Jump to content


Photo

Moonkin PvE Discussion, 3.3.3


  • This topic is locked This topic is locked
328 replies to this topic

#21 Yellowsix

Yellowsix

    Piston Honda

  • Members
  • 166 posts

Posted 08 March 2010 - 11:09 PM

A side note on the Team Robot simulator: I'm going to be updating/improving the error range that is calculated for a set of simulation runs. My statistics were a bit rusty when I wrote that code, and it definitely needs a round of review.

With regards to how latency works, I've had several discussions with people in the druid and shaman forums. I will keep an updated post with the currently implemented latency/delay mechanics at: How Does Mr. Robot Handle Latency and Reaction Time?.

Some of that information might help to shed light on differences between the simulator's results and WrathCalcs. Feel free to respond to that thread if you feel that the current implementation is either incorrect or not working as advertised.

#22 Videl

Videl

    Piston Honda

  • Members
  • 111 posts

Posted 08 March 2010 - 11:28 PM

Did you compare moonfire/starfire/starfall to IS/starfire/starfall ace?

#23 Balancemoon

Balancemoon

    Piston Honda

  • Members
  • 129 posts

Posted 08 March 2010 - 11:54 PM

At least from this patch on, spamming moonfire while moving might actually be not quite so much a dps loss since we won't be using the glyph anymore, the initial portion damage, will be much higher and should count for something.

Remember in the days before moonfire glyph, you would sometimes spam it while moving simply because of the instant damage portion.. might be high time to remember that Moonfire isn't relaly a fully fledged DoT like Insect Swarm.

I was hoping blizzard would give a second Moonfire and IS glyph just like they gave a second Rejuv glyph. The MF glyph would focus on a heavier intiial damage and probably slap on a 3 second cooldown, whiles the IS glyph will allow IS to crit. Now try making 3 glyph suggestion :) Probably Starfall, new MF glyph and new IS glyph: Perhaps in cataclysm, a better designed druid talent trees will enable a balance druid to emphasise on stronger DoTs [possibly by a certain route down the balance tree couple with modified feral talents that are now have useful bits for balance] or heavier nuke damage through a slightly different balance path and resto off spec.That way you can choose your 4 glyph slots to as IS, new IS, MF and Starfire glyphs if you are going DoT heavy,

#24 Guest_aceofsween_*

Guest_aceofsween_*
  • Guests

Posted 09 March 2010 - 06:48 AM

Did you compare moonfire/starfire/starfall to IS/starfire/starfall ace?


Yes, I did. I actually started with that combo before moving on to the one I described above. Like I said, it was strange to me that not casting IS was giving me lower results, even without the glyph. When I put the IS glyph on in place of the MF glyph, the difference was even greater. And like I said before, amusingly the difference between Starfire and Focus seems to be about 30 dps.

#25 Erdluf

Erdluf

    Great Tiger

  • Members
  • 968 posts

Posted 09 March 2010 - 01:29 PM

The default Lag setting for Robot is 150 (if memory serves). My latency tends to be in the 75 to 50 range, so in both Wrathcalcs and Robot, I've modified the value to reflect this. I'm aware of the 4t10 bug (as I'm the one who brought it up), but I wasn't aware of the way NG worked with IS as far as WrathCalcs goes. Even still, I don't believe the difference is so great that it would generate a swing of nearly 200 dps between the two models.


The Mr Robot settings page has a Latency, which is essentially meaningless for us. It has two Lag settings, Queue Lag and GCD Lag that do matter.

Queue Lag (default is zero, I think), applies to cast-time spells. GCD Lag (default is 50 ms) applies to instant-cast spells. From his write-up, I'd expect Wrath to use the queue-lag. From looking at his logs, Wrath (when under 1s) actually uses GCD Lag.

Note that if your DPS is in the neighborhood of 10k, wasting 20ms every second is a loss of 200 DPS. That is a much bigger effect than minor variations in DoT strategy.

Running Robot with

Latency: 150 ms
Reaction: 200 ms
Queue Lag: 0 ms
GCD Lag: 100 ms

and then filtering to just show Wrath, I consistently see Wrath casts 1.1s apart (=1s gcd + 100ms lag). That is the same gap I see between the end of one SF, and the start of the next SF, with an IS in between. When chain-casting SF, there is no delay between the end of one SF and the start of the next.

