EF & You?

If you’ve been paying attention to the PTR patch notes, you’ve probably noticed that there’s a big change coming in 5.4.  I’m not talking about the change to Grand Crusader, which is mostly irrelevant to how we gear and spec.  I’m talking about the 30% nerf to Sacred Shield and the 40% buff to Eternal Flame.

The nerf to Sacred Shield is understandable – it’s been head and shoulders above our other level 45 talents, and it scales very well with the extreme amounts of haste we’ve been stacking.  Blizzard probably didn’t anticipate the level of haste-stacking we’d be doing (though to be fair, neither did most of the community), which explains why they have to dial it back a bit to keep it in line.  The buff to Eternal Flame was less expected, but makes the tier much more interesting again.  We’ve mostly ignored Eternal Flame in previous patches because it couldn’t keep up with the raw mitigation of Sacred Shield.

What many people don’t realize that Eternal Flame has interacted with Bastion of Glory (BoG) since around patch 5.1 via an undocumented change.  The 100% buff when self-cast in 5.2 was another fairly large buff in addition to the BoG interaction.  In patch 5.2 and 5.3, Eternal Flame’s HoT produced a little more healing than Sacred Shield provided in absorption.  But as I said at the time, the healing would have to be significantly larger than the absorb effect to offset the opportunity cost of having to spend holy power, and thus, sacrifice Shield of the Righteous uptime.

The 5.4 changes move the goalposts, though.  First, let’s note that the sheer size of these buffs and nerfs are pretty huge.  This isn’t a minor rebalancing; Sacred Shield is getting slashed by nearly a third, while Eternal Flame is growing substantially.  When casting EF with 5 stacks of BoG, each EF tick will be almost twice as large as a Sacred Shield absorption bubble.  And remember, you’ll get two EF ticks in the same time period it takes to generate one Sacred Shield bubble.  So the EF HoT has about 3-4 times the throughput of the Sacred Shield absorb.  We’re no longer comparing equal amounts of healing and absorption.

The second change that comes into play is the Tier 16 4-piece set bonus, which removes the opportunity cost of Word of Glory and Eternal Flame if we have 3+ stacks of Bastion of Glory.  That’s a big deal.  Sacrificing a SotR for a WoG is a tough trade when it comes to smoothness, and that’s traditionally kept Eternal Flame from contending.  But being able to use EF without sacrificing SotR uptime – and even increasing SotR uptime if Divine Purpose is talented – is a game-changer.

Either one of these effects may have been enough to make EF a contender in 5.4.  Together, it’s hard to imagine even speccing into Sacred Shield.  But we can do better than speculation; we can fire up SimC and see exactly how much of an effect these changes will have.

A Note on Overhealing

If you ask a random paladin why Sacred Shield was stronger than Eternal Flame during most of Mists of Pandaria, one of the most likely answers you’ll get is some variation on “because absorption is better than healing” or “preventing damage is better than taking it and healing it up.”  Traditionally, this mindset has been ingrained in tanks and tank theorycrafting for the better part of the game’s life cycle.

Meloree summarizes most of these points rather well in this forum post, but in short, absorption effects tend to be more efficiently utilized than healing. Absorbs aren’t used up on avoided attacks and partial absorbs can apply to subsequent attacks, while partial overheals are “wasted.”  And of course, since absorbs apply before the damage is dealt, they act as a first line of defense that often makes subsequent healing unnecessary.

But Mel will be the first person to tell you that there’s a hidden assumption in there – all of those conditions apply when the absorption and healing is roughly the same magnitude.  Absorbs are better than healing point for point, because they tend to do a better job of being there when you need them, but a really large heal can still do more for your survivability than a weak absorb.

Also, if we’re rigorously analyzing the situation there are a lot of external effects that we need to consider.  For example, do those absorbs affect how your healer plays, or would they be tossing a heal on you with that global cooldown anyway?  If the latter, then it’s tough to say whether your absorb was really “efficiently utilized.”  As far as World of Logs is concerned, your absorb applied first, so it was very efficient.  But if it just creates larger overhealing for your healer, was it really a significant survivability gain?  If you have several HoT ticks that overheal because of a full Sacred Shield absorb, is it really fair to say that the absorb was 100% efficient and the HoT ticks were 100% overheal?

In practice, Sacred Shield’s efficiency is a bit overstated due to the way logging works, which is related to how absorption effects are consumed.  It’s often reported as highly efficient because the overheal it creates is shifted into other heals.  But it’s probably fairer to recognize that a good chunk of Sacred Shield ticks are irrelevant, and share some overheal burden with the healing effects that they preempt.

Further, there’s a general attitude that overhealing is a bad thing and should be avoided – that those heals were “wasted.”  While mana and/or GCD efficiency ties into that argument, I think it’s not fair to toss them aside so casually.  To illustrate that idea, if you had a Lay-on-Hands-esque HoT that healed you for half of your health every 3 seconds, it would probably generate an awful lot of overhealing – possibly 70% to 80%. Would that make it a bad ability? Of course not, it would still be massively overpowered.  While the majority might be overhealing, the fact that you effectively received a Lay on Hands during every 6-second period would also make you damn near invincible.  Any sudden spike would automatically be countered with a Lay on Hands, without any effort on your part!

The lesson here is that steady-state overheal measurements are about as useful as steady-state damage intake measurements – that is to say, not very useful at all.  Overhealing is only part of the assessment; we also need to consider throughput, and more importantly when that throughput occurs.  I’d easily take the Lay on Hands HoT over a weaker absorption effect because even though it’s high overheal, it’s also high throughput during a spike event.  It doesn’t matter if the absorption effect is 100% efficient, because it does less for me during those dangerous spikes than the overpowered HoT.

And that’s essentially the choice we’re making between Sacred Shield and Eternal Flame.  Sacred Shield is the steady, efficient absorption effect.  Eternal Flame is the overpowered HoT.  The only question is whether the magnitude of difference is great enough to make us prefer Eternal Flame over Sacred Shield.

TMI to the Rescue

Luckily, theorycrafting has come a long way this expansion.  In previous expansions, we might have to estimate an amount of overhealing for EF and do some hand-waving math to figure out whether that’s more effective.  In fact, we’ve had to do similar things earlier this expansion with the MATLAB code.  But those times are behind us, because we have TMI.

While it may not be immediately apparent, TMI essentially ignores gratuitous overhealing.  Why?  Well, consider what happens if you avoid a bunch of attacks in a row.  Your damage taken during that period is essentially zero, but you have lots of HoTs or healing effects happening.  When you perform the TMI calculation, you take the moving average and get a negative number.  Then you subtract one to get an even more negative number.  Then you raise 3 to that negative number and get a really small value, which you add to your TMI collector.  So events where you have a lot of overhealing contribute very little to your overall TMI.

On the other hand, if you take a lot of damage, such that even after those HoTs and healing effects the moving average is near 100% of your health or larger, that exponentiation generates a very large number.  In fact, the bulk of your TMI score is likely due to exactly these sorts of events: a handful of 6- to 10-second segments of each iteration where you experienced a spike.  Throwing extra healing at these spike periods makes a big difference in your TMI.  But the rest of the iteration produces a nearly-negligible background that doesn’t change much no matter what amount of overhealing you throw at it.

So in some ways, TMI is the ideal metric through which to filter overhealing.  It basically ignores the overhealing that occurs during safe times, but properly counts the healing that occurs during a spike event.  An ability that generates 90% overhealing will still perform well in a TMI measurement if it’s really good at saving you from or preventing spikes.

“Dumb” EF Usage

First, let’s consider some results I generated a few weeks ago.  I call this section “Dumb” EF usage because in these sims I’ve made no attempt to tailor EF’s usage to incoming damage.  We want to compare a player keeping Sacred Shield active to a player that uses Eternal Flame as a similar maintenance HoT, and blindly refreshes that HoT when it’s near expiration.

For these sims I used the trunk build of Simcraft, which should now be equivalent to v530-7.  The build I used had all of the 5.4 changes implemented through PTR build 17116.

