A model for quantifying tanking trinket proc/on-use values

Hi all.

We all know how to quantify the DPS trinket procs: normalize proc strength over expected uptime, and voila! you got a passive stat equivalent. However, I have never seen any model for quantifying tanking trinket procs.

Well, today on forums.worldofwarcraft.com I accidentally stumbled into an argument about trinkets; and being challenged on my casual statement that the Glyph of Indomitability on-use bonus is better than the Black Heart proc, I came up with a quantification technique to show exactly why that is so (which point seemed rather obvious to me, so I never bothered formalizing it before).

I would now like to propose that model -- and if anyone had done similar work before, please do not hesitate to point it out to me; I have not been able to find anything of the sort. I will keep using the Black Heart proc and Glyph on-use, simply because I think it makes for a very instructive (and apparently unexpected to most) comparison, and serves as a rather simple illustration of my model.

To start with, I think it rather obvious that unlike DPS trinkets, tanking trinkets' procs and uses are expected to kick in at the "oh crap!" moment. As such, you don't want to spam it on CD -- you want to fire it off in the "tank killer" situation.

What constitutes a tank killer? A common situation is when the tank must eat a sequence of strong attacks, while not getting healed, or getting insufficiently healed, due to fight mechanics (not getting healed due to healer incompetence should not be considered). An example I saw recently was Soul Reaper on LK: it fired off right after an unlucky Defile merged with the previous one, separating the healer from the tank by more than 40 yards. In such a situation the tank had to do what they can to survive the flurry of attacks while the healer, being a shaman with no "run through defile safely" CDs, had to go around it.

So the basic concept here is a flurry -- a tank must be able to survive a flurry of attacks, by avoiding/mitigating some percentage of the incoming damage. This is the time to blow your trinket's on-use, or to pray/hope/meditate that your trinket has procced at the right time. The shortest possible flurry is 2 attacks (since decent tanks never get one-shotted except through someone's gross incompetence, e.g. dying inside Frostmourne, which also should not be considered here).

So, how do the Glyph's on-use, and the Heart's proc, contribute to the tank being able to survive, say, a 2-hit flurry without getting gibbed? In the above Soul Reaper scenario, it's actually a 4-hit flurry occurring in a 2-hit time period, but let's keep it simple). We will first consider the basic two-attack flurry case.

A typical decently geared tank will have around 60% avoidance, or 40% in ICC. Let us say that any meaningful avoidance or mitigation on the flurry will permit tank survival -- i.e. that the extra armor from the Heart is as good as 100% damage reduction in terms of letting the tank survive until the next heal.

Then the basic chance to dodge/parry at least one attack from the 2-hit flurry, without either trinket, is: 1 - 0.6^2 = 0.64. (0.6^2 being the chance that both blows connect).

Now let's factor in the Glyph. At such gear level it will give you about additional 6% dodge; so the chance of avoiding at least one attack becomes: 1 - 0.54^2 = 0.71: a 7% improvement over the baseline case. Normalizing, your chance to die has been cut from 36% to 29%, or by 20%.

Now let us instead consider the Heart. it has 17% uptime. We will set aside for now the rather more complicated question of whether its proc will be up for the entire duration of the flurry. If the proc is up, let's assume that survival is guaranteed regardless of avoidance; if it's down, it's back to the baseline survival chance. SO the survival chance with the Heart proc becomes: 0.83 * 0.64 + 0.17 = 0.70, a 6% improvement over the baseline case; about 1% weaker than the Glyph's Dodge ability. Normalizing, your chance to die has been cut from 36% to 30%, or by 17%.

Now let us instead consider the Soul Reaper case, where 4 attacks will occur in the duration of 2, in addition to the Shadow damage. If you have to dodge at least one of the four attacks (which for the Glyph is now nominally close to about 25% DR bonus you get from the Heart proc), then the value of the Glyph on-use becomes: 1 - 0.54^4 = 0.915 -- compared to the baseline survival chance of 1 - 0.6^4 = 0.83, a nominal increase of 8.5%. Normalizing, your chance to get insta-gibbed has been cut from 17% to 8.5%, a whopping 50% improvement.

The value of the Heart would remain unchanged, or even go down as flurries increase in duration, because it's gonna get less and less likely that the proc will be up for the entire duration of the flurry. In fact, with Soul Reaper, the chance of the Heart proc being up for the entire 5-second duration is a measly 8.3%.

