I'm using rough round numbers in this post:

Look at it this way. Suppose you didn't have eclipse, and your rotation was MF+SF. If you delay refreshing MF for 24s, then for the fight you lose 1 MF, but you gain one GCD of SF. For arguments sake, lets call that 2/3 of a SF (which had a 50% chance to be a crit). You also lose (24s-GCD-0.5 SF casts) of the IIS bonus to SF. 20s of 3% increased SF crit is roughly 3/10 of a non-crit SF. The penalty for your decision was (MF damage + 3/10 SF - 2/3*3/2 SF non-crit damage = MF Damage - 7/10 SF damage).

If you delayed refreshing MF for 6s, I'd argue that the average penalty is 1/4 of that = MF/4 - SF*7/40.

Now consider Eclipse. I'll assume that over the fight you average 65% SF crits.

If you delay a MF refresh for about 6s during Eclipse then you lose 1/4 MF for the fight. There is a 2/3 chance that you will convert an SF from non-eclipse to eclipse in the current eclipse, and there is a 1/6 chance that you will gain an SF during the current fight. You also lose IIS for about 2.5 SF casts (over the course of the fight). Using round numbers (average MF=11000, non-crit SF=6000, no CSD, average SF (crit+non-crit over fight duration)=9900):

Lose 1/4 MF + IIS bonus on 2.5 SF casts

Loss = 11000/4 + .03*2.5*6000 = 3200

Gain Eclipse bonus on 2/3 SF casts. Also gain a partial GCD over the fight: 1/4*2/3 = 1/6 SF cast over fight duration

Gain = .3*2/3*6000 + 1/6*9900 = 1200 + 1650= 2850

All of these numbers, except the "1200" on the line above scale with time. So if MF is down during Eclipse, and you wait for Eclipse to end in x seconds, the damage you gain is

1200 - x/6*(3200-1650) = 1200 - x*258

which gives a break-even at about 5s. Obviously this number is very sensitive to your actual damage numbers and timings.

For IS refreshes (assume IS does 7000 damage, t7 bonus+glyph), we lose the IIS term, and it has a GCD/14s. The Loss and Gain terms become

Loss = 7000 * x/14 = 500*x
Gain = .3*2/3*6000 + 2/3*x/14*9900

Gain-Loss (for delaying IS) = 1200 - x*29, which means never refresh IS during Lunar Eclipse.

So does this mean that refreshing MF and/or IS becomes more valuable at lower levels of crit? For instance, in my example, I have lower SP and my crit is only about 60% for SF when I get the proc. This means RNG is even less in my favor than your 80% example when it comes to providing the damage boost. My non-crit Starfires hit for 4100 average. My average crit is 8600. Now check my math on this:

My average Starfire during Eclipse proc: .4(4100) + .6(8600) = 6800
My average Starfire hits for: .7(4100) + .3(8600) = 5450

My standard cast time on Starfire is 2.8 due to my haste. This means my NG SFs are 2.3

.4(2.8) + .6(2.3) = 2.5s

That means that a 1.5s GCD would need to yield 4080 damage to make up for the average loss of time during eclipse? I would normally cast it right at the beginning of Eclipse proc my way, or at the beginning of the iCD if I didn't refresh during proc, so only the 15s of the proc would count. That's 5 ticks of MF + my hit. My avg moonfire tick is 921dmg, or 4605 damage. My average hit is about 160, so thats 4765 damage during eclipse? Right now, at 289 haste rating, I believe my GCD should be 1.38, which would make the needed amount lower.

This would change as you got higher crit, but at my current level of crit, MF is almost always past the point of being refreshed by a SF by the time eclipse procs, because it take 9s on average for me to hit a proc from Wrath at 30% crit. Perhaps I should change my cycle to IFF -> IS -> MF to give more opportunity for MF to be extended by SF without having to refresh? The issue with this is that it would seem to reduce the potential uptime of IS.

So it seems that higher crit and SP lowers the value of refreshing DoTs, while higher haste would increase the value?

Am I barking up the right tree here?

Graylo, your conclusions are not right.

