1) Not sure what you're referring to in the first paragraph. I picked out the denominators for Eclipse uptime to only include Nuke casting time. And your description is odd; DoT's do benefit from Eclipse.

2) This was an issue going all the way back to the WLK sheet. There's no great way of handling time-dependent issues like Moonfire extension in the sheet. Ultimately I could find some clever way of averaging around it like I do everything else, but it would be another little project. For the moment, since it assumes all Moonfires are extended and Glyphs of MF and SF come out so close, it's a good enough sign to use Glyph of MF (esp. when Lunar Shower comes into play). As to eking out more damage, these are the sorts of reasons I think the rotation is already a lot more intricate than everyone is giving it credit for. As I keep saying, I'm not planning on plotting some detailed spell sequence because it would have no practical application anyway. The goal is to answer questions like when it's good to extend an Eclipse or recast various spells. I think good players will wind up juggling factors like that on the fly when figuring out what to cast.

3) I've always been aware of that, but the goal is to analyze the current system. Especially in the beta, when we first have to figure out whether there's even a problem to begin with. Only after all of this work over the past few days do I really understand the potential rotation issues enough that I could even write a comment to GC on it.

4) Mastery is very weak with the current beta numbers; there isn't much else to it.

Thorns Mana

Thorns mana is not being included in the "Rotation" mana. That is an expensive spell, and using it on cooldown (glyphed) may make us thirsty.

Cooldown Uptime

I'm don't know if there is a real good closed-form solution for figuring out how often you will manage to cast a low-priority cooldown, but I'm pretty sure the current spreadsheet is a bit too pessimistic.

Thorns and MF "want" to use up 5% and 7% of your casting time (cast time ratio, on the Rotation page). They are showing up in the rotation only 4% and 5% of the time (spell breakdown on the Main page).

For single target rotations (where DoTs and cooldowns usually won't all be available at once) it might be better to assume that cooldowns are typically cast about 1 GCD after becoming available and DoTs (which can be refreshed slightly early with no penalty) are typically cast about 1/2 GCD late. Both those number could be on the main page (up near the latency numbers), so people could play with them.

Always Be Casting

It won't be as much of an issue in cata, since we won't be haste capped, but we have enough instant-casts that I think using zero latency is too optimistic and should not be the default. I'd be happier with something like .03 for cast-time spells, and .1 or so for instants.

With Nature's Grace out of the picture, we'll have more consistent cast times, but I suspect we still won't see many logs with time-spent-casting over 97% or so, even on Patchwerk (or target-dummy) fights.


I'd welcome more discussion (or better yet testing) on these issues.

With the new instant cast spell queue, I would think the only reason you shouldn't be casting is if you make a mistake.

Perhaps Solar eclipse will be stronger but with the high cost of Thorns (and good damage) there will be incentives to keep Thorns up as much as possible but due to high mana cost we need to proc elipse more often (mana return from Euphoria) so prolonging Solar eclipse will make us go OOM on longer fights.

Another reason could be that we don´t have to Innervate ourselfs but are able to buff someone else with it (CR + innervate or OOM healer).

It's true that for the moment, even in the sheet, the 4T11 bonus causes Standard Eclipse cycles to come out on top again.

1) Will fix. Yeah, Thorns is quite a mana problem, maybe that's a balancing factor for the spell that I hadn't thought about much. Could make for interesting decisions on live.

2) On Cooldown spells don't always get all the cast time they want. Some have to take priority over others--physically, think of the fact that sometimes two spells are up at the same time, or mathematically, that the "remaining cast time" left after processing each On Cooldown spell has to combine multiplicatively, not additively. The current priority in the sheet is:
FoN > Starfall > Starsurge > IS > MF > Thorns.
I'd like to find a way to make it customizable, but it would take something pretty complex in the rotations page.

3) At least in the case of DoT's, there's almost no reason for any penalty. Since you can refresh up to a full tick early at no loss (I think right now, at least), you'll almost always be able to do so unless your next spell is SF and the DoT is IS. I'm not sure it's worth worrying about. I could add 1GCD to other spells, although the only one that's short enough to really matter is SS.

In the sheet I'm looking at (should be very close to the 913 download):

I divide "Time Spent On Thorns" by InstantCast and get 2.06, so there are 2.06 casts per rotation.
The rotation length is 72 seconds. 72/2.06 = 35. Thorns are being cast every 35s. The cooldown is 25s. Thorns are being delayed, on average, 10s.

The cast could be delayed by FoN, Starfall, IS, SS, MF, and there might have been an SF already in progress at the time Thorns came off cooldown.

That is 7.33 GCD's. My GCD is 1.3s, so the largest possible delay should be under 10s, and the average should be considerably less.

Hmm, that does sound low. What it does right now is essentially
1) FoN claims its desired cast time, as a percentage of total cast time.
2) Starfall claims its desired cast time, as a percentage of everything that's left.
3) Starsurge claims its cast time from everything that's left.
etc.

The sheet doesn't actually compute in that order, but it should be mathematically equivalent. I'll have to see there's an error somewhere or if that algorithm is bad for some reason.

Yeah, realized around when I was typing it--that does take way too much cast time from lower-priority spells. I think it just never mattered in the past because the only two spells were MF and IS and their cast time ratios were so low. It might take a significant change to fix, although hopefully the new setup where I broke out the all the time subtotals onto separate lines means that changes can be made easily.

Don't yet have a formula for the correct cast time of the lower-ranked spell though.

Ran a simulation (cast times only). 17% haste, zero latency, 24 hour run. I think I followed your priority list. I'm not sure why starfall ended up doing better than FoN.

