Avoidance is also helpful for reducing the only random damage spikes left: parry-gibs and plain consecutive hits. A useful calculation would be not only to see your chance to die over X seconds with no heals, but also how often you're put in that situation over the length of a simulated boss fight. Unless I misunderstand and that's what your resulting percentage is.

EDIT: ok, that is what you're doing, kinda. However, since you're not doing any damage besides auto attacks, I feel 5s is a faulty time interval since it isn't a multiple of the boss's swing timer. You're going to have intervals with differing amounts of auto attacks, and that will obviously favor sta gearing. However, since your calculation is no heals and auto attack damage, it's easy to see that avoidance and expertise would be favored until you have enough sta to survive another hit. Side note: how much variance is there on average boss hit and max boss hit?

I disagree completely as to what level of play you can expect from your healers. You take 42k damage over 5 seconds and you're going to consider a scenario where a single healer hits you once for 18k during that period? That is not a tank fail, that's a healer fail, plain and simple(specific fight mechanics withstanding). You cannot completely change your tank gearing paradigm for that big of a healer fail.

I'm not denying disconnects and distractions occur, but you can't change your playstyle for it. If you are, why not go the whole 9 yards and raid with 25 druids? Yes, a ridiculous and extreme example, but why not stack 25 brez's and innervates to account for any sort of player death and healer oom? How about a melee dps misses a kick on Vezax? Guess you should wear FR gear for that fight. Or a mage goes nuts on Hodir and forgets to invis during a flash freeze? Might as well glyph growl for when he pulls agro(and doesn't ice block in time too). I consider those situations as equal to 42k damage over 5 seconds and only 18k in heals from one healer when you most likely had 2-3 healing ONLY you.

Yes obviosly 5 sec is an arbitrary value. You can easly rerun sim for other values or anything you want. What we should focus is that there is always a trade of. Obviously avoidance is good but what you miss? Obviously health is good, but what you miss? What I'm trying to stress is that as a tank what you really want is to make your incoming damage as predictable as possible (in the same way I usually blame when our healers go for crit instead of spell power/haste because I actually prefer a quick and constant healing as a tank than a RNG based healing and that's why I really love disc priest, shields > everything).

Given the BiS gear you have very few choce between different gear and the great component is gemming.

Looking at the report I've previously posted going from stam to agi you are trading:
- 5624 hp
+3.95% avoidance
+648 armor (0.39% DR)

Actually (differently from you) the only fight where I'll wear an avoidance set is Vezax for the simple reason that mana is a problem in that fight and reducing dmg taken will reduce mana expendure. For istance on a fight like Vezax I'll aim at the following:
I think I should have enough stamina to resist 2 hits and enogh expertise to completly avoid parry gibbs then the healer know that they can heal me in that 2 seconds window after I get an hit but usually I'll get less hits.

But in a normal fight mana is not a problem so, apart from healer "noobness" or particual fights effects like silence and so, what can kill a tank? Basically nothing. Suppose that you have 50% avoidance and I have 25% avoidance, I'll take on average double damage, but if the healers are constantly spamming heals on me and or on you and they can spam heals on me without going oom, what's the difference? The only difference could be that the healers doesn't have enough HPS to keep me alive. But it's the same for you. We can both have a double or triple hit sequence but they should have enough HPS to substain it or you will die as I will die, yes I'm more likely to generate that event but this is not important, when I tank I must be 100% sure that I can survive also if RNG goes shit and I take 10 hit in a row and this can happen with 0% avoidance as with 99% avoidance.

Anecdotically, last time we did Hodir we had only 5 healers for that night, the first attept was made with our warror MT, he has around 41-42k HP in frost res gear. He toke a 23k+23k+26k hit -> dead, in that few second he recived few heals because our healers were occuped to don't let our raid die from AoE damages. We did it again with a druid. 55k hp in frost set. There were 0 problems.

As a last thing: many fights involve aoe damage or magical damage, the only way to survive them is stamina, you can't avoid them.