#26 Yellowsix

Yellowsix

    Piston Honda

  • Members
  • 166 posts

Posted 09 March 2010 - 07:15 PM

I've updated the explanation on the Team Robot site to reflect what Erdulf said: spells like wrath that are limited by the GCD rather than their cast time will incur the GCD Lag penalty before the next action can be performed, just like chain-casting an instant like moonfire.

#27 MÃ nze

MÃ nze

    Von Kaiser

  • Members
  • 50 posts

Posted 10 March 2010 - 09:57 AM

2) If you do not that have the 2T9 bonus (for example, if you've upgraded to 4T10):
2A) If you do not need the Insect Swarm debuff (you are the second Moonkin or are doing content for which it's unnecessary), Glyph Starfall/Focus/Insect Swarm, and leave Moonfire out of your rotation entirely.

Does this imply that we need 3% additional crit (137.7 critrating) to hit the lunar critcap, since theres no MF on the target ? Would be nice to see this mentioned in the guide. Does Wrathcalcs takes this into account, while calculating the scale values?

2B) If you do need the Insect Swarm debuff, Glyph Starfall/Focus/Starfire, unless Focus causes range problems at the fight you're doing. In that case, Glyph Starfall/Starfire/Moonfire.

So the rotation stays the same as in 2a) ? Keeping IS up nearly all the time and cast MF only while moving?

Thanks in advance.

#28 Hamlet

Hamlet

    Mike Tyson

  • • Guide Author
  • 11581 posts

Posted 10 March 2010 - 02:24 PM

Does this imply that we need 3% additional crit (137.7 critrating) to hit the lunar critcap, since theres no MF on the target ? Would be nice to see this mentioned in the guide. Does Wrathcalcs takes this into account, while calculating the scale values?


Yeah, I'll have to change the sheet. Should be easy, although I might wait to see exactly what rotations seem worth including in the next version.

The provisional version with a quick "one DoT after Eclipse" rotation is still here: http://elitistjerks....34/#post1574000 . You can use that to play with different glyphs/rotations. I should probably try to put in something a bit more final before the patch goes live.

The discussion has been really good since the new thread by the way, I've just been doing other stuff for a few days. To suggest one thing for people who are already playing with rotations in Robot (I haven't gotten around to it yet)--it should be able to take into account the "NG clairvoyance" effect in a way that WC can't really easily. It would be nice to set up a rotation that refreshes DoT's after the final nuke of Eclipse and then checks NG during the GCD to see whether to swap nukes immediately or not. Also, people trying out rotations in Robot should post the actual rotation they're using.

#29 Lord BEEF

Lord BEEF

    Soda Popinski

  • Guild Members
  • 3826 posts

Posted 10 March 2010 - 08:46 PM

After upgrading to 4t10 and seeing how little moonfire comprises of my total DPS since I just use it during movement I think I may actually drop the moonfire glyph for the rebirth glyph on some fights. Being able to res immediately instead of waiting for a boss ability could definitely be a raid DPS increase on a few fights versus the moonfire glyph.

#30 Erdluf

Erdluf

    Great Tiger

  • Members
  • 968 posts

Posted 11 March 2010 - 12:32 AM

To get people started on Mr Robot actions:

To use IS only before, or near the beginning of Solar:

Ability IS only if(NOT(target has IS) AND ((Solar is ready) OR (target has Solar and <= 5s has elapsed)))

To try switching to SF when you're trying for Lunar, and have a Nature's Grace that is probably (not certainly) from a Wrath currently traveling, put this ahead of the current Wrath action. You'll keep the simple SF at the end of the action list:

Ability SF only if((Lunar is ready) AND (NOT(the previous action was SF) AND (target has Nature's Grace and <= 1.3 sec has elapsed) AND NOT (target has Starfall)))

I know that our Target's don't get Nature's Grace or Starfall buffs, but Mr Robot is a bit imprecise. These rules seem to do the right thing.

#31 gannonjf

gannonjf

    Piston Honda

  • Members
  • 125 posts

Posted 11 March 2010 - 03:34 AM

Focus just went out the window.