I used the following action priority list:

actions=/auto_attack
actions+=/avenging_wrath
actions+=/holy_avenger,if=talent.holy_avenger.enabled
actions+=/divine_protection
actions+=/eternal_flame,if=talent.eternal_flame.enabled&dot.eternal_flame.remains<2&buff.bastion_of_glory.react>3
actions+=/shield_of_the_righteous,if=(holy_power>=5)|(buff.divine_purpose.react)|(incoming_damage_1500ms>=health.max*0.3)
actions+=/hammer_of_the_righteous,if=target.debuff.weakened_blows.down
actions+=/crusader_strike
actions+=/judgment,if=cooldown.crusader_strike.remains>=0.5
actions+=/avengers_shield,if=cooldown.crusader_strike.remains>=0.5
actions+=/sacred_shield,if=talent.sacred_shield.enabled&((target.dot.sacred_shield.remains<5)&(cooldown.crusader_strike.remains>=0.5))
actions+=/hammer_of_wrath,if=cooldown.crusader_strike.remains>=0.5
actions+=/execution_sentence,if=talent.execution_sentence.enabled&cooldown.crusader_strike.remains>=0.5
actions+=/lights_hammer,if=talent.lights_hammer.enabled&cooldown.crusader_strike.remains>=0.5
actions+=/holy_prism,if=talent.holy_prism.enabled&cooldown.crusader_strike.remains>=0.5
actions+=/holy_wrath,if=cooldown.crusader_strike.remains>=0.5
actions+=/consecration,if=(target.debuff.flying.down&!ticking)&(cooldown.crusader_strike.remains>=0.5)
actions+=/sacred_shield,if=talent.sacred_shield.enabled&cooldown.crusader_strike.remains>=0.5

which is just the SimC default with a line appended to maintain Eternal Flame, but only recasts if we have 4+ stacks of BoG. I used Slootbag’s character as our test subject and pitted him against the T15H25 boss. The only thing I changed is which L45 talent he had selected (EF or SS).

Here are the links to the html output in case anyone wants to pick through them with a fine-toothed comb:
Sloot – EF – 5.3
Sloot – SS – 5.3
Sloot – EF – 5.4
Sloot – SS – 5.4
Sloot – EF – 5.4 w/ 2T16
Sloot – SS – 5.4 w/ 2T16
Sloot – EF – 5.4 w/ 4T16
Sloot – SS – 5.4 w/ 4T16

For those who want the TLDR summary, here it is.

5.3 Results, 4T15:

Talent DPS DTPS HPS TMI
EF 313.6k 106.1k 105.7k 340
SS 315.4k 55.4k 99.8k 150

5.4 Results, 4T15:

Talent DPS DTPS HPS TMI
EF 257.6k 107.7k 106.4k 3980
SS 258.8k 74.5k 46.1k 10950

Note that this is with his current gear, i.e. without 4T16 (but with both T15 set bonuses), so it’s a simulation of what will more accurately reflect the first week or two of progression. Also note that due to how SimC does its HPS accounting, the Sacred Shield absorption is being included in the HPS value while also reducing DTPS.

The buff to EF and nerf to SS clearly shift the balance in favor of EF by a fairly large margin, even without the effects of 4T16. It’s also worth noting that Sloot’s EF uptime in these sims is 97% or better.  Since we’re only casting EF if we have 4+ stacks of BoG, this means he’s got enough haste to generate 4+ stacks every 30 seconds (more on that later).

We can artificially disable T15 set bonuses and enable T16 set bonuses using the code:

tier15_2pc_tank=0
tier15_4pc_tank=0
tier16_2pc_tank=1
tier16_4pc_tank=1

Doing that, we get:

5.4 Results, 2T16:

Talent DPS DTPS HPS TMI
EF 252.1k 113.8k 111.3k 9624
SS 253.6k 80.5k 52.4k 16518

5.4 Results, 4T16:

Talent DPS DTPS HPS TMI
EF 258.9k 106.0k 105.7k 4098
SS 253.8k 80.4k 52.4k 16476

Note that TMI goes up by disabling the T15 bonuses, moreso for the EF setup since it’s getting a significant benefit from the 2-piece (~45% uptime). It looks like EF is stronger even without the 4-piece, but the 4-piece clearly makes it a lot stronger while having no effect on SS.

However, this action priority list doesn’t include any line to simulate emergency WoG usage, so the comparison isn’t entirely fair.  We can do a little better.

“Smart” EF Usage

We could try and include emergency WoG usage with a line like

actions+=/word_of_glory,if=buff.bastion_of_glory.react>3&incoming_damage_5s>health.max*0.8

Which would fire off a WoG if you took over 80% of your health in damage in the last 5 seconds.  The idea is that this simulates emergency WoG usage as well as possible within the confines of the simulation.  While we can’t accurately model healer reactions to your health bar or your own go/no-go decision making based on incoming healing bars, in both cases because we don’t have a healer in these sims, we can at least try to minimize TMI by burning a self-WoG whenever we’re in the midst of a spike.

This also should help simulate the penalty we pay by not having those BoG stacks banked for emergency heals.  When we refresh Eternal Flame, we’re putting ourselves in a position where we don’t have a large emergency heal banked for the next 10-20 seconds, and that should have some sort of associated opportunity cost in terms of TMI.  In theory, we’ll be able to see that with this sort of conditional added.

If we put that line directly after Eternal Flame in the action priority list, we get results that look like this.  I didn’t have the simc file from the first set of tests handy, so I ran two baseline sims without that line for comparison.  The setup should be relatively similar to the first, though.

Sloot2 – EF – 5.4 PTR
Sloot2 – SS – 5.4 PTR
Sloot2 – EF – 5.4 PTR – emergency WoG
Sloot2 – SS – 5.4 PTR – emergency WoG

Talent DPS DTPS HPS TMI
EF 250.8k 107.7k 106.6k 2878
SS 251.8k 74.4k 52.6k 8499
EF+EW 249.5k 109.2k 107.7k 5827
SS+EW 248.5k 78.3k 64.3k 13006

It looks like including those emergency WoGs has helped narrowed the gap between the two results, even accounting for the Divine Protection “squish.”  Eternal Flame’s TMI actually went up here because we’re casting it more often, sometimes with only 4 stacks of BoG, and thus negating some of the efficiency we had before.   But Eternal Flame is still coming out ahead, even without the T16 4-piece effect.  It seems like the loss of our ability to emergency heal is more than offset by the sheer throughput we have in the EF HoT.  Being able to throw off a 500k WoG in an emergency isn’t as big of a deal when you’re getting 300k or more of that through a HoT.

And of course, if we artificially enable the T16 bonuses and disable the T15 ones, EF continues its dominance:

Talent DPS DTPS HPS TMI
EF 253.0k 105.1k 105.8k 1826
SS 248.2k 78.4k 77.1k 5963

We could try to optimize this even further by constraining EF to only be cast immediately after we take a melee hit, ensuring that it’s applying the base heal effectively.  In my limited testing, I wasn’t able to produce a strong TMI change with that (essentially adding an “incoming_damage_1s>health.max*0.4″ conditional and extending the allowed refresh period).  I think I could produce some improvement with finer tuning of the two parameters, but probably not enough to be too significant.

But this also highlights a major difference between the sim and real raiding.  Since we don’t have a healer, it hurts a little more to delay EF until a big spike, because it deprives us of that passive HoT ticking during some of the smaller spikes (or creates smaller spikes because the HoT isn’t there to counter them, depending on how you want to think about it).  In a real raiding situation, you have other HoTs and healing sources to cover that damage, so we should get a little more effectiveness out of sitting on EF and using it to instantly respond to the next melee attack.  In other words, delay it for a second or two (or longer if you get a string of avoids) to make sure that the base heal is efficiently used, which may save your healer some GCDs over the course of an encounter.

In any event, it’s looking like EF is going to be stronger than SS in most situations in 5.4

Haste

I do want to mention that there’s a minor caveat here: in all of these simulations, Slootbag had enough haste to ensure that he had another 4- or 5-stack of BoG with which to refresh Eternal Flame before the time came to refresh it.  If he didn’t have that much haste, one would assume that EF loses some potency.

However, in my simulations that didn’t really seem to be the case.  While lowering Sloot’s haste artificially, both TMI values went up, and Eternal Flame uptime fluctuated within the 90% decade.  But Eternal Flame was still consistently beating Sacred Shield by a large margin even as low as 24% melee haste.  It’s hard to imagine having less haste than that, since you can already exceed that value with full stamina gemming/gearing.  So there may be a relevant haste threshold beneath which Sacred Shield becomes preferable, but it’s not likely to be relevant to anyone stepping into normal- or heroic-mode T16 content on day 1.

Note also that all of these sims were with Divine Purpose talented; it’s entirely possible that DP is what keeps EF afloat at 24% haste, and that speccing Holy Avenger or Sanctified Wrath will reverse the paradigm.  I just didn’t have time to test all of those permutations thoroughly.

Conclusions

The simulations I’ve run here seem to be strong evidence that Eternal Flame is the new hotness, and Sacred Shield will merely be an also-ran in 5.4.  While that’s a fairly accurate statement for the bosses modeled here, it’s worth noting that real bosses vary.  In particular, effects that test your instantaneous effective health will tend to reward absorption effects more than healing, because those absorption effects are a temporary effective health boost.

For example, one-shot effects like Talon Rake and Decapitate may still lead you to prefer Sacred Shield.  There’s a reasonable chance you’ll be topped off before the effect, and the extra absorption may just be enough to survive an otherwise-fatal blow.

What we’ve generally modeled with TMI bosses is slower trickle-down deaths from successive melees, which tends to favor the raw throughput of 5.4′s version of EF, as we’ve seen. And while I think that’s a more reasonable model of tank death for most progression tanks, it may not match your most common death scenarios.

