Elitist Jerks Power Word: Shield & Recount

 03/31/09, 1:28 PM #1 inorlive Glass Joe   inorlive Undead Priest   Draenor Power Word: Shield & Recount Recount is one of the best addon written for Warcraft, however it does have some weaknesses. The most relevant one for my Priest is the inability to track damaged reduced from Power Word: Shield (PWS). Unlike many other Priest healers that I know, I like using PWS as a main part of my rotation. I am both speced (discipline talent) and glyphed for it. Unbuffed PWS absorbs 2564 damage for 888 mana, as a reference Flash Heal (FH) heals for 1896-2203 costing 695 mana. Some random thoughts: * Casting PWS absorbs the first 2500 damage for each hit. * PWS last for 30 seconds, so this can be a significant amount of damage absorption depending on the number of mobs and their attack speed. * PWS is a great compliment to other buffs and all forms of healing. * If there are two priests in a group, it is probably not efficient for both to use PWS in their rotation. * As a negative, it doesn't crit. We all know people who are Recount Junkies staring at the DPS or Healing meters. My view is best expressed by a fellow healer, "healers should not look at Recount, but it is a great tool for DPS". However, after being the lowest healer on the Recount meter, and having been told such, I felt compelled to do some analysis to help defend the Priests who rely on PWS. Methodology: I use Glyph of Power Word: Shield. It heals for 20% of all absorbed damage. Since we know the amount healed, we can determine the damage absorbed. I took screenshots of both Healing and Overhealing from Recount after running two 5 man heroic instances (see below - top Instance) . As you can see, the Glyph healed for 30523 and overhealed for 16048, or a total of 46571. Since this is 20% of the damage absorbed, PWS absorbed 232855 of damage. The total of the healing was 980489, so healing + damage absorbed was 1213344, or 23.7% more then the healing meters indicate. I ran a second instance and got an identical ratio (Bottom Instance) For those of you who insist on comparing healers in a raid to one another, aside from shame on you, I provide the following analysis (see spreadsheet below). Lets assume there are 5 healers in a raid and Recount shows them to have healed 23%, 21%, 20%, 19%, and the Priest with 17%. If you top up the Priest by 24%, then readjust the numbers to total 100%, you get that the Priest should display as 2nd on the charts. I have not found many posts on this topic, but the few I have found suggest that the healing from a Priest that uses PWS is understated from between 20-35% depending on the mob/bosses. This is supported by my analysis. Reader Feedback. I have posted this on Elitist Jerks, because I have found the analysis on this site to be superior to all other reference sits. Please comment on this post, I welcome and encourage your views. * Do you agree with the analysis ? * What are your thoughts about using PWS as a primary part of your rotation ? Other Related Reading: RezTheWeak: New Recount MMO-Champion Article Tales Of A Priest: DiscRecount Wowhead: Discipline Priest Recount
 03/31/09, 1:47 PM #2 TheDoctor Piston Honda   Vherna Human Priest   Arathor I am not sure that this needed to be an additional thread, and could have been easily made a part of the Disc discussion or Holy discussion, depending on which spec you are targetting your information towards. I am really unsure of the fact that you read the threads here or not because the Disc Compendium and the WotLK Healing Compendium threads both have a reasonable amount of applicable discussion. In fact methods for tracking "mitigation" have been discussed as well as the value of mitigation in relation to total/effective/over healing. So far as your analysis of PW:S I don't see any major issues. Though it would be important to note that in actuality the Glyph of PW:S heals for less than 20% of the PW:S amount because the calculation isn't so straight forward. It also isn't indicitive of the amount absorbed so much as the amount applied because the glyph's heal/overheal amounts are calculated at the application of the shield not during its consumption. Also, if you are going to look at this from the aspect of Discipline healing then DA should be incorporated. From the Disc perspective PW:S should be used often (see Disc thread). For Holy it is not quite as good, due to few talents to improve its performance.
03/31/09, 2:12 PM   #3
Squeakster
Piston Honda

Dwarf Priest

Eitrigg
Like TheDoctor said, this has been discussed several times in the Discipline thread. Unless Blizzard changes the way the combat log works to also list the source of tha shield that is being used, it will be impossible to determine the exact amount of damage absorption a priest is doing. A while back Promethia did a pretty in depth analysis of a combat log and was able to reliably determine damage absorbed in most cases but there were still some events that were ambiguous.

