Stamina: Keeping You Up Longer

What’s the most misunderstood element of tanking?  There are a myriad of topics you could offer as guesses.  But the question I field more often than any others isn’t about rotations, talents, glyphs, or cooldowns.  It’s not even about which secondary stats to focus on, even though this expansion has thrown us for a loop in that department as well.  Without a doubt, the question I get asked most often is “How should I value stamina?”

And I think it’s fair to say that this question is one of the great mysteries of tank theorycrafting.  It’s spawned discussions, arguments, and more than a few blog posts.  Everyone seems to have an opinion on the topic, but few seem to have based that opinion on evidence.

More to the point, I think a lot of the opinions I’ve seen offered are flat-out wrong.  One commonly-held opinion, which came up again a few weeks ago, is that the best survivability strategy is to stack stamina until you have “enough” and then switch to other, better survivability stats like haste.

Now, there are plenty of holes in that argument.  But the biggest, in my mind, is the errant assertion that there are better survivability stats than stamina.  The entire strategy relies on the idea that stamina becomes less valuable the more you have of it, but more importantly that this drop-off is rapid and happens at an attainable threshold (the nebulous “enough” value).  I think there’s no question that the first part is true – going from being able to survive two boss attacks to three boss attacks is certainly a larger survivability gain than going from ten to eleven boss attacks, and there has to be a fairly continuous progression between those two points.

But the second part is misguided.  I could see this argument if we were interested in total damage reduction (TDR), where stamina does draw the short straw. However, I’ve yet to see someone employ the argument to get “enough” stamina and then stack avoidance, which would be the optimal TDR strategy.  And to be frank, TDR is an afterthought in the current tanking environment – anyone trying to optimize for it is doing their raid a disservice to begin with.

No, in every case I’ve seen, the person giving advice was suggesting that the player stop stacking stamina and focus on a smoothing stat.  Generally haste, but on occasion I’ve seen mastery mentioned.  Buried in that advice is the unstated assumption that stamina isn’t as good at smoothing as other secondary stats.   And that assumption is just flat-out wrong.

Since this is a fairly pervasive misunderstanding, it’s about time we proved it.

Model Development

The first question we need to ask is exactly how we’re going to measure the effect stamina has on damage smoothing.  You may remember that in an earlier blog post, I said that we’d run into a number of problems with adding stamina (or health) to the simulation.  If we wanted to be truly rigorous about it, we’d need to include healers, health tracking, and some sort of AI that models how a healer reacts to damage spikes.  While I think all of that is true in the strict sense, I also think we can cheat a little bit.

For example, in our recent simulations we haven’t included the differences in healer reactions for a haste tank versus a mastery tank or an avoidance tank.  Your healers certainly do react differently to each of those gearing strategies, but we haven’t even considered it.  We narrowed our view down to a metric we can easily deduce: damage spike size.  It stands to reason that we can do the same with stamina.  The only caveat is that when comparing haste and mastery, healer reaction differences is a fairly small effect, so we can reasonably expect our comparison to be a pretty close approximation to the real world.  Stamina has a much larger effect on healer reaction than either of the other stats though, so by throwing out any pretense of modeling healers, we are ignoring a significant portion of stamina’s benefit.

In other words, we can estimate stamina’s worth with this sort of simulation, but in doing so we necessarily undervalue stamina.  We get a lower bound on its value, but the true value is going to be higher in practice.  That’s OK though – having a lower bound is better than having nothing at all.

Updating the model is fairly easy otherwise.  The formulas that convert stamina to hit points are well known, and were easy to insert into the sim.  I needed to make some changes to the gear sets, which we’ll discuss shortly, but those weren’t difficult either.  The biggest change that needed to be made was, ironically enough, updating the data reporting scheme.

Data Reporting Changes

The data reporting scheme we’ve been using up until now will not be sufficient once we include stamina.  We’ve been reporting spike representation as a function of “maximum throughput” – comparing the size of a N-attack spike to a string of N full hits.  But that metric doesn’t include health at all, so the impact stamina has on spike damage is transparent to it.  We need to come up with a different way to report the data that is a little more universal.  And to be fair, I think this change has been a long time coming.  While working in normalized units is convenient for the mathematician, it’s not always very intuitive.  I think that many readers have had trouble parsing the data tables on these posts because of the “unnaturalness” of the metric.