Glyph of Focus: Now increases the damage done by Starfall by 10%, down from 20%.

#32 Eilt

Eilt

    Piston Honda

  • Members
  • 179 posts

Posted 11 March 2010 - 03:34 AM

Already changed Glyph of Focus on us to lessen the impact of the Starfall buff:

Glyph of Focus: Now increases the damage done by Starfall by 10%, down from 20%.


Edit: I don't think it makes it "out the window" but on single target fights it may not be worth it, but on fights with adds at regular intervals I still see focus as a good option.

#33 Hamlet

Hamlet

    Mike Tyson

  • • Guide Author
  • 11581 posts

Posted 11 March 2010 - 03:45 AM

I'll have to check, but it probably kills Focus for single-target.

#34 Hamlet

Hamlet

    Mike Tyson

  • • Guide Author
  • 11581 posts

Posted 11 March 2010 - 05:15 AM

Spreadsheet is up. I think we can pretty safely call S/SF/MF the best Glyph setup for ordinary situations. There's not really even much of a potential DPS gain to be found in Glyph of IS. We'll still have the discussion about DoT refreshing schemes, but at the moment I think the "one DoT after each Eclipse" plan looks strongest.

e: I forgot to change the crit cap to account for non-Moonfire rotations, but that's probably not too urgent since it looks like non-Moonfire rotations aren't going to come up anymore.

#35 Guest_aceofsween_*

Guest_aceofsween_*
  • Guests

Posted 11 March 2010 - 09:01 AM

I did finally manage to get the Robot simulator to show me the results I was looking for and I'm pretty impressed with them so far.

Glyphs: Starfall, Starfire, Moonfire
Moonfire: Target does not have Moonfire DoT and you do not have Solar Eclipse Buff
Insect Swarm: Target does not have Insect Swarm DoT and you do not have Lunar Eclipse buff and Solar Eclipse is ready.

Result: 10893.7 DPS +/- 26.6
74.9% uptime of Moonfire (roughly 1 application per rotation)
40.2% uptime of Insect Swarm (applied immediately after Lunar Eclipse ends).

Note: the Beta simulator still does not correctly value the 4t10 bonus OR Nibelung. Nibelung's damage should be roughly twice what it is (to reflect the 2% proc rate vs. a 1% proc rate). The 4t10 bonus is off by easily 4% of the damage done, which is a difference of ~425 dps if my math is correct.

That's the best I could come up with. I originally thought that it'd be better to drop IS from the rotation, but it turns out that the overall damage you gain while in Solar Eclipse is enough to make it worth casting before the Eclipse begins (but not worth refreshing during Eclipse after it falls off). Something I found interesting though... Using this rotation that I've mapped out, I'm starting to see Wrath and Starfire fight for the most damage done in these simulations where previously Starfire has always beaten out Wrath and usually by a fairly consistent margin.

Edit to add: Starfall/Moonfire/Insect Swarm is still giving me higher numbers (about 40 DPS) if you set them up so they do not refresh during the opposite Eclipse (Not Lunar for IS, Not Solar for MF). Still something to consider, at least...

#36 Hyperion

Hyperion

    Glass Joe

  • Members
  • 21 posts

Posted 11 March 2010 - 06:33 PM

Quote from Blizzard staff

Glyph of Focus: Now increases the damage done by Starfall by 10%, down from 20%.

Yep. Looks like they don't want us dropping our DoTs after all, but don't want to buff them either. I guess this is just a deterrence of sorts. Based on previous math, and the issue with the range as well, I'd say this glyph is no longer worth having.

#37 Yellowsix

Yellowsix

    Piston Honda

  • Members
  • 166 posts

Posted 11 March 2010 - 07:10 PM

I've updated the Team Robot beta simulator with the new Starfall glyph. 4T10 and the nibelung proc rate should be correct on the beta site now too.

#38 conghaile

conghaile

    Glass Joe

  • Members
  • 8 posts

Posted 11 March 2010 - 10:01 PM

Wanted to report my experiences with Team Robot and to compare to aceofsween's results. A bit of academic rigor trying to reproduce results, a little bit rumination.

