This may be slightly irrelevant, but I'd like to make one small point here. Raids that were hitting, for example, Saurfang the first week out were usually attempting him with ToC level gear, possibly some hard mode gear as well. A group thats just now attempting him for the first time is likely going to have a significant amount of ilvl 264 badge gear and tier pieces from 5 months of daily heroics and weekly raids as well as the 15% Hellscream/Wrynn buff. In other words, they're likely to outgear the boss before they even get to attempt it. That doesn't mean the mechanics aren't new to them and aren't going to require practice, but its still a significant difference.Originally Posted by Bodasafa
True, however EHP will still make it easier for them. Once they have it down for sure then do other things if they're confident, maybe switch out gear, but it won't matter either way. Once they're done with saurfang they'll be moving on to Festergut most likely. Between Festergut, Putricide, Sindragosa, and even rotface, they will be facing those situations. That's not even counting how friggin' hard TLK hits.
[Today 09:38 AM] Reev: The older I get, the more I think those Greek philosophers were just annoying hipsters.
Satorri, I was going to write a response, and I got delayed with RL, but I think you somewhat covered some of the points I was going to cover in a better fashion, so I will leave it at that.
@ Agg.
In your last post, you agreed with some of the points I had mentioned about healer strategies, but then reduced it down to "HoTs are good". However, that is very analagous to saying "Stamina is good" without really understanding why it is good, a need which you addressed.
Myself, time permitting, I make a point of trying every tank type, every healer type, and I am struggling with every DPS type (pet classes such as warlocks and hunters still elude me, even when they are destro or marksman). However, Tanks and Healers, I have played a plenty, and I can relate to both sides of the equation equallly, of ALL flavours.
When a tank dies, invariably either that tank is blamed for not being adequate (be it gameplay, spec or gearing), or the healer (again, for the same reasons). Being Tankspot, I am going to assume that everybody understands the nuances of tanking, including how to diagnose and fix things. However, try as Aliena may, not everybody here has the benefit of having played a healer, most especially all the types, and the combined 5 healings specs form a spectrum from raid heals to tank healers, which everybody understands, but the IMPLICATIONS are best understood through experiences. My previous post attempted to bring that perspective into the discussion.
When a tank dies, but before the full analysis is complete, every healer goes through a process to figure out what went wrong. Some will dismiss the tank out of hand, being supremely confident in their skills and performance to the point where they are certain it wasn't their fault. Othertimes, they look at a solid tank, say with unbuffed 30++k armor and 45k HP, and they just KNOW it can't be tank's fault unless he forgot a defensive cooldown or it was a network issue or whatever.
An EHP tank may in fact take more damage than an absurdly well geared, but it has 2 key points going for it: It's predictable that most every hit is registered and a rhythm for hits is established, and it's manageable with sufficient mitigation (armor/resists).
An avoidance geared tank, setting aside the work you have outlined on the tank side, is damn near impossible to predict. RNG notwithstanding, there is NOTHING a healer can do, except react after a hit comes through. If they still continue to chain cast regardless of hits, an avoidance tank, even a very well geared one, just generates a lot more overheals. That's why healers don't want avoidance tanks, unless the avoidance streak can be chained so long as to become significant, something that can never happen between DR on avoidance stats and the ICC debuff.
