[Hypatia]

Wow. If those formulas are really predicting the values (and I trust that they are), something is seriously screwed up. *ponder*

Oh. Or I'm stupid and didn't realize that for some reason they're expressing 1% as 1.00 and not as 0.01. :X Whoops.

[Hypatia]

I've looked in a bit more detail now, and it's interesting. I'll have to do some fiddling around and see what the implications are.

On a related note, I saw a talk by a bio-stats guy today, and he mentioned a model of "reliability" that I hadn't seen before (since I'm not really in the field of stats, or operations research specifically). Reliability = MTBF / (MTBF + MTTR). Reliability is the mean time between failures, divided by the mean time between failures plus the mean time to recover. That MTTR part could lead to some interesting new avenues of analysis.

Er. Sorry. Got a little sidetracked.

[Satrina]

Source: Hypatia That MTTR part could lead to some interesting new avenues of analysis.

Yes. Yes, it could. Hmm...

[Kazeyonoma]

I saw Hypatia post a formula and my brain shut off.

[Del]

Source: phaze I think the goal is for avoidance to mirror the behavior of armor: Avoidance stats are subject to diminishing returns, but total avoided damage is not.

My arguement is that if avoidance is merely linear in terms of survivabiliy ( burst time increase), the unpredictable nature will ensure that it's always inferior to stam, since it's a sure thing.

There was still a real tension between gearing for EH/avoidance even with the superlinear scailing on avoicance. As an avoidance tank, it really pains me to see how quickly gemming is going to become entirely one sided in terms of stam.

[Antioch]

In all honesty I think like they said earlier this is more to address Blizz having to use gimmicks like Sunwell Radiance, etc. and not as much as something we'll constantly have to gear around, but rather just keep in the back of our minds. Pretty sure they'll balance and itemize stuff that by the time we're raiding at 80 you won't really notice too much difference. Again, I think they're just doing it so you won't eventually have a Warrior with over 35% Dodge alone...

[Bayho]

Was this implemented in anyway? I have been trying to make a point of writing down my stats after each patch, and recently both my parry and dodge fell some, with the same talents and gear. Dodge even went from 27.64% to 26.01%, and I have something like 250 Dodge Rating.

[Tenraiel]

Source: Hypatia On a related note, I saw a talk by a bio-stats guy today, and he mentioned a model of "reliability" that I hadn't seen before (since I'm not really in the field of stats, or operations research specifically). Reliability = MTBF / (MTBF + MTTR). Reliability is the mean time between failures, divided by the mean time between failures plus the mean time to recover. That MTTR part could lead to some interesting new avenues of analysis.

Ooohh, ... shiny ... I like it ...

This would imply that we would need to have some fairly detailed information on what our healers are capable of and how they perform in order to get the MTTR yes? This seems to tell me that reliability is more of a function of the tank-healer synergy and would therefore be kinda useless to me when gearing, since the only thing that I could change would be the MTBF.

[edit] However, if we were able to obtain this detailed information for our healers, able to gauge the MTTR, we could have a much more specific (accurate) "Raid Zone Minimum" type resource.

Unless of course I completely botched that.

[Hypatia]

Well, having thought about it some more, the idea is more useful as a thought-seed about "what kinds of things can we think about measuring" than anything else. In order to come up with the "reliability" number, you have to define what MTBF and MTTR mean. If MTBF is "time until you got hit once", MTTR is presumably "how long it takes to get healed back to full." In that case, what reliability actually measures is "what percentage of the time are you at full health?" Which implies other ideas about how "failure" and "recovery" might be defined.

But MTBF could also be "time until you get splattered" and MTTR "how long it takes to rez up and rebuff for the next attempt." That's a very hard idea of failure. (And in the case of raiding, it's much harder to measure!) "Reliability" here is "what percentage of the time is the raid actually fighting?"


So, it's food for thought, but not applicable without some serious thought.


A final insight, based on what was said during the talk:

There are two ways to make reliability approach 1.0 (100%). One way is to increase the mean time between failures. As MTBF approaches infinity, reliability approaches 100%. The other way is to decrease the mean time to recover. As MTTR approaches zero, reliability approaches 100%.

