Kishkegelt

I'm wondering whether I'm understanding the rotation DPS figures correctly. There's a pure rotation, an "effective" rotation, and an LB spam. At my current stats and latency, the LB spam figure is only about 100 dps less than the effective rotation. At the moment I'm pulling probably 1000 dps less than this, and I see that I am missing or delayed casts due to, I assume, latency. Would it be reasonable to suppose that I might come closer to the theoretical dps by spamming LB and eliminating a couple of spell "transitions"?

 

• Binkenstein

Effective = fight damage/time assuming you adjust your CL use to avoid going OOM.

 

• Binkenstein

Bah.

Left the old overload multipliers on the CL calculations (hadn't touched them in ages) which overstated the gain from using CL. Now looking at 67 for After LvB, 158 for On CD. New version soon...ish.

 

• Binkenstein

New stuff... too lazy to go into details atm. Hopefully better and stuff.

Attached Files

SEIC 3.2.2 r20.xls (1.14 MB, 733 views)

 

• Suggestive

Re:r20
Clearcast uptime is not self-updating (had the same problem with r19 actually). I'm not sure if this is intentional or not.
Turning the Totem of Wrath glyph on/off has no effect on DPS.

 

• Binkenstein

See previous comment about "if it's blank it's not back in yet".

Replace the buffs spellpower cell (D4) with =MAX(IF(I35="Yes",0.1*J35,0),IF(I42="Yes",280,0))+211*(1+0.1*M22)+glyphtow*84

[e]Fixed version + more notes: http://elitistjerks.com/blogs/binken...ion_stuff.html

 

Caelwynar

I see that in the latest version, my haste DEP has pretty much caught up to spellpower (and exceeded it, if I count the 350 flat SP of my two trinkets). Is that a result of the rotation tweaks you made?

In my case, a Reckless Ametrine is worth almost exactly the same amount of DEP as a Runed Cardinal Ruby, so I went ahead and regemmed to pick up a couple of smaller socket bonuses wherever I had been ignoring yellow sockets.

Looks like Rawr and SEIC are closer to being in agreement about stat weights in this version.

 

Souai

Be sure to check the latency value on this push if you're using this sheet. A lot of the default values got changed including latency from 150 to 0. That greatly increases the given value for haste. People using this spreadsheet should probably make sure to fill out the fields to their stats to get the most out of it, unless there was a reason I'm not aware of to use a latency value of 0. The CL calculation fix did not magically make the haste weight almost eclipse spellpower gemming.

 

Caelwynar

Ah, nice catch. When I enter 150 into the latency box, the numbers go back to roughly their previous values (~1.5 for sp, ~1.4 for haste). Turns out I was better off gemming for those socket bonuses anyway, though. The 4 haste and 5 sp socket bonuses make up for the slightly lower value from the gem (roughly a 2.6 DEP difference between a Reckless Ametrine and a Runed Cardinal Ruby before the socket bonuses are factored in).

 

• Binkenstein

I changed it to zero to compare the results with simcraft. Obviously people shouldn't use the default settings blindly.

 

dedmonwakeen

Slight clarification: SimulationCraft uses three different lag values depending upon the previous spell:
(1) prev spell has a cast-time enabling next spell to be queued
(2) prev spell was instant, triggering gcd
(3) prev spell was channeled, requiring a little extra time to ensure the last tick fires

Default values are:
queue_lag=75ms
gcd_lag=150ms
channel_lag=250ms

Actual values used are a random range centered on those values. (Random lag values are key to smoothing out the discontinuities in Haste analysis.)

http://code.google.com/p/simulationcraft

 

Jesi

Not earth shattering but Pandora's Plea is 751 SP:

=depint*108+depspell*(850*10/trinkintcd)

The original tooltip said 850 which didn't match the proc and was corrected.

 

Ajaxx

Is there a place that explains why Haste is so good and why Crit is practically half as good as Haste? I understand that part of Crit being worth less is because of the LavaBurst, wasn't sure if there was more reasoning.

 

Moshne

You answered the crit portion for yourself. Every caster class scales fairly poorly with crit due to the nature of the stat. (Yes, even shadowpriests, they just scale more poorly with haste.) The question is why we scale relatively better with haste.

First, we scale pretty poorly with spellpower, this is because of the cast times on our spells, we get a lower coefficient. All of our spells are this way. With the Shamanism buff, we are in a bit better shape, but still not as good as some of our caster counterparts.