No, I agree with your gearing option for Vezax, I just shorthand it as "more sta" most of the time.

Right, I guess what I'm saying avoidance & mitigation has more marginal utility than sta after you reach the hp soft cap(X max boss hits). If 5k more HP means the boss can still put you under in the same amount of consecutive hits, I don't feel that helps that much. As we've established, that extra HP mainly helps in catastrophic healer failure and you receive much less healing than you expected. The extra avoidance however will reduce minor healer failure situations. Since you're taking much less 'consecutive hits' situations, that healer can get lazy or bored and heal someone else(I'm not advocating a healer should get lazy/bored). Or just spam flash heal while he watches TV. Or get chased around by Kolo's eye beams. And those situations do occur a lot. Lots of fight complexity is designed around making healers stop heals for a brief second or two. If it's longer, or you don't have any hot healers, etc, then yes, wear more sta.

The Hodir example is kinda skewed(not saying it's invalid). Hodir is THE feral fight. We can up the FrR and sta and not worry about crit immunity. I'm assuming that's why the warrior had such low HP with FrR gear on. And again, that's another situation you could analyze before the fight started. Short on healers, they might have to divert attention to heal the raid, I should wear more sta.

And I'll agree about magic damage, more sta for those fights.

That's why I prefer to gear for agi as a baseline and then work in a few pieces of stamina as needed. Avoidance & mitigation take in a lot of stats, so it's best to use your best gear for that. But more hp only needs more sta, so you can use "suboptimal" pieces like old T7 gear or Polar since you don't care about the mitigation on those.

Deathwing, your "minor healer failure" situations are exactly the ones where stam plays a part: when you receive less than 100% of the healing capacity necessary to stay alive. But it's not an absolute all-healers-or-no-healers scenario. Unless your healers are bunched up (strategy fail), you won't see situations where multiple healers are stopped at the same time (hodir ice/Vezax cloud jumping/Kolo lasers/Mimiron rockets or lasers).

Also, sometimes bosses will land multiple weak hits. Say a boss's swing can end up being 16k-20k. Sure, you may not be able to lift your health to 60k to sustain 3 max hits. But raising health from 46k to 50k means you can survive three weak hits without heals. It's eliminating parts of the RNG you can't affect with avoidance.

I definitely think there's no definite answer about where health and avoidance are eventually ranked, and it changes every fight. That's why it would probably be more useful to run lots of simulations with different time periods (2s/4s/6s/60s), different boss damage (60k/80k/100k), and healing simulations (0-4 healers active) and graph all of the results before you can say one is definitively better than the other. I think it's fair to say that at this point that they both have uses, and choosing one over the other is definitely a matter of situation/preference until the math gets done more precisely.

At this point though, I think we're just throwing hypothetical/anecdotal situations around without real concepts of how often one or the other will save us.

I actually meant "minor healer failure" as in you receive less than 100% expected, but still enough to stay alive. I see where sta and avoidance can help there.

I did ask about the damage range on bosses, it's not something I've fully considered.

I agree we're just throwing out hypotheticals, some robust simulations if Nightcrowler is gracious enough to indulge us would be fantastic.

If i'm not completely mistaken, Shred outperforms Rake DPE-wise (even considering 2t8) with high ArPen and either Grim Toll, Mjolnir Runestone or both active. So maybe you could add another condition - as long as your current ArPen > 100% (this is the break even point +- a few percent depending on your crit/ap) you ignore rake and go for shred instead.

This could even be amplified if you save energy when your trinket's hidden GCD is almost over (not screwing up your rotation in the process of course). This might be almost impossible to code though.

Yes, but be aware that the only thing that will trigger that in which the healers are unaware are parry gibs. Every other spike damage the healers and tank should be ready for. Based on Nightcrowler's sim, only 1.3% of the boss's attacks are parry hasted. So your options are:

1. Gem/enchant for sta to minorly help the case that the boss parry hastes AND your healers aren't prepared to deal with it(0.013 * <chance your tank healers aren't full on spamming you>)

2. Gem/enchant for expertise and agility to really reduce that 1.3% from happening and then from connecting with you too.

