I can go make a blog in 2 minutes

Last week, a few people alerted me to a thread on the official WoW forums in which some players were arguing about the usefulness of AskMrRobot, and in particular its default stat weights.  As loyal readers, you probably already know that I developed those default stat weights.  My response in that forum thread was long and contained quite a bit of discussion about the thought processes that went into those stat weights, and it occurred to me that it would make for a good “quickie” blog post.

Much of this will look familiar to regular readers.  We’ve discussed smoothness at great length.  We’ve talked about raw TDR stat weights.  And we’ve talked about the difficulties of valuating Stamina, and how it’s often better at saving healer mana than TDR stats like dodge and parry are.  The only “new” part is the glimpse at how all of that influences the choice of default stat weights.

I wouldn’t normally post on a Sunday, but this is likely to be a busy week at Sacred Duty, so I wanted to get this up quickly.  Stay tuned for some reasonably large and important blog posts later in the week!


More seriously: I don’t usually wade into the official forums, because the signal-to-noise ratio is pretty awful around here, and this thread is no exception.  That said, there’s some particularly pervasive and inaccurate noise being repeated here, so let’s stamp that out.

1. You will never need more stam than what comes on gear. You are not doing heroics this week. It is silly to have such a high default weight for stam. It is silly to gem solid cuts.

I completely disagree, for several reasons.  The first is that you don’t have the proper perspective to make that statement.  Looking at your armory, you’ve cleared a few heroic bosses in 5.0/5.1 content, and cleared all of the normal modes numerous times.  That means you are going into Throne of Thunder with above-average gear for the first few normal modes.  In other words, you may have more than enough stamina on your gear because you overgear the fights.

However, the average raider hasn’t cleared all of the 5.1 content.  They are going into ToT significantly undergeared for even the first few bosses.  And when you are undergeared, your absolute, 100% guaranteed, there-really-is-no-argument-to-be-had-here-so-don’t-even-try stat is STAMINA.  Get the health you need to survive long enough to receive heals.  Period. This is exactly the reason that hard-mode progression tanks stack Stamina early on – to meet the EH checks they run into when running content they significantly undergear.

I’ll note that if you had pulled the last few bosses of ToT in week one, you may have gotten a very different impression.  The first 5 or 6 do not significantly challenge EH, but the last few do.  Even in ~507ish ilvl gear, Lei Shen hits like a truck.  From a random Lei Shen 10N log:

[19:20:35.738] Lei Shen hits Yogurto 163283
[19:20:37.252] Lei Shen hits Yogurto 169082

170k every 1.5 seconds is a lot to deal with on a boss with a lot of movement.  This whole myth of “get enough health to survive 3 attacks and you’re fine” doesn’t cut it, because that’s not how most tank deaths work.  If you die in 2-3 hits, its because you failed a cooldown for something predictable.  What you’re more likely to do is die to a string of 5-6 hits, or 3-4 plus incidental damage or a special, during a period where your healing is interrupted.  And nothing guards against that better than stamina.

2. The best way to reduce stress on healers is by lowering incoming damage. Stamina does not lower incoming damage at all. You take more damage if you forgo tanking stats for stam. The term manasponge came about for a reason. It is not a term of endearment.

This is another falsehood that gets tossed about by inexperienced tanks.  The term “mana-sponge” stopped being relevant somewhere around Burning Crusade.  Yes, a tank stacking Stamina does take more damage than a tank stacking, say, mastery.  Does that make them a worse tank? I’d say that the answer is unequivocally “No.”

If your goal is taking the absolute least damage possible, you should be stacking avoidance.  Pound for pound, avoidance gives more “total damage reduction” (TDR) than any other stat.  You know what our absolute worst TDR stats are? Hit, expertise, and haste.

Yet looking at your armory, you seem to have reforged much of your dodge and parry into hit, expertise, and haste.  So we are confronted with a conundrum: either

1) you think that TDR is something tanks should seek, in which case you are reforging and gemming completely wrong and shouldn’t be giving advice.


2) you don’t think that TDR is worth seeking, and instead gem for “smoothness” stats, but you’re telling people on the forums they should be seeking TDR stats anyway.  In which case you also shouldn’t be giving advice.

The only constant in those two statements is that you really, really shouldn’t be giving advice. :)

The fact is that TDR hasn’t been relevant since BC or Wrath, depending on who you ask.  Potentially very briefly at the beginning of Cataclysm, when healers were undergeared, but that quickly ended once they started acquiring even a little bit of normal-mode raid loot.  Healers simply do not run out of mana healing tanks nowadays, even in 25-man.  They run out of mana for all sorts of other reasons – excessive raid damage being the primary culprit – but if you take any half-decent healer and tell them their only job is to heal any half-decent tank, mana will never be an issue.