CD's were cast .8 to 1.8s late. DoTs averaged .1 to .2s late (DoTs could be refreshed any time after second-to-last tick).

First number shows average time between casts. Second column shows hasted DoT duration.

Nuke pattern was fixed at SF*2, Wr*3

Spell	Time between casts	DoT duration
nukes	3.55779
thorns	26.8827
mf	12.9498	12.8205
is	12.1401	11.9658
ss	18.0039
sfl	60.8457
fon	181.134

Second table had MF and IS thrown at two additional mobs, just to lower the nuke percentage, and see how long thorns would have to wait.

Spell	Time between casts	DoT duration
nukes	9.62031
thorns	30.0313
mf	13.2821	12.8205
mf2	13.4454	12.8205
mf3	13.9063	12.8205
is	12.128	11.9658
is2	12.1485	11.9658
is3	12.1742	11.9658
ss	17.5325
sfl	60.7595
fon	180.753

Edit: The time between casts computation was actually (casts/fight length). For a one day (1440 minute) fight, if FoN was just about to come off cooldown at the end of the fight, that would be enough to inflate its number by 3/1440*180 = 0.375s

Edit2: It may be better to just ignore this issue. Consider that on a 500s fight you probably get three full FoN, even though the spreadsheet says 2.8. The difference between 180s and 181s is a much smaller error than ignoring the fight length entirely.

--Some minor bug fixes since last upload.
--The cast priority issue. You're right, assuming they don't clash at all is actually a lot closer to reality than what I was doing. I'm going to do that for now (i.e. assume all cast time ratios combine additively) until I come up with something better.
--As a result, there is no spell priority and nukes have less cast time than before.


Okay, the new Nature's Torment and Shooting Stars were an enormous pain to model. I apologize in advance to the people who like to look into the sheet.

Basically, there was no good way to insert these effects into the existing model without horrible circularities. So I had to do something I'd always tried to avoid in the past with clever computation--the sheet iterates through the entire cast-time computation twice. Basic description of algorithm:
1) Guess zeroth-order values of Nature's Torrent uptime and Shooting Stars proc frequency (I tried to get good starting values in here).
2) Use (1) to compute mean cast times of spells. On the spell sheets, these preliminary cast times are marked with a **.
3) Use (2) to compute total cast time breakdown and rotation length (the new second column in the rotations page).
4) Use that rotation (called the "first iteration" or "guess" rotation at various points in the sheet) to recompute better values of NT uptime and Shooting Stars frequency.
5) Use (4) to recompute cast times and rotation length.
6) Use (5) to compute final damage.

Hopefully that guide will be enough to help the people who like to look into the sheet figure out what's going on. I'll try to explain better, but this all took forever today and I have to go out soon.

-----

Other stuff from this patch:
--Assumed that Sunfire talent is equivalent to "Moonfire is now affected by Solar Eclipse." i.g. affected by all the same buffs as MF and only one can be up at a time.
--Nature's Torrent assumed to up for 15s out of each Eclipse. This seems like a reasonable rule for all cycles I could think of.
--The various patchwork of buffs and procs in the cycle now make it hard to decide what individual spell data to put on the front page. For now there's a DPET column without Eclipse or NT, and one with both. This seemed reasonable since Eclipse and NT will go together a lot.
--I'm thinking of nixing some of the custom DoT/SS timing options. They're becoming incredibly unwieldy to support as the sheet gets so complex, and they're seeming increasingly less likely to be useful. Perhaps after live we'll be in a better place to see if any specific spell timings are really worth looking into. For now I'm most concerned with making sure that the model looks accurate for Standard Eclipses and on-cooldown spell usages.
--For now DoT's just use mean NT uptime. There's room to change that to account for the fact that the first DoT after Eclipse always [will/won't] be affected by NT.

There were significant changes to sheet again and I don't doubt some things need to be fixed. Hopefully people can poke and not be too confused by what's going on now.

-----

Only major comment is that, counterintuitively, the sheet still shows Solar-only to be good. This could still be due to a modeling error, but it perhaps makes sense:
--The new Nature's Torrent doesn't actually punish you that hard for doing it. Ordinarily, NT uptime is around 30/75. If you go Solar only, NT uptime is still 15/45. Not a huge difference.
--Moonfire is affected by Solar now, so you get quite a bonus for full Solar uptime: 100% Eclipse on both DoT's.

Wrath with a 13.33 generation gives us equal energy/second for Wrath and Starfire, eliminating the differences between the two.

With regard to AoE... they could tie Hurricane into Lunar Eclipse through the Sunfire talent as well, but aside from that the only real answer is to give us an Arcane AoE spell, which I find unlikely. I'm not really upset over the lack of symmetry from AoE though. Given that Eclipse no longer has a duration, we simply have to proc Solar Eclipse and go to town.

Edit to add:
I'm having a hard time believing that some of these recent developments will stick. It's almost as if they've reverted back to an earlier Eclipse model, particularly with regard to Starsurge. Given that GC has explicitly stated that they do not want us to be able to build up Eclipse Energy before a fight, I don't see how it's going to work out with Eclipse not decaying to 0 out of combat. On the other hand, I have always suspected that we would be either forced or heavily encouraged to oscillate quickly between Lunar and Solar Eclipses and this build certainly reflects my thoughts. At the same time though, I find this to be almost like taking 1 step back and then 2 steps forward and I think some of these new changes with regard to Eclipse Energy will end up being fixed.

It's a little odd though that Starsurge is still effected by Euphoria. That if nothing else should imply that there will be a change to come.