But now we're going in circles(not meant as an insult). Better models are definitely needed to generate conclusions.

I just don't agree with this. First of all, it implies that the healers are totally 100% aware of what you are doing and what the boss is doing at any given instant, which isn't in practice reasonable. It also implies that a healer is going to be aware that you're 20k down and that you're going to take another 20k hit right away no matter what - which again, not a reasonable assumption. The big assumption is that it assumes that the healer is always casting the right heal at exactly the right time and isn't doing something else; you're assuming optimal healing at optimal time, and that's really not fair - especially if you're not using totally optimal healers for a job. For example, most holy priests aren't going to be casting gheals all the time; they'll let the paladin do it. So they might be assuming that you'll be healed in 4 seconds for 10k, and they'll continue to heal with smaller heals for a bit until they realize that things are Really Bad.

So you're down 40k, then instead of going to full the paladin moves and the priest heals a 10k heal at best. Now you can be killed in 4 hits without parry haste just because the healers weren't perfect.

It's that level of fuzziness that really needs to be dealt with in simulators. My gut feeling is that anything that can remove that chance at all is going to be 'easier' to heal than having raw stam, but I'd like to see a suboptimal healing model.

If mana isn't an issue, why can't we make those assumptions? All hots, highest throughput, spamming them on the tank. And what do healers have to be aware of besides boss cooldowns and abilities? I think that's a perfectly fine assumption. 1 MT healer missing plasma burst is a HARD wipe. Not a "oops, if only he had 4k more HP" wipe. As for things that are truly random like kolo's stone grip, then yes, the tank should change his gear accordingly.

EDIT: I know what you mean, I'm more playing devil's advocate with this response. Players don't play 100% perfect, I'm not denying that. But that's a slippery slope, changing your gear for other players goofups. As I said above, if you take even minor goofups into account(minor in execution, not consequences) there would a whole slew of options you'd have to consider for each fight.

I feel it's better to reduce the truly random damage spikes through more effective means(expertise and agility) and hope your healers learn from their mistakes.

@Malazaar: the simulator (cat) already have a switch to spam shred instead of rake for high Arpen values. From simulated data I've seen that the right spot for the switch is when:

Rake DPE < 0.9 * Shred DPE (basically you need shred to be 10% more efficiently than rake to be e better choce because also if slightly less efficient having to spent less energy compensate for that).


As for the addon, it's already included in the latest (not still downloadable) release (still working on it).

@Deathwing: mm, I really don't know how much sense should have running a simulation that much complicated but I could try it out, to reach enough data point will be a little pain time-wise but it could be done.
I don't agree with you when you say that agility reduces damage spikes. Expertise reduces damage spikes reducing the randomness but agility increase randomness. Actually if you could have 0% avoidance the damage will be less random (and spiky) than with 50% avoidance.

A better way to come up with good hp numbers will be studing one by one every fight in ulduar setting up special abilities and so on and find out at wich hp level you "can't die" on that fight. For this reason if anyone want to contribute will be welcome (I need value pre-armor and also specials, silences and so on).

It looks like armor penetration will cap out at 100% soon. The question is, will it be better to gear high armor penetration or gear for roughly 50% and use 1 armor penetration trinket + wrathstone for maximum dps?

This topic was beat to death back during Brutallus. I'm sure the threads can still be found somewhere. The exact same issues were argued at the time. Some people wanted to stack stamina, but it was basically determined that avoidance actually provided the best survivability. While we do need to plan for worst-case scenarios, it ends up better to reduce the frequency those cases occur.