What kills tanks is not their healer running out of mana, but running out of time.  Spike damage is what kills tanks, and it’s what has been killing tanks since time immemorial.  Taking a few too many large hits in a row while your healer is distracted, stunned, moving, or what have you.  That is how tanks die.

If you look at most logs, it’s not hard to see the truth in either of these assertions.  Most healers overheal by 30% or more when they heal tanks.  That’s not because they have mana to waste, but because it’s hard to avoid doing so with all of the self-healing and cross-healing going on.  And more importantly, because they have to maintain some level of overhealing to combat spikes, because during a spike that extra throughput becomes immensely valuable, so much so that it’s worth overhealing by 30%-40% most of the time.  Because it means that extra 30%-40% throughput is there during a spike, when it matters.

And if we’ve accepted that spike damage is what’s relevant (well, maybe you haven’t, but I’m going to go out on a limb and assert that it is), then the question of stat priorities gets turned on its head.  All of the sudden, “reliable’ stats start to win out.  Having hit, expertise, and haste ensures higher SotR uptime and thus fewer/smaller spikes.  And what’s the most reliable stat of them all?  STAMINA.  it is there, always, all the time, preventing spikes.  Because having 10% more health makes every spike 10% smaller, and easier to heal.

It’s even worth noting that from a healer’s perspective, stamina is a mana-saving stat.  This talk of “mana sponges” suggests a fairly critical oversight in your thinking.   Healers are not robots or calculators, they’re people.  They heal based on a few factors, and one important one is reacting to changes of your health bar.  A larger health pool does a few things for them: it makes each attack a smaller proportion of that bar, which makes those hits feel less dangerous to the healer (because they are).  it also gives them more time to react, because they’re thinking in terms of percentages of your health bar, not hit points.

More importantly, it lets them plan more effectively.  If they know that the next hit is going to be a relatively smaller portion of your health bar, they cay comfortably start casting their slow heal, or allow that slow heal to finish if they’re mid-cast.  Doing that actually saves them mana in the long run, because they need to cast fewer of their quick, mana-inefficient “panic” heals.  This isn’t even conjecture, by the way – it’s pretty well-established healer theory.

Aside: There are actually a lot of parallels between healer theory and network queuing theory.  It’s one topic I rather enjoy discussing with a colleague of mine who is both a computer engineering professor and a priest healer.  Crazy fun stuff if you’re into that sort of thing.

3. Aggro is something you should never need to worry about. Vengeance does that job for you. If you lose threat to a non tank then its not a gear issue.

On this, at least, we agree.

4. While tank DPS mattering vs not mattering is an argument for another thread, “DPS” stats are most certainly important for a prot paladin. Hit, expertise, haste and mastery help with point 2 in reducing incoming damage more than even dodge or parry. Stamina does not ever help reduce incoming damage. (Keep your bloody DK comments to yourselves)

False.  Flat-out false.  Hit, expertise, haste, and mastery are ALL inferior to avoidance in terms of reducing damage taken.  I have tested this extensively, as have others.  If you’re suggesting all of that work is in error, I encourage you to present your thoroughly-detailed, carefully-tested mathematical models to the community for peer review.  After all, I have, it’s only fair that you do the same.

Stop telling people they need “enough” stam to not get two shot. They already have it.

Stop telling people they already have “enough,” because frankly, they don’t.

Seriously, consider this “average” tank you keep referring to.  First of all, let’s clear up some misconceptions.

1) The average tank is relatively undergeared for normal raid content.  Hence, they are already in a position where stamina is probably (if not definitely) their best stat.  Running LFR doesn’t count, you could run LFR in greens gemmed with spirit and still perform decently well if you had half a clue.

2) Unfortunately, the average tank doesn’t have a clue.  They don’t follow a tight rotation.  They stumble through it, push back holy power generators, and have large amounts of “real-life” latency because they aren’t hitting a spell immediately upon the GCD ending.  They don’t time their SotR according to the boss’ swing timer.  They don’t pool holy power so that they can double-up on SotR during a dangerous spike.  In many cases, they macro SotR to Crusader Strike.  I’ve even seen tanks go a full 20+ seconds hitting NOTHING BUT CRUSADER STRIKE (and, of course, their SotR which was macro’d to Crusader Strike).

The average tank plays considerably below their theoretical potential based on gear.

In short, the average tank is actually pretty terrible.

