Friday, May 15, 2009

Moonkin Stats and My New Model

Update: Found an error in my model. Crit is worth a little more then I previously stated. It has been updated.

Ahhhh. Remember the good old days in BC when the moonkin Rotation consisted of Moonfire and casting Starfire 4 or 5 times. That was when theorycrafting was easy. If you figure out how each stat affects each spell you can come up with a good estimate of how much each stat is worth realtively quickly.

That is all ancient history with the introduction of Eclipse, new set bonuses and Glyphs. If you look around the various forums you will see a lot of people saying how much each stat is worth and all the different versions are not as close as they used to be. As a result, I have built a new model to try and determine the value of each of DPS stats for a Lunar Rotation. Well, the results are in.

I'm going to do this post a little bit backwards then I normally do. I'm going to talk about each of the stats and give my results first. Then I'm going to talk about how I got the results in the end.

The Stats:
First off all, these results are for a perfect Lunar rotation with no lag, and no movement on a single target. I have also excluded both Force of Nature and Starfall from the analysis, and the results are based purely on DPS. Also, these stats are based upon a pre-3.1 BiS gear set with 4T7 set bonuses. I expect these numbers to change as our gear improves since our SP will go way up and our Crit and Haste will stay realtively the same. I will redo this analysis after we know more about what a BIS 3.1 Moonkin looks like.

The results aren't all that shocking. The values I came up with still fall into the standard Hit > Spell Power > Haste > Crit explanation.

Hit Rating: Relative Value = 1.54 - Hit Rating has the highest per point value of all the DPS stats assuming you are below the hit cap. This valuation just shows you how important it is to be hit capped. In my opinion, all raiding Moonkin should have the talents of Balance of Power and Improved Faerie Fire. This means we need to pickup an additional 10% hit chance from gear and other buffs to be hit capped.

So, what is the hit cap? 263 for Tauren and Night Elves with out a Draenei in their party. 236 for Night Elves with a Draenei in their party.

Spell Power: Relative Value = 1.00 - Spell Power is the standard by which all other stats are measured. After you are hit capped, Spell Power is the best DPS Stat on a point for point basis. This is why, it is always recommended to use Spell Power gems and food when possible.

Haste Rating: Relative Value = 0.80 - This is where the debate usually begins. Haste Rating is a great DPS stat the makes your spells cast faster and lowers the global cooldown. The down side to Haste Rating is that it also makes you consume mana faster and if you have mana issues the Haste will make them worse. Also, spells with short cast times like Wrath and the DoTs can reach the minimum global cooldown fairly easily with reasonable amounts of Haste Rating. You can find my post on the Haste Cap here.

Crit Rating: Relative Value = 0.62 - Crit Rating is the least valuable of the "pure" DPS stats, but that doesn't mean it isn't a good stat. Crit Rating also has the side benefit of helping you to generate mana using the Mana on Crit mechanic.

There are two important things to remember about crit. The first is that even though you don't want to necessarily stack Crit Rating, you don't want to completely avoid it either. The second is that you get a lot of Crit Chance from other Sources like talents and raid buffs. This is the main reason why Crit Rating is ranked so low.

Spirit: Relative Value = 0.34 - In patch 3.1.2, Improved Moonkin Form is being buffed so that your Spell Power will be increased by 30% (up from 15%) of your Spirit. Also, Spirit will help you regen mana if you have the talent Intensity.

As a general rule Spirit is a stat to be avoided, because Haste Rating and Crit Rating are so much better in terms of DPS. However, Spirit is becoming more and more unavoidable. Most caster items now have Spirit, and majority of the ones that don't have Hit Rating. It is fine to have spirit on your gear, but never intentionally stack it.

Intellect: Relative Value = 0.35 - Intellect is not a good stat in terms of DPS but it is almost completely unavoidable. Virtually all caster items have some amount of Intellect on them. Intellect has a lot of side benefits though in Mana Regen since it impacts all of our major mana regen talents. You will pick up enough Intellect just by making commonsense gear choices, therefore you should never intentionally stack Int.

(Edit: I updated the Spirit and Int numbers to take into account Stat Multipliers like Blessing of Kings. Thanks to Antonetz for pointing this out.)

Other Stats: Relative Value = 0.00 - All other stats have very little direct impact on your DPS. Stamina is the only other stat that you need in any significant quantity, but you should pick up plenty making normal gearing choices. Mp5 can also be helpful if you have mana issues. However, most moonkin don't have big mana issues that can't be solved with talents. Avoid Mp5 if possible. Spell Penetration and Resilience are useless in a PvE environment and should be avoided.

