Printable View
I'm not sure either, I haven't tanked anything past T5 content, so I can't tell ya, but from what I've read brutallus does have a beastly amount of outgoing damage, and gearing for avoidance against him is a good way to ensure he doesn't get 3 straight hits in a row. But threat is likewise of equal importance against him so. I'll wait for someone who has downed him to respond first cuz this is interesting.
This also brings up the issue, if the definition of EH minimum is not possible for certain bosses, does this mean the definition needs to be reviewed?
Prince Malchezaar is the classical example, actually. Thrash can punch through any EH that a Karazhan-geared tank has.Quote:
Originally Posted by Foolishness
For another example, Jan'alai can do 27k damage to an appropriately geared tank in 3.2 seconds. The chance is minuscule, but it is non-zero.
Jan'alai? You mean halazzi?
More or less. Until you outgear it, of course.Quote:
Originally Posted by Foolishness
Of course, you don't exactly "hope" for it. You pick your gear to minimize the chance of catastrophic failure, ideally reducing it to a very low value, such as a fraction of a percent. Many fights that go you beyond a practically attainable EH requirement also have a fairly low chance for high burst damage that breaks through it.
No, it means that naive EH theory is too simplistic a model. You need Markov processes to adequately model tank survivability in a boss fight. Look for posts by Tamral, he has laid out some of the basic ideas.Quote:
Originally Posted by Foolishness
I mean Jan'alai. In enrage mode, he hits with two attacks every 1.6 seconds (faster if he isn't debuffed with Thunder Clap). Thus, he can fit up to eight attacks in a Shield Block cycle, allowing for six successive crushing blows in the worst case. His main hand will crush for ~6k, his off-hand will crush for ~3k (assuming a ZA-geared warrior tank with Ironshield potions and Demoralizing Shout up). That adds up to 27k over 3.2 seconds. If you consider parry thrash, things can get worse.Quote:
Originally Posted by Alent
Obviously, this particular chain of events is extremely unlikely (though you can get pretty close more frequently).
How are EH minimums usually computed? Is there a formula? I always thought they were purely anecdotal.
Anecdotal, and combined with an avoidance score.Quote:
Originally Posted by Pandaburn
Raid Minimums
Wonderful read. Thank you very much for your contributions to the Warrior community.
Ciderhelm, is there any appetite to add this to the EH Calculator?
It would be awesome to have that incorporated into the calculator. Think of the amazing output...
It would tell you EH, Avoidance, and from there, if you added another field for Max Boss Damage, it could output the burst damage.
Additionally the calculator could also provide the information necessaary to interpret for further investment into EH or Avoidance.
Sincerely Mog - Zuljin Server.
I've moved this to the Library.
First of all, thanks Hypathia for the interesting read.
At the end of the day, however, there seems to be a significant flaw in your approach in that you never really reconciled the continuous nature of you Burst Time metric with that fact that, when push comes to shove, the damage we're taking comes in discrete chunks. This come through especially with regard to your healing model: if you're getting topped off after every burst, stamina is valuable if and only if is allow you to survive an extra hit. I'm sure you folow my point, but to be a bit more explicit, if there's zero variability in a mobs damage and it hits for 6500, it take the follow to kill you:
0-6500 HP -- 1 hit
6501-13000 HP -- 2 hits
13001-19500 HP -- 3 hits
19501-26000 HP -- 4 hits
Now of course mobs don't hit for the exact same amount every time, so somewhat more realistically let's consider the mob to instead hit for 6000-7000 damage. Our survivability window now looks as follows:
0-6000 -- 1 hit
6001-7000 -- 1-2 hits
7001-12000 -- 2 hits
12001-14000 -- 2-3 hits
14001-18000 -- 3 hits
18001-21000 -- 3-4 hits
21001-24000 -- 4 hits
Now in this situation, you still have a plateau problem. That is, given the "no heals during burst, topped off after every burst" model, going from say 15000 HP to 16000 HP has no effect on tank survivability in such a scenario. Increasing HP in the range 19000-20000, for example, does however...by increasing the probability of being able to survive 4 rather than 3 hits.
I guess what I'm getting at is that in your model, the value of stamina really shouldn't be as continuous as you have it mapped out.
Changing the healing model can have a big impact of the marginal value of stamina, but it's not clear at all that you'll wind up with this nicely averaged smooth curve as a result, and it certainly seems far from obvious that the results of a model based off of continuous inputs and a "heal only between bursts" healing model will translate meaningfully to of discretized model of damage with healing taking place at more frequent intervals.