Second, all of our spells have a cast time that can be reduced by haste. Flame Shock's DPS gain from haste is less than the others because the tick speed doesn't change, but FS is an insignificant portion of our DPS rotation. LB/LvB/CL all increase in value with haste until you get the 1.0 casting speed mark. This is very hard to obtain until you get the most top end of gear, and even at that point, LB will still continue to gain. Although, at that point haste devalues pretty quickly, for the same reason that Crit devalues from the beginning.

Hopefully that answers the concern, but the long story short is that "math" is the reason why. You have to get into the nitty-gritty coefficient and rotation dynamics to fully get it, and I'm not sure I can really show that to you without you getting your hands dirty yourself.

It's much easier to make strategic moves when you aren't trying to hold your intestines in all the time. - Juice

 

Zuathar

note: posting here because I an trying to access the "SEIC 3.2.2 r20.1 plus more" which is realated to this topic.

Bink, for some strange reason I cannot access your blog anymore "Zuathar, you do not have permission to access this page. This could be due to one of several reasons..." any idea what's happened with the rights?

 

wckddudeman

I have same problem as zuathar, but also I am curious if you will share, why you are discontinuing your spreadsheet.

 

Kelorius

By the look of things, it appears as if Bink has taken a leave of absence from the game. As for reasons, I doubt they will be shared here.

 

rellikz

I was wondering if there was anything i could do to make the load times when i click the wowhead link go faster because for some reason when i click it for the gear list not incl or incl hit it takes forever to load or doesn't load at all :-/ like i can have my browser open and the tab will pop up minutes after I clicked and my computer is very good so I don't think its a hardware issue. Does anyone else have this issue and sorry thats all one sentence.

 

Blyker

When you click the link in the spreadsheet its all coming from your pc to wowhead. Now wowhead might not be the fastest with this big of a search, so i might have something to do with that otherwise i dont know.

 

Shegokigo

So any idea what we're doing now with BINK's spreadsheet not even visible? I recently switched to Elemental and was hoping to read up on the past few months of changes in his blog, but that's not even accessable either. What now?

 

Souai

You can use Simulationcraft. I think it may model some things better than Blink's spreadsheet (trinkets, lag, etc), however it has a bit of a learning curve. You don't want to use their sample output and you don't want to use their default values for step function. So many items are essentially unreachable before Icecrown that a more helpful profile would not have the cloak, only at max two pieces of 258T9 and one Reign of the dead, etc.

Some of the information you would need is clearly in the How to Run section, but it doesn't mention the defaults for scale_ fuctions which are 150 (probably too high, gives some funky returns, particularly in their sampleoutput) except for hit which is done using -100.

There's a comment that links to another comment in Graylo's blog which has some further settings for figuring out stat weights, like putting on center_scale_weights which takes your scale value for that stat and goes up and down halfway on our current stat. It takes twice as long to run but is more indicative of the average level of benefit for your current stat level. I personally like using 50 rating for the step values, the variance was really low (within 1-2%) on the dozen or so test runs I did with 10k iterations and there were never any outlandish results like with larger step values.

Simulationcraft is actually pretty powerful, I think a lot of the reason why people don't use it is they're not sure how and it's not as easy as plugging in your unbuffed stats or getting an immediate return on.

 

Caelwynar

As I took a closer look at the workings of Bink's spreadsheet and the Rawr elemental module's source code, I was surprised to see that both of their models apparently only cover the duration of a single flame shock. I don't know if that causes their stat weights to be less reliable than simcraft, but I would tend to trust the program that models an entire fight. I know that when I'm raiding, I often come across situations where my flame shock would end before Lava Burst finished casting, so I need to decide between clipping the final DoT tick or delaying the next LvB cast (I clip the DoT, personally). It looks to me like neither Rawr nor SEIC were taking those bumps in the rotation into account, but simcraft might.

Rawr and SEIC were much more user-friendly, but I think I'll be using simcraft from now on.

 

Lord Helmchen

That's the difference between model and simulation. In the model we assume that you recast the FS before the LvB, because it is a very good estimate.

 

bone123

hm i can't attach the file. so here the last version which is no longer available in his blog:
SEIC_3.2.2_r20.1.xls

anybody else may upload it as an attachment

 

Moshne

I'm going to play with the sheet a bit later in the week when I find some time. Once I'm confident I've got it up to date I'll get it rehosted.

While I'm at it, if there is anything that you feel might be useful in it, let me know and I'll see what I can do. I can't make any promises though.

It's much easier to make strategic moves when you aren't trying to hold your intestines in all the time. - Juice