For example saying that IS would have to do more than 3761.49dmg during eclipse to justify refreshing it is wrong.
You dont lose 3761.49dmg by using GCD during eclipse for IS because the dot ticks for 7000dmg for one GCD after all!
Only thing you possibly lose is 30% extra crit from the last eclipsed starfire which is aprox 1800dmg with t7.5 gear and like I wrote earlier, that is covered by 2 extra ticks of MF or 2-3 extra ticks of IS

One starfire with 80% to crit chance instead of the same starfire with 50% crit chance VS certain amount of free dot ticks.

No. In his example, his IS does around 4500 damage(326.09DPS*14sec). The assumption is that as long as eclipse isn't up, all dots are up. The only time where the DoTs should run off is during Eclipse procs.

Given that, DoTs can only get 15 seconds more uptime than they would otherwise. So everything is counted as a fraction of that.

For most DoTS, if they are refreshed at the beginning of Eclipse, they will have to be refreshed anyway (14s for IS and 24s for MF, if I'm not mistaken) during the cooldown, so you're not technically gaining a cool down. That's why they are only counting the timing during the 15s of Eclipse, because the question that's being asked is:

Is 15s of X DoT more DPS than 1.xs of Average eclipsed SF damage.

This is also, I believe, why you can't just look at the difference. You won't make up that 1.xs of Global cooldown later in the fight if you are following this rotation. In fact, if anything, the difference would be looked at the other way. If refreshing MF would clip one SF cast at the end, then you would likely have to figure the value of that damage difference in your calculations. Though I'm not sure about this 100%.

For a top end example, look at stats from a 6kDPS Patchwerk kill. Assume 6200 average noncrit SF, 12900 avg crit, 18% haste, 2.5s cast, 2s NG cast. Base raid crit. 1325 Avg Moonfire tick, 323 average hit. GCD would be 1.27.

.2(6200) + .8(12900) = 11560 SF damage during eclipse
.5(6200) + .5(12900) = 9550 damage outside eclipse

.2(2.5) + .8(2) = 2.1

This is ~5500 DPS for Eclipsed SF spam. MF would need to do 6990 in 15s of Eclipse to outpace this DPS. 5 * 1325 + 323 = 6948. So, it SEEMS to me like it would take a high level of gear to make MF NOT worth refreshing.

Also, it seems to me that we're not taking into account movement. If you are unable to remain stationary and cast during the entire 15s of Eclipse, the value of refreshing MF would seem to go up, AND IS can be refreshed on movement.

Put differently: the ratio of MF ticks to GCD's used on MF is fixed (subject to assumptions of steady-state, full Starfire extension, and no clipping (and note that the latter two assumption holds for single-Moonfire cycles that would allow it to expire in the first place)). You can't increase the number of MF ticks in a fight without increasing the number of GCD's spent casting MF.

So, everyone keeps trying to say that we want to determine the balance between
(1 GCD of Eclipsed Starfire) vs. (1 GCD of unEclipsed Starfire) + (X seconds of Moonfire Ticks) + (X seconds of IIS benefit during Eclipse),
but this analysis implicitly assumes that the X seconds of MF uptime come at 0 net time cost. Which is impossible.

It's just that the time cost appears in a subtle way. If you cast MF during Eclipse, then when Eclipse ends, you have a partial-duration Moonfire ticking instead of a freshly-cast full-duration one. This means you'll have to refresh it earlier, and then refresh the next one earlier, etc. I'm not exactly sure how to best quantify this, but I think the "net present value" of the cumulative delay is best expressed as a fraction of a GCD lost in the present.

In other words, the comparison we should be making is:
(1 GCD of Eclipsed Starfire) vs. (1 GCD of unEclipsed Starfire) + (X seconds of Moonfire Ticks) + (X seconds of IIS benefit during Eclipse) - (X/24 of a GCD).

e: Adoriele, do you get what I'm saying here? Maybe you can just grab some numbers from WC if you think this makes sense. Everytime we start talking about DoT refreshing a jillion people start reinventing the wheel of computing the DPET of all our spells.

e2: X/24 of a GCD casting what? I don't know. If the cycle we're purporting to analyze is "refresh MF whenever it drops except in the final Y seconds of Eclipse," then some refreshes cost an Eclipsed GCD and some cost an UnEclipsed GCD. And the hidden time loss that this post describes will be some combination of the two--but I have no idea how to estimate the ratio yet.