Another, maybe simpler, way to measure the usage of absorb effects by a Discipline priest is by analyzing Rapture returns. I posted this analysis in the Discipline thread, but will link it here.

 Originally Posted by Squeakster Until 3.1 hits you can get a pretty good idea of effective healing + absorption from the Rapture returns. Take this WWS from my guild's recent naxx/sarth clear for example It shows me as having done about 4.36 M effective healing over all the boss fights. My Rapture returns over those fights was 364446. Using the equation from the first post of this thread, (ManaReturn)*11460/(0.025*MaxMana) = Effective Healing+Absorption That means I did (364446*11460)/(0.025*25000) = 6.68 million effective healing+absoption from the Rapture spells. This is not exact however, because any portion of a heal that was effective past 11460 healing is not counted (so a Greater Heal crit that was mostly effective), so 6.68 million is an underestimation. That doesn't include heals other than Penance, Gheal and Flash Heal, so those have to be added manually. PoM did 1,009,974 healing and was 76% effective, so it did 0.75*1009974 = 757481 PoH did 603558 at 50% effectiveness, so 0.5*603558 = 301779 Renew did 284253 at 91% effectiveness, so 0.91*284253 = 258670 Glyph of PWS did 190170 at 38% effectiveness, so 0.38*190170 = 72265 Binding Heal did 51087 at 43% effectiveness, so 0.43*51087 = 21967 Glyph PoH did 49156 at 88% effectiveness, so 0.88*49156 = 43257 Divine Hymn did 28452 at 95% effectiveness, so 0.95*28452 = 27029 Add up those numbers and the number derived from the Rapture returns and you get 8.16 million effective healing+absorption, rather than the 4.36 million healing WWS shows. That means 46.6% of my healing was done through shields, which sounds a little high but I don't believe I've made any math errors. The 8.16 million is actually a low-end estimate too, because any Gheal crit that has more than 11460 effective healing isn't being counted fully.

 03/31/09, 2:16 PM #4 RootBreaker Piston Honda   Rootbreaker Troll Priest   Detheroc First of all, no one's shields are absorbing 2500 or flash heals hitting for 2k. Those numbers are pre-spellpower, and the spells have different spellpower coefficients. 20-35% is a pretty wide range, but I suspect that it isn't wide enough. While the enemies you fight certainly significantly affect the amount of your healing that's power word shield absorbs, the way that power word: shield is used is much more important. Are you just casting it on the tank or spreading it around the raid? Are you casting it every time its 4 second cooldown ends, every time weakened soul comes off the tank, or are you using it more as an emergency heal or a buffer if you have nothing else to do? What fights does your guild care about your recount position? Something like Noth can probably be solo-healed even on 25 man, so if no one drops below 90% health ever, you'd probably be better of smiting or focusing on cripple dispells than spamming unnecessary heals into your tank. I would ignore healing recount for fights like this. Patchwerk is a fight where spamming overheals into your hateful tank is a good use of your time. A recount analysis of who's doing well on Patchwerk should probably include overhealing. If you only look at effective healing, and take credit for all your power word shields, than your kind of cheating, since it's impossible for the shield to overheal (unless you cast it on someone who doesn't get hit for the duration). A heal that lands 0.05 seconds after an effective heal is just as effective at saving the life of the tank it hit, even if recount gives it no credit. On sapphiron, it's certainly reasonable to use effective healing to judge how much each healer is contributing, and also reasonable to use your method to determine how much damage your shields are absorbing. On 3-drake sarth, where there are more unique healing assignments than the average fight, your healing assignment will have a lot more of an effect on the healing meters than how good of a healer you are. As far as using power word: shield as a holy priest goes: -Only if you're raiding without a discipline priest. Their shields are much more powerful and important. -It's generally inefficient and low throughput compared to other heals -There are certain places where it's useful like Sartharion breaths - attacks that you have a warning for that can be close to one-shotting your tank
 03/31/09, 2:18 PM #5 TheDoctor Piston Honda   Vherna Human Priest   Arathor Though I wouldn't recommend spending much time moving forward with the Rapture method as it goes away once 3.1 goes Live.
 03/31/09, 2:36 PM #6 Exemplar Bald Bull   Exemplar Human Paladin   Scarlet Crusade As a (non-Priest) raid leader who also runs Recount, I must fall in the camp of your friend - don't judge healers based on meters. Or, if you absolutely must use meters, you should dive into ridiculous depths for accurate information. A quick and easy example - we had a healer top Naxx 25 charts this last week. Looked real good on the old effective heal meter. Great for them, right? Except that the top three healed targets were pets. Two hunter pets and an imp. Go back to BT days (and I'm sure in Ulduar) and I bet the chart toppers will be raid healers while the MT healers are far lower with greater overheal - it's all that AOE damage and smart heals. That doesn't suddenly make the MT healers bad players. The fact the MT survives means they're at least doing their job properly - at the very least. Healing remains a relatively thankless job and the best healers can often look the worst on charts. Anyone denigrating your location on healing meters as a Discipline Priest should understand the roles and abilities of various healers, raid assignments (MT vs raid healing), and damage sources from any given boss fight. If they cannot understand the whole picture they have no right issuing judgements. Rock: "We're sub-standard DPS. Nerf Paper, Scissors are fine." Paper: "OMG, WTF, Scissors!" Scissors: "Rock is OP and Paper are QQers. We need PvP buffs."
03/31/09, 10:36 PM   #7
Promethia
Piston Honda