Settings:
420 second fight, 1000 runs, 120 latency, 1500ms missile travel time, 0ms queue lag, 50ms GCD lag. intervals.

Glyphs: Starfall, Starfire, Moonfire
Moonfire: Target does not have Moonfire DoT and you do not have Solar Eclipse Buff
Insect Swarm: Target does not have Insect Swarm DoT and you do not have Lunar Eclipse buff and Solar Eclipse is ready.


The above is rotation "1." I tested the above rotation against:
2. Always refresh IS (strict 3% hit debuff duty), but MF under the same tactics 1.
3. Always refresh both DoTs (MF priority).
4. Not using IS at all, MF under the same tactics as 1.
5. Not using IS at all, always refresh MF.
6. Use of IS as in 1., Always refresh MF.

If you look at my armory profile, you can see I use normal Frozen Bonespike/Scourgelord's baton (as I do in practice), and Idol of Lunar Fury as the idol of Lunar Eclipse is the last piece of emblem gear I want to get (and I haven't gotten that many emblems). I would like to use the normal Nibelung I own, but I would have to re-gem for some hit in my current gear to use it, and the fights where the staff is not ideal make this undesirable. I initially ran tests for rotations 1 through 4 with my current setup, and also with Nibelung (still Idol of Lunar Fury though) but decided not to post them when I discovered at some point in the middle of running everything I got set to "randomize fight duration" which would have made my results a bit inaccurate. Also, I figure parses using Idol of Lunar Fury aren't relevant enough to this community to spend the time re-running the 8 cases. I will report in general terms that rotation 1 was the best at 9975dps (w/ Nibelung), rotations 3 and 4 were about 100 dps behind, and rotation 2 was another 25 dps behind with error ranges of about 35dps.

Now for some hard numbers!

When using Nibelung and IoLE, I get:

Rotation 1, IoLE, Nibelung: 10019.7 +-25.7dps
Rotation 2, IoLE, Nibelung: 9918.6 +-24.8dps
Rotation 3, IoLE, Nibelung: 9911 +-24.9dps
Rotation 4, IoLE, Nibelung: 9920.5 +-25.3dps
Rotation 5, IoLE, Nibelung: 9932.5 +-25.8dps
Rotation 6, IoLE, Nibelung: 10018.5 +-25.6dps

Looking at my results, the judicious use of IS as described by aceofsween is best practice for single target if insect swarm is unglyphed. The best way to use moonfire is unclear to me. It was mentioned before, either in this thread or in the old thread, that a moonfire in the middle of a solar eclipse still has a pretty good chance of getting fully extended; maybe that has something to do with it. I'd test for that case specifically, but I can't think of a good way to implement it (going over to Team Robot thread to ask). Something notable is that in rotation 1, insect swarm uptime was ~33%; strictly keeping the 3% hit debuff up per rotation 2 comes at a 100dps loss (ignoring error). That makes the tradeoff about 2% more misses on the tank for a 1% personal dps loss.

#39 Hamlet

Hamlet

    Mike Tyson

  • • Guide Author
  • 11581 posts

Posted 11 March 2010 - 10:46 PM

Phylactery spellpower values on 3.3.3 PTR are 1073 and 1206. It seems that the ICD is 100s though, not 90s.

Also, my current mean values of the spellpower proc used a 4s proc time, which is based on 5/6 DoT ticks per second at 30% (i.e. both DoT's up full-time). We're going to have lower DoT uptimes in 3.3.3; I'll use a 6s proc time for now (the effect of this is tiny anyway).

New mean spellpower values:
1073*20/106 = 202.5
1206*20/106 = 227.5

Prior estimated values were 195 and 220. So the buff is actually quite small (since we probably had the ICD wrong this whole time). I'll try to run a spreadsheet comparison soon to see how it compares to Reign/DFO when under/over the crit cap (or someone else can).

#40 gannonjf

gannonjf

    Piston Honda

  • Members
  • 125 posts

Posted 11 March 2010 - 11:11 PM

Phylactery spellpower values on 3.3.3 PTR are 1073 and 1206. It seems that the ICD is 100s though, not 90s.



Uhm, maybe I missed something but isn't that a buff compared to where they are at now live?




0 user(s) are reading this topic

0 members, 0 guests, 0 anonymous users