I do think there are break-points though. When starting out you need to get your raw survival high enough to live through the big spikes. This reduces your avoidance too much and causes those spikes to occur more often. However, since you need a minimal amount of health the best option is to further reduce your avoidance and increase health even more. This makes you a bigger mana sponge, but gets the job done. Now, as your gear improves again you reach a break point where the spike frequency is reduced sufficiently to lower your health again. Where those break points occur I don't have pinpointed (and would be encounter specific).

There really aren't many fights in Ulduar that test the survivability of a tank like Sunwell did. Pre-nerf Ignis, hard-mode IC, and Thorim with high-stacked buff are examples I can think of. Most of the hard modes are actually increasing raid complexity (not tank) or just beating a dps timer. Tank healing is rarely the fail point now.

@Ramenchef:
actually it's better to don't change basically anything except switching one trinket (either Grim Toll or Mjolnir Runestone).

If you look at the main page with ulduar BiS gear (gemming everything except metasocket needs for 16 arpen) we have the following values:

Armor Penetration: 54%
Boss Damage Reduction (after debuff and penetration): 20.35%

without trinkets.

Using one of the arpen trinket will push you (for the proc time) to:
Armor Penetration: 103.7%
Boss Damage Reduction (after debuff and penetration): 0.19%

I don't know if they will push arpen to cap to 100% or Boss armor to cap to 0 in the real implementation. If they follow the first rule will still be better to use 1 arpen trinket and swap some arpen gem to agility in order to reach exactly 100% reduction with the proc. When the change goes live and the new formula will be releaved I'll rerun the simulation with the new BiS gear.

@Mijae:
I completly agree with your analysis, I was one of those high dodge druids back in TBC, my point is that at the moment (and probably also in the next tier) due to diminushing return we can't push avoidance high enough to make it a valuable stat. We still need an fight to fight analysis of the problem, as I said before if someone wants to help me giving me raw boss data (also via private message if we don't want to increase topic lenght too much) will be welcome.

After reading a lot in this forum now my first Post
At first nice Job Nightcrowler. This Thread is/was a big help for me to understand the feral ( last time feral: Season 1/2 ) =)
I´ve one question about the Bear-Gear.
Why is [Belt of the Twilight Assassin] better than the crafted [Death-warmed Belt] ? Because of Crit/expertise?

@jonny: until I have a more stable release of the addon I'll not put it on curse/wowinterface because I prefere a "more expirienced" group of people to try it out and find bugs.

@Bloodvire: Death-warmed is better than Twilight assasin, but if you try to create a perfect set you should think about hit/expertise cap. Basically with the gear I've choced death-warmed overcap expertise so twilight becames better, removing excess expertise by changing other gear (for example the chest) is a worst options because between death-warmed and twilight the difference is minimal but until you go over the exp cap death warmed is obviously better.

Comparing the two for different gemming schemes (ignoring the extra slot you get from a belt buckle):

[Death-warmed Belt] + 24 stam / 8 agi-12stam gems (getting 6 agi socket bonus) = 109 stam / 83 agi
[Death-warmed Belt] + 2x 24 stam = 121 stam / 69 agi

[Belt of the Twilight Assassin] = 84 stam/ 84 agi

So your gains depending on gem choice are 25 stam or 37 stam while losing 15 agi when compared to gaining 57 expertise. With the shear amount of expertise on the [Belt of the Twilight Assassin], this will get you a long way towards not being parry-gibbed, so that would be my guess as to why it's ranked higher, as well.

<Edit> Looks like nightcrowler beat me to it.

Personally I think it's a much grayer issue than it was in BC. Without Sunwell radiance it was a no-brainer, but even *with* Sunwell radiance our avoidance in BC was so much higher than it is now, with exponential returns the more we got.

Reducing the frequency at which spikes occur is great, but in wrath, you simply cannot hit an avoidance so high that they don't still occur multiple times in a fight, and you need to survive that. In BC, it was easy to gear such that even getting hit 3 times in a row was an uncommon occurence. Now however, if choosing some amount of avoidance over stam means such a string will occur on average 12 times in some particular fight rather than 20 times, your overall "chance to survive" (the ultimate measure) may or may not be going up.