Blood Elf Priest

Kilrogg
 Originally Posted by Squeakster A while back Promethia did a pretty in depth analysis of a combat log and was able to reliably determine damage absorbed in most cases but there were still some events that were ambiguous.
Yes, although I believe in over 500,000 events, only 2 were ambiguous. And "ambiguous" only meant I could not guarantee an exact value: I could still make very informed estimates based on prior events (e.g. the previous crit heal size for DA and previous shield sizes from the same priest). So I feel confident the overall error was definitely < 1%.

More recently, I've come to think that while getting actual absorbed damage data would be nice, it probably isn't completely necessary because within a specific combat encounter:
- the significant majority of shields are absorbed anyway
- not all effective healing ends up absorbing damage either

So if you limit analysis to "in combat" events (which might be tricky), then it may not be unreasonable just to add in total shield size and ignore unabsorbed shields. That method of accounting is exploitable, so you have to believe no one is deliberately throwing shields on targets who won't take damage, but otherwise it is fair enough to do that.

 Originally Posted by RootBreaker 20-35% is a pretty wide range, but I suspect that it isn't wide enough.
Yes, when I looked at that in detail, I saw the contribution of PW: shield varied very greatly depending on the encounter as well. So I've seen it range from about 10% all the way up to ~70%. 20-35% is a reasonable average across several boss encounters, BUT you cannot simply take a value and say that it is a reasonable estimate for a single encounter. It might not be. It could be wildly off even. That sucks, but that's the truth.

On the other hand, the contribution from DA shields seems more "stable" and was generally between 10-20% of effective healing done, depending on crit rate and overhealing. A reasonable estimate is:

$DA_{total}= \frac{0.45\cdot crit\%}{(1-overheal\%)(1 + 0.5\cdot crit\%)}Healing_{effective}$

 Originally Posted by Exemplar As a (non-Priest) raid leader who also runs Recount, I must fall in the camp of your friend - don't judge healers based on meters. Or, if you absolutely must use meters, you should dive into ridiculous depths for accurate information.
I think the problem is not so much about the accuracy of the information as much as it about the meaning of it all. The biggest predictor of healing done is damage taken, so the most direct meaning of healing done is that lots of healing was needed. That's kind of a "yeah, duh" except then it doesn't make much sense to judge healer by healing done at all. More healing done than usual is a better predictor that your raid is screwing up, standing in fire, etc.

If you have healing assignments doled out, then differences in healing done are also mostly a reflection of how much damage their respective targets were taking. This seems like a simple point, but it is somehow lost on the masses. If you have two hateful strike tanks on patchwerks and one tank takes 20% more damage than the other, then healer(s) on that tank will heal 20% more. Go check it out: this is reliably true. But what does it mean about healer performance?

Not much really, at least assuming no one dies. Your goal as a raid leader is to have enough healing capacity on each tank (and/or the raid, depending on the circumstance) but not too much. Unfortunately, healing done doesn't really tell you if your healers are performing near their maximum capacity. You only discover that by reducing healers and seeing if people start dying.

Fortunately, I think most players and especially most raid leaders know that if no one is dying, then healing isn't a big problem. It might not be maxed out, and you could possibly trim your number of healers, but it isn't "broken". If people die, then you have to investigate why. It might be people just taking more damage than they should, but it could also be weak healers. But healing meter data just does not help you sort that out very much.

