Ezareth

On the subject of Chain Lightning versus Lightning Bolt:

I can see instances of extreme latency making Lightning Bolt Spam preferable to rotating in Chain lightning.

Remember the whole advantage of chain lightning is the fact is casts 33% faster.

Lets exaggerate latency to an extreme level to clarify my point. Lets give each spell a 1 second latency between casts.

Chain Lightning now has a 2.5 second cast time, Lightning bolt has a 3 second cast time. Due to this latency the bonus cast time chain lightning has over lightning bolt was reduced from 33% to 17% effectively making it poorer DPS.

After patch I won't be casting chain lightning except for instances where I need a quick spell to get off my lava burst in time.

The buffs added to lightning bolt and increasing lag on the instance servers necessitates this.

 

tarrick

Has there been any definitive testing on the proc rate of LO on single-target CL? I have been under the assumption that it was 20% and applying a 10% DPS modifier to CL in my calculations. However, I may possibly need to reconsider as it seems that CL divides the proc chance over the number of maximum targets regardless of how many are hit. This greatly lowers the overall DPS of CL in my calculations and makes LB stand out even pre-3.0.8.

I was curious if anyone else has data to support this, though I couldn't find any through some forum searches. My current working sample is rather small: 173 CL casts with only 14 LO procs (single target CL casts only). I'll try to put together a larger sample.

 

Kishkegelt

So further to Ezareth's point about using CL if LB would delay LvB too much - my UI doesn't provide sufficient information for me to make this choice. That is, OmniCC counts down in full second increments, and since the difference in time between LB and CL is less than half a second, I really have to guess which one to cast (unless I've been able to simply stand and deliver and know exactly where I am in my rotation).

(Also, frankly, I'm not 100% sure OmniCC's countdown is accurate, as sometimes I seem to be able to launch a LvB before the button clears, and sometimes not until after).

Am I wrong that this is a problem, or has anyone got a good solution?

 

Phatpharm

Agreed. It is about rotations to a degree, but certainly not about theory-crafting rotations.

My apologies.

If there is a relevant thread, or the concensus that a new thread be made, that would be great!

Honestly, the blizzard UI has a cooldown timer built-in to the spell icons. ( counter-clock wise opaqueness )

You know the CD of LvB is 8 seconds. Therefore each "quadrant" is 2s. If your last cast is complete and it is 3/4 cooled off then use LB, if less CL. Watching this can give you a feel of what time you have left to "fill"

Hope this helps.

 

Sjera

On the subject of rotations, I've been recently looking at rotations under Bloodlust and the effect that has on stat weights.
Note: These calculations are based on patch 3.0.8 mechanics

For anyone with a reasonable amount of haste, Bloodlust will cause most of our spells to hit the 1 second global cooldown limit, especially if it happens to coincide with a trinket proc. This means that very quickly the average dps of Lightning Bolt will exceed the average dps of Chain Lightning, and the effects of haste will become irrelevant to Flame Shock, Lava Burst and Chain Lightning.

Thus under Bloodlust we switch to Lightning Bolt spam (while still being careful to keep Lava Burst cast as often as possible). This causes a change in stat weighting

We would expect to spend at least 10% of any fight under bloodlust.
There is nothing really new or surprising here, I just wanted to get some relative figures - and figured I may as well post what I found.

For final weightings, I'm now using a weighted average of Normal and Bloodlust values.

 

• Binkenstein

Well, using the new r5 of SEIC, I get the following:

FS/LvB/LB with FS,Lava,LB glyphs: 4918
Same with CL after every LvB: 4971 (1.07% gain)
Same with CL as often as possible: 4991 (1.49% gain)

Glyphs (using the other two)
Lava: 86
LB: 124
FS: 212

Swapping Lava for Shocking we get:
No ES: 4874
ES: 4849

I assumed that shocking would recude the GCD to 1 + the casting "delay" (used 150ms)

So, what does this mean? The Earth Shock idea is pretty much dead in the water now, CL gives very little extra damage, and we're looking fairly decent in terms of DPS (I used my base stats for this, 2098 sp, 559 crit, 409 haste)

 

Souai

I was looking at the value of haste for Aggressive CL use and No CL use using r5 of Blinkenstein's spreadsheet and got some interesting results. I looked at a 3 minute fight with blood lust starting on the pull and 150ms latency and adjusted my haste rating from 500 to 700 in 10 haste increments.