Now consider that this is the tank AskMrRobot is trying to help.  You cannot rely on them running a proper rotation, so trying to tell them to stack haste and mastery is a waste – they’re not timing SotR to cover boss swings, and in some cases, they’re not getting maximum uptime on SotR.  Giving them more haste often just creates more empty space rather than more ability usage, because they’re used to a 1.5-second GCD from playing other characters and just stick with that inner metronome.

In other words, buffing active mitigation stats for a tank that uses Active Mitigation poorly is not efficient.

What is efficient? Passive mitigation.  Avoidance.  STAMINA.  And to a lesser extent Mastery, thanks to the block component.  These are all stats that are always on, helping the weak tank survive through content they are undergeared (and often under-skilled) for.  When you try and optimize for a player that is guaranteed to screw up a lot, you give them the stats they cannot screw up.


This was a long post, but if nothing else, here’s what you should take away from it.  You may find that the baseline stats on AMR do not suit you very well.  In most cases, that means one of two things:

1) You’re not running 25-man content, which generally stresses EH more than 10-man content.  I’m not making a 10 vs 25 difficulty argument here, this is simple facts based on tuning.  10-man bosses don’t melee as hard as 25-man bosses because of healer throughput considerations, and smaller hits generally make EH less important.


2) You’re already at a skill or gear level that is above-average.  This may come as a shock to you, because everyone tends to think they’re the “average” tank.  I even think so sometimes, until I look at logs or progression rates.  As such, the “average” settings may not work as well for you.  However, you should also be intelligent and experienced enough to realize this fact, and adjust the stat weights to suit your needs.

I actually heavily de-emphasized stamina in the current set of stat weights compared to what I gave AMR in 5.0/5.1 based on feedback from 10-man raiders.  While I still think that stamina is a GREAT stat in all formats (seriously, I think that even in 10N, more stamina is rarely a bad thing, it’s just often not as useful as more haste thanks to the DPS contribution – but STAM is still the better survivability stat), I recognize that a lot of 10-man players really don’t care for stam stacking that much.  So this time around, the stat weights were chosen such that it favored hybrid haste/stam or exp/stam gems in nearly every non-blue slot.

Lowering the stam weight from 1.5 to 1.3 should shift it into an all-out haste (or mastery) mode that would be appropriate for 10N tanks which are overgearing content or feel comfortable with their current stam level.  I don’t develop AMR, so I can’t modify it to have a stam “cap,” but doing so wouldn’t be a bad option either.  Another possibility is to offer a 10/25 toggle or a different set of weights for 10 vs. 25, but again, that’s up to the people who make the tool.  I just do the math.

Also note: I’m not affiliated with AMR, even if I find their tool incredibly useful in my own work.  I complain about it a lot too, because I think there are areas it could be better.  But I generally make suggestions to them about how to improve it rather than taking a giant dump on the tool on forums, because the latter isn’t very productive.  As Wrathblood said, it’s like hating on a slide rule.

This entry was posted in Tanking, Theck's Pounding Headaches, Theorycrafting and tagged , , , , , , , . Bookmark the permalink.

56 Responses to I can go make a blog in 2 minutes

  1. Kalon says:

    Great post.

    I do wish you’d push armor as a good EH stat as well. Or resistance, if you could get it any more. Armor gives huge amounts of EH but also significantly helps out healing (I’ve written about it before, but the short way to look at it is look at the case where you get heals that aren’t enough to heal a tank up and see which is better – stam or armor – when looking at EH) and makes healing much better per point. It’s not on gear all that often – but when it is, it’s a giant win. I suspect strongly that if you did analysis, you’d find that it’s always better to get a higher tier gear even with fairly meh stats because of the armor boost.

    • Theck says:

      I do push armor as a good EH stat. It’s one of the best, though I still think Stamina is better. Armor is great against physical, but useless against magical. Stamina works everywhere. There are very few fights that challenge the tank with purely physical damage anymore, so that tends to devalue armor a bit compared to Stamina in my mind.

      But they’re both excellent stats. The primary reason I don’t mention it more often is that you can’t gem or reforge for it, and Blizzard hasn’t been itemizing bonus armor since early Cataclysm. So for most of your day-to-day gearing decisions, it doesn’t come into play. But it’s definitely the reason that higher-ilvl gear is generally preferable, and also why I gave it such a high value on AMR (if I recall correctly, the stam weight was 1.5 and the armor weight was 1.0; only hit and expertise exceeded armor’s stat weight).

    • Mechaninja says:

      are there ways to get more armor short of increasing your ilevel? Theck seems to be focused on things you can actually improve.

      I remember in BC, someone told me I couldn’t tank even Kara on my bear until I got that trinket from Blade’s Edge. I read more, and he was wrong, but he wasn’t FAR wrong. You could do something about more armor back then.