Last edited by Promethia : 04/01/09 at 3:06 AM. Reason: typographical error

 04/01/09, 12:32 AM #8 GIJebus Cauterize with Holy Fire     Mhedic Blood Elf Priest   Mal'Ganis As stated every other post, the only way to measure a healer is if the raid doesn't die, especially when something goes wrong. Besides all recount would say is that I spend too much time spamming power word: sprint on CD.
04/01/09, 7:08 AM   #9
Xiv
Von Kaiser

Troll Priest

<Ave>
Magtheridon (EU)
 Originally Posted by GIJebus As stated every other post, the only way to measure a healer is if the raid doesn't die.
Thats ofcourse not true.

By diving into the so called 'ridiculous depths' (even though an insane amount of possibilities can indeed take effect, it isn't always that complicated), you can often easily judge a healer by analyzing meters. Break it down, check how well a player reacts and if he's using his full arsenal of spells correctly for his personal priority-assignment, but also monitor how well he does in assisting outside his personal assigment whenever allowed.

Ofcourse its not as easy as judging dps, and in most cases not as accurate. But I think you can judge a healer way better with meters, rather than replacing and see if you feel theres much difference. Tbh that last thing is even less acurate and a stupid way for judging a healer.

[e] Also on top of that, with grid you can monitor the use of any important skill like Guardian Spirit, Hand of Sacrifice etc.

04/01/09, 11:56 AM   #10
Jer (Outofmana)
Glass Joe

 Originally Posted by Exemplar Healing remains a relatively thankless job and the best healers can often look the worst on charts. Anyone denigrating your location on healing meters as a Discipline Priest should understand the roles and abilities of various healers, raid assignments (MT vs raid healing), and damage sources from any given boss fight. If they cannot understand the whole picture they have no right issuing judgements.
While part of me wants to cheer at this statement what comes to mind are the people who look bad on meters and are not doing their assigned healing. This can't be used as an excuse without good reason.

 04/01/09, 1:09 PM #11 Snagawaga Glass Joe   Maeby Human Priest   Kazzak (EU) My thoughts on recount: Don't worry about it. Recount for DPSers is cool because competing with each other can increase performance of DPSers. But healers competing on meters will likely DECREASE their performance. Healers need to figure out how to best work as a team. The rest of the healers will affect your performance, and your peformance affects everyone elses. Performance in healing means timing, foresight, reaction, mana-management, teamwork, assignements, prioritizing targets, and so on. It can not be judged by numbers. Nor by how many people are alive at the end. Example: A first kill with 90% of DPSers dead is better than a wipe at low boss health because the tank died and the boss raped the rest of the raid rather quickly.
04/09/09, 3:35 AM   #12
brandyx
Glass Joe

Tarren Mill (EU)
I'm using RecountGuessedAbsorbs. It definitely making mistakes, but it looks like i'm getteng pretty close numbers and i'm happy with them.
 Do you agree with the analysis ?
Yeah, it looks like right. But you rely on the glyph procs, which sometimes crit. It can make numbers of absorption a little wrong.
It IS main part of my rotation now. The shield is always on MT and OT (for additional crit from Renewed Hope) and sometimes on some melee DPS.
Actually in 3.1 PW:S will become even more important for me because of new Rapture and T8(4) bonus.

 My thoughts on recount: Don't worry about it. Recount for DPSers is cool because competing with each other can increase performance of DPSers. But healers competing on meters will likely DECREASE their performance.
/agree, but it's really sad to be on the last place in parse. But with absorb plugin for Recount u can always defend your raid slot. Yesterdays Naxx25:
Healing Done (6th place)

But there were some absorbs :

Last edited by brandyx : 04/09/09 at 3:47 AM.

04/10/09, 4:47 AM   #13
Promethia
Piston Honda

Blood Elf Priest

Kilrogg
 Originally Posted by brandyx I'm using RecountGuessedAbsorbs. It definitely making mistakes, but it looks like i'm getteng pretty close numbers and i'm happy with them. ... /agree, but it's really sad to be on the last place in parse. But with absorb plugin for Recount u can always defend your raid slot.
Yes, I suppose you sometimes need something to combat the prevailing ignorance out there...

I like how it separates out absorbed damage for everyone. It should be a different statistic that is not the same as healing. Even though both healing and shielding attempt to prevent death, they work through different mechanisms to attain that end.