So this is how I propose we can quantify various tanking trinket procs and on-use abilities -- by determining how they will affect a typical tank's survival probability in a representative tank-killer situation. The quantitative value of the proc becomes the improvement in survival chances under such a baseline scenario. If nobody has proposed such a model before, I would like to call it TSB -- Tank-killer Survival Bonus model.

Not surprisingly to me, but apparently surprisingly to many others, the absolutely random chance of the Heart proc basically kills the advantage of the proc itself. Armor is great, but godly armor at random times is much less useful -- dodge is far less reliable than armor, but dodge on demand, because it applies to every attack individually rather than to all attacks as a group, is more reliable than a flat armor proc chance, as the value of dodge will very rapidly improve with the increase in the flurry duration. Even for the basic 2-blow flurry the Glyph controllable dodge bonus offers a slightly better survival chance than the Heart randomly times armor bonus.

Yet to be modeled is the value of the "on hits which take you below 35%"-type procs, like CTC. My sense is that while CTC's passive bonus is crap, its proc is absolutely godly -- with half-way competent healers, that proc, with its tiny 30-second ICD and 10-second duration, should be up for every single tank-killer situation on many, perhaps most, fights. That proc is nearly as good as Will of the Necropolis talent. It will give you more damage reduction , but only against physical damage (no big deal) and on 30-second ICD (not very pleasant) with 10-second duration (excellent). Again, in Soul Reaper scenario, with competent healers the CTC proc should be available for every single Soul Reaper, thus being nearly as good as passive 5.7k of armor -- because of course the time when you need that extra armor is when the tank gets hit the hardest, during the light-hitting times the extra armor would mostly serve to make overhealing even higher.

It would also be nice if someone more comfortable with the EHP model would more precisely quantify the effect on the proc on the tank-killer survival, as in the above example I basically used back-of-the-envelope math; but the basic idea -- quantifying the proc value as the improvement in the chances to survive a tank-killing flurry of attacks -- should be sound.

Thoughts? criticisms? rotten tomatoes?

First of all, congratulations on a worthy Wall of Text. =) Once my eyes are done falling out I shall applaud you properly.


The point is a fine one, the trick is simpler than pure numbers:
Black Heart = Armor proc at random
Glyph = Dodge boost on-use

The initial challenge is comparing avoidance to armor. The short-hand of this is comparing guaranteed small cuts off of every hit, compared to completely removing random (but few) hits off the chart. The only way to compare them equitably is on a long-scale average for total damage reduction. If you want to look at micro-examples there are two simple considerations:
1.) Avoidance can vary heavily between *didn't happen* and sickeningly overpowered because you just dodged/parried 4 swings in a row.
2.) The damage reduction of a bump in armor is very small on any given hit, but it is guaranteed to happen.

The avoidance is enough of an X-factor that you can't be certain it will help, but when it does it can be a major effect. Many people simply don't like the question mark, but that doesn't *actually* reduce its effectiveness which is the downstream assumption. The armor on the other hand is easy to say it will definitely help (in your example, if it is proc'd at the right time), but you can question the scale of whether or not it will really save you since the health margin it improves is small enough that only things right on the razor's edge as determined by the healer will it be the deciding factor.

You can do the math, as you did, to decide which has the overall largest damage reduction, but if you pick your specific test cases, the difference in how they work will make all the difference. Black Heart is an always-on ability, in effect. It will not always grant the armor bonus, but you cannot use it so you can only rely on it as an average value. Glyph has the on-use ability making it a burst CD, so you will definitely have it when it matters most, so for that case something like the LK would make the proc portion better.

If you want to consider the entire trinket though, you have to question whether the armor on-equip is a superior value to the stamina on-equip. There is a lot of magic damage that gets thrown around, and that additional health could suit you rather better than the armor.

The final place I would land on it is this, for your specific example (LK), not in general:
Black Heart would likely be a superior value overall for pure average protection (largely due to Stamina)
Glyph would be a less total value, but if the place you are most concerned about is strategic protection, it may be the superior choice for you in the context of your group.

This is all very vague and general though without the context of the rest of your gear set, your strategic choices (I swap the LK on his Soul Reapers with my tank counterpart), and the strengths and weakness of your group, in general, and in each specific attempt.

Quote:

---

Originally Posted by Satorri

First of all, congratulations on a worthy Wall of Text. =) Once my eyes are done falling out I shall applaud you properly.

---

Sorry about making it long, I thought this was an issue deserving of exploring in some depth.

However, I don't think I did a good job of explaining what exactly I was trying to accomplish.