As always, your mileage may vary.  You’ll have to make the decision about which talent to take based on fight mechanics; obviously, if there’s a mechanic that strongly favors SS, don’t hesitate to take it.  It’s not terrible in 5.4, it’s just nowhere near as good as it is on live servers currently.

There doesn’t seem to be any meaningful haste threshold at which EF eclipses SS, at least when Divine Purpose talented.  I’m sure that we’ll revisit that topic down the road, though the T16 4-piece interacts so strongly with Divine Purpose that we may see most progression raiders taking it anyway.

One thought that keeps coming to mind as I review these results is just how far ahead EF seems to be.  It feels like the 30% Sacred Shield nerf or the 40% Eternal Flame buff would have been sufficient to make the two equally valid choices in that talent tier.  But the combination of both effects seems to just swap the two; rather than Sacred Shield being the hands-down, no-brainer choice that all paladins take, 5.4 just puts Eternal Flame in that spot.

It’s sort of disappointing in that sense.  While it’s nice to have a new mechanic to fool around with in the last tier, the ideal goal of the talent system is to make all three choices viable (or at least in this case, EF and SS, since we generally ignore Selfless Healer).  But rather than having an interesting choice between EF and SS, the 5.4 changes spin the wheel too far in the other direction.  We still don’t have equally valid and interesting choices in that tier.

Although I’m not sure we’ll ever see that happen due to the nature of the three talents in that tier.  I could imagine balancing Eternal Flame and Selfless Healer, because they both provide different ways to produce extra healing.  But balancing either with Sacred Shield is tougher, because it’s an absorb.  Even if they all produced identical TMI values, we’d probably lean towards Sacred Shield – it’s an absorb, essentially passive, generally higher DPS, less reliant on perfect play.  It’s just safer than the other options if they produce similar results. I still hold out hope that next expansion, Sacred Shield will become baseline for Protection and we’ll get a third healing option in that tier.  I think that we have a much better chance of getting three interesting and equally viable options in that situation.

This entry was posted in Tanking, Theck's Pounding Headaches, Theorycrafting and tagged , , , , , , , , , , , , , . Bookmark the permalink.

75 Responses to EF & You?

  1. Angelflavor says:

    Thank you for this. I was wondering about this!

  2. Slootbag says:

    Hell, it’s about time :P

    Well done Theck! Very glad to see my crusade for EF is strong and supported with another excellent post.

    It is worth noting that there are other benefits to EF probably covered by some of the Sim that people maybe haven’t figured out:

    -EF scales with Mastery and Haste, and we need a new stat after reachable Haste cap (Stam is another option), SS does not scale with anything but Haste for secondary stats.
    -EF benefits from crit as each individual HoT has its own chance to crit, and makes it so that we can in fact benefit from all 3 stats on the amp trinket.
    -EF has more effective off healing for the raid than WoG, as you can also use it to blanket members, which helps with high vengeance. SS no bueno in that department.

    I’m sure there’s more I’m leaving out too.

    • Theck says:

      All solid points. I think that if we were to sim in a T16 gear set with higher mastery values, the discrepancy would be even greater. Conversely, in a starter gear set (low haste/mastery, i.e. someone newly-gearing through LFR/T14) I suspect Sacred Shield would come out ahead.

      One thing that I did (intentionally) ignore was the possibility of double-SSing a tank. If you have two paladin tanks, you can double up on Sacred Shield coverage. In other words, off-tank taunts, refreshes SS as soon as Vengeance spikes, and the main tank also casts his SS on the off-tank. This is, in my opinion, probably still more effective than each tank using Eternal Flame as using EF on the other tank won’t benefit from BoG.

      I didn’t include it because a) it’s tougher to sim accurately, and b) it only works for the small percentage of guilds that actually have two paladin main tanks.

    • Meloree says:

      Just to get this out of the way: I’m not disputing at all that EF is going to be better in the general case in 5.4

      I am going to quibble a bit with your first bullet point, though. SS scales with mastery, because absorbs scale with all damage reduction – like armor, and other absorbs, and damage reduction cooldowns. In addition, SS scales with avoidance stats.

      SS almost certainly doesn’t scale as well with mastery as EF. But it’s not correct to say that it doesn’t scale.

      • Slootbag says:

        That’s fair and I agree, perhaps scales indirectly would be more accurate on my part. The bigger point was the (better) scaling of Mastery on EF though.

        Also I’m hesitant about avoidance scaling for SS, I’ve heard that said a lot too. If you avoid something you never take the damage, so the absorb or the HoT wouldn’t have done anything anyways. The advantage (and an argument that was made for SS) is that of course the absorb isn’t consumed. But that is also within a 6 second window (more like 4 with haste builds) where the SS absorb would reapply anyways, and it’s not as if the absorb stacks. So ultimately avoiding more attacks or avoiding 1 in 2 attacks doesn’t really do as much as you think for SS.

        Let’s say every 2 melee swings is a SS absorb re-application. If you avoid both, you sit at your (for number’s sake) 100k absorb at all times. but you also never lose HP and need to heal it with EF.

        Same thing but 1 melee in 2 lands every time, best case SS scenario. Even melee hits = no health change. Odd melee hits land and consume SS (every time) and some HP, or land and are basically fully healed by EF.

        No avoidance obviously favors EF more. 1 in 2 will have SS absorb some of the damage and the other hit for full. But both hits will be healed by EF due to it’s tick frequency.

        You either cap out at the X absorb for SS or it is consumed. Or you avoid the damage or EF heals it, which is what this whole argument is about, EF heal vs SS absorb. So maybe here too, it’s not a matter of scaling, rather a matter of benefiting more from avoidance than would a direct heal, if at all. If SS absorb stacked every application to maximum %HP then sure absolutely avoidance would be a big deal. But given it’s nerfed form, a melee either consumes it or it doesn’t right now. And we just come back to the EF vs SS argument that is going down atm.

        • Theck says:

          To elaborate on Mel’s point: if you avoid 1 in 2 (say the first), the SS absorb sticks around to be consumed by the second. As a result, it’s less likely that the SS absorb gets wasted (because you need to avoid both attacks in a row).

          With EF, one avoid generally means that EF tick is just overheal. In the scenario above (avoid the first of two), you get the full SS absorption applied to the second attack, but only one EF tick out of the two that occur in the same time span.

          That’s the scaling with avoidance that Mel’s talking about, and he’s absolutely correct in that regard.

          • Slootbag says:

            Indeed, I must say I mostly misunderstood what he was implying then.

            I suppose in my head what I was getting to was a situation where theoretically every melee that landed had a full SS on it in the ideal avoidance scenario. Even with the raw numbers (SS nerf, EF buff) it seemed to me that EF was still a high contender via sheer magnitude in numbers now, regardless of the ideal avoidance scenario where every melee hit was soaked with a SS absorb. This in my mind would make the benefit of avoidance scaling for SS a bit more irrelevant (but also depends on many other things too).

            For an elementary example, a 500k melee hit that always had 100k absorbed and hit you for 400k more with SS.
            Or the same hit that was healed for ~160k via EF leaving you with higher net HP.

            So the question there of course becomes whether or not the nature and magnitude of the absorb is better than the extra healing from EF, which starts to (as mentioned in parenthesis above as well) incorporate a bunch of other factors, like health pool.

            Of course these scenarios as we all know are pretty unrealistic as in the real raiding environment, where there is an abundance of other damage sources eating up that SS absorb or demanding the EF tick.

          • Thels says:

            Meloree is obviously right about Absorbs and Avoidance having a much better synergy than Heals do. That’s one of the reasons Avoidance is better for DKs, as an avoid keeps their Blood Shield alive for the next hit, whereas SotR doesn’t care if a hit is avoided or not, the impact on the next hit remains identical.

            However, this is an advantage to SS in situations where we are already fully avoiding an attack, and thus taking less damage than on average. Slootbag’s avoidance is just under 30%. For each string of four boss attacks, there’s a 25% chance he won’t avoid any of the four attacks. That situation comes up plenty enough in a fight that we can’t really rely on Avoidance as part of the equasion for survivability.

            So while the synergy between Absorbs and Avoidance certainly being present, the impact it has on the end result is not that strong.

          • Meloree says:

            Again, just so we’re all on the same page: I’m not attempting to argue that SS is better than EF in the general case. EF is probably quite a bit stronger right now. I was merely pointing out that SS does in fact scale with quite a lot of things. The effect of the scaling with avoidance may be small (it is), but it categorically exists. That said, it’s stochastic, and you don’t have to have read many articles here to have seen Theck quote a soundbite of mine about stochastic effects.

            One thing to bear in mind is that while TMI is bar none the best tool we’ve had for evaluating tank survival, it still has it’s weaknesses, and I think it generally undervalues absorbs compared to heals. I don’t think it does that by enough to make SS better than EF, but here’s a couple of factors that the sim abstracts out which may matter in practice:

            1) Order of operations matters. Absorbs come before damage comes before healing, and you can’t heal from dead.
            2) Absorbs scale with other absorbs (and all forms of damage reduction). SS makes your Disc priest(s) and Holy Paladin(s) better.
            3) Opportunity cost. Seriously, the sim is against a simple boss doing simple things. EF has an opportunity cost that I don’t think the sim adequately evaluates. With EF applications you’re both reducing your flexibility in using HP, and reducing your reserve BoG stacks for reacting to things. It’s something you can generally play around, I think, but that has a further opportunity cost in general.

