Pungent

NEW RELEASE: 3/14/07

The promised update is complete and can be downloaded: HERE

Changes / Updates:
* +hit is now factored into the sample. Level based scaling included.
* The use of Shadowfiend and Manapots are now available. These are averaged into MP/5.
* An additional % damage modifier is included for specialty fights.
* Improved VE talent added.

In Progress:
* Inner Focus inclusion in sample.

Please keep me posted on any discrepancies or requests for future releases. Thanks, and enjoy.

 

Trepidati0n

You are not including spell level vs + damage. Right now you are giving a rank 1 spell full + dam. I see you sorta have this started, just not finished. Otherwise, having a hard time initally blowing holes in it.

 

Pungent

Thanks for the response. Downranking is something I've considered a few times but never really decided how should be implemented. For the sake of my own sanity I've avoided it in the spell simulation and assumed max-rank spell casting.

In the next update I'll probably include downranking as an option for individual spell data but we'll see what other recommendations / corrections get brought up before uploading a revised copy. The feedback received here and within my guild will largely determine what features get added or changed but I hope others will at least find the data useful.

 

nataku

I believe Trepidation was trying to say you are giving a low level spell it's level 70 benefit from spell damage gear when in practice, there is a penalty for using spell ranks that were trained below level 50 when your character is level 70.

 

Pungent

Perhaps I wasn't clear. I understand the issue being discussed but haven't put in a system to calculate coefficient changes based on character level. This change is a simple matter and would correct the readout for downranking but I haven't decided how I want to do it.

The initial idea was to put in current gear levels and see how an individual spell would scale with that gear as I leveled - without the need to change additional variables. As you can see, that initial idea has grown a bit. So at the initial conception, downranking was a distant thought. This will be corrected in the next release.

 

Pungent

I'm having a great deal of trouble modeling the coefficient for SW:P downranking. Does anyone know the actual formula used? The one below doesn't work.

untalented: coefficient * ((level learned for next highest spell rank - 1 + 6)/70)
talented: (coefficient + (0.185 * rank of imp. SW:P)) * ((level learned for next highest spell rank - 1 + 6)/70)

 

LucidityAxel

Your formula looks exactly like the one that Blizzard published when they announced the downranking penalties. What are you seeing that makes you believe it is inaccurate?

On a different note, does your spreadsheet model hit penalty? I didn't see any values for it anywhere.

If you don't want to go to the trouble of modeling player vs target level, then you could simply assume the player is at the hit cap and factor in a static 1% miss rate.

 

Trepidati0n

Do me a favor. Take the default spreadsheet and turn VT on and off in terms of the queue. The delta is ~100DPS. However, the spreadsheet says 200DPS as an individual spell. How is there such a huge difference?

The reason is that because we have so many choices of what to cast that we actually lose DPS because we can't use all our spells effectively at the same time. I always knew this to be true....just never realized how big the effect was (all 50% or 100% depending upon which way you look at it).

The fact that mages can chain their spells with no lost time due to DoT casting, they can much better approximate their DPS change from 1 point of +damage while locks/priests have to actually model a cycle in order to figure out what is going on.

 

Northerner

This I find to be fascinating actually. (I also quite agree by the way.)

Mages seem to be increasingly encouraged to move towards an AB(x)/main-nuke(y) rotation at the very least and quite possibly will end up running some sort of AB/AM/fireball/scorch weave by the end of the expansion. I don't think this is at all accidental though and imagine a good part of that is obfuscation of the casting model on Blizzard's part.

I don't mean this in a cloak-and-dagger or tinfoil-hat sort of way at all. I just mean that Blizzard wants casting to be more involved obviously and more 'skill-based' than gear-based. I don't know if that end will be met but I do find it terribly amusing that actual simulations are needed to find optimal casting cycles. What I don't find quite as interesting is how many casters will be far behind simply because they don't have access to the math that they need to make good decisions.

 

Pungent

NEW RELEASE:

The promised update is complete and can be downloaded: HERE

Changes / Updates:
* Threat and VE calculations have been updated to correspond with the latest patch.
* SW:P coefficient has been updated to correctly reflect talent scaling.
* Spell cycle updated to accommodate level changes. Unavailable spells will be dropped from rotation.
* Downranking has been modeled across all spells (exception is Devouring Plague which scales ... oddly)

Interesting Discoveries: (based on in-game testing)
* SW:P suffers no degradation from downranking until Rank 3 or lower.
* The SW:P coefficient nerf was not a flat 9%. The improved talent scales this to a 12%.
* Devouring Plague scales strangely when downranked. Normal scaling is 80% but drops to around 65% at the next lowest rank. It then scales back upwards to 80% at Rank 1.

Please keep me posted on any discrepancies or requests for future releases. Thanks, and enjoy.

EDIT: In response to the question about including +hit in the calculations. It's something I've put a lot of thought into but neglected for now. I wanted to get this build pushed out ASAP in order to get feedback on the underlying spell formulas before adding more variables. It's something I'm looking into, but don't know how doable it will be with Excel. I'm already hitting the upper bounds on logic formulas in a lot of instances.

 

Asmo

Thanks for this. Great work!

One interesting point is assuming that optimal cast sequence and 100ms lag or less only points 1 and 4 in imp MB does anything. Having 3 instead of 1 point apparently never results in a situation where you gain any extra mindblasts in the spell rotation. Which does seem to confirm my own calculations I tried to make based on the spell rotations I believe Shalas posted earlier in the other shadowpriest thread.

So at least theoretically with my gear the answer to the question of where to put the 3 points I have left for imp MB or shadow power is optimally spent 1 in imp MB and 2 in shadow power.