I was kind of surprised that there would be such a difference. I'm not sure how much of this is casting CL while blood lusted, the effect of having BL take up such a large portion of the fight, or artifacts of the method that's used to generate the DPS estimations or if this is actually how haste is valued between the two approaches.

I looked at the long term into gear that will not likely be obtainable in the game:


The trend doesn't continue. Aggressive CL use EP moves up and No CL EP moves down until they are about equal.

Also, for me, the average difference between Aggressive CL use and No CL use over the interval of gear realistically obtainable was 20 spellpower or 26 dps. Over the long term the average difference was 25 spellpower or 32.7 dps.

 

MatsT

Would be interesting to see a diagram with "optimal CL" in comparison, meaning that it wouldn't use chain lightning if it didn't help you cast more spells before next LvB.

 

baldr

I am new to party/raid DPS altogether, having mostly tanked for the past few years.

I just got into heroics and I think I am following the proper priority system, but my DPS isn't anything like what MaxDPS says it should be. The best I did in my first few heroics was 1500 dps on a boss. My avg was about 1100. Unless I am misinterpreting the intent of the presentation, MaxDPS says I should be getting 2136 DPS against lvl 82 (MaxDPS.com - Elemental Shaman DPS Gear Rankings).

I am trying to cast in order of priority, hitting FS > LvB > CL > LB, whichever is available. Even though I'm hitting them right when they're available, it seems to often end up being a cycle anyway, going down the priority list, doing two LBs, one CL, and then back up to the top. Once in a while when I know a CL or LvB is almost available, I hesitate and cast it. I am using cast bars to let me know when the GCD is up so that I can jump to the next spell before the current one is done.

My question is: is this 2136 a reasonable target, given my current stats (available in the link above)?

The only thing not apparent are my glyphs - I have the two main ones, FS and LB, but haven't decided yet between Lava and FT for the third). Assuming that estimate is valid, what sort of changes could boost my DPS? The changes I can think of so far are: getting the third glyph, alternating between FS and Earth Shock, and getting better at timing my casts.

 

Armisael

Since the patch isn't out yet, I would go with these glyphs: FS, LvB, and FT. This is because the LB glyph hasn't been changed so it isn't as good over those other glyphs yet.

 

gothic

This may be a silly question, but are you reapplying Flameshock right as the FS cooldown expires? From your text, it sounded as if you were, but I wanted to be clear.

Also, I've found the numbers MaxDPS present to me when I use it to be a little on the high side. I use Bink's old spreadsheet (or did), MaxDPS, and a program of my own to get a general idea of DPS, gear upgrades, etc. It never hurts to use the information available to you, just don't take them as gospel, and try to get multiple "opinions".

 

Kishkegelt

Very nice graphs, would you consider also modeling a middle ground such as 1 CL following every LvB, as Bink does in his SEIC?

 

baldr

Thanks for this advice.

I am trying to re-apply it right as it expires. I am not getting that right on every fight.

Thus my request for your opinion. ;) Thank you for it.

I am still trying to learn how to use the spreadsheets that are posted here. So far I've used them to help me update my current gear and plan my gear targets. I'm also re-reading all of the posts here, and I'm understanding more as I get more hands-on experience.

 

Kumiko

Was bored in class, so I tried to think of rotation improvements to truly min/max dps. Then it came to me:
Why aren't we casting 2 lightning spells after FS and before LvB?

This only has merit for 3.0.8, but to maximize the chance that we get the 10% bonus damage from Oath/CC, we should cast 2 of our highest crit spells just before LvB. Now the chunk of the rotation looks like "...FS,LB,LB,LvB..." but the LB's could be CL if it's off cooldown.

I realize this is on the minutest scale for dps increase, but it IS an improvement. At say, 35% crit (40% lightning), the chance of LvB getting Oath/CC bonus is 3% better with 2LB than with LB+FS as the preceding spells. If LvB is roughly 25% of your damage, this turns into a 0.075% dps increase.

Thoughts?

 

Agash

Considering your LvB in isolation, it is a 0.075% (~4 DPS) increase. However, there are DPS losses that you haven't taken into account:

1. You haven't calculated the DPS loss from your first 2 LBs possibly not being effected by CC when in the current rotation they always will be. In other words, whilst your average LvB damage will be higher, your average LB damage will be lower.

