Resto gemming question

[lynx 19]

I'm new to resto and don't understand something, why all restos I see gem with pure int? Aren't hybrids better?

If I'm going for the crit build, by the simple logic that 2crit > 1int the 80int+160crit hybrid should be better than 160int.

Or my assumption that 2crit > 1int is wrong?

[Drafty53 17]

It may just be the resto's YOU see or are inspecting.  I gem hybrid on my resto as do most that I see.  The two "consultants" if you will, to the resto class guide here at IcyVeins, Zetian and Sonie (two of the best resto shamans in the world by the way) both gem hyrbid or even straight secondary stats with yellow crit gems when they can.   You cannot always just go off a few shamans you inspect.  

 

Crit is currently our strongest stat after the most attainable haste breakpoint (very important nowadays with how good rushing stream and HR are for us as well as the LMG proc).  Mastery can still be very strong in progression environments as opposed to the rng that can occur with crit, but the fact that crit also doubles as a mana regen stat gives it that extra boost this tier.  It allows us to not carry absurd spirit levels, further buffing SP and secondaries.  The other slight problem with mastery, even in progression environments is that a lot of other healer classes have some sort of absorb mechanic or strong HoT mechanics.  By the time you get your GHW heal off on a person with a rejuv or soothing mists, your mastery buff to your heal as already diminished.  Same logic goes to playing with disc priests...shields prevent a lot of players from getting low enough for our mastery to really pack the punch it is intended to.

 

Now to the actual question posed....The one conventional thought with int gems as a healer is that often times, healers secondary stats end up adding "over-healing" (Whether intended to or not).  In a lot of cases 1 heal is enough for the target you are healing.  Unlike dps classes who do not "over-dps", heals can often times be "wasted".  So the argument becomes would you rather your heals be stronger on average (the first cast) or have them randomly be double as strong, often times resulting in additional over-heals since players hp pools are MUCH smaller than boss hp pools.  

 

Even with that said, Most people prefer the hybrids because over-healing is going to happen regardless and crit will help off-set a lot of the mastery resto shamans are currently favoring to reforge out of...Add in the fact that crit's return a portion of some mana and its a very strong stat right now.

 

IMO I would say go with the hybrids...just one mans opinion though...Hopefully Stoove and some other's will chime in.

Edited January 6, 2014 by Drafty53

[Stoove 347]

Well I think my response comes all down to this;

 

2crit > 1int

 

I, personally, prefer taking Intellect. When I set up my stat weights in Ask Mr Robot, I always have me secondaries below 0.5 (usually 0.45) so that I don't gem for them. Largely, that's because Intellect gives Spell Power, and as Drafty has pointed out that's more consistent healing than the secondaries.

 

I have to say that I usually feel a negative impact when I'm lower on Intelllect, so I gem for it where possible. I'm actually not sure in terms of numbers whether 2 Crit really does give more numbers than 1 Intellect. I suppose I shall experiment with some theory on that one.

 

The same goes for Trinkets - I actually prefer static Intellect to procs because the healing is more reliable and more constant.

 

Also I'm old fashioned, and I never really got out of the Cataclysm mindset of "all the Intellect you can eat".

[lynx 19]

Yes, I indeed come from the world of dps where hybrids are much better than pures simply because "over dps" is rarely a consideration at all (definitely meaningless against the boss). But now I see what you mean with consistent healing, and it does make sense.

I'll try to put my dps theorycrafting skills to healing to make better sense of it all, as well as try out different variations in game to see which one works better.

 

Thanks for the replies!

[Stoove 347]

It shouldn't take much effort to produce the results, i's just a matter of calculating it for all the different spells (because spell power and Crit give different amounts of benefit to different spells).

[lynx 19]

I made a small check and it seems that for balanced gear hybrids are a little bit better for expected output.

 

If a spell's effect is computed as: A+B*SP

Fully buffed intellect without gems in red slots is I.

Crit rating from gear without gems in red slots is C.

Weapon's SP is WSP.

A total of M red slots.