  2. viktoryls says:

    Paying domain names is hard.

    As a healer, “smoothness” is very important and saves a ton of healer mana. The less spiky the encounter is, the smarter I can heal and make judicious use of my entire toolkit, not just the fastest heal every melee swing. Yes, spikes should be mitigated by cool downs, but if encounters and mechanics were executed properly all the time, every guild would be farther progressed than they are today.

  3. Weebey says:

    Small nitpick about the TDR value of avoidance vs other stats: it is certainly true that avoidance BUILDS take less overall damage than any others. However, if you interpret the question in terms of the marginal TDR value of a stat, i.e. the TDR impact of one additional point of that stat conditioning on the rest of one’s stats, avoidance is no longer necessarily the strongest. For instance, running my own character through Theck’s 5.1 spreadsheet with DP enabled mastery is significantly stronger at TDR than avoidance (~.47 vs ~.37). (Not that I normally run with DP enabled; it was just easier to get that number than averaging the HA and non-HA values.) This is of course a reflection of the synergy between accuracy stats and mastery (and haste), and doesn’t take into account the Divine Purpose changes. This isn’t news to Theck, who I am fairly certain pointed this out when he was running his original TDR sims back in beta.

    As for Ask Mr Robot, I certainly like the idea, but I think there are still some issues with their optimization code. Often it will output a “solution” for which very obvious manual changes will result in increasing the objective function (e.g. it reforges from a non-capped stat to another non-capped stat with lower weight.)

    • Theck says:

      The synergy between mastery/hit/expertise/haste can push mastery ahead of dodge and parry for TDR when you’re considering small changes, but I also think that it’s a little misleading to say that makes mastery “significantly stronger.” Sure, those 10 points of mastery are stronger than adding 10 points of dodge or parry, but only because you’ve already sunk 7650 rating into hit and expertise and a bunch more into haste, all of which are terrible TDR stats.

      What really matters is the whole of your gear set, and how that affects your overall damage taken. That’s partly why I’ve stopped generating differential scores. They only have meaning for a particular gear set, which limits their applicability. By using realistic “gear set” compositions, we get a more accurate picture of how an overall gearing strategy works.

      I also find that AMR’s solutions are not very robust. It’s not uncommon for me to find that I’m able to tweak things by hand to improve them slightly. That said, this is a discussion I’ve had with one of AMR’s developers (yellowfive) on several occasions. It ends up being a trade-off of speed versus accuracy. They have a vested interest in having their optimizer run quickly to minimize server load and wait times for users, so their tolerances are a little lax compared to what you or I might like. While I’d personally like a more robust algorithm, I can’t fault them for that decision either.

      When there are egregious errors, like the algorithm failing to cap stats properly or clearly reforging a sub-optimal stat to an optimal stat (like haste to mastery or dodge in a control/haste build), I report it. Those usually get fixed pretty quickly.

  4. zoopercat says:

    Weeby. I’m from Ask Mr. Robot. When you see odd cases, can you send them my way? Twitter or email are great (feedback -at -askmrrobot dotz com). Sometimes it’s just a matter of explaining why Mr. Robot is doing what he’s doing, other times there are bugs (gasp!!!)

    • Zoroth says:

      Amazed to see you can still handle the extraordinary (and exponentially increasing) volumes of bullshit, Z. Hope all is well.

  5. Matty says:

    I love Mr. Robot, and Mr. Robot loves me. Nuff said.

  6. Shoot says:

    Lol Mr. Robot doesn’t like to make me expertise capped. Therefore I don’t like it anymore. I cannot figure it out either. It only happens when going for the control/haste build.

    • bryjoered says:

      There is a force cap option in AMR. If you are not reaching the caps, likely you don’t have the secondary stat budget for them. Mr. Robot prefers that you have stamina on all your gems before sticking a big fat and probably sub-optimal pure expertise gem in your gear. It’s really a very good tool, and is only limited by the time it takes them to research and subsequently update every class and spec. It’s not always right, but it corrects those mistakes over time. Even so, if you think you know the optimal stat weights that AMR doesn’t have you can set your own custom weights and it’s STILL an exceptional tool for that.

      • Shoot says:

        It’s not that. If I choose the haste build, it puts me at 9% expertise, even with enforcing the cap. If I choose the mastery build, it’ll get me expertise cap, but for some reason it favors haste when reforging. It really doesn’t make any sense and none of the stat weights are off either.

        • bryjoered says:

          Yeah, it’s buggy sometimes. The best thing to do is to hit update from armory and make sure you have the right preset. If you are still having problems, post on their forums. They are EXTREMELY helpful and prompt in their replies.