TL:DR
Hit = 1.54, SP = 1.00, Haste = 0.80, Crit = 0.62, Spirit = 0.34, Int = 0.35

The Model:
If you want to look at the model and check my work you can find it here. However, I want you to keep a few things in mind. 1. It is a very large file. Download at your own risk. 2. It does have a couple of macros on it but you don't have to enable them for the spreadsheet to work. 3. This is NOT a tool to help you model your DPS. If you want something like that go check out WrathCalcs. 4. In conjunction with #3, it is not userfriendly. It even confuses me and I built the damn thing.

How it Works: I built the model to choose which spell to cast based on the current situation. It uses the random number generator to determine crits, misses, and procs. The model assumes the linked Talent/Glyph build. The model assumes the following Cast priority:

1. If FF is not up, cast FF.
2. If IS is not up, cast IS.
3. If MF is not up, cast MF.
4. If Eclipse is not on cooldown, cast Wrath.
5. If Eclipse is on cooldown, cast SF.

This means that FF and the DoTs will be refreshed during Eclipse. I won't go into detail about it now, but I now think this will give you better DPS then weighting for Eclipse to run out.

The model uses a 60,000 cast sequence. This is equivalent to about 26 hours of continuous casting. This may seem like overkill but I wanted to smooth out as much of the randomness as I could.

Using base stats of 2900 spell power, 41% crit chance, 16.5% haste from gear, and 100% hit chance, I ran the model 1,000 times to find the average total damage, average total cast time, and average DPS for a control group. I then adjusted each of the stats by significant amounts and reran the model another 1,000 times for each adjustment to see how the DPS changed.

The adjustments were made in significant amounts and then averaged to a per point value. For example, to find the value of Spell Power, I increased the spell power used by 50 to 2950. This resulted in an average increase of 61.81 DPS or 1.24 DPS per point average. To evaluate Crit and Haste I increased both by 1%. To evaluate Hit Rating I decreased hit chance by 1%.

The stats where then weighted relative to each other, with Spell Power being the standard. These new values will be applied to my gear ranking and I hope to have it updated soon.

Whats Next: I am kind of excited about what I can do with this model. I plan to create additional models to evaluate a solar rotation. I would also like to add lag into the model to see the impact. I would also like to build a model that takes movement into account if possible. Finally I will be using the model to evaluate new set bonuses, glyphs and talents.

26 comments:

Echoris said...

Excellent post!

After a 5 month hiatus from WoW, it's good to be back to blowing stuff up again. : )

Anonymous said...

I'd love to the this for the Solar Rotation, cause I really have enough of casting SF all the time. I hade to do so in Sunwell for too long.
Spamming Wrath with a cast time of just 1 second and make 10k critical hits is much more my thing.

By the way...will you rewrite your gear list in order to adjust for the 30% spirit buff and maybe also with focus on solar rotation?

greetings
Nightedahs from Khaz'Goroth (EU)

Maestro said...

Heya Graylo ~

This looks great, I'm downloading the excel spreadsheet now. I'm looking forward to messing with it to model a solar set, although I'm sure I will have to take some time and learn what you did here. I'm sorta a newb when it comes to these things.

@ Nightedahs
The spirit buff does very little to impact the gear list. When I added the spirit buff into the calculations for my gear list as well as a couple gear lists using both Murmurs' and Graylo's scales, nothing but the relative values changed. The increase is minimal on a per item basis.

~Relevart

Boize said...

You based your models assuming:
"2900 spell power, 41% crit chance, 16.5% haste from gear, and 100% hit chance".

However, would it be the same for, say, lower crit/spellpower values?

With my current gear, self-buffed, I am sitting at ~2100 spell power, 404 haste rating, and am hit capped, but have a measly 21.13% crit chance.

On top of this, due to being "out of the loop" in terms of raiding pre-3.1, I am also lacking the T7-4pc bonus.

In this scenario, would crit scale to be more important than haste? Or would haste still trump crit? (I have a couple of Ulduar items that I can swap out which would increase my haste further, but at the cost of some crit.)

- Boize
Dath'Remar-US

Float said...

Thanks for the new stats, Graylo

Do you find that the value of haste remains constant (or nearly so) even after you hit the haste cap?

Nightedahs said...