My goal was to reduce the trinket procs to a single universal value which could then be compared across different trinkets, kinda like DPS trinket procs can be reduced to EP; to quantify not merely the trinket proc value in EP terms, but also the proc nature (on use vs. random vs. -- at some point -- "on hits below 35%"). However, because of how tanks tend to use the trinkets, I don't think the baseline EHP model addresses this issue -- it doesn't take into account the controllability factor, and AFAICT it can't do so in its baseline form.

Quote:

---

The only way to compare them equitably is on a long-scale average for total damage reduction.

---

But this is where I think the analysis goes wrong -- this approach doesn't even try to quantify the immense value of the fact that the Glyph proc can be controlled.

I mean, of course we could go the DPS route and simply convert the proc into EP points, but that's not how tanks use trinkets, right? the EP model ignores the fact that for a tank, a random proc is A LOT less valuable than a controllable proc or a smart ("on hits which take you below 35%") proc. I attempted to quantify exactly this difference, i.e. how valuable the controllability is -- to turn that qualitative difference into quantitative difference.

Do you think this is the wrong way to go?

Quote:

---

You can do the math, as you did, to decide which has the overall largest damage reduction, but if you pick your specific test cases, the difference in how they work will make all the difference.

---

That is true. I tried to pick a baseline case, the two-strike tank-killer situation -- i don't think a typical three- or four-strike tank-killer qualifies as "tank killer", with very rare exceptions like Soul Reaper (due to haste).

If you think there are better representative scenarios to plug into the model, it would be great to try that.

Quote:

---

If you want to consider the entire trinket though, you have to question whether the armor on-equip is a superior value to the stamina on-equip. There is a lot of magic damage that gets thrown around, and that additional health could suit you rather better than the armor.

---

Right, Stam > Armor > everything else. But to compare the entire trinket, we would have to have a way to quantify the procs -- and not just their stat/uptime, but the way they would help an actual tank in an actual hot spot.

Now at the moment the trinket proc is still not reduced to the universal tank medium of exchange, the EHP -- instead we have the % chance to survive a tank-killer; but this is IMO a step in the right direction, as now we have only two values to aggregate, instead of wildly disparate and incomparable procs.

Perhaps a way to convert the % value into EHP would be to determine how much EHP would be needed to survive the equivalent tank-killer flurry, and then prorate it for the percentile improvement? E.g. if the Glyph gives you +7% survival chance, and +25% EHP would be needed to survive the same tank-killer without the Glyph, then the Glyph's proc would be worth 0.25*0.7=1.75% of your EHP for comparison purposes?..

That is the problem with survival stats that is at the baseline of why there isn't a definitive mathematical proof available for the Avoidance vs Health argument.

There are 3 main tank survival elements influenced by gear:
1.) Health
2.) Armor/Physical Mitigation
3.) Avoidance (Dodge/Parry/Miss)
(4.) Shield Block and Savage Defense are relevant, though rarely focused on)

Each element plays a distinct role and works in a specific fashion. You can't really create a universal standard that applies in every situation for a few reasons.
1.) The relevance of health is proportional always to the scale of healing, scale of damage, and the filter of mitigation (thus EH reducing health and mitigation into one value that works so long as you have established proportional values of mitigate-able damage to non-mitigate-able damage, or assume that one or the other is not present).
2.) Avoidance is stochastic, which is a popular word here lately, meaning that if you take anything other than a long statistical average you cannot be certain it will have a specific effect. You can have 20% dodge and dodge 5 swings in a row, or not dodge anything in 5 swings. If you take it as a long average, it is fair to assume that you will dodge 1 of those swings.

Trying to compare trinkets on a single standardized value is the same as trying to compare health, armor, and dodge rating on a single value, which doesn't work because on top of the inherently apple/orange nature of how they work, they will also have different significances at different levels of interaction and in different scenarios.

You can surely try to compare them or create a standard for comparison, but my main advice is this: identify the assumptions you make so that it is clear to readers what situation the analysis will apply in. Be clear, and show your work, or people won't have anything to provide feedback on, other than "lawl I hear avoidance is bad!"

And above all else, try to find the simplest way to represent what you want to demonstrate. You're not in unexplored water, but maybe you can find something good to offer. It is always worth trying.

Quote:

---

Originally Posted by Satorri

That is the problem with survival stats that is at the baseline of why there isn't a definitive mathematical proof available for the Avoidance vs Health argument.