2. You haven't calculated the DPS loss from the possibility that you will lose 1 LvB from the fight overall. This will occur approximately 35% of the time (depending on haste). If you cast 30 LvB in a boss fight now, you will on average cast 29.65. Against this needs to be weighed that you should squeeze in an extra LB on those fights.

It's difficult to put exact numbers to these losses without going into it in great detail but if we assume an average 3500 damage from LB and 9000 from LvB, these losses could be along the lines of:

1. 0.65 x 3500 x 0.1 + 0.14 x 3500 x 0.1 = 276.5 damage

2. 0.35 x (9000-3500) = 1925 damage

So, a total damage loss of around 2202 to balance against your 4 DPS gain. So your proposal would only increase overall damage in a fight lasting over 9 mins.

 

Kumiko

This makes perfect sense. It might be considered trying "FS,LvB........LB,FS,LB,LvB..........LB,FS,LB,LB,LvB...", shifting back the flame shock at the end of the rotation by 1 cast each go-through. This way, you don't cast those 2 Bolts without CC, don't lose anything in the first shift-back, and don't step on the toes of your FS DoT. Yes, it's a little silly considering how every movement would reset this, but dps is dps. ^^

This I'm not so sure about.

The shift-back method would also fix this, but I'm not certain it's even an issue:

All that's being added is the 4 unhasted seconds of casting at the beginning of the rotation. It seems to me that no fight time can be predicted within 4 seconds of accuracy. With an even distribution of fight finish time (relative to 4 seconds), combat could stop at any point during the rotation equally. This is the same as without the 2 Bolts.

 

Ezareth

That MaxDPS site has garbage calculations. It isn't even close. The site looks pretty however, and is almost exactly what I'm designing for my new site DPSSIM.com.

You want to get your Lava Burst rolling as soon as humanly possible. After the first Lava Burst the spells preceeding lava burst will always be 2 lightning spells if you can help it for just this reason.

Every time you reset your rotation(in a moving environment) the same rules apply, your DPS is always increased the sooner you get your lava burst off.

The only thing you are gaining by putting 2 lightning spells after the flame shock before your first lava burst is a slightly higher chance at getting a 10% bonus to that lava burst all while costing you at minimum 33% of your lava burst damage(2 lightning spells / cooldown of lava burst).

I was thinking down your lines earlier until I thought more on it.

 

LITTLER0TT

That kinda defeats the purpose of Lava burst.

Lava burst is to be casted as fast as possible. Those 4(ish) seconds of Lightning bolt could have been spent casting a Lava burst and being 2 seconds into its cooldown.
(ezareth beat me to it)

 

Agash

Whilst over the duration of the fight, 4 secs isn't much you need to bear in mind that you are casting a LvB every 10 secs or so through the fight. So the fight will always end within 10 secs of a LvB cast. Sometimes it will be <1 sec after the LvB sometimes it will be 10 secs. So, whilst we can't predict when the fight will end, it is a reasonable assumption that it will finish after somewhere between 0 and 10 secs of a LvB cast and there is no reason to assume that the distribution across that range should be anything other than linear.

By pushing back the cast of the first one by 4 secs, any fights which would have finished within 4 secs of a LvB cast will, under your new rotation have that final LvB clipped off the end. That will be about 40% of fights unhasted or probably about 35% including haste.

Looking at your 'shift back' method, it looks like you are delaying your FS cast by 2 spell cast times. This will effectively lose you 3-4 secs (depending on haste) of DoTs. If we assume 3 secs which equals 1 tick of the DoT, that will mean losing around 700 damage through the fight. So again, your system would only increase DPS when 4DPS > 700 damage which is after about 3 mins. But I haven't accounted yet for the increased damage on the flameshock DoT from both the EO change and the S,E & F change. These might take your average FS DoT to about 800 and mean that the break even is around 3.3 mins.

So your 'shift back' method does appear better than your original suggestion and might show a 1.2 DPS increase on fights of about 5 mins though would lose DPS on a fight under 3.5 mins.

 

Kumiko

Yes...that is what moving flame shock back two spells will accomplish: putting 2 lightning spells before Lava Burst. See my revised rotation in the post above yours.

Only on the first LvB, which is fixed with the FS shifting.