Well I agree that the change from 15% to 30% of spirit as spellpower wont change that much in the calcs that rated spirit with 15% before.
But as I read in the gear list spirit isn rated at all and some items would gain 20 or even more spellpower out of this now.
And that might change the rating of the item itself.

Graylo said...

@Anon1

I will update the Lunar Gear list soon. I do plan to build a Solar Model and provide a solar gear list at some point.

@Boize

If you increased all of my stat values by 5% or 10% and then redid the analysis then the conclustions would probably be very similar.

Now if you kept the spell power constant, but reduced the Crit and haste values then they would go up. Since you do not have the 4T7 then Crit will be much more valuable for you.

@Float

If the GCD didn't have a minimum value then my haste value would be higher. However, if your using a Lunar rotation you can forget about the haste cap. Haste is still very valuable for a Lunar rotation even when haste has little impact on Wrath.

@Nightedahs

I don't know if your refering to my gear list or not, but in my list the dps value of Spirit is used to rate the items, but the mana benifits are ignored.

David said...

just to clear something up, are you suggesting to use SF over Wrath after the lunar eclipse is over? the longer cast time and crit chance of both would seem that wrath spam after that rotation. i dont know. i dont crunch numbers i just blow shit up and see what works.

Jormundgard said...

Graylo,

Is this basically the output of Monte Carlo modeling then?

Anonymous said...

Do these models take into consideration full raid buffs and IMotW? If so spirit and intellect should see some gains from stat multipliers.

Sorry if I missed something; I just browsed this and I'm tired :P

-A drowsy Antonetz

Naturi said...

Graylo A few things,

Whys is your haste and crit values so different to Murrums(In terms of the amount they are about)?

And Considering we will be aleast getting Tier 8 2 set, does that increase crit somewhat since we loss 5%?

Graylo said...

@David

I use SF as my filler spell. Wrath techically has more DPS in theory but it has other problems with lag and Spell Queuing. Plus I have the SF idol equiped for Eclipse. Therefore I think SF makes more sense for a Lunar rotation.

@Jormundgard

Though I didn't know the techical name for it, I think you are correct. It has been over 10 years since I took any kind of Statistics so I don't know the margin of error and all that fun stuff, but I assumed 1000 trials would be enough. If some one with more stats experiance then I would like to run a regression analysis on it and give me the results that would be great.

@Antonetz

Good Point. I will update my results soon.

@Naturi

I have several issues with Murmurs numbers and posted a big post on TMR.

I think there are some big issues with his crit number. First of all using a starting value of 30% is pretty low given 15% haste and 2700 SP. A moonkin with that quality of gear is likely to have 4T7. Therefore he would only be getting 10% crit from the base, Int and Crit rating. I think his starting number should be at least 35% if not higher.

The second big issue with his crit number is he doesn't take Eclipse into account. Eclipse actually decreases the value of crit because it provides such a large amount of it, and 2T8 will make that even worse.

I also didn't like the way he talked about haste. I thought it was missleading, and there were some issues with how he calculated it in my opinion.

None of this is to say my numbers are 100% correct. I expect these numbers to increase by the end of ulduar. Our SP numbers will go up quite a bit, and our Crit and haste numbers will actually go down a little bit. This means that Crit and Haste will improve dramatically in value over this patch. I will update this analysis when more is known about what a 3.1 BiS geared moonkin looks like.

Kai said...

Another great post Gray!!

Alrdy D/L both links you have here. Plan to play with them 2nite. One question for you and the gallery..given that the "best" gear while upgrading is always in flux what site, mod and/or program do you recommend to calculate and recalcuate (after an upgrade) what loot is best for each slot. (ei. like the weight scale in wowhead.)

Thx again for all the info love this site :).

-Kai

Anonymous said...

Any intuition for why int and spirit have the same value? It is the one part of the numbers that I can't rationalize quickly --- how does the 30% of spirit to damage come out to the same ranking as a little bonus crit and bonus damage from int? I understand if that's what the model spits out (and if it does, it means blizzard has done a remarkable job of balancing the dps values of the two stats), but it is still very surprising.

Graylo said...

@Kai

I don't really like any of them. I do use WoWheads comparison sometimes but I use my own scale. This will sound self serving but I would check back here.

@Anon3

The spirit number is easy to calculate. Its basically 0.30 * (1+0.1+0.02) = 0.34 rounded. The 0.1 and 0.02 are Kings and Imp MOTW.

