The issue I have with the current definition of spikiness is that it doesn't really reflect how we are spiky. Instead of an overall damage variance I think it should reflect or quantify in some way the difference in incoming damage in and out of shield block.
Alternatively a sort of TTL calculation reflecting worst case damage in shield block (ceiling(6/bossswing) +1) blocked hits and out of shield block (ceiling((6*(1-sbuptime%)/sbuptime%)/(bossswing))+1 unblocked hits), and the average damage taken per boss swing over time might be more useful. You can do the calculations using floor, but then your condition is less likely to represent the maximum number of boss swings on a tank given a streak of bad rng, not to mention any boss stun mechanics that interfere with shield block timing.
maybe just ((damage_worst_case-avg_damage_per_swing)*num_worst_case_swings/avg_damage_per_swing?
This would represent the amount of additional damage the tank would take in a worst-case scenario, normalized to the average damage per swing the healers are seeing. For the representative tank in the default spreadsheet, this number is 2.11- i.e. the tank is essentially eating 2.11 additional "swings" of average damage in this time period.
This could also be utter bunk, I'll look at it again in tomorrow.
I have it on fairly good authority that TTL can only really be simmed, and that calculating it with any sort of accuracy is somewhat problematic.
At the beginning of cata, there was a very early point where in terms of TDR our secondary stat gearing was Parry>Dodge>Mastery, However most warrior tanks realised that TDR was not actually all that important at the start of cata. It was more important to have have a steady predictable damage income, even if it meant you took more damage overall and mastery was always the premier stat for that metric whatever you want to call it, In MoP all of our relevant secondary stat in some way go towards doing what mastery did in cata; increasing our block chance, and i want to know how vaiarable, or spiky my damage intake is gonna be and how my stats affect that.
Would it be possible to define spikiness as how often you take an unavoided/unblocked hit? That's where I would say most of the (physical) damage spikes that we typically think of come from. In Cata, it was easy to gear for this as you would just max CTC, but with the block changes in MoP, hit and expertise need to be taken into account. Doing this could also allow us to come up with some rough scale factors to be able to compare the relative effectiveness of our various damage reduction stats (i.e. how much dodge/parry/mastery is actually equivalent to hit/expertise).
You can define spikiness however you want, I shall do the same
there is a section at the bottom which allows you to see what affect adding x amount of a particular stat does to your total damage reduction and incoming damage variance. for maximising Total damage reduction you want Mastery>parry>dodge>exp=hit, for decreasing incoming damage variation you want Mastery>exp=hit>parry>dodge
I know what your spreadsheet shows, but to be honest I think variance and TDR separately are quite useless in determining how to gear properly since quite often when one goes up the other goes down yet they're both important. Also, as you mentioned in your post, TDR and variance have two different stat priorities. How then do you know for sure if a piece of gear is an upgrade? "You have to judge for yourself" is a terrible answer, and one I find unacceptable.
Basically, I'm not asking you to change your opinion or how you wish to define spikiness, but I think a good addition to the spreadsheet would be a cell for like "Percentage of hits not avoided or blocked" or "average ctc" or something of the sort. I think if this spreadsheet is going to be a truly useful tool there needs to be some more ways to quantify our damage reduction.
Well unfortunately you can't just resolve tank survivabilty into a single number like DPS. You may find it unacceptable but thats how it is, what may make you easier to heal may mean you need more healing this was true at the start of cata where smoothness of damage influx was priority or wrath where TDR was merely a function of EH and we were more than willing to give up TDR if it allowed us a bigger EHP. You are going to have to make trade offs between the three. thats what makes gearing interesting.
Originally Posted by Eetabeetay
You cant accurately quantify survivability with a simple formula. There's just too much things to take into account. I tried it for a little while and gave up - you end up with a headache as a result most of the time.
TDR will always increase with every one of our defensive stats (except for Stam), and now even with Hit and Expertise, and in a very predictable and ''formulable'' way. So calculating it is fairly straightforward to do.
Spikiness is somewhat a bit more abstract and less clear. Intuitively when a melee hit bring us from 90% HP to 20%, we would want to call that HP variation, a pretty high spiky variation.
However, we know that technically that doesn't even necessary qualify as a spiky variation, as in order to make that qualification you'll always need something to relate your damage variation to.
If the tank's life is constantly dropping from 90% to 20% in a single swing, during the whole fight, then it would be incorrect to stay there was a spiky variation in any of those times he dropped from 90 to 20%.
This forces us to have something to relate ''spikiness'' to, and without that we can't have any accurate mathematical representation of that word.
So IMO, the first step is finding the most appropriate damage variation we'll be using as a baseline, and anything above it would then be qualified as a spiky variation.
Once we have that we could link TDR to Spikiness, and perhaps have only ''one'' stat priority instead of two.
Last edited by kebess; 08-20-2012 at 05:50 PM.
Unfortunately when you are talking about spikey damage you also are negatively impacted by avoidance. Similarly to why druids where considered very spikey to heal and mana sponges for most of the early expansions. Variance from a norm works both ways, plus and minus. Another reason why, even in a world where mastery and parry are close in usefulness, mastery is going to be a more sought after stat in mist.
As long as you understand what the value represents that teng is giving us their isn't much more we really need. If you really want to know what your average ctc is it's fairly easy to figure that out just by looking at the numbers on the spreadsheet... but keep in mind it is simply that, average ctc. It's not exactly a useful number though. Generally you can just look at it as 1-3 hits between each shield block have a chance to hit you full on (minus avoidance... almost no point to even hope for blocks at that point)
I would expect rage generation to increase if you lower the GCD. Dropping the GCD to 1 decreases rage generation significantly in .1.6.