To illustrate why this is, note that in the last post we were discussing damage “spikes” of 30%-40% maximum throughput.  This led to a discussion between Mel and I about what really constitutes “spike damage” in the first place.  Your healers are going to have to deal with strings of 2-3 full hits in a row from time to time, so they are definitely going to plan their healing setup around having enough throughput to keep up with the occasional back-to-back or back-to-back-to-back hit.  Their throughput healing is almost guaranteed to exceed 50% of the boss’s maximum throughput, because on occasion they need to be able to heal through 100% throughput.  So is a string of attacks that doesn’t exceed 30% of maximum throughput really a spike? Or is it something we shouldn’t even be concerned with?

I think Mel makes a strong argument here that the categories we cared about were generally the highest one or two in each column rather than the bulk of the distribution.  And that argument makes haste an even more solid victor in the last round of simulations, though I think it was a pretty clear winner anyway.  But even I got a little too focused on the numbers and didn’t give enough thought to what the numbers meant, and I think that primarily happened because the metric isn’t as clear as it could be.

So it’s about time we came up with a better, more intuitive metric that people can look at and instantly understand.  And the logical way to do that is to think about the data the same way a healer would.  What a healer notices when you take damage is the size of the chunk that disappears from your health bar.  In other words, they care about how big a hit is compared to your maximum (or current) health.  As such, it seems logical to normalize our data the same way.  Thus, in the tables in this and future posts will list the size of the spike in terms of percentage of total health.

This metric has two major advantages.  First, it’s very intuitive – if I see that a spike is 100% of my max health, it’s very clearly deadly, while a 30% spike probably isn’t.  Second, it automatically includes stamina, because increasing your total health reduces the relative spike size accordingly.  And it’s a logical way to incorporate stamina, as turning a 100% spike into a 90% spike by stacking stamina has a very clear meaning (namely that you can now survive something you couldn’t before).

Gear

With the metric “fixed,” we need to consider how to modify our gear setups.  I could include stamina options for each of the different strategies, but that seems like overkill.  The Control/Haste set has been dominating the charts, so it seems reasonable to build our stamina set out of that one.  Stamina doesn’t have any sort of synergy with any of our secondary stats, so there’s little point in trying different versions of stamina strategies if our trial case is a clear winner or loser.

We do, however, need to assume a baseline stamina value for all of the sets.  After doing a little research, I settled on a value of around 28k stamina from gear.  Note that this is before the modifiers for Plate Specialization and Guarded by the Light, so your character sheet will generally show a value that’s about 31% higher than this.  That 28k stamina will look like around 37k stamina on your character sheet once base stamina is included.  And of course, Power Word: Fortitude and Flask of the Earth will increase that value further.  The simulation assumes all of these effects are present to generate the player’s hit points.  Once all of that is factored in, this gives the player 755k health fully buffed.

For our stamina set, we’ll stay conservative.  We’ll only sacrifice 4000 of the 12k haste that the Control/Haste set is packing, which gives us an additional 6000 stamina.  Our stamina set, which we’ll call “C/StH” to indicate that it’s a Control/Stam+Haste set, then has a total of 34k stamina before flask and raid buff are considered, or a little less than 45k stamina on the character sheet.  This is equivalent to 876k health fully buffed.

The stats for all of the gear sets are given in the table below.  Note that the Strength value is after buffs, not before (unlike stamina).  Also note that at Blizzhoof’s insistence, I’ve added the 3000 mastery raid buff to the simulation, though it doesn’t show up on the table below.

|    Set: |  C/Ha | C/StH |  C/Ma |  C/Av | C/Bal |  C/HM | Avoid | Av/Mas | Mas/Av |
|     Str | 15000 | 15000 | 15000 | 15000 | 15000 | 15000 | 15000 |  15000 |  15000 |
|     Sta | 28000 | 34000 | 28000 | 28000 | 28000 | 28000 | 28000 |  28000 |  28000 |
|   Parry |  1500 |  1500 |  1500 |  7500 |  4125 |  1500 | 10825 |   7717 |   4000 |
|   Dodge |  1500 |  1500 |  1500 |  7500 |  4125 |  1500 | 10825 |   7717 |   4000 |
| Mastery |  1500 |  1500 | 13500 |  1500 |  4125 |  6750 |  1500 |   7716 |  15150 |
|     Hit |  2550 |  2550 |  2550 |  2550 |  2550 |  2550 |   500 |    500 |    500 |
|     Exp |  5100 |  5100 |  5100 |  5100 |  5100 |  5100 |   500 |    500 |    500 |
|   Haste | 12000 |  8000 |     0 |     0 |  4125 |  6750 |     0 |      0 |      0 |

 

Data Tables and Simulation Details

The new tables are a little longer, but contain most of the same information.  The top section still lists the mean damage intake, the standard deviation of damage intake (using a 5-attack moving average), and average SotR uptime, “S%”.  However, it now also gives us two new rows describing hit points.  The first is fully-buffed hit points in thousands, “HP”.  The second is that value normalized by the size of a boss attack, “nHP”.  In other words, if you  have 600k hit points and the boss hits you for 150k on average, “HP” is 600k and “nHP” is 4.