And a final final thought: While my first thought was the these values (MTBF, MTTR) should be single values, analysis becomes potentially easier and more precise if you instead use probability distributions. Specifically, how do single values for MTBF and MTTR allow you to think about "how big was the hit, how many times were you hit in a row?" vs "how much healing can be done?"

Instead, let's say "X" is a random variable that varies over the "types of failure", where P(X) is the probability of X occurring. Now we can take the weighted sum Reliability = Sum (x in X) of MTBF(X) / (MTBF(X) + MTTR(X)) / P(X) to find the "sum over all possibilities". You could possibly use a method like this to analyze different *levels* of failure based on their likelihood. For example, if you know that the expected time between hits is A, and that it takes T_A time to heal up from a normal hit, and that the expected time between crushing blows is B, and that it takes T_B time to heal up from a crushing blow, you can use those together to calculate the "reliability" across both of those kinds of failure.


But you still have to think about how you want to define failure and recovery, and what reliability means given your choices.



Okay. Wall of text done.

[Tenraiel]

Muahaha, you're much better at this than I but I'm following you. I really like the idea. Personally when I saw this I was thinking MTBF as the average time that it took the boss you're fighting to take your health from full, to dead without heals. A typical time to live function. But the problem I was coming up with was how I would define MTTR. I like the idea of the probability distribution, it does in fact seem like you could get more from it. Hrm ... /thinking cap. I sure wish I took more than business stats >_<

[Kazeyonoma]

THE BRAINZ THEY IZ HURTINGZ..... NUUUUU

<3 Hypatia.

[Glomgore]

Kaze, I hear you there.

Hypatia, I've been trolling these forums for about an hour now, and everytime I see your name my eyes glaze over and my brain hurts, and I find myself reaching for my mousewheel. Your too smart. ^_^

On a related note, it looks like avoidance will scale in a simular way armor now.

Now, take into consideration that dodge gives more per point of dodge rating than parry, the real question is when does that NO longer happen, or will all those old numbers be thrown out the window?

Blizzard, with threat, and armor, and now avoidance scaling I think is trying to tell us to STFU with the theorycrafting, and just play the damn game LOL.

[OnourisofRavencrest]

Source: Glomgore Blizzard, with threat, and armor, and now avoidance scaling I think is trying to tell us to STFU with the theorycrafting, and just play the damn game LOL.

If that is the case, it's a tragedy. I've never been good at theorycrafting but it's fun to read and certainly makes the game more interesting than it would be alone.

Back to the topic of the thread: just to be sure, 5/5 Anticipation will always give me more 5% dodge regardless of what my gear is, correct? It's not as though Anticipation would lose value as I gained avoidance from gear is it?

[Krashtork]

~~~>[[[[[]]> <----- this threat

(O_O) <----- My head

[Domathan]

Source: OnourisofRavencrest Back to the topic of the thread: just to be sure, 5/5 Anticipation will always give me more 5% dodge regardless of what my gear is, correct? It's not as though Anticipation would lose value as I gained avoidance from gear is it?

Anticipation would not naturally lose its value, but I feel that there could be circumstances in which you prefer to you not spent your talent points in it.

My Assumptions:
With T10 Gear and Avoidance Diminishing Returns I have the fealing that the Item-Creators at Blizzard screw up horribly and add more and more avoidance-stats.

But as DRs are kicking in at this level of Gear many Itemstat-Points are wasted for some tiny bits of avoidance.

And that could mean that you could try to not skill the talent so that your item-stats arent wasted.

Sure your avoidance will still be lower by some 0,x% but the 5 Extra talent points could be well worth it.

Just my 50cents ^^

[Thortok]

So, is this recalculation hitting with 3.0, along with new talents and re-jiggers to bosses?

[Satrina]

Yep.

[Glomgore]

Ok, so with 3.0, i can expect my def to go down, my dodge to go down, but my atk power and block to go up, got it ^_^ Dom, that's a very interesting theory, and could solve some of our warrior problems of getting everything we want in our talent trees.