        • zoopercat says:

          Hey, zoop from AMR here :) What is your character name and realm, I’ll see what’s going on. One thing to note is that Haste affects the RPPM trinkets, enchants and meta gems. So depending on what gear you have, Mr. Robot might decide (based on math) that Haste is better.

          Let me explain a bit more: Haste increases the proc chance of the RPPM ‘things.’ So if you get extra haste and extra procs, Mr. Robot scores the optimizations by adding in both haste (multiplied by the haste weight) PLUS proc stats (multiplied by their respective stat weights). If that score outweights the score of Mastery alone, Mr. Robot will pick Haste+extra procs.

          I’m not sure if that is happening with your character until I see it. So I want to check up on it and make sure there isn’t something else going on.

          • Shoot says:

            Shootmeagain on Ner’zhul. Something is bugged with it and I’m not sure what. I also get an error sometimes when it tries to locate my helm.

          • Shoot says:

            Thanks in advance!

          • zoopercat says:

            I don’t see it reforging to haste, out of the exp cap. Can you save your profile (blue save button) and send me the URL? That will let me see exactly what you’re seeing – same options and everything.

            Here’s what I see for you – so I’m guessing we have different options selected: http://www.askmrrobot.com/wow/gear/ed95207e-a610-4be2-8b0a-25525ae4d5c9

            Also, I think that helm wasn’t in the databases yet, so it should be added now. Let me know if you still get errors, and what the error says.

          • Shoot says:

            It’s working from the link you gave me. I went onto a different browser to to make sure there wasn’t any saved settings or something and it did the same thing. After raid I’ll see if it does it again.

  7. Talarian says:

    I love this post so, so much.

    “The only constant in those two statements is that you really, really shouldn’t be giving advice.”
    Nothing like shutting down a poster with cold, hard logic. Normally I like to be pretty polite, but folks throwing out false information and then attempting to make the appeal to authority that they ARE the authority always makes me both sad and giggle a little.

    “Aside: There are actually a lot of parallels between healer theory and network queuing theory. It’s one topic I rather enjoy discussing with a colleague of mine who is both a computer engineering professor and a priest healer. Crazy fun stuff if you’re into that sort of thing.”
    Would it ever be possible to delve into that aside? I’m a comp sci major myself, and would love to see such a discussion.

    But yes, as a healer, anything that adds stability and predictability to a fight is better than taking less damage. And all that extra stamina gives me, the healer, a buffer to react to big changes. If it means you die in 4 hits rather than 3, or even 5 hits, all the better. I can pop cooldowns to catch up on throughput, but I can’t catch up if the tank is dead.

    • Theck says:

      It would be possible to delve into that aside, but frankly I’m not well-versed enough in it to do so. I can only provide you with basics because I don’t know much about network queuing theory.

      My (limited) understanding is that as a priest healer, he chooses spells based on how long the tank has until death. He’s planning 6-7 seconds in the future at all times – if the tank won’t die in the next 2-3 GCDs, he’ll use a slow, mana-efficient heal. If that risk becomes possible/likely, he’ll use a faster/larger heal. That dovetails into my work because the information I produce about “is a X-hit string exceeding Y% probable” lets him plan more effectively. The analogy to queuing theory is that you’re constantly gauging and predicting incoming network traffic (damage patterns) so that you can choose effective routes (heals or spell choices) that minimize packet loss, wait times, and latency (deaths).

      I’m sure he would cry foul at how blunt and inaccurate that description is, but that’s about the limit of my understanding since I’m not well-versed in queuing theory.

      • flosch says:

        The combination sure gives a new, if very punny meaning, to RED queue management. Which stands for “random early drop”. Whoops, there goes the tank, sorry!

  8. Steve says:

    Good post.

    I have to admit to stacking hit/exp/haste over stam, but for me it’s for dps output and being able to get holy power faster for when I know the healers are in trouble. Self healing seems to have come such a major tool for us. I assume battle healer also benefits from haste.

    However I am in the 10man camp and have stam trinkets etc in my bags ready should the fight need it.

    Optimal maybe not, it just feels for 10s, tank dps is often more important than rare damage mitigation.

    • Theck says:

      I wouldn’t call that “not optimal.” Optimal is whatever it takes to kill a boss more effectively. In the case of many 10-mans, that doesn’t mean “stack more survivability.” Tank DPS is a very real and very significant factor in a 10-man raid. So I think it’s entirely fair to say that meeting a minimum survivability requirement and then stacking haste is optimal for 10-mans.