A devastate that procs snb at 1.5s is less valuable than a devastate that procs at 1.45s, right?
Unless i'm mistake prot warriors are incapable of shortening their GCDs, we're stuck with 1.5s and no amount of haste on earth will change that, As such the sheet isn't set up for such fantasies.
Lust/Hero? In theory we could see survivability increases due to lust/hero, which I think healers would welcome- those tend to be heavy damage phases.
AFAIK even Lust/hero don't affect our GCD.
Nothing (of those 'normal' buffs (I don't know, if there are any... 'special', encounterspecific buffs, which decrease it)) affects the GCD for Warriors.
Also, you shouldn't overvalue the influence of the GCD on rage generation. There are still some things which aren't this dependent on it (Anger management, Shouts, basically Revenge (Sometimes you can use it earlier, but... it still depends mostly on enemy attackspeed/number of enemies and it's own cooldown, rather than if you can use it .5s earlier), berserker rage, critblock.
Because of this, the only 'real' increase you could see would be out of SS and SnB.
The decrease in the sheet is because of the calculation. The overall RPS stays the same, as those chains for the "Dist of rage" aren't changed to fit a 1.0 GCD (for example: dev - dev - dev - SS with 1.5 GCD would be dev - dev - dev - dev - dev - SS @ 1.0 GCD (with all possible implications (SnB procs...))), but as the sheet is, the only thing which changes is rev-proc-chance, as it's calculated by hits per GCD, and, as there aren't more revenge uses / use-possibilites, the overall RPS decreases. (And as a result, the critical block value decreases, as overall DPS, so SB-Uptime decreases, which decreases the RPS further)
Spikiness =/= variance. As you pointed out, going from 90% to 20% could be very consistent damage. Going from 90% to 20% is still very spiky though. Your health pool is going up and down in huge chunks, that is what spikiness is. It is some function of both variance and damage taken.
Originally Posted by kebess
I know there isn't one overarching formula for survival, but I think we can all agree that our main goal is to make us easier to heal. This revolves around minimizing how often the healers have to use their big, or mana-inefficient heals. Which to me means reducing our spikiness, or how often we get "chunked". Yes, we would have to simply pick what exactly qualifies as a chunk, but maybe someone with a little more healer knowledge than I could shed some light on what kind of damage intake justifies using their less mana efficient heals. That's the direction I was looking when I thought of using "average ctc." The less unavoided/unblocked hits, the less mana the healers have to use. If we were somehow able to reach ctc (i know we cant), our damage intake could still be highly variable yet we would be incredibly easier to heal and I doubt any healer would call our damage "spiky."
I feel like you guys are settling for having to use guesswork because we can't have an exact number like dps do. While there's no way currently to unite everything under one "survivability" number, I feel there are ways we can better approximate which way to head with gearing. As it stands right now, I would have to make some arbitrary decision as to what is a reasonable tradeoff between TDR and Variance.
thats merely a function of magnitude and while yes that;s fine for establishing how much EHP you need. if you;re taking 70% og your health each hit I don;t think thats spiky, that's clear indication of not having enough EHP, and you are going to straining your healers to point that mana effieiceny is a non issue. the problem comes when say you;re takingsay 70% of your health and have incoming damage variesso that whilst on average you tak 70% per hit your sometime only gonna take 50% and like wise somr times gonna take 90% anf thats the difference which will catch your healers out, sometimes they'll be lulled into think your taking 50% damage, and then suddely you eat a 90% hit and they throw a massive inefficient heal to get back up and then another and thenyou take a 50% and then they've wasted there mana overhealing
I was just using the 90% to 20% as an example since that is what kebess used. Since cataclysm, EH hasn't been an issue since we can survive taking 3-4 hits in a row just from the stamina and armor we naturally get from our gear. Yet we can still be chunked for a portion of our hp large enough to make healers use their big heals. I'm not sure how much testing you got to do on the new bosses, but from my experience there were several bosses where, though the damage intake was consistent, healers were going oom because they were consistently needing to use their big heals. The numbers you listed, for example, are much higher than what I experienced on beta and would easily run your healers oom. Typically what I saw was that a 250k hit (which would be around 50% of your hp) was about the top end for damage from a single hit. That's why reducing variance is simply not enough by itself.
Reducing variance by itself is not an issue when the bosses are average hitting you for half your health becuase its an EHP issue.
Half your HP is the top end of the damage, not average.
edit: not to mention 250k damage in one hit means your healers have to start using their big heals on you, how is that an EHP issue. Your healers will need to use their big heals no matter what your health pool is.
Last edited by Eetabeetay; 08-21-2012 at 05:10 PM.
becuase damage reduction is part of EHP?
I didn't say Variance was the same as Spikiness. I used the term ''spiky variation'' in my post, rather than ''spikiness'' because it speak more for itself. A spiky variation simply means that your Variance is high.
Originally Posted by Eetabeetay
What I was pointing out in that post, as you understood it, is that you can't have a general definition of a high Variance. You'll need to relate that to something else, that you'd call a ''normal Variance''.
Originally Posted by Eetabeetay
Only reducing your variance is not enough, otherwise Dodge/Parry would have no benefit for us, since they increase TDR but also increase Variance.
If your optimization is correct, and the boss still hits you for 50% of your HP with unblocked attacks, then you simply need more gear. I don't see many other ways around it.
If you do have enough gear, and your gameplay and stat optimization is correct, and you're still getting meleed for 50% of HP, then Blizz needs to nerf that boss...
Last edited by kebess; 08-21-2012 at 05:26 PM.