A stam tank does in fact becomes the mana sponge you mentioned, and yeah they do exist, but not as a mana issue but an attention issue. When I see statements here that "mana is not an issue", it frustrates me to no end because in some cases, it's a foregone conclusion rather than an appreciation that it requires effort and management, something a healer must make time for in between healing a tank's arse, much as some healers take for granted the careful timing of tank CDs. And while with current gearing levels and general expected healer competency does allow for mana sponges to exist and clear hard modes, it doesn't negate the fact that those sponges are an actual increased burden on healers, leading to comparisons of how one tank is easier to heal over another, or making it harder to tackle other problems such as fight movement or other gimmicks. A case in point is when Toravon first came out, on my bear I willingly traded stamina (about 10k Hp worth) for absurd quantities of frost resist. Now both are EH values, and by sheer budget I was losing more stats because I used lol-polar gear instead of ilvl 245 gear, but the tradeoff was that I was able to with stand upwards of 15 stacks of his debuffs while the DPS farted around. Yes, I had severe snap threat issues, solved only with misdirects, since I was pulling only 2-3k threat instead of the usual 5-6k at the time, and really the fault lay with the DPS not getting their act together and downing the frost orbs, but that was about the last time we really could pull that trick off (I haven't found anything in ICC where I can pull that trick). Gearing for stamina is really about giving that CUSHION to somebody, that healers had more time to react, or forgiveness to miss a heal, or DPS to spend more time on a target. That's it, and it's really inferior to mitigation, the holy grail for us tanks. Because despite the diminishing returns on armor and resists, it's still better pound for pound than taking a full hit, something you mentioned in your discussion with avoidance gearing. What any tank who has tried to gear for this realizes is that we just don't have options, really. Everybody has access to the cloak and rings with extra armor, Plate tanks have about 3 pieces with extra armor, and that's about it (Bear tanks be screwed, AFAIK, but luckily they already have a lot of armor anyways). Beyond that, a metagem, cloak enchant and gloves are the only other perks, which leaves us to stamina everywhere else. But from a healer's perspective, you appear to be "easier" to heal because you have reduced the incoming heals to a more manageable rate and the extra stamina where you couldn't pack armor or resistance are just the extra cushions which are really demanded by the increasing content challenge.
Which segways into preventative healing vs active healing, which I characterized as HoT/Shielding vs active heals. I discount shielding (disc priests) because while I can't quantify this accurately, there is no comparison to the preparation burst heal value of rolling 3-4 HoTs on a tank vs a PW:S with possible Aegis (a pretty weak mechanic), and I won't even go to Sacred Shield or ES or Prayer o' Mending, despite their significant contributions in the log run. And it also explains the 4 druid example that was taken so badly out of context earlier, because you can stack 12 HoT's but only one shield can exist. Either way, the attention required for rolling HoTs on a tank is minimal. If your EHP is insufficient for the task, then the preventative healers need to start supplementing with active heals (penance, swiftmend, nourish, etc.), which takes them out of their envelope.
First: I love shields for the sole reason that they are 100% effective healing with no "overheal" unless the shield is still on the tank when the boss dies. Shields are fanatstic mechanics and I have loved disc priests for them.
Secondly, I understand what you are saying about armor/resistance vs. raw stam, however, after talking with all of my healers they still prefer whichever allows me to take the most damage without dying. If the small amount of mitigation isn't more than the increase in hit points, then the buffer the healers have to heal me up in time is shorter.
I will agree that tanks with low armor bother me a lot, and yes, they obviously do take a lot more damage, however this is because they have made bad gearing choices where the choices with armor give a LOT more overall EHP than the stam pieces. In current ICC gearing content, armor is a HUGE contributor to EHP. I think there are only 2 places where someone can adequately and reasonably contend an actual stam vs. armor debate that is realistic, not just theoretical gains (since, you can't gem for armor and I advocate armor to gloves and cloak anyways as per my prot war guide).
That is:
Rings, Trinkets
I think rings are a great place to make up an armor loss, and the simple fact of the matter is that most of the rings out there with armor, even at lower ivalues (actually all 3 except the devium's ring or however it's spelled are ilevel 245) they provide more EHP than their 264 counterparts. Here my statement of "pick the one with more EHP" holds true again, and it just happens to be a vote for armor (H-Band of the Twin Valks over Juggernaut's band is the main example) so really if people are following my advice they'll be stacking armor here anyways.
The only place where I would say go with stam over armor in a general sense is in trinket selection. Glyph of Indominability is worth roughly 170stam worth of EHP. However, I would recommend The Skeleton Key over the Glyph of Indominability because 1) I think the absorb effect isn't as useless as people make it out to be, especially since it got buffed and it works great in conjunction with the warrior T10 4pc bonus if you choose that gearing route and 2) without considering modifiers, The Skeleton Key is almost 60 stam more worth of EHP. That's like putting 2 gem slots in glyph of indom, or an entire profession worth of EHP better than glyph of indominability. While the tank does lose the marginal benefit of the damage being mitigated rather than needing to be healed, his overall buffer for the healers goes up.
So really trinket selection is the only thing we might actually disagree on when looking at gear.