It might be interresting to look at it from a whole fight perspective, because at that level you can make the assumption that the number of GCD's used is equal whatever cycle you are implementing (whereas looking at a cycle only has problems with all the dots affecting the next cycle).

From a full fight perspective, you will indeed spend more GCDs keeping MF up (which results in a higher uptime) which will in turn diminish the number's of SF you cast during the fight, ie there is an impact further down the road than just getting one SF out of the Eclipse. While i initially thought that that cost might be negligible, 1 GCD worth of SF represents a non-negligible amount of damage (considering that we aren't going to gain many MF's either).
One MF per cycle at fixed intervals should amount to 60-70% uptime. Refreshing MF ASAP except at the middle/end of the Eclipse proc won't result in 100% uptime either, and i don't quite know how to guestimate the resulting uptime. I'm going to use 90% for the time being.

In a 5 minute fight, supposing 35s Eclipse cycles (lucky situation but whatever) 90% MF uptime means we have spent 3 more GCD's on MF. We get 20% more MF uptime (+3 full MF damage) at the cost of 3 GCD's worth of SF and at the potential cost of 9 Eclipse'd SF's (for those 9 eclipses, you won't always have to refresh MF during the proc, and even while refreshing you won't necessarily push an SF out).
If you are less lucky on average cycle time, the 1 MF case loses even more MF uptime (bigger damage increase from higher uptime) and there are less SF's to lose at the end of the proc, but you lose more SF GCD's too.

Using the numbers i get from Wrathcalc, this would put the "refresh MF" scenario at an advantage (yay, 1300 more damage over 5 minutes) even in the worst case of 9 lost Eclipse'd SF's. The thing i'm somewhat iffy about is the MF uptime guestimation because the net gain isn't great (even if this is a worst case scenario regarding lost Eclipse'd SF's).

Edit : didn't include iIS effect which should be a net gain for the refresh MF scenario

I saw you changed moonfire options but i wasn't sure how it was modelled and not motivated to dig through the formulas.
Edit : i get a dps decrease by switching to 100% MF, and i don't think that reflects what happens in a "refresh as often as possible but not at the end of an Eclipse" case.

The whole thing results in a ~1k-20k damage increase (+ iIS which shouldn't add more than 10k) depending on the number of SF's actually pushed out. Let's take a middle ground of a total 15k damage increase (10k on base and 5k on iIS). Supposing we do ~1M5 damage on a 5 minute fight (5k dps), we are looking at an optimistic ~1% dps increase. Probably not worth the additional attention to MF vs Eclipse timing, but still an improvement.

(i didn't take into account NG which probably favors 1 MF/cycle more due to more SF's during Eclipse, so the difference will be reduced some more)

My lectures this morning was cancelled so figured i'd spend my time on something useful, & decided to play abit with how potions affect owls.

The too-long-didn't read version:
Gain from [Potion of Speed]
With MF up: 10723.8 damage
Without MF: 10575.33

Gain from [Potion of Wild Magic]
With MF up: 9006 damage
Without MF up: 8825 damage



Current gear used for numbers.
Haste 611, specced for both imp. MkAura & Celestrial Focus.
All numbers assume Wrath of Air totem.


[Potion of Wild Magic]
Used during heroism, at Eclipse proc.
Casttime of SF is 1.75 at that point. Crit chance of roughly 80% during Eclipse, resulting in 4/5 SF's being boosted by NG.
NG boosted SF during heroism is 1.46seconds if I understand how the various haste effects stack.

This gives an average SF casttime of 1.46*0.8+1.75*0.2 = 1.52 seconds. Adding a latency of 0.1 to all numbers results in an expected casttime of 1.62seconds.

