PvE HPally Theorycrafting Pt.2
07/20/2011 10:01 AMPosted by WtfboomEloderung is 6/7H and has submitted plenty of logs to show the validity and performance of the build. There's nothing my 4/7H pally can show any better than he already has.
so.... no 4/7 heroic guilds on sen'jin good try though keep lying
INC: ITS MY ALT IM NOT ON SEN'JIN
I agree that fun is important but I also try to maximize my character at all times. The basic concept I am using for this low spirit build is to use the HPM benefits of Mastery and Crit to make up for the loss of Mp5 (obvious I know).
I have absolutely zero interest in gimping myself so that is really where the questioning is coming from.
Spirit and haste are still the best 2 stats for heroic 10-man Firelands, in my opinion.
From a purely numerical perspective, haste provides the best throughput out of the three non-spirit secondary stats.
- 128 haste rating provides 1% haste, which in turn provides 1% more healing over time.
- 179 crit rating provides 1% crit, which provides 1% more healing over time. Admittedly, crit has the added bonus of doing its extra healing for free - but this greatly diminished by its poor stat conversion and relative unreliability, and in my opinion, makes it a stat not worth stacking.
- 179 mastery rating provides 1.5% extra healing on all heals minus Beacon of Light and Holy Radiance. After a quick look at a bunch of my recent parses, the contribution of these two spells to my total healing output falls somewhere between 35% and 43%. This means mastery contributes slightly less than two-thirds of my heals, making its healing contribution point-for-point roughly equal to (but better than) that of crit.
So, with mastery and haste being the closest on paper, the next things I look at are some of the intangibles of 10-man healing (and this is precisely why I'm very weary of considering EJ, and especially these forums, as a source of healing recommendations; a lot of people seem to only theorycraft without taking "real-world" situations and experience into consideration): the value of reduced reaction time, and the value of potentially increased health pools via mastery bubbles.
The first thing I listed is pretty simple: obviously, when progressing through content, people are bound to make mistakes. In my experiences getting to 6/7 heroic, I've seen my raid take a ton of avoidable or reducible damage - the difference between our first few heroic kills and our kills this week is like night and day.
Having high haste is extremely important and powerful since it affords your players more room for error; people say they've reforged haste off their gear and their heal cast times increased only by a few tenths of a second - in progression content, when people are making mistakes, this can easily be the difference between life and death. Again, this is difficult to quantify, but in my opinion haste is irreplaceable simply because of this fact.
Secondly, the value of mastery bubbles increasing maximum HP. In my experience, a lot of damage goes out pretty constantly in Firelands encounters, so there's very little opportunity to stack it (except maybe pre-pull on the tank). Therefore, I'm not convinced that it's fair to give mastery extra value because of its EHP possibilities, simply because individual bubbles tend to be pretty small, and because they're almost constantly being used (with very little "overheal" - bubbles falling off due to 15 seconds of no damage taken). While it's good that mastery bubbles are almost constantly getting used (and thus adds legitimacy to the consideration of stacking mastery) the nature of damage in 10-mans this tier invalidates the argument that mastery provides any notable EHP increases.
So, because of the points I listed above, I consider mastery against haste purely by how much healing output is added per rating - that is, not giving bonus consideration to mastery due to EHP increases. And, since they're somewhat similar point-for-point, with haste already being a little better, I'll take haste (since it provides faster reaction times) in a heartbeat for the reasons I listed above.
I know a lot of the points I make, and the evidence I provide to back them up, are mostly anecdotal, but healing is definitely the hardest-to-quantify role. I feel when it comes to healing, real-world experience almost always outweighs theorycrafting, and hopefully my progression gives credence to my opinions.
TL; DR: Since players are bound to make mistakes and play sloppily during heroic progression, haste gives you the reaction time necessary to make a mistake non-fatal, and spirit provides the regeneration necessary to ensure you're not completely screwed when you're forced to cast inefficient spells.
Spirit and haste afford you and your raid the most room for error.