I might agree with you completely (btw the really big paragraph is the main one I don't agree with, for the most part I agree with what you've said) if it wasn't for a few things.
1st) Overhealing meters. they get as high as 70% or sometimes even more. This tells me as long as I am giving my healers a better buffer, they'll be able to heal me back up and the extra EHP isn't wasted on heals.
2) Max HPS and HPS spikes. Our holy pally can spike up to 20k HPS (or at least that's the highest I've seen, but I don't go that in depth in our logs, it might be higher) and that in conjunction with other heals means I can always be healed to full, as long as the buffer is big enough.
3) MaxDTPS is not constant. The reason why minTTD and minTTDwHeals calcs exist in the first place is because eventually when avoidance breaks down as the fight duration increases, they will take MaxDTPS, and this is when healers have to kick it up a notch so that their normal ~5kHPS routine has to catch up to the increase DTPS by the tank. While yes it is true that armor/mitigation both reduces DTPS and increases EHP, if that decrease in DTPS is not equivalent to the increase in EHP provided by the item, then I contend the stamina is better. Again, it's about buffer windows when talking about actual gear selection. If armor gives that better window, then okay.
4) Now, this is one I'm not sure about, but it seems to me that the ICC buff probably helps healers out in this fashion a lot more. They were already over healing us, and now that we have even more HPs, they can also heal even more, so I think this buff is to ensure that the scenario you talk about isn't really an issue in progression tanking (again, this is what this guide is aimed at). I mean, paladins already weren't taking divinity because the extra healing just turned into overheal anyways.
For all these reasons, in realistic gear picking situations in ICC, I stand by my statements that you take whatever gives you more EHP, be it armor, stam, or magic resistance. I would also contend that in your Toravon example you probably had more EHP with the IR gear, though I don't know exactly what your setup was to run the numbers.
Edit: I apparently had the wrong pally talent, I meant divinity
Last edited by Aggathon; 05-07-2010 at 10:36 AM.
[Today 09:38 AM] Reev: The older I get, the more I think those Greek philosophers were just annoying hipsters.
I've been trying to come up with a way to present an approachable representation of survival tools, I had a first pass come and go months back.
I suppose the beta is a good time to resurrect this project. In doing that though, there is something that sort of came to light for me that might be meaningful to you Agg or anyone else who is in on the discussion.
EHP is actually kind of quasi-counter-intuitive to a practical understanding of your survivability. Here's how I figure. When you are actually tanking what do you see?
If you are sharp at watching numbers you can even see the effects of your CDs, and if your addons allow it you can see how much a block or bubble reduces a hit by when it isn't fully absorbed.
- Miss/dodge/parry
- damage caused by a successful hit
- your current and max health
So if you want to understand the effects of things like health and armor, EHP is actually a bit backwards. Your EHP value shows the health that would be compared to unmitigated damage. As a player you don't see unmitigated damage while tanking, really ever. So, EHP is not intuitive in the sense of understanding the practical application of your survival abilities.
What EHP *is* good for making a simple standard to compare added values of armor vs health. It is a valuable way to make a single comparable value of mitigation vs health for making choices on that topic, but that is the limit of the metric. Unless you go and back-calculate the unmitigated damage of a boss, the number is meaningless once you set foot in a raid (unless you want to measure your tepeen against another tank); Not that that couldn't be a fun project too.
EHP is a fun metric, and there are other nice metrics as well like TTL/TTD, Burst time, etc, but my quest is to find a simple representation or set of metrics that can adequately express all the survival bits in a comprehensible package.
The (Old) Book on Death Knight Tanking
The New Testament on Death Knight Tanking
-----------------------------------------
New really isn't new anymore. There will be a new wave of tanks with success in progressive raids, even heroic raids, using avoidance gear over EH gear that will have more than acceptable EH levels due to the stacking buff. There WAS a flat minimum that you needed for survivability on certain encounters, either from sustained hits or burst magical damage, that will be trivialized due to the nature of the zone.
I think people will try to argue the EH is win portion of your post simply because ICC25 is a horrendous zone implementation for the EH argument. Progression guilds can use tanks in half DPS gear at this point for encounters that were bleeding edge a mere 3 months ago.