After that comes the usual breakdown of spike representation, but this time it’s given in terms of percentage of maximum health.  So for example, if a row is labeled  “40%” in the “4-attack moving average” section, it is telling  you what percentage of 4-attack sequences exceeded 40% of your total health.  For clarity, the new tables automatically filter out rows that are all zeros, so you can safely assume that rows above the highest shown are trivial (i.e. all zeros).  It also suppresses rows where each set has greater than 50% representation – this is our semi-arbitrary cut-off for “spike damage.”  The theory being that your healers can trivially handle anything spike that’s less than 50% of your maximum health, so they’re not worth considering either.

I’m using the SH1 finisher because it generally performs better for C/Ha than the SH2 finisher does. I’ve simmed out both just to check, and this is still the case in these simulations.  I’m also including Sacred Shield in the model.  Again, I’m trying to give Control/Haste every possible advantage it can get just to make sure we’re getting a fair comparison.

Data – 150k Swings

Since there’s a large disparity in damage intake between both 10- and 25-man raids as well as normal and heroic modes, we’ll look at a few different boss attack sizes.  We’ll start with a boss that hits the tank for 150k damage per swing after armor mitigation.  This is our model for a “10-man normal mode” boss.  Many people suggest that since 10-man bosses hit for so much less than 25-man bosses, stamina is far less valuable in that format.  If that’s true, then this simulation should reveal that weakness.  In terms of normalized health, we’re looking at values a little over 5 boss attacks for the usual sets and about 5.8 boss attacks for the stamina set.  Note that this is intentionally shy of a new breakpoint; the stamina set will not survive 6 boss attacks, so we’re not seeing any discrete effects due to being able to survive one more full boss hit (not that those effects are very strong or relevant anyway, but some people seem to erroneously think they are, so we may as well indulge them just in case).

Here’s what the data looks like:

