[Moonkin] Guide :: 5.4 Edition

[Deleted]

No, opinion plays no role in it. Only math. And math says "no". The timeframe is so short that you cannot interrupt your spells in time for it to be a DPS gain.

There are so many different ways that your wrong to just say no its impossible to state them.

[Enjeh]

Just a quick question, what dps did balance do around iLvl 497?
I took 90 two weeks ago so I have no idea what dps I should be doing right now.

I don't got any logs so as for reference, I did about 105k-ish dps on Gara-jal in LFR, which died before I could get in a second round of Incarnation/Celestial Alignment. also just one extremely bad UVLS proc that gave me only two starsurge procs from moonfire/starfire.

Just want to know If I'm around the dps I should be doing, or that I'm just under preforming atm. : D

http://eu.battle.net/wow/en/characte...iyako/advanced

[Othinsson]

Hey guys,

New boomkin here. Can someone enlighten me on what the Solar opener is? I'd like to be using it if it's a dps increase on single target fights. It's not on the front page, though.

[huth]

It doesn't affect whether interrupting is worth it. Getting more procs doesn't make you react faster.

[Zareth]

Just a quick question, what dps did balance do around iLvl 497?

If you're curious to see your potential DPS, put your character into WrathCalcs.

Old T14 BiS (~509 ilvl) showed a DPS of about 110k I think.

Edit: Your character showed about 95k simmed DPS. Keep in mind that with a very short fight, where UVLS + other stuff will proc more than usual (RPPM on pull) and Heroism having a really high uptime will show your DPS higher than usual.

[huth]

Seeing as how neither of specified a number of targets or haste scenarios, or even UVLS like Slippy mentioned, I don't see how you get off just saying "no", as if that makes you right.

If you want to do a fight like heroic Council without cancelling be my guest, though. You're going to lose a ton of procs and have a bad opener.

Having more targets doesn't make you react faster. Which is what you would have to change to make it worthwile. Neither does having UVLS.

And Juvencus isn't slippy.

[Slippykins]

Having more targets doesn't make you react faster. Which is what you would have to change to make it worthwile. Neither does having UVLS.

And Juvencus isn't slippy.

He could be... ;)

Just kidding. Juvencus is much cooler than me.

[Deleted]

No, opinion plays no role in it. Only math. And math says "no". The timeframe is so short that you cannot interrupt your spells in time for it to be a DPS gain.

Please enlighten me how math is the only answer?
Like Fountaiin said try do Council of Elders without ever cancelling a cast and see how it plays out.

Scenarios do matter whether you like it or not. Some of us have experimented with this alot and found out that it is a DPS gain in certain situations if you do it properly.

[Aboubacar]

Please enlighten me how math is the only answer?

It's quite often incorrect(by a lot) as well. I'm not familar with any simcraft or wrathcalcs data on cancel-casting, but there's a decent margin of error for what these tools actually model.

Tbh should primarily be using "math" (guessing you're referring to wrathcalcs) for gearing decisions. It's very reliable for that. It has some very good applications and has found some very useful data in the past, but the problem is you think that you can find general rules of play by modelling a handful of scenarios in excel, which is a ridiculous idea.

[Lilija]

Tbh this discussion about cancelling seems rather pointless the way you are pressenting it. Noone so far showed any real arguments on which solution is better. Huth is right however, that latency and human reaction plays a role here. Untill I see some math includin latency and human reaction or reliable test results I'll stay with not cancelling because my experience with not so stable latency shows I can hardly gain anything by interrupting. Not to metion that in certain scenarios (like Council opener) I just have a proc after proc so I can spam SS by default :P Plus openers more often than not are with Hero/BL that means your SF casts are pretty damn fast anyway.

I do believe that unless you have stable latency at some low level you shouldn't even start thinking about cancelling any casts.

The basic model would go around: latency+reaction time to cancel + reaction time between cancelling and pressing SS. If this is close to the size of window that makes cancelling worth it, don't even bother. I don't yet how to model the window tho.

You could minimse the last by using stopcasting macro within your SS, however, with the queuing mechanic that is rather dangerous solution. And if you cancel by jumping there will always be some time between you cancel and press SS.