 

Then the optimal amount of pure gems is (before rounding and clamping to [0,M]):

229.587 - 0.00541126*A/B + 0.00356861*C - 0.00456628*I + 0.141956*M - 0.00595238*WSP

From several tests I've run for CH and HR this turns out negative for higher ilvls and greater than M for lower ilvls. For mid ilvls it usually sits in [0,M].

With that said, the raw effect of the expected heal didn't change by more than 1% when going only for pure or hybrids, hence the appeal of pure gems is apparent.

[Stoove 347]

I'm not that used to reading the format you used there Lynx, so I can't check your maths myself. It helps to show how you derived the formula you used, since that is just a string of numbers

 

I found it easier to calculate whether the gain in healing for some spells was larger when Crit increased by 100 than when Intellect increased by 50. I used some approximate figures (base 25% crit and 45k spellpower) and calculated the spreadsheet you can find here.

 

The result was that on no spells did an increase of 100 Crit increase the healing by more than the effect of 50 Intellect. From this, we know that at typical gear levels 1 Int > 2 Crit and it is therefore probably better to gem for primaries unless you're trying to hi a breakpoint.

 

Also: WOO! 800th post! Now everyone can see just how sad I am :)

[lynx 19]

Sure. I threw away global resto coefficients because they're just global, and mastery is just constant here so it doesn't matter as well.

 

Assumes that all red slots are empty, and INT is fully raid buffed.

CH without crit:

H = 7590 + (INT + WSP*1.1 + (RED*160 + ORANGE*80)*1.05*1.1)*0.6876

 

Total crit probability, where C is the crit rating from gear:

PrC = 0.022 + (INT + (RED*160 + ORANGE*80)*1.05)/253400 + (C + ORANGE*160)/60000 + 0.05

 

So the expected heal (without globals and mastery, which are constants so they don't matter) is:

E(H) = H*(1 + PrC)

 

Since RED + ORANGE is a constant, call it M. Replace ORANGE with M - RED, and derive by RED. Equate to zero and solve for RED. That's basically what's written in the post above.

Edited January 6, 2014 by lynx

[Stoove 347]

Since RED + ORANGE is a constant, call it M. Replace ORANGE with M - RED, and derive by RED. Equate to zero and solve for RED. That's basically what's written in the post above.

 

Ah. Now I understand how you got there!

 

If I do what you explain there, and with the assumption M approximates 10, and the spellpower coefficient approximates 0.1*7, I find that the number R is positive given reasonable assumptions for intellect, WSP, and so on. That coincides with my calculations in the spreadsheet given above.

 

I updated my calculation (same link) to also use Sonie's stats (straight from his armoury) as an example of high ilvl. I came to the same conclusion; at Sonie's stat level, increasing Crit by 100 has less of an effect than increasing Intellect by 50.

 

Actually, I'm only counting Spell Power in my Intellect calculations because the contribution to Crit from Intellect is fairly small. Counting the Crit gain from Intellect would just drive up the value of Intellect more.

 

So I'm still reasonably well convinced that 1 Int > 2 Crit, even in high ilvl cases.

 

The only reason I can think that I would take Crit in gem slots is to ensure that I get lots of Ancestral Awakening procs, which are undoubtedly useful in a 25man environment. Although those are included in my calculations, the splash healing might outweigh the raw gain.

 

Update: Corrected my coefficient for Ancestral Awakening to 2.6 from 2.8. Silly me, but it doesn't change the results.

[Stoove 347]

Incidentally, I am genuinely fascinated by this topic.

[lynx 19]

I found an error in my calculation of raw heal, it should be:
H = 7590 + ( (INT - 10 + 2873 + WSP)*1.1 + (RED*160 + ORANGE*80)*1.05*1.1)*0.6876

I just forgot about earthliving, -10 for SP, and 1.1 buff coefficient on it all.

 

I totally forgot about AA, but this is me being noob in resto 

 

After adding it:

E(H) = H*(1 + 1.6*PrC)

Which further increases the value of crit.

 

Now, to the point. I agree that with the non constrained calculation 1int > 2crit, however since all calculations must be constrained to currently equipped gear (we can't just increase int or crit out of the blue  ) it must be taken into consideration.

