Resto Stat Weights for WoD

[Stoove 347]

One of the big discussion points before WoD has been stat weights for gearing. Everyone wants to know how to gear when they start running Dungeons in Warlords, so the time has come to really pay attention to what stat weights we should be using in a week's time. I've been spending a lot of time recently thinking carefully about the WoD stats, so I thought I'd share my conclusions.

 

The discussion is often dominated by quite simple mathematics and a heavy amount of reasoning regarding how people "feel" about a stat - something I like to call "touchyfeelycraft" because it is quite disconnected from actual mathematics. By approaching things in this way, I think we lose some important information about our stats. I set out (over a year ago now) to take a much more mathematical approach to stat weights. That doesn't (shouldn't) prevent us from applying our understanding of raid and healing mechanics to the results, but hopefully we will be better informed than previously. I previously posted my analysis and results on my blog (here), and I'm making this post to give a quick breakdown of the consequences.

 

Here are the basic results of my theorycraft;

1. Mastery is by far and away the stat with the highest potential, and will definitely be the best stat for Resto Shaman in progress raids in WoD.
2. Haste's throughput value has previously been understated by a lot of other discussions. It's about as good as Mastery when healing players around 55% health.
3. Despite what some people think, Multistrike actually gives you slightly higher throughput than Crit. Both of these stats are worse than the other two by a nontrivial margin.
4. Overall, the stats are quite close over a whole rotation; taking a piece with Crit instead of Mastery won't affect you massively, but there is definitely a benefit to having Mastery.
5. We won't be wanting to gem for Mastery or Haste over Intellect until about ilvl 700, because the relative value of secondary stats compared to Intellect only reaches 0.5 at about this point.
6. Don't make an effort to "balance" your stats; diminishing returns don't affect you as strongly as you might think.

The analysis I did only looked at throughput in a rotational setting, so there is no consideration for Haste's negative influence on your longevity in a fight. However, such mathematics is currently very challenging - until someone manages to do it, you should decide for yourself whether you want Haste to be your second or third favorite stat.

 

I made this table of mean ilvl and stat weight, which you can use to look up what I would suggest as a starting point for stat weights used in Ask Mr Robot's custom settings;

 

 

If you are somewhere between ilvl columns, you can just take the average of the stat weights either side.

 

Personally, I would recommend using Haste as your second favorite stat. I think that the throughput increase is enough to offset any mana worries you might have about it, but that is somewhat of a personal choice.

[dean771 3]

1. We won't be wanting to gem for Mastery or Haste over Intellect until about ilvl 700, because the relative value of secondary stats compared to Intellect only reaches 0.5 at about this point.

I was under the impression gems were secondary stats only for some reason.

 

What average hp % did you use when you weighted mastery?

[Stoove 347]

I was under the impression gems were secondary stats only for some reason.

 

What average hp % did you use when you weighted mastery?

 

Yes, I just realized this last night. Never mind that point :)

 

 

What average hp % did you use when you weighted mastery?

 

50%, I stated that in the blog post but it didn't seem important in the summary. It seems like a reasonable place to put it, and I can tell you that at about 55% Haste and Mastery become roughly equal.

[Murilo 0]

Excellent post Stoove, thanks for putting in so much work.  Was curious about the weighted mastery numbers, are those results including our 5% passive bonus?  Sorry if it's written somewhere and I've missed it.

 

I watched some of the Blizzcon raid, was hoping for the new triage model in action, but it seemed much closer to SoO than early Cata healing.  I'm not sure what kind of gear they had though.  Watching it made me think more about weighing mastery, and what a fair hp% constant is.

 

All factors point to stacking mastery being the correct strategy for us, I'm just a bit concerned that it's not going to end up being as effective on live as the Beta made it appear.  I was sticking with Panda for the +105 mastery buff, but recently went Dreanei for the +65 INT instead.  Fingers crossed hoping our mastery shines on progression, but we'll see. I got my hopes up that disc would get majorly Nerfed after how silly ToT & SoO were, and that didn't quite pan out (atleast compared to 3-4 months ago).  Anyway thanks again man! Bravo.

[Stoove 347]

Thanks very much for the compliments! :)

 

Excellent post Stoove, thanks for putting in so much work.  Was curious about the weighted mastery numbers, are those results including our 5% passive bonus?  Sorry if it's written somewhere and I've missed it.

 

Yes. It's written in the original post on my blog and if you want to dig and check, I linked to the spreadsheet which I used to determine what coefficients to apply. If you're super crazy I'll send you my code and you can check it for me!  =3

 

 

Fingers crossed hoping our mastery shines on progression, but we'll see. I got my hopes up that disc would get majorly Nerfed after how silly ToT & SoO were, and that didn't quite pan out (atleast compared to 3-4 months ago).  Anyway thanks again man! Bravo.

 

All the indicators currently show Holy as the by far the most powerful Priest spec, and actually probably one of the strongest specs out there for healing.

 

I watched some of the Blizzcon raid, was hoping for the new triage model in action, but it seemed much closer to SoO than early Cata healing.  I'm not sure what kind of gear they had though.  Watching it made me think more about weighing mastery, and what a fair hp% constant is.