[asharia]

Hey guys,

New boomkin here. Can someone enlighten me on what the Solar opener is? I'd like to be using it if it's a dps increase on single target fights. It's not on the front page, though.

start anywhere in solar, with arrow pointing left
starfall
inc + CA
dots (one MF or SuF)
SS
SF spam
refresh starfall
refresh dots before CA falls
try and hold one SS proc (or pray for one) so you only use 3 spells to get to lunar
starfall
SF/SS spam
refresh dots before Inc falls off <-- actually not 100% sure if that's correct

i think the reasoning behind opening this way, is if you have LMG and RPPM trinkets, they will almost always proc straight away. so with the lunar opening, you're actually getting to the middle (spending all lunar energy) way too quickly, and end up with several seconds of Inc left over while you're inbetween eclipses (heading towards solar).

i might be off on a couple of things, someone should be able to correct me

[Fountaiin]

Well, as far as I'm concerned "Just spamming starsurge hoping for a proc" on council is the same as cancelling; you're disregarding starfire when you have meta gem lust berserking ca dots on 4 targets with 65% crit and you expect incoming procs.

I'm not great with math, but I strongly feel like this one one of those situations where the math is wrong. It's the same thing with UVLS. After the first week, even though it was "bis" and simming highest, most of the top raiding moonkins came to the conclusion that it was an awful trinket right away, and most of us stuck with Light of Cosmos until we got Cha-ye/Breath. I didn't have math to back it up and I had people disagreeing with me on the forums in the BIS list, but soon people realized the math was wrong because it was just common sense.

Don't get me wrong, I love that we have models to math out everything, and I understand that when you have math vs no math you can only go so far with assumptions, but this is one of those scenarios where I think comments like "It is never good to cancel casts" is just going to hurt someone's play. I know arguments from authority are lame and all, but there's a reason every top multidotting parse has people cancelling (Super, Lorgok, Pacteh, Zoomkin, Celebrity, etc.). I couldn't imagine not cancelling on a fight like tortos opener or council or animus or twins even.

That's all I've really got to say on the matter, though. More dps for me .

[Aboubacar]

It's the same thing with UVLS. After the first week, even though it was "bis" and simming highest, most of the top raiding moonkins came to the conclusion that it was an awful trinket right away, and most of us stuck with Light of Cosmos until we got Cha-ye/Breath.

This is the point where I started using simcraft+Wrathcalcs more sparingly. Heroic LoC is far above UVLS in every way, especially for progress. The program needs a rewrite. Every day far less reliable for information than the last.

[Lilija]

Well, as far as I'm concerned "Just spamming starsurge hoping for a proc" on council is the same as cancelling; you're disregarding starfire when you have meta gem lust berserking ca dots on 4 targets with 65% crit and you expect incoming procs.

I'm not expecting procs - I get them before I even think about switching to SF ^^

I'm not great with math, but I strongly feel like this one one of those situations where the math is wrong. It's the same thing with UVLS. After the first week, even though it was "bis" and simming highest, most of the top raiding moonkins came to the conclusion that it was an awful trinket right away, and most of us stuck with Light of Cosmos until we got Cha-ye/Breath. I didn't have math to back it up and I had people disagreeing with me on the forums in the BIS list, but soon people realized the math was wrong because it was just common sense.

Don't get me wrong, I love that we have models to math out everything, and I understand that when you have math vs no math you can only go so far with assumptions, but this is one of those scenarios where I think comments like "It is never good to cancel casts" is just going to hurt someone's play. I know arguments from authority are lame and all, but there's a reason every top multidotting parse has people cancelling (Super, Lorgok, Pacteh, Zoomkin, Celebrity, etc.). I couldn't imagine not cancelling on a fight like tortos opener or council or animus or twins even.

That's all I've really got to say on the matter, though. More dps for me .

Telling people that cancelling is always a dps gain is also wrong. And well, yeah, when you have both dots applied by CA multiple targets during the opener you can easily assume you will spam SS. But I don't see much justification for it in a normal rotation. Also, I personally find jump to interrupt+press SS incredibly slow so unless someone shows me some more reliable way to do that I'll hesitate it's a good solution.

[Fountaiin]

When did I say cancelling was always a gain? I even specified multi-dotting fights. Please don't straw-man my argument. Huth said "Pretty much never", and I said I strongly disagree and that it depends on the scenario.

Also you don't jump, that would be silly. You just strafe or double-click mouse to move forward an inch.

[Slippykins]

Well, as far as I'm concerned "Just spamming starsurge hoping for a proc" on council is the same as cancelling; you're disregarding starfire when you have meta gem lust berserking ca dots on 4 targets with 65% crit and you expect incoming procs.