I lean towards the opinion that avoidance is only truly a survival stat when it becomes reliable, and at current avoidance levels we're simply still going to see a significant amount of spikes that need to be survived with stamina, armor, and heals.

Still, it's a complex issue and the only "accurate" simulation would be one that emulates exactly what happens in game - realistic healing simulation with some randomness that includes healers dealing with fight mechanics, gaps in their MT heals, choices between small and large heals, crits and hits, etc, all with different numbers/classes of healers and assignments. The "optimum" balance of avoidance/stam which maximizes chance of survival over many iterations would change drastically depending on how this incoming healing changes, so there's never going to be one answer even for a single boss.

Our preferences are ultimately educated opinions with no means of proof one way or the other, and from a raiding perspective my own "solution" is just to diversify all your tanks as much as possible. If you've got a high-avoidance tank in your guild, don't just mimic him but come up short, gear uniquely with your first set so that you have options for different fights and raid compositions. In the end I think this diversity is raising your "chance to kill boss", the measure that's even more important than your own "chance to survive" on any given fight.

Brutallus was also a very specific case where he had a very fast swing timer, and what killed tanks was often the stacking of special attacks along with regular attacks. For stamina to be effective, you need incoming healing, and he could hit you four times in 2 seconds. Avoidance gets better the longer the string of attacks you have to consider.

Another way to look at it:

Situation 1-- you need to avoid one hit to live in a 2-second period. There are 4 hits in 2 seconds. If your avoidance is 50%. That's a .5^4 = 6.25% chance to die to 4 consecutive swings. If you increase to 60% avoidance, the chance to receive 4 consecutive hits is .4^4 = 2.56%. Ten percent avoidance cut your likelihood of dying by over half.

Situation 2-- you need to avoid one hit to live in a 2-second period, but you only have a 25% chance of dying in that scenario (to make things even with situation 1). There are 2 hits every 2 seconds. If your avoidance is 50%, there is a 25% x 25% = 6.25% chance to die to 2 consecutive swings. If you increase to 60% avoidance, the chance to receive 2 consecutive hits is 16% x 25% = 4%. Ten percent avoidance cut your likelihood of dying by less than half.

Brutallus showed us that avoidance is more effective against frequent hits than it is against slow hits.

Hey FBN, i just saw this posted on MMO and was wondering what are the implications of this ArP nurf and how can we plan for it. I notice that it isn't into affect yet; however, your thoughts would be nice. I have around 32% ArP with no buffs and when my Runestone procs, I have around 83%. I haven't lost enough hit yet to wear the Grim Toll, but was wondering if the value of ArP will decrease since there will be a "cap" at 100%. Thanks.

Hey, kinda wondering the same, Im on about 30% unbuffed and dont have any ArP proc trinkets.
Does this mean that im free to stack as much ArP gear as i want, since I wont ever hit 100% with gear alone?
Or does ArP count from other sources too, e.g. ff, sunder?

Armor reduction from other sources (sunder, faerie fire, shattering throw) is applied separately such that it would not make sense for them to count toward this cap.

Sadly i cannot post on the N/A forums where the "whole" ArP discussion is going, bus as this is basically the same discussion I'm gonna ask here.
How does sunder stack with ff and ArP from gear and how much ArP% is needed to nullify bosses armor?
As far as I understood the formula its: 10643/((10643+16635)/3)=~1.17 ~117% ArP (10643 being boss armor and 16635 that lvl83 constant)
Gemming for ArP I can attain ~40% from equip + 50% grim toll/runestone procc + 20% sunder and 5% ff which would put me at 115%, pretty much the ArP cap if my calculation above is correct.
Is there anything wrong with my assumption or is there anything i overlooked?
edit: shouldn't do so much other things while writing posts, thanks for clarification at the 2 posters above. This rises another question though; gem for ArP using only 1 proc and never reach ArP cap or gem agi, using both trinkets and let part of the second proc go to waste?