account for part of the difference between simulators. Travel time means it takes longer for Robot to proc the first Lunar, on average, but Solar->Lunar gaps should be smaller on average (since wraths completed during Solar can still proc Lunar).

SimulationCraft models Wrath with a 21 meter/yard per second travel time with Eclipse, T10-4pc, and E&M being triggered on impact.

Cdin · 03/24/10, 9:25 AM

Erdluf,

Are you sure you used the correct spec? If you had to change GoIS to GoStarfall then you were likely using my Saurfang/Heroic Anub spec that does not have Improved IS. Unfortunately I forgot to switch back last night before I logged out, but if you used my Typhoon spec then that would explain some of the differences in our results.

The IotLE stacking may also be part of the issue. Does anyone know how to start a fight in SimulationCraft with both DoTs but not increase their priority in the spell rotation?

EDIT:

I tried to run the simulation with Team Robot using my standard spec.

In the first run the casting conditions were:

MF: Not ( Target has MF)
IS: Not (target has IS)

This run resulted in a DPS of 10746.7

In the second run the casting conditions were:

MF: Not ( Target has MF) & Lunar is ready & Solar is Cooled Down
IS: Not (target has IS) & Solar is ready & Lunar is Cooled Down
No other modifications were made to the default rotation

This run resulted in a DPS of 10063.7

I haven't worked with Team Robot much so I may have made a mistake somewhere, but the Team Robot sim seems to confirm my results from SimulationCraft

Erdluf · 03/24/10, 9:38 AM

Originally Posted by dedmonwakeen

SimulationCraft models Wrath with a 21 meter/yard per second travel time with Eclipse, T10-4pc, and E&M being triggered on impact.

So both simulators should be similar in that regard.

I think sinulationcraft has a slightly more optimistic latency model, so I re-ran Robot with both lags set at zero (Graylo's armory had changed so that he already had GoSfl, I don't know if there were other changes). I've also increased the number of runs to 10k, to reduce the reported error margin to around +/-9

Full DoTs: 10746.5
One DoT: 10763.2

Note that one-dot, as modeled above, has the advantage of not wasting as many ticks at the end of the fight. I modified both rotations to not cast IS after 286s, or MF after 285s. Results:

Full DoTs: 10773.8
One DoT: 10768.4

That last comparison is a difference of just 1/20 of one percent, about the equivalent of blinking at the start of the fight, and taking an extra 1/7s for your first cast.

Edit: For these later tests you do not have IIS. You do have GoSfl. I'll add IIS (taking points from Typhoon and GW) and run again.

Cdin, for your later post, this might work for simcraft. This assumes Simc has both eclipses on cooldown if either eclipse is active.

actions+=/moonfire,if=!ticking&!buff.lunar_eclipse.cooldown_remains
actions+=/insect_swarm,if=!ticking&!buff.solar_eclipse.cooldown_remains

Edit2: My One DoT logic is

MF only if (NOT (Target has MF) and (Lunar Is Ready))
IS only if (NOT (Target has IS) and (Solar Is Ready))

So both DoTs get applied at the start of the fight (or in rare circumstances where it takes more than 15s to proc eclipse).

Starfox · 03/24/10, 12:07 PM

Just wanted to point somethin out here quick:
SC in latest release does not have the 15s cd on the other eclipse if you proc it, just realised it and added that.
So only using cooldown_remains may not bringt the results you'd expect. :S

Erdluf · 03/24/10, 12:20 PM

Team Robot, 10k runs, Using Graylo (5/5 t10.264) with 3/3 IIS, glyphs of MF, SF, and Sfall, no end-of-fight logic, 0 delay on all casts:

Full DoTs: 10944
One DoT (both at start): 10868

Changing latency to 30ms queued, 100ms gcd I get:

Full DoTs: 10280
One DoT (both at start): 10281

The difference is pretty small, and the "winner" likely depends on things that we aren't measuring here (differences in gear, movement, pushback, ...)

aceofsween · 03/24/10, 12:32 PM

That was pretty much the conclusion I came to myself Erdluf. Although, if you're going to use Full DoTs, Insect Swarm will probably provide better results at the cost of the hit debuff.

As far as which setup is actually better, I think movement (and obviously gear, but that's something we can measure) is going to be the biggest thing to take into consideration, and I'm really not sure which is better to be honest. Perhaps the 1 DoT setup, simply because you're not going to be clipping anything during movement, but actually applying a damage source that wouldn't otherwise be getting use.

However, in the end that's speculation and I'm really not sure.

Cdin · 03/24/10, 12:57 PM

Originally Posted by Erdluf

The difference is pretty small, and the "winner" likely depends on things that we aren't measuring here (differences in gear, movement, pushback, ...)

I agree. My main point is that I don't think we can definitively say what the best strategy for refreshing DoTs is.

I get a lot of emails that say things like "Murmurs says we should cast IS when Solar is up," or "EJ says I should only cast one DoT." When I test these things with the tools available I can't match the claim.