            I think 4T16 probably puts the nail in the coffin, but before that, I think the case for SS is a lot closer than the raw TMI scores would show.

            Again, I suspect that EF wins out regardless. I don’t think it wins out by an enormous margin. It’s worth noting that TMI is new enough that nobody really knows how much better 5k is than 10k. That part hasn’t really been nailed down yet. All you can really say is that it’s “better”.

  3. Dornex says:

    One thing worth noing for SS is that it now can be maintained on multiple targets. For fights where both tanks are tanking continuously that could be quite powerful.

  4. Geodew says:

    Doesn’t the HoT of EF depend on how much Holy Power you consumed (It did when I tried Holy Pal in 5.2)? Shouldn’t you only refresh it with 3 holy power if it’s strong enough?

    • Theck says:

      Yes, and I probably should have included a holy power conditional in there to ensure it was only cast at 3 Holy Power. Looking at the results though, it seems that the average holy power spent per cast is almost identical to SotR, suggesting that we were almost always casting it at 3+ HP anyway. The combination of Divine Purpose and the banking logic on SotR keeps us at 2-3 Holy Power almost all the time, barring special cases where we have an emergency.

  5. Wrathblood says:

    Theck, I’ve run a significant number of sims (I should put the data someplace so people can look at it) looking at different circumstances, and a couple things popped up. A big one is that, in current gear and without our 4p, HA is very powerful compared with DP because the sim assumes that you’re single tanking whatever you’re fighting, meaning DP always has 10-12 seconds of awful at the start of every pull as it gets up to speed, HA (and I assume SW as well, though I didn’t test it) don’t have this weakness. I was thinking a way of working around that would be to sim a bunch of ~30 second fights, testing out the best way to mitigate the pull (0 BoG stack EFs you overwrite as quickly as possible? Hold out for 3 stacks? 1 HoPo EFs? Many possibilities).

    The trade-off, of course, of HA vs DP is that with HA (with current levels of gear and no t16 4p) you *never* get 5 BoG stacks on your EF outside of HA’s activation period. With DP you always get 3 stacks, and in fact you virtually always get 4 stacks, sometimes getting 5 (depending on the minimum refresh time you set, I played with 0 through 5 seconds). With HA you get 1 5 BoG stack EFs during its duration, and then you’re looking at a mix of 3s and 4s until EF comes up again.

    Surprisingly to me, however, by far the biggest driver in keeping TMI down was EF HoT up-time. Appearing to beat even ShoR up-time handily. Trading ShoRs for EFs (resulting in increased HoT uptime) appeared to consistently win for TMI, though I assume trading SHoRs for EFs that *don’t* increase your uptime (which will happen a lot when we get 4p) will then swing to a loss.

    • Wrathblood says:

      Answered my own question on the ShoR vs EF (with no change in HoT up-time) by running with HA and cranking the allowed HoT time remaining to re-cast EF all the way up to 10 seconds. Up-time was unchanged, and it just converted 2.5 ShoRs into EFs. It increased my dtps by about 1500 though it also increased by hps by about 970 (and reduced my dps by about 1200). Net result (against T16Q boss) was jumping my TMI from 294K to 553K.

  6. blizzhoof says:

    I’m terrible with Weakauras. If you (or anybody else) make one for reapplying EF and tracking BoG stacks, can you post a link to it? It would be a great help. Thanks.

    • Theck says:

      I’ve updated my pastebin WeakAuras to include EF tracking. So far the text auras that tell you whether it’s a gain to re-cast EF only take into account Vengeance/AP, not Holy Power and BoG. I plan on adding those soon, probably this weekend since I’ll have some time to work on it.

  7. Will be nice to have 2 talent choices instead of one.
    It may increase the value of Glyph of Word of Glory.

    Im not sure about this but Boss skills may affect this choice.
    Skills like Serious Wound (TOT Durumu) will probably make SS better than EF.
    I believe in Siege we will have increased healing with some purify skill.

  8. Thiron says:

    I wonder what helps more for second tank though, since EF will be without Bastion in this case. I guess it might depend on class too…

    • Thels says:

      The vast power from EF comes from the bonuses from self-casting it. When casting it on others, EF drops substantially in effect. Not only do you lose the BoG benefit, you also lose the 100% bonus for selfcasting, which already halves the healing on it’s own. That’s exactly why running with 2 paladins, SS might be better on tank switch fights. However, keep in mind that if you are both tanking a mob, you can actually keep EF up on each other for a small loss in SotR uptime, whereas you cannot share SS at all.

  9. Wrathblood says:

    Issue, first noted by Schroom over at Maintankadin but its worth expanding on. TMI increases rapidly as tanks take larger and larger %s of their health on hits (as one would expect to happen as we get to the end of the expansion) but its already spiraling high enough to make it daunting to grasp. The T16Q boss only does 20% more damage than the T15H25 boss, yet TMI is spiking up very sharply against it. My gear using SS against it gives a TMI of 13 million.

    It could be a bug in which case its worth looking into. But its possible that its just a function of how TMI scales, in which case it might be worth trying to ratchet it down a bit.

    • Theck says:

      Not a bug, that’s just how TMI scales. The boss hits for 20% more of your health, which means the worst case is probably about 40%-50% worse (assume two full hits and maybe another partially mitigated hit). That means it’s going to be $3^4=81$ or $3^5=243$ times worse for TMI due to the exponential scaling.

      Remember that the T16Q boss is just a rough guess at T16H25 Garrosh, and that you’re (probably) attempting it in T15 gear, so you’re considerably undergeared for the encounter. You’d expect the TMI to be pretty huge if you were attempting heroic 25-man Garrosh today.

      There’s really no reason to “ratchet it down a bit.” We could reduce the HDF from 3 to 2 (such that it would be $2^4=16$ or $2^5=32$ times worse), but that comes at the expense of fidelity at the low end (i.e. where most people play, not significantly under- or over-geared for the encounter). There are various ways to “cap” the result, but again, all that does is throw away useful information.

      It makes more sense to me to preserve all of that information than to throw it away. If you get a TMI of 13 million, it tells you that you’re significantly undergeared for the boss, whereas a TMI of 50k might mean you’re only slightly undergeared. If we put in some sort of cap, you might get 50k in both of those scenarios, which isn’t very helpful.

      • Theck says:

        Edited that for clarity. If the boss is only hitting 20% harder *than before* rather than, say, going from 20% of your health to 40% of your health per swing, the worst-case should only go up 20% and thus be ~$3^2=9$ times worse.

        • Wrathblood says:

          That’s the thing, dtps jumps about 20% from T15N25 to T15H25, and TMI goes up about an order of magnitude. Going from T15H25 to T16Q, dtps jumps roughly another 20%, yet TMI goes up two orders of magnitude.

          Its not a big issue since we’re looking at the final boss of a tier, but a metric that can reasonably expected to range from 100 (T15LFR25) to 13,000,000 (T16Q) in the same gear might be hard for people wrap their brains around, especially since dtps only doubles.

          • Wrathblood says:

            To be clear, I get that the increased damage is vastly scarier as a % of tank health, and then you add expected spikes on top of that and it gets even more scary and spiky. But a metric that scales over 5-6 orders of magnitude is hard to intuitively follow.

          • Çapncrunch says:

            Yeah, but you’re jumping over many difficulty layers there:

            T15LFR
            T15N
            T15H
            T16LFR
            T16Flex (maybe? depends how that scales)
            T16N
            T16H

            I don’t know, jumping 5 or 6 difficulties coming out to 5-6 orders of magnitude almost seems appropriate.

            It’s also not really *that* hard to intuitively follow. Having a tmi over a certain amount essentially just means the same thing: “You Are Not Prepared!” And in the vast majority of cases that’s all the information you’ll need. Whether your tmi is 100k or 100M probably doesn’t really matter. And as Theck said, he *could* put a cap in there but what would you really gain from that? It wouldn’t be any more intuitive, you’d still be getting a number that meant “you’re way undergeared for this”, it just may be a few orders of magnitude smaller but still very large. Like he said all it’d accomplish is throwing away information that is somewhat useful just to tell you something it was already telling you.

            If anything I’d say that its the dtps growth that is non-intuitive, as a 20% growth gives you the impression of a much smaller change than it really is. As well as the impression that another 20% increase would have a similar effect, when the effect is going to be considerably more impactful than the first. Exponential growth is the reality, it’s the dtps that is misleading you (as it often does, was there ever a time when it was a useful figure?)