Also I never really understood the argument for avoidance based gear due to human error on the healer end of the spectrum. This post is about the best possible chance for survival, and when you bring RNG or human elements into a mathematical equation you muddle down simplicity and hard truth.
Either way, I've looked through it more, and I still really enjoy it. I don't think you need to add more... threat is a totally different beast that delves into a class by class analysis, and healing from a healer perspective can generally be summarized in a few words on a boss by boss basis itself. There isn't quite as much needed outside of the tanking zone because there aren't as many variables.
I'd be careful with that one. Shield can overheal, it just doesn't show up on the meters. What you say is absolutely true when you are constantly being bombarded with steady hits coming in. But if there are phases, like Marrowgar during Bonestorm, or swapping tanks in Saurfang, there's constant overhealing on the shields, but I submit it's not really a huge deal. With the notable exception of PW:S, the shielding from Aegis, Sacred Shield, etc. are small contributions and RNG based anyways. It's great if it procs, but I never count on it (or even expect it). My biggest love as a Disc priest is 1) being able to pre-shield before pull with no degradation, as a pre-HoT ticks away reducing in effectiveness significantly over time (time between pull and actual first point of incoming damage), and 2) Inspiritaion. It's just freakin' win, like vigilance or BoSanc.
I think that's kinda our common gripe, even if we don't approach it from the same point, there just isn't much choice. And really, with the non-linear value of armour, we have so much armour by ICC that additional armour really has to be scrutinized in context with you already have, where as straight stamina gains are always linear. Everybody who has enough understanding of avoidance will know that despite the seemingly low final avoidance numbers, we are so high on the curve that the value of additional avoidance has thrown it in to the next universe in terms of desirability. In my mind, the last frontier is really Resistances, where we traditionally are limited to only to whatever resistance aura is available, which is not all the high on the DR curve. Having said that, our choices are limited there too:Secondly, I understand what you are saying about armor/resistance vs. raw stam, however, after talking with all of my healers they still prefer whichever allows me to take the most damage without dying. If the small amount of mitigation isn't more than the increase in hit points, then the buffer the healers have to heal me up in time is shorter.
I will agree that tanks with low armor bother me a lot, and yes, they obviously do take a lot more damage, however this is because they have made bad gearing choices where the choices with armor give a LOT more overall EHP than the stam pieces. In current ICC gearing content, armor is a HUGE contributor to EHP. I think there are only 2 places where someone can adequately and reasonably contend an actual stam vs. armor debate that is realistic, not just theoretical gains (since, you can't gem for armor and I advocate armor to gloves and cloak anyways as per my prot war guide).
That is:
Rings, Trinkets
I think rings are a great place to make up an armor loss, and the simple fact of the matter is that most of the rings out there with armor, even at lower ivalues (actually all 3 except the devium's ring or however it's spelled are ilevel 245) they provide more EHP than their 264 counterparts. Here my statement of "pick the one with more EHP" holds true again, and it just happens to be a vote for armor (H-Band of the Twin Valks over Juggernaut's band is the main example) so really if people are following my advice they'll be stacking armor here anyways.
The only place where I would say go with stam over armor in a general sense is in trinket selection. Glyph of Indominability is worth roughly 170stam worth of EHP. However, I would recommend The Skeleton Key over the Glyph of Indominability because 1) I think the absorb effect isn't as useless as people make it out to be, especially since it got buffed and it works great in conjunction with the warrior T10 4pc bonus if you choose that gearing route and 2) without considering modifiers, The Skeleton Key is almost 60 stam more worth of EHP. That's like putting 2 gem slots in glyph of indom, or an entire profession worth of EHP better than glyph of indominability. While the tank does lose the marginal benefit of the damage being mitigated rather than needing to be healed, his overall buffer for the healers goes up.
So really trinket selection is the only thing we might actually disagree on when looking at gear.
Flasks (Stoneblood vs Lesser Resistance)
Head Enchant (at the cost of some stam)
Cloak Enchant (may have to keep one enchant for every damage type)
Wrist patches for LW (same problem as cloak enchants)
POSSIBLY the Ony25 ring
Anything beyond that and you are sacrificing too much elsewhere (resist gems are a bad idea, as are TBC resist patches). Also, the EH gains on Frost gear also needs a slight tweaking if you account armor losses for ilvl200 gear, given that even Sindragosa still has a physical component. Not that I would even remotely suggest gearing that way ICC, it was barely acceptable at best for Ulduar, I got a lot of flak for trying it on Toravon but it worked until it passed from progression to farm a long time ago.