      The error occurs when someone claims they’re doing that for survivability, because they flat-out aren’t. If survivability is all you care about, then stacking stamina to the sky is generally your best survival strategy. However, you *can* reach a point where adding more survivability does not significantly affect your likelihood to die (the murky “enough” stamina threshold everyone talks about, but is rarely well-defined). Above that point, it’s often more useful to stack something to increase DPS, and for tankadins haste is king because it does that while *also* giving you a non-trivial amount of survivability.

  9. queldan says:

    Theck is rarely angry, but like another nerdy archetype of some reputation, you won’t like him when he’s angry!

    Honestly, though, I’m under the impression that most of the trouble with AMR is people using AMR thoughtlessly. Like every program, if results don’t match expectation, re-check the base data (your config). 95% of the time, he problem is there.

    • Theck says:

      Which leads to my choice of stat weights. It seems reasonable to assume that the player using AMR thoughtlessly is not the type that is reading resource sites (EJ, MT, Tankspot, my blog, etc.) and tweaking their gear from encounter to encounter. They want a one-size-fits-all solution because that’s all the effort they want to put into it. And stamina is really the only one-size-fits-all survivability stat.

      • bryjoered says:

        I think everybody would like a one-size-fits-all solution. My only argument for Mr. Robot is that the current paladin stat weights, and warrior ones for that matter, are directly based on the hard work that Theck has done. How is that any different than going on EJ for the stat weights and doing them yourself? It might even be more optimal than doing the math in your head, surely quicker. The only point of contention is the stamina weight, which is very hard to simulate, but almost universally accepted as the best all around stat for tanking. Add this to the fact that there is both the Control/Haste build along with the Control/Mastery build and you have the two most popular tankadin presets in MOP. Obviously, I’m an AMR fan, but I just feel there is a stigma surrounding the site and players that use it. As if manually reforging and gemming based on theorycrafting is any different than what the people over at AMR are doing.

        The best tanks will change up their styles per encounter for heroic progressions sometimes stacking more mastery than usual (Sha of Fear) or More stamina (Lei Shei). For these types of players AMR is probably not the best option, but for an all around optimal build in EVERY situation, I think AMR is pretty good.

        • zoopercat says:

          Bryjoered, you say that for tanks that specialize stats for a particular fight – AMR might not be the best option. I want to ask for clarification. We made the tool to be all math-based, so if you edit your stat weights you get custom suggestions. So if you want to stack Mastery, Mr. Robot will do that. There are millions of combinations of reforges alone, so a tool like AMR does the math for you.

          So I was hoping to get clarification in case I’m missing something about the flexibility of the tool (or a bug). We love to hear from users, so lay it on me!!

          • bryjoered says:

            I was just referring to the default presets. They generally cater towards a universal solution as opposed to specialized ones. AMR stat weight editor works pretty well. I altered some weights in there to get more hit/exp before they added the hit/exp preset. I was saying that for heroic progression tanks it may not be in their best interest to always follow the default preset, as they often change up their styles a little bit per encounter.I said before that I still feel that AMR is a great tool for reforging and gemming, even if you have your own custom stat weights.

  10. Bassm says:

    Hello Theck! My name is Bassm… and… I actually do have a question (unrelated to my mediocre abilities).

    WRT Boss Swing Timers – what… is a good addon that has it? I’ve sort of glanced around in Curse and have not found one.

  11. blizzhoof says:

    My only problem with Askmrrobot is how it rates trinkets. It’s just hard to believe that the Jade Warlord Figurine is stronger than Ji-Kun’s Rising Winds or that Soul Barrier is a decent trinket at all (then again, I’m not a fan of Colossus – the enchant – either). I understand it’s based on stat weights, but is there a way to include the procs/on-use effects or to make a separate trinket ranking system?

    Then again, maybe I’m just ignorant and over-value the secondary effects of trinkets as well as under-valuing stamina. Although Ji-Kun and Durumu (350k melee swings on tanks ftw) are making me value stamina a lot more. Love the blog btw.

    • Theck says:

      Trinkets are one of those notoriously hard-to-quantify areas. Some on-use effects are great, others are lackluster. I personally like Soul Barrier, but more for the raw Stamina than the on-use. It’s very likely I’ll be using Soul Barrier + Lao-Chin’s Liquid Courage (2/2) for many heroic fights this tier.

      Also note that trinkets have a stamina bias compared to gems. You get 1.5 stamina or 1 secondary stat (haste/mastery/hit/exp) per itemization point on a trinket. You get 1.5 stamina or *2* secondary stat per point on a gem. So mathematically speaking, trinkets are a much more efficient way to pick up stamina than gems. If you’re trying to convert “excess” stamina to secondary stats, you should be converting all of your gems first before dropping stamina trinkets. (This of course ignores situations where the special effect of the trinket is the deciding factor – i.e. I’d probably take a trinket with spirit and an automatic Cheat Death effect over a stam trinket).