I'm not great with math, but I strongly feel like this one one of those situations where the math is wrong. It's the same thing with UVLS. After the first week, even though it was "bis" and simming highest, most of the top raiding moonkins came to the conclusion that it was an awful trinket right away, and most of us stuck with Light of Cosmos until we got Cha-ye/Breath. I didn't have math to back it up and I had people disagreeing with me on the forums in the BIS list, but soon people realized the math was wrong because it was just common sense.

Don't get me wrong, I love that we have models to math out everything, and I understand that when you have math vs no math you can only go so far with assumptions, but this is one of those scenarios where I think comments like "It is never good to cancel casts" is just going to hurt someone's play. I know arguments from authority are lame and all, but there's a reason every top multidotting parse has people cancelling (Super, Lorgok, Pacteh, Zoomkin, Celebrity, etc.). I couldn't imagine not cancelling on a fight like tortos opener or council or animus or twins even.

That's all I've really got to say on the matter, though. More dps for me .

While I've quoted you Fountaiin, the below is kinda general but pertains to the conversation.

I think the main reason people are saying the maths are wrong is mostly because the derivation itself is flawed. Say you try to model a very basic single-target, no movement fight in order to deduce whether or not cancelling a Starfire or Wrath cast is advantageous to stop a Shooting Stars proc from overwriting itself. Your extremely basic scenario would include Starfire cast time, lost damage from the Starfire being cancelled, lost cycle time from the cancellation, gained damage from the Starsurge, gained energy from the proc use to convert into cycle time. You would arrive at an answer for this scenario quite easily, but it is only a basic model and has a wide margin of error. Then you decide to add in additional factors, like latency and reaction time. This'll make your model more accurate, but again it's not perfect. You'd then add in the separation of DoT ticks to determine the chance of a proc occurring with the current one up, what chance each tick has to proc a Starsurge, and even then since it's random you'd need a proc distribution model to accurately describe the frequency of double procs. And remember, this is just for a stand-still, single-target fight. Adding targets, movement and other mechanics will further change how you're modelling something.

What I'm trying to get at is that modelling is impossible to get perfect. You always start with a basic model and build upon it until you get things like SimCraft and WrathCalcs. Unfortunately, because WoW is so complex and is constantly changing, it's excessively difficult to keep up and perfect these out-of-game models so that they stay relevant and correct. Even if WoW didn't change, it'd be nigh impossible to be 100% certain of anything with them.

However, that doesn't dismiss the information you receive from them. Just because it's not entirely accurate doesn't mean all the data you obtain is useless. As long as the margin of error stays relatively low, you can be relatively certain of some assumptions, or completely certain of others. If there's not much of a difference between alternatives, it's best to say you can't ascertain what is better from the current assessment. Models should always be used with some common sense, and you should approach a model with a hypothesis or contention in mind already, with the goal of proving that idea.

I've come across a few people that just abhor using models simply because they believe in-game data is much more valuable. I've stated this before, but there is a marked difference between testing something in-game and testing something using a model. Singular tests have humongous margins of error because of random chance events, like crits and procs. As everyone has noticed, sometimes you'll get 15 procs of UVLS in a fight and sometimes you wont get a proc at all. Or you'll get 60% of your Starfires being crits even if your crit chance says you only have 30%. It's these random events that cause singular tests to be completely unreliable. You can't make any conclusions from small data sets. And not many humans can sit at a target dummy for 10,000 separate 3-minute trials just to figure out if one piece of gear is better than another, or if having 1,000 additional mastery is better than 1,000 additional crit. This is clearly where models should be used, to figure out things that we just can't do in-game.

So on the topic of cancelling casts, it's difficult to test in-game without referring to very small sets of data, and currently no model is accurate enough to tell you what is the correct choice; hence the current discussion. And as you kind of said Fountaiin, referring to the top parses is often inadequate as well. The reason the top parses are top parses has a lot to do with RNG. Player skill absolutely has a big role in their parse, but instead of something like a 80/20 skill/RNG split, it's more like 60/40. You'll notice higher than average crit chances for most spells, higher than average procs, or even that procs overlapped each other more frequently than you'd expect. It's great and all to take what the top boomkins do, but all logs should be analysed with a grain of salt. Models aren't exempt from salt grains either.