I would be curious about the actual methods used. From reading the threads on WoWHead and WoWAce, it sounds like Elsia is attempting to estimate damage actually absorbed. The downside of trying to do that is you'll have some error that is hard to estimate, although probably fairly small. However, as I've posted elsewhere actual absorbed damage is definitely superior to healing as a measure of usefulness since it was actually used to absorbed damage. All you really know about "effective" healing is that it made a health bar bigger (after which the target may have hearthed and logged).

04/10/09, 10:04 AM   #14
Elimbras
Don Flamenco

Dwarf Priest

Eitrigg (EU)
 Originally Posted by Promethia However, as I've posted elsewhere actual absorbed damage is definitely superior to healing as a measure of usefulness since it was actually used to absorbed damage. All you really know about "effective" healing is that it made a health bar bigger (after which the target may have hearthed and logged).
I've never been really convinced with this idea. I acknowledge the theoretical difference, but in practice, our HP bars are small enough for most effective healing to be useful.
When there is not raid damage, at most 50k (one tank health) of your effective heal is not useful.
When there is RST, nearly all your healing on raid is useful, because it decreases the death probability, and you can't predict the next target.
When there is global aoe damage (Sapphiron), you could stop raid heal a few second before the kill, but that's not so important.
And keep in mind also that some targets are more "important" than other ones. I prefer direct healing on a target that is at 5k HP healer than shielding and preventing 5k damage on a melee dps at 25k HP, even if the healer takes no further damage.

04/13/09, 9:01 PM   #15
Promethia
Piston Honda

Blood Elf Priest

Kilrogg
 Originally Posted by Elimbras I've never been really convinced with this idea. I acknowledge the theoretical difference, but in practice, our HP bars are small enough for most effective healing to be useful. When there is not raid damage, at most 50k (one tank health) of your effective heal is not useful. When there is RST, nearly all your healing on raid is useful, because it decreases the death probability, and you can't predict the next target. When there is global aoe damage (Sapphiron), you could stop raid heal a few second before the kill, but that's not so important. And keep in mind also that some targets are more "important" than other ones. I prefer direct healing on a target that is at 5k HP healer than shielding and preventing 5k damage on a melee dps at 25k HP, even if the healer takes no further damage.
I pretty much agree there should hardly be any difference, especially when you're talking about tank healing where there is constant ongoing damage. The vast majority of healing will end up absorbing damage. What I'm really saying is that absorbed damage >= effective healing, not absorbed damage >> effective healing.

The odd inconsistency to me is that people will worry about unabsorbed shield and then simultaneously dismiss the idea that absorbed damage is a bit better than effective heals. I got thinking and (later) posting about this because of what I saw when I went through combat logs and tallied up how much of a shield actually absorbed damage. It turned out that the vast majority of shields were absorbed and usually right away (within a few seconds). The ones that were not absorbed were consistently at the end of combat or when mob aggro changed (e.g. a dps or healer gets hit and then subsequently a tank pulls the mob off).

That isn't really surprising. However, I also noticed that the same targets that were getting shielded and subsequently not taking any damage were also getting healed and not taking subsequent damage. That also shouldn't be shocking at all. If someone was worried enough to shield, then someone would be worried enough to heal as well.

But here's the killer question to me: Why then do we want to spend all this effort to find cases where shields do not absorb damage and discount that when we never do a similar thing for heals? If you spam shield the tank, at most the last shield wasn't useful. Same deal with heals. But with shields we want to subtract out that last shield that wasn't absorbed. With heals, we don't want to do that. No, it doesn't amount to a huge difference, but it is both a conceptual and procedural inconsistency IMO.

Of course, I completely agree healing (or shielding) a 5K HP healer >> than either action on 25 HP melee. This is perhaps the single biggest reason why statistics like healing done are so poor in evaluating healers: they count every point of healing as if they are all the same, when we know the bigger issue is judgement about who to heal and when. Both the who and the when elements are completely missing from healing done (although healing breakdowns can at least help one address the "who" part).

In fact about a year ago I wrote an application which weighted the value of heals based on how low the target's health bar was at the time the heal landed. That was an interesting venture that didn't really prove much. In a relative sense, raid healers didn't look quite as good by such an index since their targets tended to have higher relative health when healed, but the difference was slight. In the end all I showed was that in practice pretty much all healers do preferentially heal targets with lower health...

I feel dumb writing that now because I suppose I should have known that would be the outcome, and it took me a while to write that application. But oh well, now I know for sure.

Last edited by Promethia : 04/14/09 at 12:12 AM. Reason: Possibly confusing and unnecessary verbiage removed

 Elitist Jerks Power Word: Shield & Recount