With Sonie's stats I get a negative R for CH. His gemming is a bit trickier though, as he's ignoring spirit bonuses and gemming with pure crits in red sockets and also pure crit in prismatic sockets. This can be taken into account but will make all these calculations more difficult, so I chose to omit it for the sake of simplicity.

[Drafty53 17]

Two things:

 

Without getting into the math 10m and 25m environments do make a difference.  25m naturally favors more crit with more healers to snipe heals and diminish mastery further than  in 10mans...and you also usually have more opportunities for crit heals.

 

I know the mana return with a crit heal cannot be factored into the math for healing output but that is an important component in the argument for 2 crit > 1 int as it allows us to keep lower spirit levels and maintain more secondary stats...As a pure output calculation I can see 1 int being favored over 2 crit so I wonder if there is a way to factor that as an independent benefit to the equation that we could add on at the end.  

Edited January 7, 2014 by Drafty53

[lynx 19]

I know the mana return with a crit heal cannot be factored into the math for healing output but that is an important component in the argument for 2 crit > 1 int as it allows us to keep lower spirit levels and maintain more secondary stats...As a pure output calculation I can see 1 int being favored over 2 crit so I wonder if there is a way to factor that as an independent benefit to the equation that we could add on at the end.

In order to take resurgence into account we can introduce a new metric, mana efficiency. Namely: Heal/(Mana used).

Let's examine CH.

The mana used for CH is 13500, so we get:

E(H/ManaUsed) = H/13500*(1-PrC) + H/(13500-2946)*2.6*PrC = H/13500 * (1 + 5850/1759 * PrC) ~= H/13500 * (1 + 3.326*PrC)

So here you can get the feel of how mana efficiency scales with crit. Though this is not 100% exact reflection of what's happening, because as I wrote earlier trading int for more crit means H will become smaller.

To get better understanding you can repeat the steps from before, just for mana efficiency. You'll see that hybrids (int+crit) are becoming better than pure ints by larger margin with this metric for heals that trigger resurgence.

Edited January 7, 2014 by lynx

[Stoove 347]

Ah, I appear to have made a laughable error in calculation on my spreadsheet. When I calculated with +50 SP I also calculated at the same time for +100 Crit. So I was comparing +50SP + 100 Crit to just +100 Crit. How embarrassing.

 

So I agree with lynx here about gemming for secondaries. The only spells that I could find which were exceptions to this rule were;

• Healing Surge (with Tidal Waves)
• Healing Rain
• Riptide
• Healing Stream Totem

However, this was solely in the case of exact gemming (+160 Crit and +80 SP compared to +160 SP). When I compared raw increases (+160 Crit to +80 SP) I found that Crit always won.

 

I went further and looked at the case of choosing between a Hybrid and +60 SP socket bonus, and a pure +320 Crit gem. The results are the same: Crit outperforms Intellect/SP except for the spells I stated.

 

HOWEVER

 

I look at my logs from my raid's last Garrosh kill, and I see that the spells I listed above did over 50% of my healing on their own. Even if I look at Dark Shaman or other fights where Healing Rain tends to drop down the list, the healing done by spells which benefit more from Intellect was significantly more than otherwise.

[Stoove 347]

I look at my logs from my raid's last Garrosh kill, and I see that the spells I listed above did over 50% of my healing on their own. Even if I look at Dark Shaman or other fights where Healing Rain tends to drop down the list, the healing done by spells which benefit more from Intellect was significantly more than otherwise.

 

When I took this into account, I found that gemming for Crit using hybrids, and even gemming for Crit using pure gems, overall reduced the amount of healing I did. Calculations all (still) in the link I posted yesterday.

 

On balance, I don't think I'll be gemming for secondaries any time soon.

[lynx 19]

Thanks for checking Stoove, I was thinking along the same lines. 

Since right now I'm running with about 10k spirit and still without the LMG, I think I'll stick more to crit so that mana won't become an issue (sometimes I'm struggling with management as it is). Once I'll get the LMG I think I'll balance between pure and hybrids instead of going "all in" on one kind.

 

Overall I believe I got my answer, thanks a lot for the help!

