Elitist Jerks Moonkin Beginner Guide (obsolete; use the TTT article now (in my sig))

07/30/09, 1:00 PM   #166
Hamlet
<Druid Trainer>

Tauren Druid

Mal'Ganis
 Originally Posted by Adoriele Both very interesting articles. Probably more suited to the question of "How many spells can I cast in a given Eclipse", but still very much worth a listen for anyone who's interested in the subject. On a similar note, there's a spreadsheet I left in Efejel's old thread that did the sort of calculations in the second article, with the goal of determining how many casts you could fit into the duration of one Moonfire back in TBC when IS and Wrath weren't worth casting. Thing are orders of magnitude more complex now, hence why WC no longer even tries to determine that sort of info and goes for simple averages instead of probabilistic ones, but a good deal of work went into it for anyone who'd like to dissect its innards. I'll see if I can get a link to it here.
To be clear, I thought about this while modeling and have been working under the assumption that the high degree of randomness in our cast times (most classes have lag, but we also have NG) smooths out the probability distribution sufficiently, even in a 15 second window, to the point where the difference between continuous and discrete models is well underneath the real-world precision of the model.

07/30/09, 1:34 PM   #167
Happy October 19th!

Night Elf Druid

Dragonblight
 Originally Posted by Arawethion To be clear, I thought about this while modeling and have been working under the assumption that the high degree of randomness in our cast times (most classes have lag, but we also have NG) smooths out the probability distribution sufficiently, even in a 15 second window, to the point where the difference between continuous and discrete models is well underneath the real-world precision of the model.
Agreed. Unfortunately, I think the only way to get accurate results is with SimCraft, which isn't portable to Rawr for gear prediction. The way I see the niches of the three main models is that WC is great for mucking with the internals to figure out the best way to model things, Rawr then takes that model and makes it useful for generating gear lists/choices, and SimCraft is a great way to compare things between classes in a closer-to-real-world way, and is also good for double-checking Rawr's outputs.

 07/30/09, 1:41 PM #168 • Hamlet     Hamlet Tauren Druid   Mal'Ganis This reminds me of something else I've been pondering while working on WC/TTT stuff. We've always been modeling NG uptime as a steady-state average for each type of spell, but of course in reality there are significant edge effects since we change spells so often. Now, there's a good reason to imagine that the edge terms largely cancel. Wrath has high NG uptime for example. So when you switch to Wrath, the model slightly overestimates your haste for the first two Wraths, until you're into the steady state, but it will correspondingly underestimate your haste for the next two spells you cast after switching away from Wrath. And in the current WC I posted, I extended this rationale to instants--I give Moonfire the same NG uptime as Wrath, and IS I simply treat as having no NG effect since it can't crit. Even though this fudges the time spent casting an individual spell, it should provide a workable estimate of the eventual NG contribution from the spell. Curious if anyone has ideas for improving on this.
 07/30/09, 2:10 PM #169 dedmonwakeen Bald Bull   dedmonwakeen Undead Priest   No WoW Account Given the fact that the Eclipse CDs will be unlinked, it may very well now be possible to model an optimal Moonkin cast sequence as a markhov chain ala Rawr.Mage. I'm not saying that this reduction in complexity makes it easy..... but it may now be in the realm of possibility.
07/30/09, 4:22 PM   #170
Happy October 19th!

Night Elf Druid

Dragonblight
 Originally Posted by Arawethion This reminds me of something else I've been pondering while working on WC/TTT stuff. We've always been modeling NG uptime as a steady-state average for each type of spell, but of course in reality there are significant edge effects since we change spells so often. Now, there's a good reason to imagine that the edge terms largely cancel. Wrath has high NG uptime for example. So when you switch to Wrath, the model slightly overestimates your haste for the first two Wraths, until you're into the steady state, but it will correspondingly underestimate your haste for the next two spells you cast after switching away from Wrath. And in the current WC I posted, I extended this rationale to instants--I give Moonfire the same NG uptime as Wrath, and IS I simply treat as having no NG effect since it can't crit. Even though this fudges the time spent casting an individual spell, it should provide a workable estimate of the eventual NG contribution from the spell. Curious if anyone has ideas for improving on this.
You spurred me to work on some math that got wildly out of hand, so I put it on my Blog to keep the thread at more of a beginner's state. Link's in my sig for anyone who's interested in the math we've been talking about recently, and I attached the copy of my old spreadsheet that I mentioned earlier on. If you really want to break your brain, imagine that, in order to get truly accurate results, you'd need to generate the tree for each cast that's generated in the sheet (though you'd need a much smaller tree since, at that point, it was a lot easier to figure out which casts came before the cast in question).