There are 3 main tank survival elements influenced by gear:
1.) Health
2.) Armor/Physical Mitigation
3.) Avoidance (Dodge/Parry/Miss)
(4.) Shield Block and Savage Defense are relevant, though rarely focused on)

Each element plays a distinct role and works in a specific fashion. You can't really create a universal standard that applies in every situation

---

Right; but you can create a mathematical model which, when you plug in specific damage and healing scenarios, will give you a formula for conversion of avoidance and mitigation into EHP. This is kinda like what I am trying to do, only for the intrinsically situational procs. Of course YMMV depending on the conditions you are modeling, so the Avoidance <=> Mitigation <=> HPpool conversion was different in BC that it is now, it's different for different classes, roles, and bosses, etc. Nonetheless, there's a rough overall conversion function which people use as a rule of thumb in the current content. I am aiming for approximately similar level of "precision".

Ultimately yes, the evaluation of any two pieces of gear will be multivariate, and their ranking is a partially ordered set; but the result -- "equip this or that" -- is always a binary choice, which in turn entails a fully ordered ranking set, whether we explicitly formulate the ordering or not. Thus we simply have no option but to reduce and simplify the complexity of gear stats to a single yes-or-no evaluation. Might as well do it explicitly up-front -- the alternative is an informal, non-quantified evaluation along the lines of "LAWL avoidance sucks".

Quote:

---

You can surely try to compare them or create a standard for comparison, but my main advice is this: identify the assumptions you make so that it is clear to readers what situation the analysis will apply in.

---

OK, I thought I had, but I will enumerate them explicitly here.

1. Most importantly, I assume a specific type of tank-killer scenario, where a tank gets two- (or three- or four-) shotted due to missing or insufficient heals, and must avoid/mitigate a given fraction of that damage until the next top-ff comes. In fact your analysis of the Unholy tanking, and of the utility of Bone Shield against tank-killer, was in part what prompted this line of thought for me.
2. Less importantly, I assume a given typical level of avoidance (60%) and the ICC debuff (-20%).
3. Finally, I assume that the on-use CD will in fact be available when the tank-killer situation manifests.

Do you think I am missing some assumptions?

Quote:

---

And above all else, try to find the simplest way to represent what you want to demonstrate.

---

% increase chance to survive a tank-killer is the best and simplest representation I can come up with. It seems pretty intuitive to me, though if someone can find a better framing for it, I would accept it gladly.

Quote:

---

You're not in unexplored water, but maybe you can find something good to offer. It is always worth trying.

---

Can you give me some pointers to prior work on quantifying not just the proc value (that is a solved problem, we can convert it into EHP using the DPS-type trinket evaluation) but also the proc type (on-use vs. random) as pertaining to tanks specifically? I have never seen anyone put a numeric value to the controllability of the on-use procs, i would love to read some prior work on this issue.

Quote:

---

Right; but you can create a mathematical model which, when you plug in specific damage and healing scenarios, will give you a formula for conversion of avoidance and mitigation into EHP.

---

I think this is harder than you are assuming.

Quote:

---

I think this is harder than you are assuming.

---

Isn't that along the same lines as what Aggathon and Thegreatme spent a night working on together and ultimately just could not do it?

Quote:

---

Originally Posted by Kazeyonoma

I think this is harder than you are assuming.

---

It's a hell of a hard problem to solve because avoidance is probabilistic by its very definition. Transforming probabilities into the basically linear domain of EHP is not a trivial exercise.

Edit: If you really want to make a go of it, perhaps look to the Laplace transform. Using Laplace might let you use a first passage time model with avoidance probabilities transformed into moments mapped into it.

To condense the crux of the problem:
You could take the avoidance percentage (say 30% for argument sake), and amplify existing EHP by this value, much as armor or resistances increase the value of stamina. The fault in this lies with the fact that armor or resistances does this every time guaranteed (yes, I'm aware of how resistances are not quite guaranteed, but it's close enough), whereas avoidance is complete statistical in nature. Trying to force a probability on events that are guaranteed, or worse, two independent statistical phenomenon (boss attacks and avoidance), basically invalidates the whole point of EHP. EHP is the goal you seek, a RELIABLE measure of a tank's survivability, based on a collection of stats.

Edit: Bah, Satrina was faster and more concise.

Quote:

---

Originally Posted by Kazeyonoma

I think this is harder than you are assuming.

---

Right. The issue is if you were to properly do a model, it has to involve some sort of risk assessment based upon the healers. Not only do tanks have various CDs at their disposal so do the healers, and this leads to the question of identifying the proper "killer sequences." Except in maybe some extreme scenarios, we are left with cases that are the result of compounding factors which result in some unclear damage and healing scenarios.