How I arrive at my conclusions
First, I'd like to state that WoW is a math game. Everything can be theorycrafted and everything can be modelled accurately, with one exception, player reactions. And even then, there are some guidelines we can consider. The "gut" feelings and the "art" of healing is people being lazy and people with bad models. That said, the statistics can get pretty complicated and determining the right models can be difficult.
Difficult, not impossible.
If you do not believe this, then I can do nothing for you. You're also wrong, but you wouldn't believe me if I tried to explain why.
I'm not going to discuss HPM vs. HPS concerns at this time since quite frankly, I'm not going OOM. But the simple answer (though with admittedly somewhat weak math) is to stack spirit without sacrificing int except -10int for +20 spirit looks like an ok trade.
So since everyone and their grandmother enjoys arguing about stat weightings, let's discuss that, and then find out why none of it really matters anyways unless you're at the bleeding edge of progression.
Here's what no one is going to argue with:
11.025 Actual Int
12.125 SP w/ 10% raid bonus
4.135 mp5 from 90% replenishment uptime
In 359 gear (~8300 SP raid buffed), this is roughly a 0.0653% throughput increase.
~6.3 in combat mp5
10 Haste Rating
0.078% increased haste
10 Crit Rating
~0.0558% increased crit chance
10 Mastery Rating
Now we need to consider "real world" concerns. The easiest way to do this, is to include what I call deprecation factors.
What deprecates int throughput? Overheal. What deprecates haste throughput? Overheal. What deprecates crit throughput? Overheal.
Oh hey, isn't that pretty conveniently listed in most of our logs? So let's deprecate all the throughput bonuses by overheal.
Example. 20% overheal (lowest realistic overheal I've seen)
0.0653% -> 0.05224%
10 Haste Rating
0.078% increased haste->0.0624% throughput
10 Crit Rating
~0.0558% increased crit chance->0.04464% throughput
What deprecates mastery? First, not all abilities proc mastery. Second, shields can fall off or be capped at 30% of our max health. Shield capping should not occur outside of Baleroc type fights.
My last Rag kill. http://www.worldoflogs.com/reports/1ewcz4spwj7b63yy/details/16/?s=15062&e=15636 This is a bit weird in that I wasn't assigned to the tanks, but hey, it demonstrates the math.
Eligible Mastery Healing
<Spell> - <Useful healing> - <Overhealing>
DL - 1060489 - 914351 (46.3%)
HS - 575263 - 120869 (26.6%)
WoG - 285481 - 275385 (49.1%)
LoD - 274121 - 172330 (38.6%)
HL - 131852 - 191315 (59.2%)
LoH - 76098 - 100463 (56.9%)
FoL - 31976 - 17142 (34.9%)
Eligible Mastery Healing - 4227135
Ineligible Mastery Healing
HR - 907218 - 724469 (44.4%)
Beacon - 806804 - 1668055 (67.4%)
PotI - 284901 - 616684 (68.4%)
EJ - 31036 - 106291 (77.4%)
SoI - 4194 - 29629 (87.6%)
Ineligible Mastery Healing - 5179281
Total Healing Period - 9406416
Mastery Eligibility - 4227135/9406416 = 44.94%
0.45-0.5 Eligibility is what I usually see. 0.5 is what I tell most people to make the math easy.
Illuminated healing did: 517571
My mastery: 12.88 -> 19.32% shields.
Expected Shield Healing - 1000637
Used Shield rate - 517571/1000637 = 0.51724
Total Mastery deprecation: 0.4494*0.51724 = 0.2324
10 Mastery Rating
~.0837% shields -> 0.0195%
Note, I almost never healed the tank directly becuase the melee group was ~60 yards from the tank so I was playing games with beacon. Almost all raid healing (which mastery is god awful for).
10 Haste Rating
10 Crit Rating
10 Mastery Rating
But Lylthe! Can we really say XXX% haste/crit/shields/etc is REALLY XXX% throughput?
Sort of, but not really. We're considering XXX% increased throughput of a base case.
But the base case already includes some crit and mastery. Haste too, but since haste multiplies instead of adding like crit and mastery, we can say 5% haste from rating is actually 5% throughput.