For int you have to remember that 45.9 Crit Rating equals 1% crit, and 166.667 Int equals 1% crit. So you can take 45.9/166.667 = 0.28. Since Crit rating is worth 0.57 you can say that in terms of Crit chance Int is worth 0.16 (0.57 * 0.28 = 0.16). Then add the 12% spell power you get from Int and Int is worth 0.28. Now you have to take into account stat multipliers and you get 0.28*(1+0.1+0.1+0.02) = 0.34.

Kai said...

I argee Gray...so when can we expect to see your new list based on these numbers??

Naturi said...

Thanks for explain that Graylo.

A final question, do your model predict a rating or % of crit we should not drop under?

Kwidney@DarkspearEU said...

Hail to the king, baby!

Thanks alot for these numbers Graylo, I'm curious what these numbers will do to the gearlist..

I'm still not convinced it's worth it to lose my 4pc T7 using a lunar rotation (or solar for that matter), looking at the haste and crit we lose..

Sonak said...

I am always surprise by the values you announce speially on sp.

I often raid in 10 man and I never reach that high SP - and yet I am gully naxx25 equiped (4t7.5) with lots of best in slot items (not all anyway).

My sp is more like 2050 unbuff (so more flasks it means +- 2200). I am curious to know how you can reach 2900.

(got also a lot of e219 stuff anyway)

Rolzroyce said...

Have you changed the stat weight of spirit to accomodate for the 3.1.2 +15% sp buff from spirit or is the version currently online just from the 15%?

Thanks,

Rolzroyce

Robert said...

I'm actually finding that with 2t8 a solar rotation actually starts pulling ahead of a lunar one really fast. It would be nice to have a model of a solar rotation given 2t8. though i imagine it would just switch crit and haste.

Tinuviel said...

Thanks for the excellent post and blog Graylo - keep up the good work!

My Moonkin alt is currently 77, and I am hoping to raid Boomkin with her at 80 so your site is invaluable.

I have a question. I love my AOE - hurricane, starfall and typoon. I wanted to be able to retain that in my raid build, both for the trash and for soloing and grinding. But would I be losing a too much DPS if I stole 3 talent points to put into Gale of Winds and Typhoon, and where would be the least harmful place to steal it from in your opinion? Or can you convince me that this would be a very bad idea.

Btw my guild is a 10man raiding guild, so unless I am pugging, my moonkin is probably going to be doing heroics and lower 10mans, with my hunter main brought in for the higher level content).

You can find my over at my hunter blog - Steady Shot (http://wowwhimsy.wordpress.com/)

Graylo said...

@Naturi

I don't think there is a minimum level of crit we must have, but as our Crit chance decreased the value of crit will go up. I don't know how much Crit will go up but in the end it will balance out.

@Kwidney

I'm convinced that T8 > T7. We do lose a lot of Crit, but we gain a lot of Spell Power, and the new Set Bonuses are nice also.

@Sonak

My numbers are fully raid buffed in a 25man raid. They are based upon the numbers I've seen myself reach in game. Unfortunately it is impossible to model for every possible combination.

@Tinuviel

Since your Moonkin is not your main, you can probably spec how ever you want without issue. That said, Typhoon and Gale Winds are of marginal utility in raids. In a majority of raids (especially in T7) the opprotunity to AoE aren't that important and your just draining your mana. Other classes do AoE better.

That said, If you really want those talents I would try and pull the points out of Mana regen if you can. If not pull it out of Imp IS.

Chip said...

Do you find that you swap gear for moments when you have a Draenai in your group due to the hit buff? I'm just wondering if it's prudent to have additional gear available for any eventuality.

sonak said...

For me optimization of items based raid composition is quite dangerous.

But indeed changing a trinket/item if you have some spare seems pretty smart and feasible.

Reckah said...

Well, i have a question reguarding talents. I've noticed that raid damage is a significant part of Ulduar. That being said I am currently spec'ed into Owlkin frenzy, at the cost of moonglow and gale winds. I never go oom anymore with the buff to innervate. I normally see it proc 8-12 times a fight probably an average of 10 times. I was wondering if we could find some numbers for this. Granted it is in random parts of our rotation but i was thinking if we can pull 4.4kdps without it. That 4.4 x.10 x 100seconds = 44k extra damage in a fight. I was just wondering if anyone else has been taking this into consideration lately and maybe running some numbers to see what kind of dps increase this would be.