And as mentioned before by others, Anub is a total gimmick fight where having extra health was penalized. In that fight alone, dropping stam for resist made total sense if you negatively scored HP (or at least degraded its value), and going nuts with a NR set was OK. Even still, you can't switch gear mid fight, so you were either gearing for P1&P2 and overcomming P3 with brute healer force, or Gearing for P3 and making do for P1 and P2 (which I had done a lot of times).
Overheals totals are a bit of a misnomer until you look at a breakdown. The method of overheals is important to note. For example, for many fights, a Pally Healer beacons the OT, and unless there is a constant cleave mechanic (marrowgar), Bacon o' Light is pure overheals, there's 50% overheals right there alone, but with the increased healing scaled with gear, that overhealing is creeping upward (10% overheals on the base HL alone, + 50% with Beacon of Light, or 60% as compared to the "T9 Beacon"). You can find similar inefficiencies with each healer class, but I've found that the shaman Chain Heal still yields the lowest overheal percentage due to its powerful smart heal properties (I can bring it down to ~30%). The point of all this is that it again boils down to the discreteness of heals (which hasn't changed), but the damage profile in ICC has changed (through ICC debuff and boss design).I might agree with you completely (btw the really big paragraph is the main one I don't agree with, for the most part I agree with what you've said) if it wasn't for a few things.
1st) Overhealing meters. they get as high as 70% or sometimes even more. This tells me as long as I am giving my healers a better buffer, they'll be able to heal me back up and the extra EHP isn't wasted on heals.
I would almost wager that Tank EHP hasn't scaled to the same degree as healer HPS. One could test that theory by comparing parses with the stacking weekly buff over several weeks. While each week both tanks and healers get 5%, if there is a disparity between the increases one should expect to see increased overheals as the healers pull ahead faster, but it could also be that the EHP gains are mainly from armour which don't scale with the buff, but that is tricky as well given that the EHP value of armour goes up with increasing stamina...
When T10 came out, I was noticing 70+% overheals, until I started adjusting my heal style. I think the increase in healing burst HPS throughput offered through T10 crit and haste on most gear didn't quite line up to the damage rate, which was supposed to be the same amount but more frequent, making it actually easier to heal than the previous content. I think they overcompensated after ToC healers struggled with one-shot-dead tanks. Admittedly, I haven't done H-ICC25 LK, but assuming your figure of 56k hits is somewhat correct, I submit you wouldn't see a 70% over heal rate (but I don't have data to prove that).
A Holy Pally HPS is dead easy to calculate. Tank healers (Disc priests, Shamans, Paladins) should easily be able to crit for 20-25k on their nuke heal (Penance, HW, HL, respectively), and given the nature of Paladin Healing, they can bring down their HLs to near 1 second cast times. The other healers will be going with smaller heals (8-10k one each target, across several targets), and then the incidental little things (Prayer of Mending, Aegis Shielding, Earth shield, Sacred Shield, etc.)2) Max HPS and HPS spikes. Our holy pally can spike up to 20k HPS (or at least that's the highest I've seen, but I don't go that in depth in our logs, it might be higher) and that in conjunction with other heals means I can always be healed to full, as long as the buffer is big enough.
As I mentioned above, if all things were equal healer output should equal Tank health increases, but I suspect it isn't. <Incidentally, SA is for mana regen, not heals>3) MaxDTPS is not constant. The reason why minTTD and minTTDwHeals calcs exist in the first place is because eventually when avoidance breaks down as the fight duration increases, they will take MaxDTPS, and this is when healers have to kick it up a notch so that their normal ~5kHPS routine has to catch up to the increase DTPS by the tank. While yes it is true that armor/mitigation both reduces DTPS and increases EHP, if that decrease in DTPS is not equivalent to the increase in EHP provided by the item, then I contend the stamina is better. Again, it's about buffer windows when talking about actual gear selection. If armor gives that better window, then okay.