In the case of int, it's static throughput and easy. It scales with the base case linearly. But for easy comparative math, and assuming we don't add huge chunks of int, we can pretend that it scales at ~0.0653% throughput.
In the case of crit, no, paladins have an inherent 3.3% crit chance and whatever crit comes from int. So the 0.0558% crit chance is really 0.0558%/(1+BASE CRIT) increased throughput instead of 0.0558% throughput. Lylthe has ~15% base crit. So technically, any crit rating I get should automatically divided by 1.15 to keep it fair and scale off the same base case.
The same argument applies to mastery. I did not deprecate by the starting 12% shielding we get with 0 mastery rating. Again, mastery throughput is technically scaled by 1/(1+Base Shielding), or 1.12.
Haste multiplies so it's fine.
Now for the deprecation factors.
Problems with Crit's Deprecation Factor
But Lylthe! Not all overheal is equal. A crit is likely to overheal more than a normal heal.
Yes So crit deprecation is higher than expected.
Problems with Haste's Deprecation Factor
But Lylthe! You advocate spamming Judgement as often and as hard as possible! Judgement is considered a melee special attack and consequenctly not aided by haste! The GCD off Judgement is fixed at 1.5 sec.
You're right, so if you judge once every 8 seconds, we should deprecate haste again by (1-1.5/8) or (1-1.5/N) where N is the average time between judgements.
But Lylthe! my tank was about to go caput! I had to spam heals otherwise a 2nd attack would land and kill my tank! That overheal was necessary to keep my tank alive!
You're right. This modelling assumes a triage situation and not the tank explosion situations we're seeing in Heroic modes.
In this case we should actually be simulating a boss swing timer and abilities and tank's health and avoidance and mitigation. And try to find a new deprecation factor. But I'm lazy.
Problems with Mastery's Deprecation Factor
But Lylthe! EJ healing shouldn't be counted against mastery if you're comparing to Haste.
Yeah, but it's small and I don't want to consider all the special cases because it does scale with crit and int.
But Lylthe! my tank was about to go caput! I had to build a big shield to increase the tank's "maximum health"!
Yes you're right. See above about haste.
But Lylthe! We're running into situations where triage is a load of baloney! We have burst healing situations where we can't rely on the law of large numbers to smooth out RNG from crit and we have to respond very very quickly to heavy burst damage!
Yes, so if you're spamming the tank, stacking mastery can be a good idea and if you're spamming the raid, haste is a good idea, or split things relatively evenly.
Those with poor reflexes wont be able to utilize a haste build, and so will get more mileage from mastery/crit.
Knowing ur own strong points will let you pick what stats are best for you.
I am able to keep the gcd always triggered. The Castbars addon helps a lot with this btw. So i go haste/spirit/crit/mastery. If I found myself with downtime between the end of gcds and picking a target for my next heal, haste wouldnt be doing squat for me though.
Let's be generous and pretend you have 3000 secondary stat points to shuffle around from Tier/OffTier/reforging. That's pretty generous.
Best Case Deprecation factors for Mastery
Crit Deprecation: 0.4*(1/(1+.15)) = 0.348
Haste Deprecation: 0.4*(1-1.5/8) Judge perfectly on CD= 0.28
Mastery Deprecation: 0.5 Eligibility *1 Shield Usage*(1/1.12) = 0.446
Stack Pure Haste: 6.556% throughput bonus
Stack Pure Crit: 5.82589% throughput bonus
Stack Pure Mastery: 11.1998% Throughput Bonus
Balance all 3 1:1:1: 1.02185 (haste) * 1.0194 (crit) * 1.0373 (Mastery) = 1.0806 -> 8.06%
Best Case Deprecation factors for Haste:
Crit Deprecation: 0.8*(1/(1+.15)) = 0.696
Haste Deprecation: 0.8*(1-1.5/20) Not judging often = 0.74
Mastery Deprecation: 0.5 Eligibility *0.5 Shield Usage*(1/1.12) = 0.223
Stack Pure Haste: 17.33%
Stack Pure Crit: 11.65%
Stack Pure Mastery: 5.5998%
Balance all 3 1:1:1: 1.05776 (Haste) * 1.0388 (Crit) * 1.0186 (mastery) = 1.1193%
More realistic Deprecation Factors:
Crit Deprecation: 0.5*(1/(1+.15)) = 0.435
Haste Deprecation: 0.5*(1-1.5/15) How often I judged on Rag = 0.45
Mastery Deprecation: 0.5 Eligibility *0.75 Shield Usage*1/1.12 = 0.3348
Stack Pure Haste: 10.537%
Stack Pure Crit: 7.282%
Stack Pure Mastery: 8.4545%
Balance all 3 1:1:1: 1.03512 * 1.02427 * 1.02818 = 9.01%
So worst case, we're talking a difference of 10% at the extreme situations. More realistically, we're talking deltas of 5%. If 5% is making or breaking your raid, you are raiding at the bleeding edge, or you have bigger problems.