I feel that the only way to identify these scenarios is a good amount of data reduction of combat logs. You'd have to identify the killing sequences, then you have to properly correlate them to their causing factors. Or even how much was the healers fault as compared to the tank. Both react to the situation at hand. Or was it lag? You'll likely find killer sequences in cases where the tank survived, and then have to assess whether that was by luck, or skill based. But if you're going to go that far then you might as well assign a value based upon those statistics gathered rather than try to model them.

I had thought about doing this, and even mentioned the Aggathon I may be willing to do this, but upon later thought felt the effort of acquiring such data and reducing it was far too great. Especially given the possible lack of return on the research. Ideally you'd want some sort of spectrum of tanks of X avoidance and Y survival rate, and then it has to be normalized against healer behaviour.

The only reasonable (in terms of effort), and precise, models that I think could be done at this stage of the game are ones where your tanks and healers use their abilities in a blind/naive fashion. While precise, it wouldn't be an accurate representation of reactive players. So we're left with heuristic arguments for which Aggathon wrote a great article. While imprecise, it offers an accurate assessment of the current state of tanking.

Come Cataclysm, this may change. But that depends on how much the healing game changes.

Quote:

---

Originally Posted by Satrina

Edit: If you really want to make a go of it, perhaps look to the Laplace transform. Using Laplace might let you use a first passage time model with avoidance probabilities transformed into moments mapped into it.

---

But this leaves you with the task of identifying the appropriate (partial) differential equations doesn't it?

Quote:

---

Originally Posted by danilche

Can you give me some pointers to prior work on quantifying not just the proc value (that is a solved problem, we can convert it into EHP using the DPS-type trinket evaluation) but also the proc type (on-use vs. random) as pertaining to tanks specifically? I have never seen anyone put a numeric value to the controllability of the on-use procs, i would love to read some prior work on this issue.

---

There are people here who are far more savvy in messy math than I am, but my focus is usually on what is most approachable and comprehensible, further to the aim of simplicity.

My goal would not be to try and generalize about everything because situation is a huge variable. Find common situations or use existing fights to inform your short-term models.

Healing is a very hard thing to give a fair shake to. If you try to make it a general X incoming healing model, you miss one of the center points of what kills tanks, your tank-killer scenario. In general, healers must be able to out-heal your incoming damage or no matter how good you are, you will never win. However, if you assume a constant flow of healing it is hard to find a reasonable scenario in which you will take more damage than you can survive in the space between the pulses without finding a scenario where you should be using a CD, *and* finding a rough spot that anything else the tank does can actually make a difference.

The aggregate value of survival stats on gear will set the general tone for your survival, but very rarely will little distinctions like which trinket you use *actually* represent the margin between you winning and losing. The gains/losses are relatively very small next to many other human elements, RNG effects, and the more extreme swings of CD-effects and healer skill.

So, I would try to find the most realistic situation you can describe, or an extreme misfortune (like say no healer available for a period of time), and try to work within that. I suspect if you were to track real situations you would very rarely be able to identify trinket A vs trinket B as something that would have made the difference between living and dying, but that is partly anecdotal.

There are lots of things floating around. If you look in Satrina's Evil Empire guides directory in the Critical Guides section you will find a lot of great nuggets. Aggathon and TGM have been doing a fair bit of brain crunching as well. Theck has done a lot of work, though I expect you will find a lot more of it on maintankadin.com than here. Mathcrafters have been trying to express this relationship since BC first introduced the possibility of loading survival values in a meaningful way.

Quote:

---

Originally Posted by BruisedOoze

But this leaves you with the task of identifying the appropriate (partial) differential equations doesn't it?

---

/shudder

I did not enjoy PDE's.

Quote:

---

Originally Posted by BruisedOoze

But this leaves you with the task of identifying the appropriate (partial) differential equations doesn't it?

---

Yes. That would still probably be easier than working in the non-transformed domain, though.

Quote:

---

Originally Posted by Satorri

I did not enjoy PDE's.

---

There's a reason I never tried it =) Well, that and the fact that I am not certain that this method would solve the problem either.

I've got a reflexion, not about the maths behind the choice of a trinket but in a matter of cooldown.

I mean that in my opinion the main difference between the black heart and the glyph, is that for the first one, the RNG issue is before the effect of the trinket, and for the second one, it is after. You said so in fact, when telling that if the Black heart proc, you will survive, but if you use the glyphe you doesn't have any guarantee about that.