      • Mechaninja says:

        As I’m sure you know, trinkets can be complicated for DPS as well. Pathal and friends are going nuts trying to model Rune of Origination, for example.

        I assume Blizz does that deliberately.

        • Theck says:

          Oh sure, but with DPS trinkets you’re still trying to maximize a single metric – DPS. Tanks don’t have a single unified metric. It’s much less clear how you balance a “Last Stand” on-use proc vs. an avoidance on-use proc vs. a haste proc vs. etc.

          That doesn’t mean the theorycrafting for DPS trinkets is any easier (in fact, often I think it’s harder – see the Markov blog post from last week). It’s just a different game entirely for tanks, because there IS no clear way to arrive at a numerical result with such disparate choices.

  12. bryjoered says:

    Totally agree with you blizzhoof, I’m not a complete believer in their BIS in general. I got a ilvl 528 “Thunderforged” tank trinket off the Council of Elders, you think I’m not going to wear that over a 489 trinket? Even though, especially for pallys, dps strength gear might actually be better, in a raid environment most tanks get tank gear and don’t even have the option to roll on melee dps gear.

    • Putting too much weight on ilvl is quite dangerous and not always correct. Trinkets are situational. Our warrior tank is planning to roll with a double upgraded Shado Pan stamina trinket through T15 heroics.

    • Theck says:

      I picked up the 528 Fortitude trinket you’re speaking of as well. It was my first bonus roll loot in ToT last week. Sadly, it’s sitting in my bags, because the 489 Jade Warlord Figurine has been more valuable to me. That might change as I replace other slots with more 522 gear and get more stamina overall, at least for farm bosses or special cases (I can imagine using Fortitude for extra mastery on some of the “one-shot”-esque boss mechanics, like. Ji-Kun’s talon ability).

      • bryjoered says:

        Yeah, the on-use is actually pretty good for the trinket though, like a small last stand. I can still use lao-chin along with it or niuzao. AMR would recommend I just use lao-chin and niuzao until I get soul barrier. I dunno I really like some of these new trinkets and with a guild like mine by the time we get to heroics in this tier I’ll probably be good and geared. I’d estimate that we down lei shen about a month and a half from now. The ji-kun’s trinket looks really useful as well with it’s special ability.

        • Theck says:

          Yeah, I really like the last stand effect on Fortitude – it’s the one reason I may end up using it at all, to be honest. I can imagine that I’ll swap between Soul Barrier + Lao-Chin and Soul Barrier + Fortitude depending on whether I feel like I need the always-on extra health or whether I can make effective use of the Last Stand effect, which could give me a much larger amount of health for short durations.

          Also note that the Niuzau trinket will score highly because the on-use is massively over-budget. It gives over twice as much dodge rating as Lao-Chin gives mastery, at no cost to the stamina itemization, despite being lower ilvl.

  13. Moshne says:

    This doesn’t really impact your conclusion, but I’m not sure I buy the premise that the average tank (the person you in theory built the default stat weights for) is the average user of AMR. You are significantly oversampling advanced normal mode and heroic mode raiders when they’ve made the decision to branch out and start optimizing their gear.

    Perhaps it would be useful to link to the profile used to come up with assumed stat weights? If someone knows that a 507 average ilvl was used to compute them, they’d have a better understanding of what they are looking at. I doubt it would prevent threads like you found on the official forums, but it might be useful information for those that really care (your intended audience.)

    • Theck says:

      Oversampling, or undersampling? I’m not sure whether you mean that I’m oversampling them by making the Stamina weight so high (since that has more relevance to heroic modes), or undersampling them because a greater percentage of the “average” tanks don’t bother optimizing their gear.

      In any event, there isn’t really a gear profile involved in the stat weights. These aren’t meticulously calculated based on smoothness or TDR metrics, because that sort of accuracy just isn’t general enough for a tool like AMR. The mastery value, for example, changes drastically depending on how much hit, expertise, and haste you have.

      This is in stark contrast to the way stat weights are usually calculated for DPS classes, where the weights *are* meticulously calculated. That’s because DPS classes have a far more uniform metric to work with. DPS is DPS, and it’s easy to determine how much more effective haste is compared to mastery to a reasonably high degree of precision for an average encounter.