[Stoove 347]

Thanks for checking Stoove, I was thinking along the same lines. 

Since right now I'm running with about 10k spirit and still without the LMG, I think I'll stick more to crit so that mana won't become an issue (sometimes I'm struggling with management as it is). Once I'll get the LMG I think I'll balance between pure and hybrids instead of going "all in" on one kind.

 

Overall I believe I got my answer, thanks a lot for the help!

 

Good stuff. If you're willing, I'd like to explore this a little more so we can give less mathematically inclined people some solid advice. I asked a healer and theorycrafter I respect (Dayani, who writes the blog Healiocentric) what their logic was for running mostly Hybrids - expect an answer tomorrow.

[lynx 19]

Sure, sounds great. I'm also interested in knowing what other theorycrafters have to say about this as I feel this topic wasn't explored as thoroughly as for dps.

[Stoove 347]

Well I had a bit of a google, and I checked some of the more popular blogs out there (Healiocentric and Life in Group 5 in particular) and there's no mention of gemming strategies as far as I can see!

[Drafty53 17]

Sure, sounds great. I'm also interested in knowing what other theorycrafters have to say about this as I feel this topic wasn't explored as thoroughly as for dps.

Exactly my thoughts. So much time was put into the dps secondary strength versus SP/AP strength when it came to gemming, and none really to the effect of secondaries on healing versus the traditional "get as much int from gems as possible". Obviously as history shows, it may not result in any sort of massive shift in numbers as performance is so closely tied to the players ability, but I love that this discussion can now happen for healers. I still see most healers in top end guilds use hybrid or yellows when it comes to gemming but I could never find any articles stating why they were doing that or why. It's very possible that a lot of them followed the theorycrafting advice of their dps brethren and just assumed secondaries were twice as strong for healers as well. I know I favored hybrids on my resto sham initially because of how I gem my main warlock.

Granted it's a bit tougher on healers since we do not have an adequate simcraft type system (and yes I know simcraft is far from perfect but it does allow dps classes to have some reference for stat weights and what different reforging can possibly do)...Heals also does not have a dummy to test on as the role is so much different.  Really looking forward to see where this discussion goes and what results people may have if they were to try out the two different sets.

Edited January 7, 2014 by Drafty53

[Stoove 347]

Really looking forward to see where this discussion goes and what results people may have if they were to try out the two different sets.

 

The thought occurs that a reasonably sophisticated parsing program and  some statistical analysis might help with the latter part of that. It's just a matter of getting some people together who share a common programming language and some skill with statistics.

 

Simple, right? Riiiight?

[Stoove 347]

We have an update on this matter!

 

Dayani, who runs the blog Healiocentric, took some time out to give me a thorough explanation of the Hybrid strategy. You can find the full explanation here (via Storify).

 

I checked the maths and agree with the result, and I'll note a few things we missed which made the analysis easier;

• Expressing the healing increase as a % early on saved effort
• The "Rule of 11000" would have simplified the analysis that Lynx made

Also my spreadsheet must be wrong somewhere. I shall find out where!

 

Finally, the gains are universal - Healing Stream Totem and Healing Rain should have the same properties.

[lynx 19]

Thanks a lot for checking this even further! So far this has proven quite insightful.

 

Indeed, Dayani's points would've made the analysis easier. I'll follow the same pattern in the future.
 
I have a small question though, Dayani wrote that CH doesn't AA. Is it indeed so? I noticed that in your spreadsheet CH has 2.6 crit coefficient, so I assumed that AA works for it too.

[Raff 0]

I'm not familiar with the rule of 11,000; is there somewhere I can get more information?

[Stoove 347]

I'm not familiar with the rule of 11,000; is there somewhere I can get more information?

 

I think it's a general observation about the ratio of base heal to spellpower coefficient. I'll ask more/investigate later.

 

 

I have a small question though, Dayani wrote that CH doesn't AA. Is it indeed so? I noticed that in your spreadsheet CH has 2.6 crit coefficient, so I assumed that AA works for it too.

 

I would trust Dayani to know this, I assumed it did but haven't checked. I bow to her knowledge on that matter, will update spreadsheet.