          • Wrathblood says:

            Except that the T16Q number isn’t literally the damage that 25H Garrosh does, its just a guess. And its roughly proportionally as much over T15H25 as T15H25 is over T15N25. So it might implicitly be skipping levels, but in terms of damage output its right in line. They’re both ~20% jumps in damage done, yet one increases TMI by a factor of 10, and the other increases TMI by a factor of 100.

            And that’s the thing. While I agree that putting a literal cap on it would be ridiculous, the way it stands now its kinda meaningless. Sure, 100 is overgearing the content and 13,000,000 is undergearing the content, but that’s an *incredibly* wide range.

            If 100 is overgearing content, what is “appropriate” gearing? 500? 10,000? Is 50,000 ok? How about 250,000? That seems like a wide range, but it reflects less than 1% of the range I’ve seen in sims. Which one of these numbers reflects “appropriate” gearing?

            And what do the numbers even mean? Is 10,000 twenty times as good as 500, while being 25 times worse than 250,000? Since a 20% increase in dpts results in a 10x increase in TMI in one case and a 100x increase in another case we can’t even say.

            For a metric to be meaningful, people need to be able to derive some sense of meaning and proportionality at a glance and TMI doesn’t deliver that atm. One could argue that a clearer bounding of when TMI is a useful metric (“Won’t produce useful numbers if your gear isn’t within about a tier of the target ilevel for the content”) but that’s way too narrow a range to be useful and hard for a casual player to grasp without a fair bit of handholding.

          • Meloree says:

            To be blunt, it’s a pretty new metric, and the answer to most of those questions is simply “We don’t really know yet”, which is why I, at least, tend to argue against anything that removes information. We don’t really know what’s meaningful yet. We could take the log of TMI, to just represent orders of magnitude, but we don’t know what we lose. The range where fidelity is most important may well be between 1000 and 1,000,000, and then just dividing TMI by 1,000 to make the numbers more manageable might be the best representational solution. We just don’t really know the answers yet.

            The point of the metric was to find a way to represent “big spikes bad, throughput damage not-bad” in a simple metric. It’s been pretty successful at that, but we haven’t yet managed to nail down what given numbers really mean.

            This tier, frankly, will be really helpful for that, because it’s the first time that we’ll be able to track TMI through a progression tier. We can snapshot some gearsets now for typical progression, and snapshot the gear a few people kill bosses in, and see how the TMI progresses. That will give a pretty good indication of what appropriate TMI is for the Treckies and Slootbags of the world in progression, as well as what’s more reasonable for everyone else.

            So, to summarize: TMI isn’t meaningless now, but it’s only meaningful in relative terms. 10 is better than 100. 10,000 is better than 13,000,000. It’s tough to say that you’re twice as likely to die, or 300 times as likely to die, because a lot of that depends on external factors. You’re spikier, your healers will have to work harder… that’s what we have right now. More work needs to be done to make the numbers more meaningful, but it’s not a simple matter of bounding, either. The number 13million is as useful as the number 100. Right now both say *something*, we just haven’t nailed down exactly what.

          • Meloree says:

            Replying to myself, here, but I just want to be a bit realistic about the state of TMI.

            A few months ago, Theck would fire up his sim-machine after making an improvement, and send out 600 columns of numbers to the rest of us at Sacred Duty HQ. And then we’d have some legendary fights about what they actually meant. “Clearly avoidance is looking stronger now that we’ve incorporated ” “Are you a crazy person? It has representation two full decades above haste!!” “Yeah, but like 0.1%, that’s a corner case.” “Corner cases are the only things that ever kill you.”

            TMI came about because Theck (mostly Theck, it’s pretty flattering that I get my name in it, but I don’t really deserve it) wanted to build some kind of simple metric that summed up all our various arguments… and weighted them.

            So we hopped on vent, and we fired some ideas back and forth. Theck laid out what he was thinking, I complained about it, he defended it, I whined some more, he gave a little ground, and we ended up hammering out a framework for making “spikiness” into a number. Again, mostly Theck, I was a sounding board more than anything. So he built an early iteration based around the idea that “taking 10% more damage in a period of time is about twice as dangerous”. We ended up settling on 3 times as dangerous – he explained why in the blog post about TMI.

            But that’s the state it’s in right now – it’s a brand new metric, and assorted implementation bugs have been hammered out, but right now is the very first time it will get to be realistically tested in progression at all levels of play. Now is the first chance we’ll really have to start trying to derive meaning from the numbers and relating them to reality in a better way than “less is better than more”.

          • Theck says:

            Responding to Wrathblood: I agree about meaning and proportion, I just disagree that TMI is inherently hard to understand in that context.

            Certainly the range is large, but nobody seems to have trouble handling money even though you can have as little as $\$10$ or as much as $\$100$ million. Similarly, DTPS can range from as little as 0 to a million or higher under the right circumstances.

            And remember that the metric is founded on the idea that a spike which is 10% larger is 3x worse. So in some sense, TMI is a direct reflection of that – a TMI of 100k is 10x worse than a TMI of 10k, and it means you’re taking roughly 20% more damage (because $3^2=9$, which is close to 10).

            You’re getting a little too hung up on the exact value, as if saying “my TMI is X” has some inherent meaning, like having 10k TMI is bad and 5k TMI is good. But that’s no more logical than saying that 100k DTPS is bad and 50k DTPS is good. One is clearly worse than the other, but it’s not clear that either is strictly “bad” – if you’re receiving 150k HPS then either DTPS value is fine, and if you’re receiving 30k HPS then both DTPS values are bad.

            Of course, I’ve tried to normalize it such that there’s a useful range – i.e. the 1k to 10k spread – but remember that the normalization was done under 5.3 conditions with 5.3 bosses. We had a big Vengeance nerf, which affects our AM, which subsequently shifts the goalposts a lot. We also had a lot of mechanics changes. I may end up re-normalizing, or I may just leave it as is and say that 10k-50k is the “acceptable” range. It’s not clear what the most straightforward solution is yet.

            I would still argue that the TMI metric is better than the alternative. Before we quantified this, we essentially had no good way to truly measure smoothness. Me posting tables of numbers and divining analysis from them just doesn’t cut it for a large scale. That doesn’t mean there isn’t room for improvement, but it’s also a very new idea, and it takes time to get accustomed to new ideas. I’d rather let it settle in and give people a chance to get used to it before making any drastic changes. We may find that 6 months from now, the concept of having a 13M TMI is just understood to be “undergeared” and not interpreted as some sort of bug or error.

          • Wrathblood says:

            To Mel and Theck,

            I certainly get what both of you are saying, and in particular to Mel, I recognize the newness of TMI and am in no way intending to imply that TMI is worthless because it failed to spring fully-formed from Theck’s head like a quantitative Athena.

            My intent is actually much the opposite. I think TMI is fantastic both in understanding gearing, as well as being a necessary first step towards truly measuring tank survivability balance across classes. I would like it to thrive and slowly conquer us all, eventually turning us into little TMI pod-people.

            But to get there, it has to be accessible to the masses, and the intent of my comments here (and my little back and forth with CapnCruch) has been to help thrash through some of the issues around TMI and help shove it a little farther along its path to maturity.

            So, at the moment, I’ve got two points worth making. First, based on Theck’s comments, I’m a bit concerned over the T16Q boss mechanics. It only appears to do 20% more dtps to the tank, yet the tank’s TMI jumps by 100 rather than by 10. Does it intentionally have bigger spikes than the T15H25 boss or some other explanation?

            Second, from a marketing standpoint, having a metric that gives people results like 13,000,000 isn’t really very easy for people to follow. Even worse, I’m in 540 gear. Imagine someone in like green questing gear running it for them tanking T16Q? They’d probably get a result in scientific notation.

            I have no idea what the “correct” approach is, and certainly T16Q might spike the tank 1,000x “more dangerously” than T15H25, and the TMI accurately reflects that. But wouldn’t it be easier to get people to adopt a number that was a little easier to follow? I mean, sure, people get $\$10M$ vs $\$1$, but they don’t *really* know what $10M implies. Private jet? Penthouse in Manhattan? They understand the number, but not the implications. As TMI gets thrashed out, consider the value of it being functional rather than accurate, especially in a case where “accurate” (or “fidelity at the low end”) doesn’t really mean anything.

          • Theck says:

            Regarding the 10x vs 100x issue, can you give me the sim information (version, armory, etc.) so I can test it myself? I highly doubt it’s a bug – the TMI calculation is pretty insular, and works properly – but it’s worth double-checking. Maybe the T16Q boss is doing something funky I didn’t expect?

            Regarding usability/marketing: Yeah, the metric is sort of abstract. That’s something I struggled with at first, because the entire concept of smoothness is sort of abstract. Unlike DTPS, there’s no obvious set of units in which you’d measure smoothness. Maybe the variance of DTPS/health, which is essentially what we’re doing (albeit weighted), but that’s already pretty abstract.