A realistic timeline will be something in the lines of this (T=time in seconds).
T=0 - The W proccing eclipse leaves your hands.
T=1.1 - Next W is cast.
T=2.2 - You notice Eclipse proc, but next W will be in queue / allready casting.
T=3.3 - Potion activated, first SF firing. Eclipse will have 13-14seconds left by the time you start your first SF.
T=3.3->16.26 - 8x SF during Eclipse+Pot.
T=16.26->18.1 - 1 SF during Pot.

Eclipse roughly lasts from T=2 -> T=17.
Potion lasts from T=3.3 -> T=18.3 - Benefitting only casts finished before 18.3.

In my example you are going to land 8 SF's during Eclipse + Pot & one SF outside of Eclipse but still during pot.

Using Wrathcalc my DpET is the following - 1st row is my stats, 2nd is my stats with the potion active.
SF = 4615.28 -> 5030.31
SF (MF) = 4697.62 -> 5118
SF (E) = 5912.88 -> 6412.58
SF (E+MF) = 6004.40 -> 6510.11

Gain from potion if MF is up for full duration:
2*(8*(6510.11-6004.40)+1*(5118-4697.62)) = 9006 damage.

Gain if MF is not up:
2*(8*(6412.58-5912.88)+1*(5030.31-4615.28)) = 8825 damage.

-------
[Potion of Speed]
Used outside of heroism, at Eclipse proc.
Casttime of SF is 2.26, 1.89 with NG. Assuming a crit chance of 75% gives an averaged SF casttime of = 1.983 seconds. 2.083 when adjusting for 0.1sec latency.

During Potion effect this translates into, 2.01 & 1.68 during NG. Average of 1.763, 1.863 with latency.

T=0 - The W proccing eclipse leaves your hands.
T=1.2 - Next W is cast.
T=2.3 - You notice Eclipse proc, but next W will be in queue / allready casting.
T=3.5 - Potion activated, first SF firing. Eclipse will have 13-14seconds left by the time you start your first SF.
T=3.5->16.5 - 7x SF landed during Eclipse.
T=16.5->18.4 - 1 SF fired still benefitting from higher NG chance.
T=18.4 - Potential chance of 1 more SF benefitting from potion, ignored in examples below.

Eclipse roughly lasts from T=2 -> T=17.
Potion lasts from T=3.5 -> T=18.5 - Benefitting any cast started before 18.5.


Without potion:
T=0 - The W proccing eclipse leaves your hands.
T=1.2 - Next W is cast.
T=2.3 - You notice Eclipse proc, but next W will be in queue / allready casting.
T=3.5 - First SF firing. Eclipse will have 13-14seconds left by the time you start your first SF.
T=3.5->16 - 6x SF landed during Eclipse.
T=16->18.1 - 1 SF fired still benefitting from higher NG chance.


The gain of using the Haste potion was:
1 additional SF landed during Eclipse. Loosing 0.3seconds of time for next part of your cycle.

SF (E) = 5912.88 --- 1 SF being worth 11837.76 damage.
SF (E+MF) = 6004.40 --- 1 SF being worth 12008.8 damage.

Loss of 0.3seconds costs:
SF at 45% crit = 2.19 casttime averaged incl. latency. Meaning you loose 13.7% of a starfire.
SF = 4615.28. --- 13.7% of a SF being 1262.43 damage.
SF (MF) = 4697.62. --- 1285.96 damage.

Gain from potion:
With MF up: 10723.8 damage.
Without MF: 10575.33



--------
Comments?
Difference seems too large to realistically be overcome with gear changes during Wotlk, allthough spellpower (& to some extend crit) will probably scale abit better than the way Potion of Speed adds its haste.

In new content I see myself saving potion CD for [Runic Mana Potion] for the first couple of unstable kills as 10k damage is practically nothing for a class with 5k+ dps. Fights could force you to use them for focused dps at specific points though.

I still have a doubt, about rotation.

What worths more?
Wrath to proc -> Starfire
Starfire to proc -> Wrath

My rotation is like:
IS - MF - (If in a "not-AoE" boss) Treants - Wrath (until proc) - {proc} - Starfire - {Eclipse ends} (back to beginning)

Read this thread: Moonkin Guide for Dummies!

If you're still in doubt, feel free to ask a question that isn't already answered a dozen times.