      That’s just not easy (or in my mind reasonable) to do for tanks. Partly due to mechanics – sometimes you care about short spikes, sometimes long spikes, sometimes magic damage, and sometimes single-shot abilities. Further, the gear variations will generally be larger for tanks than for DPS at fixed ilvl – there’s a variety of gearing strategies that are viable, and even within a single strategy we frequently “make do” with sub-optimal pieces (i.e. dodge/parry) because they dropped, they’re a slight upgrade in several departments, and nobody else will use them anyway.

      Instead, I chose stat weights that tried to match the overall gearing strategy. I.e., “I want to focus on stats X/Y/Z, but will take pieces with V/W on them if they’re a large enough upgrade in other departments.” I also tweaked them based on gemming constraints – for example, if I want to pick up socket bonuses with hybrid gems, I need to make sure that 160 secondary < 120 stam < 160 secondary+120 secondary. And so on.

      • Moshne says:

        I’m saying that there is a bell curve of “average raider” (1) and there is a bell curve of “average user of theorycrafting sites” (2) and the curve of (2) is almost entirely centered in the right end tail of (1). There are surely bad players out there who theorycraft, but I’m not sure they are helped considerably by fudged stat weights in their favor. Like I said, it doesn’t really impact your conclusions considerably, but I think the premise that your defaults should be chosen when the average raider in mind is flawed, as the average raider is unlikely to ever see it, you really want the average AMR user.

        I’m mostly just suggesting some context is in order, outside of having to come to your blog to understand why certain stats were done as a default. This is more a criticism of AMR though, as I don’t think the site really supports letting you explain why you chose what you did.

        • Theck says:

          Ah, I understand your argument now. And I think I agree with you about the overlap of those two distributions. Though I’d be curious to see statistics from AMR themselves.

          That said, the difference in stat weights is pretty minimal. For example, I actually use the Control/Haste preset when I optimize my own gear. It’s perfectly applicable to a 25-man normal/heroic raider. I also agree with the argument that it’s less applicable to a stronger 10N raider, as the survivability threshold i slower in 10N. But I think that’s more easily solved with a stamina cap than by tweaking the stat weights. The survivability stat weights don’t really change, what changes is the point at which you say “that’s enough survivability, now let me boost DPS with the rest of my itemization.”

          • zoopercat says:

            Ask, and you shall receive. Here’s data from the most recent 500,000 characters loaded into our site:
            >=450: 89.73 %
            >=455: 87.95 %
            >=460: 85.66 %
            >=465: 81.78 %
            >=470: 76.68 %
            >=475: 69.26 %
            >=480: 58.81 %
            >=485: 45.53 %
            >=490: 30.32 %
            >=495: 16.61 %
            >=500: 8.19 %
            >=505: 3.57 %
            >=510: 1.31 %
            >=515: 0.31 %
            >=520: 0.02 %
            >=525: 0.00 %
            >=530: 0.00 %
            >=535: 0.00 %
            >=540: 0.00 %

          • Zoroth says:

            Is anyone else SUPER curious who the few world-first-level raiders rocking the 520+ average ilvl using AMR this week are?

          • Moshne says:

            Is this players who forced themselves into the site via a refresh? My character was loaded in without me ever asking it to be. That appears to be the gear distribution for level 90 players in general.

          • Zoroth – I’m not world first, just very lucky, and I’m in the 520 numbers. If you armory Anafielle, I’m at 522 (although that’s counting my offspec gear and mainspec gear together). And I was extensively using AMR last night to play with my reforging and export to Simcraft.

            AMR is a good tool, even if you aren’t trusting its dps numbers. I like to use it like I used to use Chardev.

  14. bryjoered says:

    HIs point is that AMR is built behind the premise of making theorycrafting accessible to everyone. Most heroic progression tanks reforge and gem themselves with the help of theorycrafting. I think people would be surprised how similar the gearing strategy of those progression raiders and AMR is. In most cases, AMR gives adequate stat weights, and very rarely gives you flat out wrong and terrible ones. If you’re truly a heroic progression raider than you shouldn’t need the assistance of a tool like AMR anyways. That being said, I think at the very least, it’s a great tool for reforging, even for the most advanced raiders.

  15. Wrathblood says:

    What’s this “Busy week at SacredDuty” stuff?

    • Hype!

      He’s got some good stuff to post this week…….. get excited. :)

    • Theck says:

      Well, if I can force myself to sit down and write a post this evening, you’ll have your T15 2-piece sim results tomorrow.

      I have a different surprise in store for Thursday or Friday.

      • Wrathblood says:

        Oh, well, if it ended up being an idea for a useful blog piece then I’m happy if I helped. But if you’ve sat down and cranked out 2,000 words just telling me that I need to let it go and come to terms with it being crap, then I’m going to feel guilty.

Leave a Reply to bryjoered Cancel reply