4) Now, this is one I'm not sure about, but it seems to me that the ICC buff probably helps healers out in this fashion a lot more. They were already over healing us, and now that we have even more HPs, they can also heal even more, so I think this buff is to ensure that the scenario you talk about isn't really an issue in progression tanking (again, this is what this guide is aimed at). I mean, paladins already weren't taking 2/2 SA (only 1/2) because the extra healing just turned into overheal anyways.
I've never paid attention to the physical component of Toravon's hits, but needless to say it's less than all the frost damage flying about, which is why the EHP for that fight was greater with a polar setup. Again, this did drop my TPS from 5-6k to 2-3k with the same rotation and skill, which was bad, as this forced me to seek constant help from hunters and rogues. 2-3k TPS is not enough to compete with (at the time) 9-10k DPS I was seeing.For all these reasons, in realistic gear picking situations in ICC, I stand by my statements that you take whatever gives you more EHP, be it armor, stam, or magic resistance. I would also contend that in your Toravon example you probably had more EHP with the IR gear, though I don't know exactly what your setup was to run the numbers.
Avoidance worsens human error, because you just can't predict it. Increasing EHP decreases the effect of human error on healer side by increasing the cushion provided.
That's exactly what I just said :\
Thanks for this post, just coming back after a year off and it's nice to reinforce that what has been true, remain true.
I salute you for the work done here, but I'd like to bring up a few points (although I do agree with the conclusion).
On how you measure the benefit of avoidance :
As you explained very well, at current levels, you WILL most likely be hit N times (let N be 5 in your text, or 3 as was proposed in some other post).
I think one thing that matters more than the effect of avoidance on this likelihood, is how many of these streaks you should expect.
Why?
Because the less streaks, the less chance you will die, and...
Because of cooldowns.
Of course, if the fights already forces you to time your cooldowns, you will use them proactively.
But if it doesn't, you will most likely use them (or try) on one of those streaks.
This is why I think the streak frequency is more useful to value avoidance than its probability :
if streaks happen often, you'll find yourself cooldown-starved, unable to react to a streak.
if they are much less likely, you'll have more breathing room.
A little bit of calculus on the subject :
Chance of taking a k-long hit streak :
(1-avoid)^k
Probability of it happening on a fight where the boss strikes N times :
((1-avoid)^k)^(N/k)
(roughly)
Expected value on the same fight :
(N/k)*((1-avoid)^k)
If you want to relate this to the duration of cooldowns, you could just quotient out boss attack speed to said duration, and get rid of the fight length
Unfortunately, I don't have any numbers to illustrate those formulas, otherwise I'd be happy to graph it.
On a side note, about your paragraph on the stochastic nature of boss hits :
Could you point to me any data showing its gaussian nature, or hinting at it?
(as it could very well be any other simple statistic law, and it would make sense)
Thank you for reading
The problem with avoidance is a conflict between the random nature RNG and the stability required while tanking.
Here's an idea I've had for some time now:
1. Leave only one RNG avoidance mechanic as a case of low probability. This should be done in a way that either changes a bit with gear, or doesn't change at all. Let's call this one Dodge.
2. The other avoidance element - Parry - isn't RNG based but a non-GCD skill.
3. Parry: Deflects the next incoming melee hit. Duration 5 seconds (or so...).
4. Avoidance: effects the CD of the Parry ability.
From now on it's just tweaking the CD numbers and the effect avoidance has on them.
I know it changes the whole nature of avoidance as RNG, but I think that's where we have a problem. The alternative would be making heals, and incoming damage very low compared to the tank's HP. More hit attempts and less spikiness would increase the value of RNG.
I don't actually, that was a bit outside of the range of what I wanted to do. I can say it "feels" like it is a lot of the time, for some fights a lot of the magic damage is constant. The point overall is though, regardless of if the curve is gaussian, it is stochastic, and increasing EHP pushes the "z" value further off the graph. You could make a sampling argument that due to the central limit theorem regardless of what the curve looks like with enough samples it becomes Gaussian and therefore at any given sampling (aka over the course of a boss fight or several boss fights) it is still a valid assumption.
where are you getting the N/k from? I think it's still a runs problem and you need a simulator to get more accurate data, it would probably have to be represented by a histogram because I can't imagine it would be consistent and you'd have to determine the parameters for what an N+1 streak counts as too. And due to its RNG nature, it's really hard to nail down a consistent frequency over time because that distribution would be Gaussian and you might take several hits in a row or you might have a minute long break, it's impossible to tell and rely on that number. Maybe if you could explain what you're doing with the math a little more in depth I might see what you're talking about better.