 07/30/09, 5:29 PM #171 • Hamlet     Hamlet Tauren Druid   Mal'Ganis Updated fully for 3.2.
07/30/09, 6:28 PM   #172
ehakam
Glass Joe

Draenei Shaman

Aerie Peak
 Originally Posted by Arawethion Correct. In fact, one of the things to examine more in the 3.2 TC is the potential gain from switching to SF immediately when you see that a Wrath has procced NG as it leaves your hands (thereby telling you that you have a 60% chance of starting an Eclipse when it lands).
Is this practical, or am I misunderstanding you? If you're trying to chain-cast, you'll have already selected your spell before you know whether or not NG proc'd. Therefore, if you have to decide which spell to cast, you introduce a reaction time delay (maybe 0.2 s). That's a pretty significant loss in dps, considering that this strategy affects all wrath casts before eclipse, not just one.

What about spell-weaving W's and SF's? It might have a slight edge at the start of a fight when both eclipse CD's are off.

07/30/09, 6:39 PM   #173
Happy October 19th!

Night Elf Druid

Dragonblight
 Originally Posted by ehakam Is this practical, or am I misunderstanding you? If you're trying to chain-cast, you'll have already selected your spell before you know whether or not NG proc'd. Therefore, if you have to decide which spell to cast, you introduce a reaction time delay (maybe 0.2 s). That's a pretty significant loss in dps, considering that this strategy affects all wrath casts before eclipse, not just one. What about spell-weaving W's and SF's? It might have a slight edge at the start of a fight when both eclipse CD's are off.
Well, evidence seems to suggest that queuing only works if the following two conditions are both false: a.) you cannot queue during the first second of the GCD, and b.) you cannot queue until some short period of time before your current cast finishes, empirically seen as .3s, but it could have something to do with your latency. With those in mind, if you have over 400 haste and NG has procced, you should be able to see NG refresh before it's possible to select the next spell in your sequence. Granted, the chance of actually doing this is slim-to-none, but it does exist.

Realistically, the spell you cast after Eclipse procs is going to be a lost cause. Due to Wrath's travel time and short cast, though, it is possible that you could be already queuing the second spell after the one that procs Eclipse by the time Eclipse appears on your buff bar. NG would have procced a non-trivial period of time before that, and it's that eventuality that watching for NG procs/refreshes hopes to prevent.

07/30/09, 8:44 PM   #174
Hamlet
<Druid Trainer>

Tauren Druid

Mal'Ganis
 Originally Posted by ehakam Is this practical, or am I misunderstanding you? If you're trying to chain-cast, you'll have already selected your spell before you know whether or not NG proc'd. Therefore, if you have to decide which spell to cast, you introduce a reaction time delay (maybe 0.2 s). That's a pretty significant loss in dps, considering that this strategy affects all wrath casts before eclipse, not just one. What about spell-weaving W's and SF's? It might have a slight edge at the start of a fight when both eclipse CD's are off.
No, I really meant, one Wrath after the one that procs NG. That's still 1-2 Wraths earlier than if you wait until after the travel time to react.