            But it’s also tricky to change. One thing I considered along the way is to just take the log of the result. That helps fix the range issue, because now you’re comparing e.g. a logarithmic TMI (or LTMI) of 3 (corresponding to ~1k) to a TMI of 6 (corresponding to ~1M)

            Numerically, it doesn’t add anything though, and it’s not as clear that an LTMI of 3.3 (TMI~2000) is actually 2x worse than an LTMI of 3. In fact, I’d argue that most people would treat that difference as negligible. Of course we could normalize that again by multiplying by a constant, i.e. to turn those into 3300 and 3000. But that still doesn’t reflect the reality of 3300 being a lot more dangerous than 3000.

            That’s why I brought up fidelity. The current version makes a lot of sense when the values are between 1k and 10k, for example. It’s clear that 5k is a lot worse than 3k, the “amount” worse is very plain (it’s about 60% worse), and because we know the HDF is 3 we can work backwards to figure out how much bigger the spikes are ( around 5% of health, because $3^{0.5}=1.73$).

            On the other extreme, it’s not very useful to have detailed information for 1M+. If you’re getting TMIs of 1M or more, you’re simming against the wrong boss or doing something wrong in the APL, period. Taking the logarithm to bring that number down to 6 doesn’t add much to our understanding, it just makes the number smaller. Ignoring LTMIs larger than 6 isn’t much different than ignoring TMIs larger than 1M, we’re just moving the goalposts.

            It could be that we just want to lower the HDF from 3 to something smaller. I’ve discussed the possibility of reducing it to 2 with several people. That reduces the range at the cost of a little fidelity, and shifts value from haste into avoidance, but the metric still works pretty well. The HDF is somewhat arbitrary anyway. But you’d still be able to get values in the millions if you sim e.g. a 500-ilvl geared player against a T16 boss. That’s just the nature of the exponential beast.

          • Wrathblood says:

            First off, I’m using the armory entry for me in my current gear http://us.battle.net/wow/en/character/drenden/Wrathblood/advanced
            I’m using Simcraft 530 – 7, and the simulate commands are exactly the base ones plus: actions+=/eternal_flame,if=talent.eternal_flame.enabled&((target.dot.eternal_flame.remains=3)&(buff.bastion_of_glory.react>=3))
            which is inserted between the Divine Protection command and the first time ShoR gets mentioned.

            I just ran it again (set to PTR), and my TMI for T15H25 was about 11k, while the same sim against the T16Q boss jumps to 417K. Hmm, that’s not as bad as it was earlier. Its a 38x jump instead of a 100x jump, and its a 23% increase in damage rather than a 20% increase. Still seems oddly large.

            Additionally, related to the aside I had before, I reran the T16Q boss only changing the time remaining on the HoT to <3 (down from <5) and TMI spiked from 417k to 11.8M even though dtps actually went slightly *down* and EF and ShoR both did roughly what I would have expected. Happens on T15H25, too. With a 5 second refresh limit, TMI is 11k. With a 3 second limit, it jumps to 108K. Bump it further down to 2 seconds and it falls a bit to 60k. I'm sure something is going on but I can't figure out what it is.

          • Wrathblood says:

            Whoops! Sorry, I cut and pasted the wrong line I added to the commands. Not sure where the came from. I’m actually adding:

            actions+=/eternal_flame,if=talent.eternal_flame.enabled&((target.dot.eternal_flame.remains=3)&(buff.bastion_of_glory.react>=3))

          • Wrathblood says:

            Huh? That’s weird. That’s not what I just copy-pasted…

            actions+=/eternal_flame,if=talent.eternal_flame.enabled&((target.dot.eternal_flame.remains=3)&(buff.bastion_of_glory.react>=3))

          • Wrathblood says:

            Ok, sorry, I’ll stop responding to myself, but its cutting out the conditional where I make it only only cast EF if it has 3 or more HoPo. I have no idea why its cutting that out of the middle of the command, but it is.

          • Theck says:

            Probably an html parsing issue with the comment system.

    • Thiron says:

      Yeah, TMI seems to be very…non-linear, so to say.
      Maybe its logarithm would be more convenient for reference?

      • Theck says:

        That’s something I considered initially, but see my response to Wrathblood. It’s nonlinear no matter how you slice it, because it’s intended to be nonlinear. Comparing a log(TMI) of 3.3 to 3 isn’t more informative than comparing TMIs of 1k and 2k, and recognizing that a log(TMI) over 6 is insane isn’t much different than recognizing that a TMI of 1M+ is insane. If anything, I think the log(TMI) makes it less clear, because the average player is more likely to recognize that a TMI of 1000000 is a lot worse than 1000 than they are to recognize that a log(TMI) of 6 is equally worse than a log(TMI) of 3.

        • Wrathblood says:

          What if you do something simpler like take the square root of the TMI? That turns the massively overgeared 100 into a 10, the reasonably geared 10,000 into a 100, and the massively undergeared 13,000,000 into 3605?

          • Wrathblood says:

            Then tell people that anything under 50 they’re overgeared for, 50 to 150 is about right, and over 150 they’re undergeared?

  10. Manstus says:

    Perhaps, although, with each tick of EF representing twice as much throughput as SS, and ticking twice as often – assuming you have SoTR up during Hard Stare, EF may top you off and get rid of the bleed by itself, whereas SS would merely cushion the blow so your healers can top you up. Could be entirely wrong of course.

  11. Manstus says:

    Was meant to be a reply to Mesala re: Durumu mechanic.

  12. Çapncrunch says:

    This is interesting, mainly in 2 parts that make me scratch my head. First is where you put in the lines regarding emergency WoGs. This is something I was actually trying to mess with in SimC the other day but I gave up when I kept getting results similar to yours: adding emergency WoG’s had a negative impact on TMI. Which I found rather non-intuitive, as it implies that tossing a large WoG on yourself when you get a big spike makes you more likely to die than if we just ignore the massive spike and continue to just use SotR. Which goes against common sense, as we use the emergency WoG to save ourselves after a near-death (obviously a SotR beforehand to prevent a spike is better, but when we do drop to 20% health a 500k+ heal is probably more likely to save us than a 60% damage reduction, or at least it seems that way). Maybe emergency WoGs just aren’t useful to start with? Or maybe the reason for this is because of the relationship between the emergency WoG condition and the shifting que condition. That is should the eWoG be responding to a “steeper” spike than the shifting que or not, and which should have a higher priority? In the handful of adjustments I tried making to my eWoG condition I couldn’t find one that improved TMI. I mean if adding an emergency WoG to the non-EF sim is hurting our TMI then doesn’t that mean the lack of an emergency WoG is a non-issue anyways? (ie we’re giving up something that the sim says hurts us anyways)

    The other thing is, is EF really performing THAT well even without the T16_4p bonus? I mean I understand the sheer size of the EF hot ticks, but at the cost of giving up a SotR in addition to SS seems like the results should at least be closer.

    As to the issue regarding whether we can have enough BoG stacks when we refresh EF, vs haste, it actually seems rather forgiving. To maintain 5 BoGs per EF means we need to use SotR every 6 seconds or less. Just looking at the standard rotation (CJFCFJCFF) we generate 5 holy power every 9 globals, which gives us 1 SotR every 5.4 globals (not even counting GC or lvl75 talents), without haste 5.4 globals is 8.1 seconds, it only takes 35% haste to shave that down to 6 seconds. So at 35% haste (and near-perfect play) you’d be able to maintain 100% uptime on a 5 BoG/3HoPo EF, and factoring in GC/talents it could be done with less.

    To maintain 4BoG EFs would mean a SotR every 7.5 seconds, which only requires 8% haste, and factoring in GC could probably be maintained with almost no haste at all. And even at 4BoG/3HoPo EF will be ticking for more than a SS bubble, twice as often. I’m still a little torn on the idea of EF without the new 4p, but it definitely seems like they went a little overkill on making SS not the automatic best choice (that or they didn’t consider the impact that the new set bonus would have on making EF free). 3Bogs would be the floor seeing as that would be 10s per SotR and 8.1 is the average at 0% haste, and again even at 3Bogs a 3hopo EF ticks slightly harder than SS in half the time.

    • Theck says:

      Regarding emergency WoG, I had the same concerns. I think the most likely explanation is that we need a slightly different formulation for the eWoG conditional. For example, what we have currently is incoming_damage_5s>health.max*0.8. That means we’ve already taken a spike that’s 80% of our health or greater before we activate the conditional. Some of the time, that extra WoG nullifies the spike. But some of the time, we heal for a chunk and then take two un-mitigated hits in a row because we couldn’t put up SotR. That means we now have an even *worse* worst-case scenario than we did before using emergency WoG. Statistically speaking, we must be opening ourselves up to more of those new worst-case scenarios than we are mitigating existing worst-cases.

      Of course, part of that comes down to the limitations of the simulation. We don’t have a healer, so there’s no clear concept of “buying time for the healer to react.” In a real encounter, that WoG might let you survive the next full hit, which may then give your healers time to land their big heals and top you off. Crisis averted! But the sim doesn’t have healers, so it can’t represent that.