Also, 3000 secondary points to shuffle around is a lot. More realistically you'll have 1000-2000 points to move around.
For mere mortals, balancing stats may not be perfectly optimal, but it has a strong showing in all situations.
07/21/2011 07:58 AMPosted by AltruisticMy question has always been is spirit really the superior stat choice over crit or mastery if we are talking regen/efficiency.
Complicated question with a somewhat convoluted answer. I need to rework the math with my new analysis of deprecation factors, but for the sake of timely answer before I leave town.
From my spreadsheet on sticky: https://spreadsheets.google.com/spreadsheet/ccc?key=0AhYqR2TzgIgadEtkM245Z05VSWtSUGl4VS1JOFFzaWc&hl=en_US#gid=1
1. I am considering only a straight DL or HL spam situation.
2. Ignoring Overheal and other deprecation factors
3. Did not consider mastery at the time because I thought crit was better. Not necessarily true with my recent work.
The situation I'm considering is a paladin spamming nothing but HL and DL, at some initial ratio. Say 1 HL->1DL->1HL->1DL for a 1:1 ratio. This ratio can scale anywhere from 0:1 to 1:0.
Now add 10 of haste or 10 of crit. Assuming the initial ratio was sufficient HPS, then you could downgrade DL->HL and find an equivalent savings in mana from spirit.
Regardless of the initial HL:DL ratio, 10 haste is always equivalent to about 1.2 spirit. Translation, if you have mana problems, haste is a crappy solution, but we knew that.
<Inital HL:DL Ratio> - <Equivalent Spirit of 10 Crit>
.1:.9 - 11.55
.2:.8 - 10.79
.3:.7 - 10.01
.4:.6 - 9.2
.5:.5 - 8.5
.6:.4 - 7.7
.7:.3 - 6.9
.8:.2 - 6.2
.9:.1 - 5.4
So from here, the more DL you cast, the better crit is for longevity. It looks like once you start casting 70% or more DL, then crit technically beats spirit for regen. HOWEVER, once you consider as little as 20% overheal, crit never beats spirit for regen.
You could do similiar math for mastery. For SINGLE TARGET ONLY, I would not be surprised if mastery beats spirit for longevity. Of course, you would need a minimum amount of spirit to sustain the initial ratio indefinitely.
If however, you find you're never going OOM, you can upgrade some HL->DL. In which case 10 spirit is equal to about 5.16 crit and 3.69 haste.
Haste and Crit are worse than spirit for longevity. Spirit is worse than haste and crit for throughput. In special circumstances (actually the one's I'm modelling here) mastery can be better than spirit for longevity. SPECIAL CIRCUMSTANCES
Anyways, please understand these are very specific "lab" situations. In real raids, it's much more complicated and I haven't figured out a great model. But the takeaway is to understand your limiting factor and gear accordingly: burst HPS, sustained HPS, or longevity.
Please report any Code of Conduct violations, including:
Threats of violence. We take these seriously and will alert the proper authorities.
Posts containing personal information about other players. This includes physical addresses, e-mail addresses, phone numbers, and inappropriate photos and/or videos.
Harassing or discriminatory language. This will not be tolerated.