 07/30/09, 10:20 PM #175 • Hamlet     Hamlet Tauren Druid   Mal'Ganis More miscellaneous questions while I work on TC. 1) General heuristic question. In the 3.2 rotation, would you rather cast an instant during Eclipse or during pre-Eclipse? Eclipse DPS: 6500/7000. Pre-Eclipse DPS: 4600/5000. Cycle length: 44s. Cycle damage: 276000. Cycle damage w/o DoTs: 233000. GCD: 1.2. Casting an instant during Lunar costs 1.2*7000/276000=3.04% DPS from that cycle. Solar: 1.2*6500/276000=2.82% Casting during pre-Eclipse: 1.2/45.2*233/276=2.24%. (it extends the cycle, but this only effects non-DoT DPS) So, to first order, casting outside Eclipse is better in 3.2. Significantly enough that it's not going to reverse with small changes in gear, but small enough that it's unlikely to be worth sacrificing DoT uptime in favor of one or the other. Conclusion: Without the post-Eclipse phase, the opportunity cost of an instant cast is similar both inside and outside Eclipse. This preponderates in favor of constant DoT refreshing. I'm going to try to work this into the spreadsheet, as noted in the OP. 2) Lunar clairvoyance--is it worth casting Starfire when you see Wrath proc NG in PL phase? 3 possibilities: A) That Wrath does in fact proc Eclipse. You've cast one more Wrath since since then and then started a Starfire. 60%. B) That Wrath does not proc Eclipse, but the next one also crits and procs. This is actually the ideal situation. 40%*proc chance = 10%. C) Neither Wrath procs. 30%. C: you waste the duration of one NG'd SF cast during PL. Damage/cast of one non-Eclipsed Starfire: 10700/2.02s. Cycle DPS w/o DoTs: 6000. Damage lost. 2*6000 - 10700 = 1300. Really very little loss from spending 2s casting uneclipsed Starfire. A: you gain some time casting Eclipsed Starfire instead of uneclipsed Wrath. Note that this can increase your effective L phase duration above 15s, if you start casting a Starfire before Eclipse procs. The tricky part is determining how much time you gain--it depends on how quickly you ordinarily manage to react to the proc. And that in turn depends on how far into a Wrath cast the proc occurs, which varies by travel time. Let's assume you typically lose 2 Wraths to travel + delay. This is conservative. The DPS gain of the early swap is (Wrath cast)*(SFEclipseDPS-WDPS)=1.2*(7100-4600) = 3000. B: The very lucky case. Twice the DPS gain of the prior case. This seems pretty clearly in favor of casting the SF and I'll try to work it into the model in more detail.
 07/30/09, 10:47 PM #176 Poromu Piston Honda   Poromu Night Elf Druid   Windrunner Adoriele This might not be the appropriate place to ask but since it is the Beginner thread and I figure it's probably a irreplaceable beginner addon - will you be updating S&A for 3.2 to show both Eclipse CD's?
 07/31/09, 1:40 AM #177 • Hamlet     Hamlet Tauren Druid   Mal'Ganis 3.2 spreadsheet in the OP updated with a constant-DoT-refreshing model. Turns out a bit better than I'd hoped. Haven't put in the Lunar clairvoyance yet. The two interact in kind of an annoying way.
07/31/09, 1:52 AM   #178
Altiris
Von Kaiser

Tauren Druid

Zul'Jin
 Originally Posted by Poromu Adoriele This might not be the appropriate place to ask but since it is the Beginner thread and I figure it's probably a irreplaceable beginner addon - will you be updating S&A for 3.2 to show both Eclipse CD's?
It's largely needless to track both, as the CD for one will be over by the time you will want to proc it, unless you are clicking off buffs. If you have it track the cooldown now, the timer will reset when you proc a different eclipse, so its current functionality is sufficient.

07/31/09, 2:43 AM   #179
Poromu
Piston Honda

Night Elf Druid

Windrunner
 Originally Posted by Altiris It's largely needless to track both, as the CD for one will be over by the time you will want to proc it, unless you are clicking off buffs. If you have it track the cooldown now, the timer will reset when you proc a different eclipse, so its current functionality is sufficient.
Ah well, if the Eclipse bar gets replaced once I proc the 2nd eclipse that'd be fine, I only want to track it in case I have 15s without an Eclipse proc or with too much movement to proc it. This way I can reset back to Lunar.

 07/31/09, 4:53 AM #180 Miim Von Kaiser     Miimzy Gnome Mage   Shadowsong (EU) Has anyone done calculations on how good the mana regen portion of the 3.2 Owlkin Frenzy is? Before you criticize someone, you should walk a mile in their shoes. That way, when you criticize them, you're a mile away and you have their shoes.

 Elitist Jerks Moonkin Beginner Guide (obsolete; use the TTT article now (in my sig))