So overall, I'm just trying to say a maths vs. no maths argument is pointless. Models are only as accurate as we make them, and only serve to improve your play. If there's some hideous inaccuracy you should report it, or try to find out why. Either way, it's important to be able to use models and extrapolate the data they provide so that you can make a meaningful in-game assumption. You can't analyse everything in-game, and nor can you analyse everything with a model.

[Juvencus]

He could be...

Just kidding. Juvencus is much cooler than me.

/gasp
/blush

Like mentioned above it's not a cancel vs not cancel or math vs experience situation. It's about the situation and the ever changing fight variables. Putting it simple, no I would not cancel any spell to fire an SS proc during normal rotation. If now I have UVLS dots on 2 targets (Megaera) and SS procs come later than expected I will probably start hard casting but if I see a proc and 1-2 refreshes I may or may not cancel or stop hard casting and sit there and wait/fire procs.

In the end it's a matter of judgement and as I'm obviously not the best boomkin out there, I would expect me to make false decisions, but I would expect from a top boomkin to almost always make the right call on the fly without him or me being constrained by stone-set rules.

[Aboubacar]

Well that's the issue. Wrathcalcs and Simcraft are not transparent in their "rotation" or AI that goes into the models. I've read bits and pieces regarding SS proc wastage being added in, but for all I know it's just mindlessly popping Incarnation every 3minutes in lunar, using all the lunar energy and popping CA. It models a "proper" rotation, but I'm guessing, correct me if I'm wrong, it doesn't change behavior based on trinket procs for example and will use items like synapse springs on cooldown. If that's the case, the small data sets from players' WoL is far more valuable than simcraft/wrathcalcs model. Again, I could be wrong, but I cannot locate actual information about how any well-know simulator plays my character.

[Slippykins]

Well that's the issue. Wrathcalcs and Simcraft are not transparent in their "rotation" or AI that goes into the models. I've read bits and pieces regarding SS proc wastage being added in, but for all I know it's just mindlessly popping Incarnation every 3minutes in lunar, using all the lunar energy and popping CA. It models a "proper" rotation, but I'm guessing, correct me if I'm wrong, it doesn't change behavior based on trinket procs for example and will use items like synapse springs on cooldown. If that's the case, the small data sets from players' WoL is far more valuable than simcraft/wrathcalcs model. Again, I could be wrong, but I cannot locate actual information about how any well-know simulator plays my character.

I agree with you on SimCraft's lack of transparency, or that might just be my inability to find all the equations behind-the-scenes, but WrathCalcs is all there for you to inspect. Everything used for its calculations is in the spreadsheet (mostly in the Rotation tab), so you just need to learn how it works to understand how each cell influences the others.

SS proc waste has migrated from a flat value the user can set, to being integrated within the model itself. Hamlet explains it better here, but basically it calculates the chance for two consecutive ticks to proc an SS and then says you'll lose the next two ticks if they're also procs. It's still in the works and might be changed to just the next tick, but we'll see. T16 gear needs to be implemented before it can be used comparatively.

But I do agree with you in that both models don't react to some of the things we do as boomkins. Like you said, reapplying DoTs or casting FoN when trinkets/procs are up for maximum effect I don't believe is in either model. It's very difficult to ascertain what contributes to what using logs though, because the spell data itself is just an average value (unless you talk about DoT ticks), so you have compounding RNG when combined with crits and changes in stat values.

I just think it'd be nice if, instead of people complaining models and maths are useless, they would help to figure out what exactly is wrong and propose changes to the model authors. That way, everyone can benefit from more accurate and reliable tools for our class.

[huth]

But I do agree with you in that both models don't react to some of the things we do as boomkins. Like you said, reapplying DoTs or casting FoN when trinkets/procs are up for maximum effect I don't believe is in either model. It's very difficult to ascertain what contributes to what using logs though, because the spell data itself is just an average value (unless you talk about DoT ticks), so you have compounding RNG when combined with crits and changes in stat values.

You can set your entire rotation yourself in SimC. It's just a bunch of work. That includes things like casting treants when procs are up or when the third charge is about to repop.

For the record, math is exact, and about the only thing in the world that is. Which leads to the problem: The tools to implement it are not, like computers. Also, math is the only thing a computer can do. Everything else you see is a result of said math.
You can math out exactly what option is best for any given situation. You may not be able to do it in any reasonable timeframe, though.