No, it improves it.
They're called rotations because they come full circle. Lava Burst can only be cast every 8 seconds, so why not make the 5.5 seconds prior to every other Lava Burst be FS,LB,LB, in that order, to maximize chances at a CC'd Lava Burst?

I can now see why you wouldn't START with FS, LB, LB, but there's no reason not to shift back the shock to get to this point.

Yes, it can be said that a fight will end withing about 10 seconds of a LvB cast, but it cannot be said that pushing back the rotation as a whole will increase the chance of dropping a LvB off the end.

Take a line of colored balls(red and blue): RBBBBRBBBBRBBBB, the reds representing LvB and blues representing Bolts. Call them all 2sec casts. This brings the range to 30 seconds, not a terrible range for estimating fight completion time. Choosing a ball at random, call it and all the balls before it "casted". Now shift it forward due to the starting bolts: BBRBBBBRBBBBRBB. Repeat picking the balls. You can see that it is equally likely in either case to cut off a Lava Burst. Yes, this is an excruciatingly simplified model, and it should in fact be a probability curve over about a minute period, include proper cast times, CL, FS, etc., but the basis is there.

You can say that because a fight took X seconds, a Burst might be cut off, but that requires knowledge of how long the fight lasts within a 4second accuracy, before the fight.

My bad, I didn't explain this very well. I didn't mean to suggest delaying Flame Shock, I meant switch its position in the ending part of the rotation with the LB after it. This way, its the same number of spells and time spent casting between Bursts, but you get those lightnings in the right spot. This won't cut your FS DoT short, and won't delay it out of its normal delay in the rotation.

 

Mmootimus

I'm all for maximising DPS, but isn't this talk of rotation order is just a big waste of time? Even under ideal circumstances the DPS gain seems incredibly small. Also, in any fight where you often have to stop casting for a time and reset your cycle, then not casting LvB ASAP is just foolish. I would imagine future content is likely to see more fights like this, where maximising your DPS in 30 second windows may be more the norm than Patchwerk style nuke-fests.

 

Agash

Your coloured balls model is a good way of explaining what I am talking about. The reason why your explanation seems to fit is because you have assumed 3 full rotations in your examples. This is not the case because as you say, you do not know where the sequence will end. Your sequence will end in one of 5 ways:

RBBBBRBBBBRBBBB
RBBBBRBBBBRBBB
RBBBBRBBBBRBB
RBBBBRBBBBRB
RBBBBRBBBBR

All examples have 3 Rs.

Now, if we add 2 Bs on the front and remove the last 2 spells to compensate:

BBRBBBBRBBBBRBB
BBRBBBBRBBBBRB
BBRBBBBRBBBBR
BBRBBBBRBBBB
BBRBBBBRBBB

Now 40% of the examples only have 2 Rs in them. It doesn't matter how long the fight goes on for - you just add in as many RBBBB sequences as you need in the middle but the end it always one of those 5 options and in 2 of them you lose a R. So you don't need to know the fight length in advance, you don't even need to know where in the sequence the fight will end. What you do know is that on 40% of permutations you will lose a red ball.

So, whereas you cannot say that a LvB WILL be cut off without knowing the fight duration you certainly can say that there is a CHANCE that one will. Because you don't know where the fight will end and because each finishing sequence has an equal chance of occuring your can definitively say that this chance is 40% in a non hasted rotation.


I'm not sure how you can swap a spell with the one after is without pushing it back. If your second FS is 18 secs after your first and you now swap it with the LB after it, it is now 20 secs after your first and you have lost 2 secs of DoTs.

 

Kumiko

I see what you were getting at now, but I find it hard to say anything definitive without a probabilistic curve and an accurate range for when fights will end.

 

Agash

To define a probability curve you need to just one assumption which is that the probability of each of the end sequences occuring is the same and are the only possible outcomes. I accept that this is not 100% accurate. As far as the the range of when a fight will end, all you need to know is that it will end with 10 secs of the final LvB.

Like Mmootimus says though, the DPS gain is minute even in a 100% nuke fight (Patchwerk). In any other fight it would be detrimental to your DPS due to the breaks in the rotation.

 

• Binkenstein

Doing some EP comparisons as haste increases. (I haven't finished it all yet... obviously)



Haste is probably jumping all over the place because I'm adding 200 haste rating to generate an EP value, but it's probably not much to worry about.