 

Brekk

that is for optimum cycle, ive found 4 in Imp MB is nice just because it puts it on the same cooldown as SW, it just makes it easier to deal with those two being on the same cooldown.

 

heel

This is an excellent spreadsheet. A few issues, though:

-According to the spreadsheet, DPS is maximized with one point in improved SW:P (1/2 produces higher DPS than 2/2). This doesn't make sense.

-Shadow Weaving is not modeled properly. Each point in your spreadsheet results in a linear increase in DPS, which is not correct. 4/5 Shadow Weaving is *very* close to 5/5 in a raid environment. I did the math a while ago, and the chance that Weaving ever falls off in a 5-minute fight is less than 3%.

 

Pungent

On SW:P - What you're noticing here is exactly the same thing commented on above by Asmo and Brekk. It's not that the DPS of the spell actually increases; it's that given the pieces to the puzzle, having only 1 point causes everything to fall into place more frequently. This is a subjective increase based on lag and spell rotation.

On Shadow Weaving - I'm not 100% sure what you're implying. Yes, it's true that in a raid situation 4/5 shadow weaving will rarely drop off the mob. What I'm doing in the spreadsheet however isn't calculating frequency of drops, only the total damage done under situations where the rank of shadow weaving in the spreadsheet corresponds to the number of stacks on the mob. The number of points you have in the talent is ultimately irrelevant, as this is simply for DPS analysis. As long as the stacks get on the mob, that's what counts.

 

Brekk

however, the shorter duration requires more recasts, and with the cost of SW:P for long duration fights 2 pts in it is better.

 

Lanthon

Very helpful spreadsheet, thanks! One request: In calculating Longetivity, it would be helpful to be able to factor in Judgment of Wisdom. This would definitely help in planning out spell sequences for different lengths of fights.

 

LucidityAxel

Another thing to consider about Imp Mind Blast is that you need not wait for the Mind Flay to actually finish channeling to before starting to cast it. You have the option of trading some mana efficiency for extra burst dps.

 

Pungent

NEW RELEASE: 3/14/07

The promised update is complete and can be downloaded: HERE

Changes / Updates:
* +hit is now factored into the sample. Level based scaling included.
* The use of Shadowfiend and Manapots are now available. These are averaged into MP/5.
* An additional % damage modifier is included for specialty fights.
* Improved VE talent added.

In Progress:
* Inner Focus inclusion in sample.

I won't be including MF chopping unless I find an easy way to add it without re-writing the underlying spell logic. The spell queue is complicated enough as is.

 

LucidityAxel

I'm not suggesting that you try to model an abbreviated channel time for Mind Flay. I was merely trying to point out that the talent has practical uses that may not show up in this particular bit of theorycrafting.

This is also isn't a criticism of your fine spreadsheet. A model doesn't have to be 100% complete to yield useful insights.

I think spell hit is a widely underappreciated factor in assembling a boss nuking gear set, especially for shadow priests who may be newer to a raid dps role than a mage or warlock. I'm glad you incorporated it.

 

Pungent

Thanks, I appreciate the feedback. I'd like to get this as accurate as possible but you're right about not needing a complete model to gather useful information. It's already shown me a lot.

Also, I've applied several little fixes throughout the day. They fix some relatively minor issues but issues nonetheless. This includes expanding the miss-scaling beyond +5 levels and the manafiend formula to correct mana return and include scaling with % modifiers.

 

Orium

Great work!

It would be nice to see the effects of Fel Mana Potions on both longevity and DPS.

 

Brekk

I knew the importance of +hit to some extent from talking with guild mages, but that spreadsheet really showed how big of an impact on DPS +hit and shadow focus make.

Just from that spreadsheet im considering dropping to 1/5 Imp MB, from 4/5 Imp MB after seeing the 3 extra pts netting me 6 dps. 1 pt for 7dps is still worth it though.

 

Myul

adding 100 spelldmg is about 44 till 51 dps (regarding your cycle)

so its about a lose of 11 dps - 12.75 dps

Regular used [Super Mana Potion] adds 2400 mana on average or 100 mp5.
Using [Fel Mana Potion] may increase your mp5 by another 33,3 (period).

I can't tell you the exact longevity because with the more mp5 you have the more it's scales and we don't know your cycle.. the easist might be to put in another 33 mp5 into the stylesheet while substracting 25 spelldmg from your gear.

eg:
1000 spelldmg
mb mf swd swp vt cycle
100 mp5 (from super mana potion but not using the potion option)
->
1.057 dps
2,93 longevity

using your choice instead
975 spelldmg
bla
133 mp5
->
1.045 dps
3,28 longevity

---
well funny thing while playing around with the numbers, just check this
w/o mb & swd

841 dps
9,68 longevity

or

830 dps
15,03 (!!) longevity

for losing 11 dps you may add 55% longevity for only 33 mp5 oO sounds bit odd for me

 

Bloodtear

Great work.
I love it that I can tweak the spell priority. And I'm even more amazed from the impact of latency. One casting rotation gives better dps with lower latency value (0.2), another for a higher latency value (0,3). Cooldown matching ftw

One minor remark. When you give lowest priority on VE (6), the spreadsheet counts as if I receive it's healing and threat while in effect it's never casted. I suspect you are using some "Will VE be casted" rather then if it's actually casted.

 

Pungent

Hmm, good catch. That's exactly what's going on. The formula is only checking that VE is in the active rotation instead of looking for an actual cast of the spell. Anything with a lesser priority than MF is basically eliminated since MF has no cooldown. I'll take a look at what can be done.