 

Not sure about this - it's a very complicated question to answer in terms of mathematics (possible, but a statistical problem which I don't yet have the resources to tackle) and additionally has a lot of touchyfeelycraft involved. Hamlet and Dayani have made good points on the matter to the effect of "it more or less doesn't strictly matter what you heal someone for when they are at 90% in the new expac, it matters much more in the 55% or lower region" which I kind of agree with. That's why I did the analysis with hp set to 55%.

[dean771 3]

Not sure about this - it's a very complicated question to answer in terms of mathematics (possible, but a statistical problem which I don't yet have the resources to tackle) and additionally has a lot of touchyfeelycraft involved. Hamlet and Dayani have made good points on the matter to the effect of "it more or less doesn't strictly matter what you heal someone for when they are at 90% in the new expac, it matters much more in the 55% or lower region" which I kind of agree with. That's why I did the analysis with hp set to 55%.

Mathematicly its fairly simple, the weighted average hp% when healed, weighted by the size of the heal, just a lot of data involved, warcraft logs does track hp over time so should be possible in theory but not something I'd want to tackle.

 

Definately agree about heals over 90%, what ever stat you stack most will overheal anyway

[Stoove 347]

Mathematicly its fairly simple.

 

I've been looking into the statistics of it. Trust me, it's somewhat more complicated in terms of a theoretical interpretation than most people would give it credit for. It depends upon your approach, but I would take a theory-first approach rather than an empirical approach which I think you are talking about. I actually have a good set of theory on the statistics behind it, but the resources behind acquiring the data is the bit that I can't get. It's too big a job for me :)

[rodrigo 0]

Excellent work and a great post!!!! this is a very helpful info for aiming the pre-raid gear... i'm just wondering how Spirit scales vs ilvl... or is it just irrelevant respect the others stats weights?

[Stoove 347]

So, from what we know it will always be your second best stat simply due to how difficult it will be to stack. Take Spirit at like 0.9 or something :)

[rodrigo 0]

let me see if I understand... the weights i'm using now are this:

 

intellect 1

spell power 0.9

spirit 0.7

mastery 0.5

haste 0.45

critical 0.35

multistrike 0.35

versatility 0.3

 

I see that you normalized the weights to spell power.. so this stats normalized would be something like:

 

intellect 1.111

spell power 1.000

spirit 0.778

mastery 0.556

haste 0.500

critical 0.389

multistrike 0.389

versatility 0.333

 

and according to your analysis, the weights of mastery, haste, etc etc are even much lower than this (and the sugestion of 0.9 spirit).. so the stats would be something like:

 

intellect 1.11

spell power 1

spirit 0.9

mastery 0.29

haste 0.28

critical 0.23

multistrike 0.23

versatility 0.15

 

am i following this or talking just nonsense here???

Edited November 18, 2014 by rodrigo

[Stoove 347]

am i following this or talking just nonsense here???

 

That all makes sense. When I did my analysis, I didn't differentiate between Spell Power and Intellect - the difference is that 1 Int gives 1.05 SP due to your Armour Skills passive. I don't know why it should be 1.1 rather than 1.05.

[domeblade 0]

I was wondering why you rate Crit so low I was thinking that Crit was going to be higher because of the amount of spirit on the dungeon gear here is how I thought the stat weight was going to be Int>Spirit>Crit>Haste=Mastery>Multi.

[Dofri 4]

I was wondering why you rate Crit so low I was thinking that Crit was going to be higher because of the amount of spirit on the dungeon gear here is how I thought the stat weight was going to be Int>Spirit>Crit>Haste=Mastery>Multi.

 

The reason is that this stat weight is based only on throughput and therefore places no value on mana positive or negative effects.

Edited November 23, 2014 by Dofri

[Stoove 347]

The reason is that this stat weight is based only on throughput and therefore places no value on mana positive or negative effects.

 

Quite so. However, one thing that Hamlet and Dayani did point out a while ago is that the marginal gain in mana efficienct from Crit (via Resurgence) is really quite small!

 

EDIT: Also, I am very actively working on a way to account for mana issues in this kind of analysis - it is quite difficult and will be done When It's Ready ™.

[Dofri 4]

Let me start by saying that i really do like this piece of theorycrafting and it is truly quite useful for what its meant for.

 

Im guessing that the mana return would depend quite heavily on the spells you are using. Of course the return from chain heal was heavily nerfed and that would have been the spell where we most need the mana return. But seeing that we need to cast a lot of riptides and healing waves simply not to go out of mana and i dont think the mana return from those were nerfed as badly.

 

Also when comparing crit to haste we are going from a stat that is mana positive (even if its not much) to a stat that is mana negative so when making the comparison it would make sense to add those 2 effects to get a total difference in mana and I would guess that this would not be a marginal difference but then i dont really have numbers behind it.

 

It would not make sense to even try to somehow fit the spirit difference into your stat comparison it would however make sense to have a different set of comparison that looks only at the mana efficiency part and then we could look at the 2 sets of data and determine for ourselves what is best.