So in a MT point of vue, when I have to take a flurry, my goal is to use the appropriate(s) cooldown(s) : juste enough to be sure that I will be alive, but not too much so that i could keep some for the next flurry. With the glyph, there is no combination of CD that alone is not sufficient but become OK if I use the glyphe with it (except for particular cases like getting more than 100% avoidance ).
On the other hand, with the Black heart there are combinations that will be OK if the trinket proc but not sufficient if not. So that the Black Heart proc become a cooldown saving.

So if I got the Black Heart, I can watch the proc and its internal CD, and keep in mind to use less personal Cooldown than usual if the proc armor is here. But if I got the Glyph, its /use ability could only help me if I feel that my healers can't get me full life, and they need some RNG by avoidance to do it.

I prefer the first option, don't you ?

It's an interesting question and I'm not saying you should be put off it but from a mathematical point of view you have made far too many assumptions. I haven't really thought about it too much but I think as someone has pointed out, it's a lot more complicated than you have assumed in your modelling. For example:

1) You have assumed that the results hold true for all gear sets, sometimes results can be reversed depending on the rest of your gear, there may well be a difference between tanks with 60% avoidance and 50%. Also, you have ignored block and block value which may play a part in the overall results.
2) You have assumed that either proc/on use will be the difference in surviving, particularly in the case of the armour proc, this will depend on your eh (I.e. your entire gear set) vs the possible damages from the swings.
3) When looking at the 2 attack flurry, you have assumed that you need to avoid one attack to live i.e. you can take one hit but in the 4 attack flurry you assume you still only need to avoid one attack but can now survive 3 hits and translated both scenarios into a 'chance to live' stat and compared them.
4) Tying in with the above, there may be another source of damage such as defile (not sure if it can be up during soul reaper but on other fights there may be other sources of damage going out in a dot fashion) or periodic healing and as a flat reduction and a total avoidance leave you in different states after healing it is somewhat complicated. There is the question of simply being able to survive but there will be some situations where both will allow you to survive 2 hits say but one will leave you in a better state afterwards.
5) You would have to take into account the overall mechanics of each fight and asses the likely hood and length of periods of not being healed directly or having no hots rolling, in certain fights this may happen and in others it should not unless your healers are afk.

You really need much more accurate numbers and to eliminate as many assumptions that do not always hold true as possible for all of the variables you are considering for it to be meaningful and even then you can only really answer a question that is part of the overall picture of survivability. That's not to say that's a bad way of doing things but I think you need to be a bit more specific about the question you are trying to answer and whether or not that translates directly into one proc being better overall than the other.

Quote:

---

Originally Posted by Menethal

So if I got the Black Heart, I can watch the proc and its internal CD, and keep in mind to use less personal Cooldown than usual if the proc armor is here. But if I got the Glyph, its /use ability could only help me if I feel that my healers can't get me full life, and they need some RNG by avoidance to do it.

---

That's a dangerous game to play in itself. It doesn't always proc as soon as the ICD is up and there's no guarantee that it will even be up at any point during an enrage phase. I'd take the Glyph over the black heart on most bosses just because it offers more baseline EH than Black Heart, and the on use is random when I want it to be, not when it wants it to be.

This is not a debate about Glyph vs The Black Heart though. While it's outdated now, you'd have to include in your model something like Commendation of Kael'thas. Those effects happen every now and then and would need to be modeled. I think one drops in Ruby Sanctum now, too. If your model can't describe that trinket then it is only marginally useful, unfortunately.

Quote:

---

Originally Posted by Petninja

That's a dangerous game to play in itself. It doesn't always proc as soon as the ICD is up and there's no guarantee that it will even be up at any point during an enrage phase. I'd take the Glyph over the black heart on most bosses just because it offers more baseline EH than Black Heart, and the on use is random when I want it to be, not when it wants it to be.

---

I was not comparing the two trinkets, but only their proc/on use effects. And as I said I prefere the one of the black heart for all the reasons above. And you saif that there's no guarantee to have the black heart proc, but there is no more guarantee that the avoidance bonus helps when I /use the glyph.

But, of course, what I said is useless if you want to put a number or an EH value to the advantage of one proc. And as Satrina said, I think that it is extremly complicated because that each boss needs a different approarch of cooldown and damage reduction ( LK needs a strict cd rotation, Halion needs a lot of panic button, etc ) that the heal team has also an importance ( for LK, 2pals 1 disc 1 holy 1 sham change a lot of things for the MT than 2 disc ) the team tank also ( war and intervention )