      Part of it also may be that we’re blowing 1- or 2-HP WoGs (and Bastion stacks) because we don’t have 3 HP when the eWoG conditionals are invoked. We could add a holy power conditional to that – or more thoroughly, a (holy_power>2|buff.divine_purpose.react|(set_bonus.tier_16_4pc_melee&buff.bastion_of_glory.react>2)) conditional – to try and keep that from happening.

      We may also want to play with the duration (is 5s the best? Should we aim lower and use 3s? If so do we keep 80% or lower that too?). I didn’t spend a lot of time optimizing that, so it’s likely that there’s a better combination than 5s/80%.

      • Wrathblood says:

        Theck, I think the lack of a HoPo conditional in the eWoG causing trouble, because its not getting overwritten later. If the eWoG happens to be 1 HoPo at 3 BoG stacks, its going to run for the next 28 or so seconds since the normal EF cast isn’t allowed to overwrite it until its less than 2 seconds of duration remaining.

        Is it possible to put in some kind of conditional on the initial EF command allowing it to overwrite weaker EFs or just EFs with lower HoPo or BoG levels (leaving aside Vengeance)?

        Also, as a minor aside, I found a weird pattern in which setting the max EF remaining duration before recasting EF to less than 1, 3 or 4 seconds resulting in strangely high TMIs, whereas 2, and 5-10 were all fine.

        • Thels says:

          Did you set a minimum HoPo requirement for that max EF? It could be that you just burned the HoPo on a SoTR, and you’re casting 1 or 2 HoPo WoGs.

          Did you also try it with 4 piece Tier 16? If the problems don’t arise there, it’s very likely to be the HoPo.

      • Çapncrunch says:

        I guess I should have shared more details regarding my experimentation with eWoG conditions, since I was trying them before reading this post (they were also done in the 5.3 engine, though the general concept of eWoGs shouldn’t be any different between the patch versions, also I had no set bonuses). My eWoG condition did have a 3 hopo condition on it, as well as a 4+ BoG condition because I figured I’d aim for just the biggest eWoGs (didn’t want to risk sacrificing a SotR because I was too lenient and wasting hopo on a wog that wouldn’t do anything). IIRC my incoming_damage condition was only looking at a 2 or 3 second window (rather than 5) and so I was also looking at smaller spikes, only in the 50~60% max health, and even tried for 70%+.

        The reason I choose to look at a smaller window rather than a 5+ second window is because I figured if I took something like 80% of my health in 5 seconds then I probably would have also taken 30%+ health in the 1.5 second window that the shifting que condition would have caught. So I was trying to catch just the very worse situations than that and let the shifting que continue to handle the smaller spikes. In other words I was working off the assumption that if I see my health drop by X I’d use SotR to try and keep the spike from getting bigger, but if the spike was almost all of my health in the same instant then I’d reach for the eWoG to get a quick heal back. So I figured I’d start out trying to just eWoG those worst case scenarios where WoG would be most worthwhile (ie even if I only used 1 WoG for a marginal improvement in TMI, and then dial it out until I found the point where it stopped helping anymore).

        But even so I was still getting worse results than ignoring eWoGs entirely. What’s more from the sims that I looked more closely at I seemed to be using just as many WoGs regardless of the conditions (you’d think changing the condition from a 50% spike to 70% would have eliminated some of the WoG casts, but that didn’t seem to be the case). For a second I thought maybe I was just simming against too strong of a boss that I was being spiked that hard constantly, but no I was simming against the T15N25 boss which gives me a non-eWoG tmi of less than 1k (usually around 500~600) but adding eWoGs made it jump up to I believe it was around 10~12k (might have been a few k smaller than that, but still a clear jump drop in performance).

        I have a couple of thoughts on this, mainly related to the way the sim works: First, when using an eWoG we still took that huge spike, so even though we healed it up wouldn’t the TMI calculation still end up counting and weighing it since in the moving 6 second window we still have a -80% health spike that gets weighted very heavily and then even if we take little damage in the following seconds those get weighted less so since the healing is higher than the damage intake? I mean it makes sense that even if the eWoG completely cancelled out the spike that it’s still “better” to have not taken the spike at all than it is to heal it off after, so maybe the calculation favors more smoothing from SotR than it does healing?

        Another thought I had was that maybe it had something to do with how the sim allows us to go into negative health instead of just ending the fight when we die. Though I can’t recall my line of thinking for how/why this would impact the usage of an eWoG when both our condition as well as the TMI calculation is more concerned in the health delta than the actual health amounts.

        @Wrathblood: that could be an issue, but notice that the issue shows up when not using EF (ie the SS+eWoG test compared to just the SS sim, the TMI still jumps, likewise my experiments were with the live mechanics still using SS as well). So while the issue of overwriting vs not overwriting stronger/weaker EFs may be effecting the EF sims, it’s not the main issue.

        • Çapncrunch says:

          I had another thought regarding the choice between EF and SS in 5.4 and it’s the dreaded “skill” question. This has come up a lot for tankadins this xpac about whether or not some of the right choices are only right if you’re skilled enough to make use of them, this being a common thought for a while regarding haste vs mastery. Although typically the skill question has been repeatedly shown to be a non-factor.

          However, as I compare SS and EF I think the skill question has merit. All SS really requires to be used effectively is that you press it every 30 seconds. failing to refresh it has a linear impact on its usefulness, and refreshing it early has almost no loss beyond the wasted gcd (and can even be a good thing if taking advantage of a vengeance spike). Whereas EF doesn’t only need to be used every 30 seconds, its effectiveness is directly tied to how well you perform the rotation (ie generating as much holy power as possible in order to get your BoG stacks), making sure you cast it with enough holy power. Refreshing it early is also a larger loss than SS since you waste both holy power and BoG stacks.

          It’s easy to see how a below-average tank could easily misuse EF and cut its effectiveness significantly compared to the relatively idiot-proof SS. Just casting EF with 2 holy power instead of 3 and only 2 BoGs instead of 4 (assuming no set bonus and no procs) would cut EF’s hot by roughly half. So I could imagine that for the cliche “not-good” LFR~normal mode tankadin may be better off using the nerfed SS over EF.

          I’m not really sure what my point is here. Just something that crossed my mind. Obviously the typical “answer” is for players to practice and get better. But it just seems like it’d be worth noting that the 2 talents scale very differently with skill. And that for a more casual tank who isn’t confident in their ability to keep track of EF properly they may consider taking SS as a “safer” option?

          • Theck says:

            Yeah, I think there’s definitely a skill-based argument to be made there. Even though “4+ stacks of BoG and 3 Holy Power” isn’t exactly a high requirement, Sacred Shield might be a safer option for players that are just learning, or aren’t confident in their ability to micromanage EF.

          • Meloree says:

            I’m always a little concerned about skill-based arguments, because they aren’t always well defined. I highly doubt that “the skill question” has ever been demonstrated to be a non-factor. That said, I’ll also happily jump into it.

            If you’re 80% efficient with your rotation, but otherwise perfect, that’s one kind of “low skill”. If you’re 100% efficient with your rotation, except that you regularly get it wrong, that’s another kind. More commonly, you’ll see both. If you aren’t running your rotation efficiently – and, seriously, nobody ever seems to wrap their head around this, but most people don’t – then your BoG stacks will be lower on average, and your EF casts will be a higher proportion of your HP spent, so there’s even less ShoRs availables to spend.

            I’d be really interested to see sims run with someone executing a rotation correctly… but slowly. I think the “skill” setting in SimC basically has you pushing the wrong button, which isn’t helpful in this case.

            I would be interested, though, to see if EF really does have harsh negative scaling with “skill”.

          • Çapncrunch says:

            When I say it’s often been demonstrated as being a non-factor, I’m mainly referring to the large jump in haste’s value after SoI was added to the model, as well as the realization that haste has just as much passive benefit as the other stats in addition to its active benefits. Which took a lot of the wind out of the skill arguments against haste.

            I’ve run a few sims to demonstrate the “non-issue” of skill vs haste doing things to the priority list to emulate different forms of poor-play aside from the built in skill setting. Such as removing SotR from the rotation entirely (since without spending holy power it doesn’t matter how often you use your generators), and haste still beat avoidance (though it did drop below stam considerably). And you can do things like enforcing a cd on abilities to simulate slow play at least with regard to some abilities. And in the end you need to do some really REALLY extreme things to knock haste below the other secondary stats. I’m think there’s a way to make the sim wait a set time between actions, but I’m not sure exactly how it works if it’d be able to enforce a gcd or not.

            And it was worth noting that in some of these things I did it pushed my normally sub-1k TMI up to something like 30k or so against the same boss with haste still sometimes winning. Which more or less showed that by the time you simulated a skill-level low enough for haste to fall behind you were probably already at a point where your choice of stats wasn’t going to change your likelihood of falling over 10s after the pull anyways.