 

Do you know if anyone has done this type of comparison? Also, is there any noticeable difference in the scaling of each spell compared to the total rotation (are there cases where stat x is best for spell y but stat x is  NOT best for spell w)?

 

 

and btw, thanks for putting all that time into the calculations, really appreciate it!

[Pandacho 286]

Well, as you already know, I always try to minimize my Spirit. In my opinion, knowing exactly what are you doing and when, can help you to trade big enough amount of Spirit for throughput stats.

 

In my current Mr.Robot weights Spirit does not present at all. I have two Spirit trinkets (615 and 630 ilvl), current Spirit is around 1200 (logged off in ele gear and can't change it from work) and ilvl was 626 yesterday. Rest of the weights - according to Stoove's table, but honestly, meanwhile I just swapped my low ilvl items with whatever I got.

 

I was struggling really hard with mana issues in the first Heroic dungeons, having mana breaks after every pack. Until understood that my healing style has to be changed drastically. Not that I didn't know that, but to stop myself from heals spamming and trying to keep the group and tank always topped - it almost gave me a heart attack

So, the current heal style named "Zen - heal slowly" - ESh on tank 100% uptime, HoTs and HST while people are above 50%, GH on those who dropped under 50% to get them to 70-75% HP. No topping! HTT when all the group / raid dropped below 50%. SLT instead of healing if tank or one other person takes constant heavy damage in HC dungeon. And so on.

With this style I don't need mana breaks anymore - still without prioritising Spirit in any way. 

 

Yesterday we did full guild MC run (not LFR) - had 10 healers in composition: 3 resto shamans, 3 holy priests, 1 disc, 2 resto druids, 1 holy paladin. Through all the run I remained at first place in Healing Done with about 17.5K hps. So this new zen style works at least for me

Edited November 24, 2014 by Pandacho

[Stoove 347]

It would not make sense to even try to somehow fit the spirit difference into your stat comparison it would however make sense to have a different set of comparison that looks only at the mana efficiency part and then we could look at the 2 sets of data and determine for ourselves what is best.

 

Do you know if anyone has done this type of comparison? Also, is there any noticeable difference in the scaling of each spell compared to the total rotation (are there cases where stat x is best for spell y but stat x is  NOT best for spell w)?

 

So, one of the main advantages of the calculations which I presented above is that they are a unified and consistent framework for looking at the relative merits of each stat. I would really like to extend this benefit to mana considerations, even if that sacrifices a little bit of precision. Provided that the correction is done in a consistent way for all stats, I see no reason why it shouldn't be a perfectly valid result.

 

Actually, "making the decision yourself" is another phrase for "applying a personal weighting", which could be done in my framework by changing a number between 0 and 1, where 0 represents "no mana consideration" and 1 represents "only mana consideration", or something like that. I will play with that idea, too.

 

The closest that I can find in terms of a spell-by-spell assessment isn't actually stat-weight related. Hamlet's "Healing Theory" parts 9 and 10 look at heal values and heal efficiencies on a spell-by-spell basis (part 10 goes into rotational questions), but there is no assessment of the stats there. It's a nice piece of work, but I don't think that it has a whole lot of long-term perspective which I'd like for looking at stat weightings.

 

 

and btw, thanks for putting all that time into the calculations, really appreciate it!

 

 

Always a pleasure! :)

[domeblade 0]

Thanks for the answer Stoove I really appreciate the quick reply

[ValiumMm 0]

Hey Stoove, Thanks for all your theory crafting so far, really appreciate it!
Are you able to update it with versatallity. So far ive read its pretty bad but would like to actually see how it fairs.
Also would versatallity scale a lot better with mastery builds?

[Stoove 347]

Hey Stoove, Thanks for all your theory crafting so far, really appreciate it!

Are you able to update it with versatallity. So far ive read its pretty bad but would like to actually see how it fairs.

Also would versatallity scale a lot better with mastery builds?

 

Always a pleasure.

 

Versatility inclusion is on my To Do list, but we do know it's very low on the prio list. When I started the theorycraft above, I wasn't interested in Versatility at all. I'm still not really interested in it, but for the sake of completeness I will update when I can. It's not a matter of pressing a button or two, though, so it might be a few weeks. I'm going to aim to include it when I do my Mana Concerns update to the same theorycraft.

 

As for the last question - in short, no. One of the things I showed in the original article was that stats scale more or less independently of each other.

[Cheze 6]

versatility is a % scalar, so it will improve relative to everything else as our ilvl improves (and we get more spellpower and other stats), similar to mastery for resto druids in MoP.  Whether it will do so by enough to actually be something we actively want is an open question.  I think it's probably not as bad as I've seen it made out to be; 'worst' stats tend to get sandbagged in general regardless of how much worse they actually are.

 

I also have to say, I'm kinda re-evaluating the decision to stack mastery based on a week of normal/heroic raiding.  Per RestoShamanStats my heal targets are around 70% life (average) even on high-damage fights like Butcher, and mastery is not that far ahead of crit in terms of throughput.  I also seem to be the new president of Low Mana Club, so I might change around to a crit/haste build for next week and see if that feels better.