@Adrael:
That's pretty much what blizzard attempted with CotT, but as you can see with the graph, more hits means you're more likely to take bad hit strings in a fight. I am kind of intrigued by the idea of a parry you can use as a cooldown, and I think shield block is kind of a variant of that. The real problem with why avoidance is so underwhelming at the moment is that healers don't have mana problems so a tank can afford to gear for situations where they are able to just take a truck load of damage at once because healers can heal them up with relatively no problems the majority of the time.
I still remember back in BC doing illidari council. Those mechanics, I think, are where things like avoidance had value. The bosses hit slow but hard, and healers had mana problems, so they had to pre-cast and stop-cast based on your damage intake. I think it took a lot more coordination and skill and that was one of my favorite bosses in TBC to tank. Unfortunately blizzard has gone away from the downranking and stop casting ideas (and are getting rid of the 5 second rule in cata because they don't think healers should be just sitting around letting mana regen), and I think that has caused some of the situation we are in now.
[Today 09:38 AM] Reev: The older I get, the more I think those Greek philosophers were just annoying hipsters.
Oh man, I wish I hadn't mostly stopped playing WoW and could get into this thread.
Lots of good effort on your part, Agg, and you make a lot of very strong points. However, many things in this area are quite variable depending on all the numbers input.
You do hit on a lot of good points, and I would mention that even a Burst Time model agrees that Stamina is almost always point-for-point better than avoidance. Depending on the size of hits, sometimes this valuation can be very close--but Stamina is almost always better in most cases. It's important to remember that even models which consider avoidance will value Stamina highly if they are well-crafted.
Maintainer of Rawr.ProtWarr theorycrafting tool. Feel free to PM suggestions or feature requests!
That's right. But what I was think of was a model in which tank's HP is so much more than inc. heals and damage that it would take maybe over 20 hits to kill a tank. Off course that's not something I'd like to see
Mana limitation is interesting. It's a feature that existed pre-WotLK but there were other features involved back then so it's hard to evaluate it's influence over healing and tanking. If mana management is to be an issue that requires avoidance, than there should be a proper reaction time for the healers to consider whether they want to use big heals, meduim heals, smalls heals or not heal. To my understanding of the statements coming from Blizzard, this is exactly what they attempt to achieve.
However, in order to achieve that they will need to make sure that there is in fact a *proper* reaction time for the healers, and either eliminate spam heals, or minimize the mana cost of minor healing spells to the extent that they can always be used as a filler without draining the mana pool.
Giving time for the healers to decide which spell they want to cast would make things a lot more strategic, yet a lot slower since it requires tanks to be able to just stand there and not die for relatively long periods. Which brings me back to the idea of tanks with a much larger health pool.
I wonder if such a change would make the game more fun and strategic, or more boring to the point where tanks 'panic buttons' will no longer be triggered by panic.
Actually that equation tells you the probability of having an N hits long streak, not the probability of hitting a string of k hits in N attempts.
What you are trying to model is a bit more complicated. It will need to be an N choose k style equation, but even a standard one won't work by itself because there are overlapping probabilities (I.E. one string my start at h1 and go to hk, while another string my start at h3 and go to hk+3, having overlap between h3 to hk).
Take Burst Time, multiply by avg time between swings, that's the avg time in between lethal streaks. Just divide total fight time with that and you got avg amount of streaks.
Protection Warrior Spreadsheet
Kojiyama! I never thought I'd say it but I've actually missed you in amongst all this =P. Even though we've disagreed a lot, you're one of the few that's been able to back it up with numbers and make me re-think this stuff. I think you're also probably better at stats than I am, lol.
[Today 09:38 AM] Reev: The older I get, the more I think those Greek philosophers were just annoying hipsters.
Interesting read, thanks. I just wanted to touch on a couple of technical points.