Ultimately, I think because of the change to Starfall and the fact that Moonkin DoTs do not scale very well has lead a lot of people to believe that DoTs are inferior to nukes at high levels of gear, and should be dropped from the regular rotation.

While modifying our DoT refresh strategy may not hurt our DPS much, I am not seeing any indication that it real helps our DPS either. Therefore, there is no real need to recommend anything different then we were doing before 3.3.3.

• Hamlet · 03/24/10, 1:10 PM

I was generally impressed by the degree to which WC and Robot have been agreeing, and have been mostly going with those results. A few pages back when people were using Robot to check rotations in detail, lower DoT uptimes were looking favorable, although the differences were small and hard to resolve. I do think the point about favoring a rotation with fewer DoT's even when they're equal on paper, because of the reduced opportunity cost of movement, is an important one.

Dropping a DoT entirely looked like it had some potential for a short period before the Focus nerf. That's why people keep asking all those question. In truth, I don't think that it was a good idea even then, it just looked like before we examined low-uptime DoT rotations more. At the moment, there's no reason to leave out either DoT. But I think the weight of the evidence is in favor of lower uptimes than before. There's also a bit of common sense here--an unglyphed IS or an unextended Moonfire have such low DPET that's they're clearly not worth casting at certain times (say at the beginning of the "wrong" Eclipse). So even independently of the difficulties of modeling this all, there are going to times when it's not optimal to refresh.

The differences are so small at that all but the highest levels of play, it might not be worth worrying about the differences. I'd still recommend refreshing only after Eclipse to newer players, because it's probably easier to do consistently on the whole, and will leave them better prepared to learn to handle movement. And for advanced players who are going to execute any rotation correctly, it looks like partial DoT uptimes produce optimal DPS by a small margin.

Yijiao · 03/24/10, 3:31 PM

Originally Posted by Erdluf

Team Robot, 10k runs, Using Graylo (5/5 t10.264) with 3/3 IIS, glyphs of MF, SF, and Sfall, no end-of-fight logic, 0 delay on all casts:

Full DoTs: 10944
One DoT (both at start): 10868

Changing latency to 30ms queued, 100ms gcd I get:

Full DoTs: 10280
One DoT (both at start): 10281

The difference is pretty small, and the "winner" likely depends on things that we aren't measuring here (differences in gear, movement, pushback, ...)

So thats an example at one gearing level, has anyone checked at other gearing levels? I notice he is a little lighter in haste, I was curious if higher amounts of haste would diminish the value of DoT refreshing relative to the nukes. Certainly, in regards to starfire, it absolutely would.

Can you run numbers on say, my gear?

aceofsween · 03/24/10, 3:52 PM

One thing that I have noticed is that these slight differences in rotations come up with less than 1% of your total DPS (as most of us are comfortably over the 10k mark). Anything that we attempt to test out in game is going to require massive amounts of carefully controlled scenarios which are quite simply impossible to actually control.

And these fractional changes in rotations are really not going to be visible. Take the glyphs issue. The difference for me comes out to be less than 5 DPS. That's 0.05% difference. Not only will human error cause that difference to be irrelevant, but so to will the nature of RNG (not to mention the margin of error for the approximations themselves).

In the end, it really comes down to personal preference and to some extent what you want to bring to the raid.

EDIT: I did want to add though that the Focus glyph may not be so insignificant on certain fights. During the Lich King attempts last night, Starfall was really doing some insane damage, particularly in phase 1 and 3. Every attempt I could usually get 2 Starfalls off in phase 1 and they would do around 300,000 damage total. That would be easily 30,000 extra damage. Is it worth it though? I don't know. For me it's not but again that's a personal preference. I prefer my max range and if you're using Focus you should also be using glyph of Insect Swarm.

Lord BEEF · 03/24/10, 4:25 PM

I've been mainly sticking with the strategy of just keeping unglyphed insect swarm up for the hit debuff and my idol, and moonfiring during movement.

I have some questions about starfall though that will impact my timing of it. How big is the difference in total damage in the following scenarios:

Single target
Two targets, spaced apart
Two targets on top of each other

I think I read starfall uses more stars if you have more than one target but I'm not sure on what the real difference is.

Starfox · 03/24/10, 4:58 PM

Originally Posted by Lord BEEF

I've been mainly sticking with the strategy of just keeping unglyphed insect swarm up for the hit debuff and my idol, and moonfiring during movement.

I have some questions about starfall though that will impact my timing of it. How big is the difference in total damage in the following scenarios:

Single target
Two targets, spaced apart
Two targets on top of each other

I think I read starfall uses more stars if you have more than one target but I'm not sure on what the real difference is.

Two targets, spaced apart

20 * ( 608 + .37*SP + 101 + 0.13*SP )
20 * ( 709 + .5*SP )
14180+ 10*SP

Two targets, spaced apart

20 * ( 608 + .37*SP + 101 + 0.13*SP*2 )
20 * ( 810 + .63*SP )
16200+ 12.6*SP

Difference
2020 + 2.6*SP

If both targets are debuffed, it is a ~3600 damage difference @ 3400SP