            But the EF vs SS question was different since a lot of the little mistakes that people can make in their rotation can have a big impact on EF’s effectiveness while they have little impact on SS. EF is set up in such a way that by itself it’s actually rather mediocre but is amplified by a number of factors in a way that if used properly it becomes very good, which isn’t a bad thing, but is something to consider when comparing it to something as simplistic as SS. EF also has the strange behavior that it can actually scale up with downtime vs refreshing it too often (since if you forget to refresh it that downtime may give you an extra BoG stack vs early refreshes that will clip the benefit of BoGs).

            I haven’t run any 5.4 sims yet, let alone any “skill” sims regarding EF, but I could imagine a few to do it. Could remove most of the refresh conditions on EF and add an enforced player cd of maybe 20s to it to simulate someone refreshing it too early/often without regard for their BoG stacks or current holy power. Could also add a player cd of over 30 seconds to simulate downtime on it (maybe 40 seconds so it only has ~75% uptime). Could remove/reduce the holy power condition too, though removing/reducing the BoG condition probably wouldn’t do anything since BoG stacks are mostly going to be a function of the time between recasts (actually overall the BoG condition could probably be removed completely for that reason alone).

    • Thels says:

      Keep in mind that, at least early on, you also need HoPo to keep the buff up. If you want a 5 BoG EF, you need 6×3=18 HoPo, not 15 HoPo. That basically means you’ll have one BoG less than you would with the T16 bonus.

      • Çapncrunch says:

        Oh right, I was assuming in my calculations of haste requirements that the T16_4p was present. Without it maintaining a 5BoG_3HP_EF would be impossible (well possible depending on GC/DP procs) since the standard rotation floors out at 5.4 seconds per SotR (5hopo per 9s or 32.4s for those 18 hopo), so instead my numbers would shift down 1Bog per level: 35% haste to maintain 4BoGs, 8% haste for 3BoGs and 2BoG regardless of haste.

        Without the set bonus you’d need procs to shrink that 5.4 average SotR time down to 5s.

  13. Zapelm says:

    I’m not expert on SimC, but I don’t see in the list any priority for GC’ed AS.. it’s a core aspect on paladin tanking and IMHO should be represented now that GC is completely unlinked from CS and buffed in procchance: just like Ret prioritize exorcism on AoW procs, Prot could potentially lose GC procs keeping it that low in priority. Even if I’m doubtful it will change much and surely will not dethronize haste, it could be interesting to keep things in line. If possible, it could be interesting to see how this scales on numbers of adds too.

    sry for my English, for being a bit offtopic, or just plain wrong :P

    just my 2c

    keep up the great work

    • Theck says:

      It has almost always been a net loss in holy power generation to prioritize Grand Crusader procs over CS or J. The 5.4 change to Grand Crusader’s triggering method shouldn’t change that against single targets. And AS is already our top filler ability without a Grand Crusader proc. So the priority would be:

      /crusader_strike
      /judgment
      /avengers_shield,if=buff.grand_crusader.react
      /avengers_shield

      Which makes the grand crusader line completely redundant. That’s why there isn’t a line specifically handling it.

      In AoE situations we follow a different rotation entirely depending on whether we’re maximizing damage or maximizing holy power, and in that case you might prioritize AS above the others, but by that point you have enough mobs attacking that every AS cast is associated with a Grand Crusader proc anyway.

    • Çapncrunch says:

      Also, iirc while earlier in the expansion ret prioritized exorcism over at least judgement, maybe CS too (note that the AoW proc itself was irrelevant, unlike GC ret gets no special benefit to casting exo during the proc), but in ToT between weapon scaling and the sword of light buff, exo is now the lowest priority holy power generator for ret as well, so really in that respect ret and prot treat their procs the same way.

  14. Kal says:

    OK, first off, a discussion between Sloot, Theck, Mel, and Thel about stats? Nerdgasm!

    I’ve always had the philosophy while playing as a healer that HoT’s are a method of slowing damage intake. In other words, HoT’s are spike mitigation. I have no qualms switching to EF in principle.

    I’ve also had the philosophy that overheals are spike mitigation as well (I’m thinking most especially of Greater Heal spam as Disc in Cata). If damage was taken between casts and was healed, all well and good, but if it wasn’t, no matter, bubble anyway. But more importantly, the potential to bring a tank up like, BAM!, like the hit never happened, that was the main motivation to spam.

    • Meloree says:

      I love HoTs, but I wouldn’t really call them spike mitigation per se. Primarily because mitigation has a really specific meaning in my particular tanking lexicon, and all healing is a response. I mean, I understand what you’re getting at, and HoTs definitely serve the purpose of smoothing damage intake, I’d just be really careful about using the term mitigation there.

      I’ve always personally put HoTs in the bucket labelled ‘incidental healing’. Over any given 2 second period, you can’t really count on incidental healing, it might not tick. But over 4 seconds, it’s immensely powerful. As you add more and more HoTs, you make it more and more likely that at least one ticks in a given short period. So HoTs are always a good safety blanket, and very strong in the aggregate, but not necessarily something that you can really count on too much in burst situations where your tank might be getting globalled – where absorbs and other pure mitigation effects are correspondingly extremely strong.

      What makes EF a lot more interesting is that paladins are running so much haste that getting a tick after any given melee swing is extremely likely.

      • Kal says:

        Understood on the mitigation nuance. I meant damage smoothing, of course. I also like that phrase “safety blanket”.

        Ultimately, though, the math wins.

        • Thels says:

          Agree with Meloree here. The thing is that EF always ticks one attack faster than SS does. That’s quite important.

          Let’s assume a decent amount of haste, that would place SS tick timer between 3 and 4.5 seconds. Let’s assume a boss with a swing timer of 1.5 seconds, who hits hard enough that a string of 3 attacks will kill us without any mitigation.

          Because SS ticks at least every 4.5 seconds, it ticks at least every 3rd boss swing. For a string of 3 attacks, it will either tick before the first swing, or in between one of the swings, and mitigate the damage from the next swing. Therefor, SS will be a reliable source of mitigation for a string of 3 attacks.

          If EF would tick as often as SS, it would not be as reliable. Remember that an SS tick before the first swing is useful. If we take an EF tick before the first swing, it would do nothing. We would then eat 3 attack swings from the boss before EF would tick again. But it wouldn’t, since we’re now dead on the floor. EF would not be a reliable source of mitigation.

          Fortunately, EF ticks twice as often. That means that with SS ticking between 3 and 4.5 seconds, EF is ticking between 1.5 and 2.25 seconds, or at least every 2nd boss swing. That means that EF will always tick between either the first and second swing, or the second and third swing, healing back up the damage caused by the last attack, and thus a reliable source of mitigation for a string of 3 attacks.

          Increasing haste to push SS below 3 seconds, and thus cause it to tick every 2nd boss swing, would push EF below 1.5 second, and thus cause it to tick every boss swing, so the same principle applies.

          Reducing haste to push SS above 4.5 seconds would mean SS tick every 4th boss swing. EF however would remain between 1.5 and 2.25 seconds, and apply every 2nd boss swing, which would make EF reliable for shorter strings of attacks than SS. However, at these haste levels, it might be hard to keep a decent amount of BoG stacks active.

          • Kal says:

            What if we matched bin size to the rate of SS or EF ticks?

          • Meloree says:

            @Thels: I’d still be really careful about calling healing “mitigation” in any context. I’ll grant that EF is more reliable (higher frequency), and thus pretty much the entirety of your post. But, again, order of operations can matter. If you die from a hit, EF can’t heal you back from that. SS can prevent the death. It’s just worth bearing in mind.

            @Kal: To what purpose? The bin-size is based on number of boss swings, adjusting it based on haste would basically serve the purpose of normalizing with respect to haste. What information does that give us?

          • Kal says:

            I meant matching bin size to boss melee swings within the window of 1 SS or 2-3 EF ticks. I don’t see how it would normalize haste, so much as adjust bin size automatically to reflect haste’s real value. Isn’t a flat 4 swing window kind of arbitrary?

    • Theck says:

      I think that the comment system is trying to interpret the less than and greater than signs as html tags. Putting them in code tags should fix it:

      actions+=/eternal_flame,if=talent.eternal_flame.enabled&dot.eternal_flame.remains<2&buff.bastion_of_glory.react>3

    • Wrathblood says:

      Interesting, Mesala. Not requiring BoG stacks actually hurts if you’re running with HA or SW, and getting out that first weak EF before waiting just a couple GCDs for a good one ends up being a minus.

      But with DP, only casting it with 3 HoPo and recasting with less than 5 seconds on the current EF HoT, is the single best TMI (on a light movement fight, with my gear, on T15H25) I’ve seen. I even tried adding on additional conditions, like only casting the 0 BoG EFs if there was no EF running at all, and also allowing it to cast EF with more time remaining if it happens to get up to 4 or 5 BoG stacks, and I couldn’t find a combination that was an improvement.

  15. Pingback: Prot [5.4] - EF You - Elitist Jerks

  16. Hìjaal says:

    Why are people so stupid and Haste stack as Tank…….stacking Mastery is the correct way.

Leave a Reply