Fundamentally, it just doesn't seem like the approach has an response to the "survive an integral number of hits, then ignore stam and go for avoidance when you can't hit the next integer" approach. Using more sophisticated healing and damage models seems like it might yield more interesting results, as in the end raid tanking tends to be about sequences of individual large events, and trying to smooth them over without providing a good probabilistic interpretation for non-integer values seems like it sacrifices a lot of accuracy.
Dorvan: The approach you describe is certainly better. Part of my goal with this was to come up with a single number that people can look at and say "Oh! That's how big a difference it makes!" when they're making off-the-cuff estimates of the strength of various stats.
In the wild, however, it's a lot more complicated, as you say. If you can push your HP enough to reach the next "plateau" of "I can survive more hits in a row", that can be pretty damned powerful. On the other hand, if you can't reach that plateau no matter how hard you try, it might make more sense to stack as much avoidance as possible without dropping yourself down one further plateau.
Basically: It's not a simple trade-off, and it never will be. I hope that I was able to make people think a bit harder about how to manage the choice between avoidance and effective health--because it's the folks who really think about how things behave that will get the most out of any of this sort of discussion.
(And: Kudos to Blizzard for making a system that has stood the test of time so well. It's impressive that after so long with so many people pounding away on the math, there are places where the choices still aren't obvious.)
Very good read, thanks for posting it.
In practice I think avoidance has one glaring problem. A gain of 1% avoidance doesn't necessarily mean you will avoid 1% more damage. EH gains are a little more of a guaranteed gain in this regard from my experience. Working on both is a good practice, and I've seen my total damage numbers coming down as I pick up more dodge and defense rating.
i love this thread Hypatia, your analysis and math are unique and quite logical, you offer a fresh perspective on gearing avoidance vs. stamina [my favorite subject ;)].
just thought i'd share this with you: Ridiculously Important Information: Socketing - TankingTips.com
i just caught wind of this website for the first time from a recent blog entry on tankspot, and i must say this is absolutely the worst thread i've ever seen. i don't know where to begin. i think the most horrifying part is that there are so many people saying, nope that extra 1% avoidance will never be better than that extra 400 hp, and everyone's like, 'ya, you're right, i socketed all stam and now my gear is so much better.' the one voice says avoidance is good (not even gemming for it, just saying avoidance is good!), and he comes down saying, 'no, sorry, stam is the way to go.'
UNBELIEVABLE
anyway, i love this thread, it's one for the ages.
In veneretio's defense, that was more of a kinda joke then it is a global sweeping truth. In general it isn't worth it to match sockets, most of them suck (lawlz +2 parry rating!), but if it's a stam one, why not, if it's a sizable avoidance sure go for it. But what veneretio is stating here is that you don't HAVE to match sockets, which most people do because they think you have to.
Depending on your gear, you really have to look at your stats and try to balance them out. When that article by veneretio was written, 2.4 and i don't think even 2.3 was out yet, so lots of the great badge gear that exists now, didn't then, and back then, most gear (look at t4) is stacked with avoidance already in it, so you had to balance your stats by stacking stam. Of course in t6 or with new badge gear, you see that your stamina levels are reaching very steady levels without having even stepped foot in a 25 man. This is where the change began for many of us to re-evaluate our gemming.
I am reading through this chain and I wonder if there are intentions to update it to 3.0 or Wrath, or if the mechanics will change?
Mechanics will change a ton. There are already some threads around the library that discuss diminishing returns, which will have a huge impact on the stacking power of what we saw with avoidance pre-3.0.
Yes. The basic principles of how a given level of *actual* avoidance helps you continue to hold. (i.e. how powerful 60% actual avoidance is compared to 40% actual avoidance). However, because of the diminishing returns on *rated* avoidance, things will be quite different.
In particular, the material comparing the value of small deltas of stamina vs. dodge rating for burst time is now completely irrelevant and wrong. (But that material was honestly always quite shaky, since the change in value from a large stamina-dodge tradeoff was different from combining many small changes.)
One question/comment and one suggestion:
Comment/question:
- Am I missing the affect of Ardent Defender? I can see the impact of the pally EH enhancing talents, but can't see the knee in the curve that I'd expect from AD kicking in for pallys.
Suggestion:
- One way to think about the burst likelihood impacting life is the probability of a burst sufficient to kill the tank happening within the number of hits implied by either an immortal or undying run. Basically given the average number of hits in a naxx 10/25man full clear there is a .10 probability of burst damage killing the tank for example.