You're misusing the central limit theorem to justify using a normal distribution to approximate boss damage. In reality we have no clue what the distribution over the range of possible raw damages for a melee swing is (I don't think); I guess I always assumed it was uniform over the range (i.e. if boss melees for 1000-2000 I assumed that each integer between 1000 and 2000 was equally probable). It may be this, it may be normal, or it may be something else entirely (although uniform would make the most sense from an efficiency perspective, since generating normally-distributed samples is done by transforming a random number which is uniform over the unit interval). The central limit theorem says that regardless of the distribution of a variate, summing a sufficient quantity of samples from a variable following any distribution will result in a roughly normal distribution of the sum (or average, since they differ only by a constant). It most certainly doesn't justify assuming the individual samples of the original variate are normal.
The observation about the frequency, as well as occurrence, of n-hit-strings is interesting, too. While the suggestion is that provided n-hit strings occur with sufficient infrequency that a cooldown is always available for them (assuming that said string is life-threatening and that using a cooldown negates the danger), there are two issues with this. Firstly, the suggestion is that one is able to reactively time cooldowns to co-incide with periods of danger, there's no way to know until the n-th hit lands that you were in a lethal situation, assuming that n-1 hits aren't going to be fatal and so you couldn't pre-empt the n-th. Secondly, just because the expected frequency were low wouldn't guarantee you wouldn't see back-to-back n-hit strings and thus not be able to time cooldowns---it's the same argument as with avoidance, expected damage taken or frequency is irrelevant since you can't survive the (potentially limited) number of times where these events do occur.
As far as the analytic expression for determining the probablility of observing at least one n-hit string from k attacks, Jere is right - it's way more complicated than what's being made out there. There probably is an analytic/combinatoric expression for it, but I see no harm in using what will likely be a much simpler Monte-Carlo style method as done above. The simulations will allow you to answer any questions you want, such as the original "What's the probability of observing a string of exactly (or at least) n hits?" as well as "With what frequency do such strings occur?" etc etc.
As far as the combat table goes, why assume that the rng is discrete (i.e. it's a 10240 sided die?). Not that it changes anything anyway, but it's sort of counter intuitive and I don't see the purpose of including it. You also don't need the whole 102.4% thing, boss level just reduces our chance to avoid attacks by a fixed amount per level. You could simply alter avoidance chances (and thus the combat table) for the level of the attacker and scale everything between 0 and 100. And it's almost certainly going to be a continuous number that's generated in that interval (all random numbers a computer generates start with a uniformly distributed continuous variate in the 0-1 interval), from which you can simply check which outcome the number generated corresponds to.
Finally, regarding the stamina Vs armour debate especially for the trinket slot, a discussion over at maintankadin (http://maintankadin.failsafedesign.c...rb_v=viewtopic) resulted in quite a lot of us rethinking our opinions in that regard. The basic point is that when healing is added into the equation, armor increases its value significantly, health doesn't. You can treat each point of non-overheal healing received in the burst scenario (maybe 3-4 seconds, corresponding to that 5-hit string for a fast hitting boss) as well as any absorbs as though it were health for the purposes of EH calculations determining the relative merits of armour and health. Since, as we all know, increasing your health increases the value of each additional point of armour, but not each additional point of health, the amount of healing you need to receive in the danger period for armour to start looking more attractive is very small. If I recall correctly, for the Skeleton Key vs ilvl245 armor trinket, ~24k healing received during the burst is sufficient to make the armour the superior choice, with the added advantage that you never take the damage rather than needing to be healed back up from it.
@Aggathon : On the gaussian
I was just being curious in this case : I'd like to know what the spread actually looks like (and numbers on the standard deviation).
Actually, that would be testable, although it would require quite a sample to be reliable.
Wouldn't do that on a boss though, more likely on a standard 80 mob (not too low to avoid rounding issues).
And I agree with bashef, I'd bet it's uniform
@Aggathon & Jere & Bashef: On the streak frequency
That was definitely napkin math, and it was wrong indeed.
I've tried finding the expression, but haven't succeeded.
I'm now working on simulations, but it's not quite finished yet, I hope I'll be able to finish it by the end of the week, and get numbers to back up those claims.
Posting Permissions
Reply With Quote
Bookmarks