Finisher = SH1, Boss Attack = 150k, data set smooth-10000-41

| Set: |   C/Ha |  C/StH |   C/Ma |   C/Av |  C/Bal |   C/HM |  Avoid | Av/Mas | Mas/Av |
| mean |  0.329 |  0.339 |  0.310 |  0.312 |  0.329 |  0.302 |  0.282 |  0.286 |  0.291 |
|  std |  0.120 |  0.120 |  0.118 |  0.142 |  0.132 |  0.114 |  0.144 |  0.135 |  0.126 |
|   S% |  0.522 |  0.482 |  0.410 |  0.419 |  0.452 |  0.471 |  0.366 |  0.362 |  0.357 |
|   HP |   755k |   876k |   755k |   755k |   755k |   755k |   755k |   755k |   755k |
|  nHP |  5.034 |  5.843 |  5.034 |  5.034 |  5.034 |  5.034 |  5.034 |  5.034 |  5.034 |
| ---- | ------ |  --- 2 | Attack | Moving | Avg.-- | ------ | ------ | ------ | ------ |
|  10% | 62.691 | 49.929 | 54.016 | 54.026 | 50.931 | 48.579 | 46.493 | 45.292 | 50.230 |
|  20% | 15.057 | 10.363 | 21.207 | 18.159 | 20.220 | 16.530 | 16.262 | 18.871 | 18.945 |
|  30% |  1.555 |  0.002 |  0.835 |  3.220 |  0.752 |  0.103 |  2.081 |  0.484 |  0.583 |
| ---- | ------ |  --- 3 | Attack | Moving | Avg.-- | ------ | ------ | ------ | ------ |
|  20% | 43.186 | 38.691 | 46.775 | 40.820 | 47.142 | 43.082 | 34.946 | 40.453 | 37.135 |
|  30% | 16.292 |  4.971 |  9.730 | 15.009 |  7.841 |  6.885 | 11.310 |  7.236 |  9.301 |
|  40% |  0.627 |  0.170 |  2.346 |  1.877 |  2.070 |  1.035 |  1.697 |  2.133 |  2.353 |
| ---- | ------ |  --- 4 | Attack | Moving | Avg.-- | ------ | ------ | ------ | ------ |
|  30% | 38.041 | 25.490 | 30.972 | 34.316 | 34.587 | 25.874 | 27.377 | 25.821 | 27.281 |
|  40% |  4.878 |  2.293 |  9.641 | 10.464 | 11.735 |  3.720 |  7.840 |  9.014 |  7.525 |
|  50% |  1.046 |  0.034 |  0.684 |  2.954 |  1.111 |  0.413 |  2.305 |  1.061 |  1.014 |
|  60% |  0.000 |  0.000 |  0.001 |  0.039 |  0.022 |  0.000 |  0.034 |  0.022 |  0.009 |
| ---- | ------ |  --- 5 | Attack | Moving | Avg.-- | ------ | ------ | ------ | ------ |
|  30% | 59.866 | 49.691 | 54.815 | 52.069 | 57.158 | 49.828 | 43.297 | 43.952 | 47.461 |
|  40% | 24.880 | 16.673 | 24.063 | 24.710 | 29.552 | 18.935 | 18.827 | 22.079 | 20.083 |
|  50% |  6.191 |  2.008 |  3.396 |  9.911 |  9.260 |  3.279 |  7.322 |  4.778 |  3.981 |
|  60% |  0.102 |  0.018 |  0.354 |  0.884 |  0.665 |  0.091 |  0.738 |  0.589 |  0.594 |
|  70% |  0.004 |  0.000 |  0.069 |  0.095 |  0.071 |  0.009 |  0.147 |  0.117 |  0.120 |
| ---- | ------ |  --- 6 | Attack | Moving | Avg.-- | ------ | ------ | ------ | ------ |
|  40% | 47.392 | 36.848 | 42.403 | 41.896 | 50.114 | 38.803 | 32.995 | 37.109 | 36.236 |
|  50% | 19.806 |  9.678 | 15.496 | 21.328 | 24.322 | 13.762 | 15.784 | 13.368 | 13.865 |
|  60% |  4.727 |  0.223 |  3.562 |  6.052 |  6.288 |  1.774 |  4.452 |  4.244 |  3.545 |
|  70% |  0.121 |  0.005 |  0.296 |  1.205 |  1.321 |  0.060 |  1.011 |  0.643 |  0.396 |
|  80% |  0.002 |  0.000 |  0.001 |  0.042 |  0.008 |  0.000 |  0.061 |  0.022 |  0.011 |
|  90% |  0.000 |  0.000 |  0.000 |  0.001 |  0.000 |  0.000 |  0.003 |  0.001 |  0.000 |
| ---- | ------ |  --- 7 | Attack | Moving | Avg.-- | ------ | ------ | ------ | ------ |
|  40% | 65.495 | 55.517 | 59.491 | 56.730 | 65.748 | 57.023 | 46.694 | 51.129 | 52.506 |
|  50% | 38.579 | 24.792 | 32.720 | 35.224 | 42.416 | 29.837 | 27.022 | 26.237 | 27.976 |
|  60% | 16.473 |  3.415 | 11.878 | 16.301 | 19.863 |  8.017 | 11.599 | 11.432 | 10.355 |
|  70% |  2.131 |  0.363 |  1.859 |  5.008 |  6.172 |  1.181 |  3.655 |  2.699 |  2.023 |
|  80% |  0.173 |  0.000 |  0.075 |  1.182 |  0.616 |  0.057 |  0.845 |  0.357 |  0.185 |
|  90% |  0.000 |  0.000 |  0.006 |  0.042 |  0.015 |  0.000 |  0.053 |  0.030 |  0.033 |
| 100% |  0.000 |  0.000 |  0.000 |  0.000 |  0.000 |  0.000 |  0.002 |  0.000 |  0.001 |

For 150k swings, none of the categories regularly exceed 100% of our health, so it’s questionable whether we can call any of these highly dangerous.  That said, there’s no doubt about it – stamina beats the pants off of everything else in terms of damage smoothing.  Even for these very weak boss hits – the regime where stamina should perform worst – converting 4k haste to 6k stamina creates a significant reduction in spike representation.  And it’s not a small reduction either; it’s often reducing spike representation by a factor of 10 or more in the highest categories, which are the ones that matter most.

The C/StH set does take about 3% more damage than the C/Ha set and sacrifices about 4% SotR uptime (absolute, about 8% relative uptime).  But the 3% higher damage intake is a small price to pay for the incredible amount of smoothing that stamina brings to the table.  We make a far less efficient trade of damage intake for smoothness by choosing C/Ha over one of the avoidance gear sets.  And the loss of SotR uptime is only relevant in the context of smoothing, so it’s not actually a loss we’re concerned about.  We’re trading the smoothing of that SotR uptime for much more smoothing through a different avenue.

Now, if you’re a long-time reader of this blog, even this data shouldn’t be surprising to you.  It’s pretty easy to see that a 10% increase in health makes every spike 10% smaller in a relative sense, and that makes those spikes significantly less dangerous.  This is the intuitive basis I’ve been using to argue that stamina was our best survivability stat all along.  All I’ve done here is quantify it within the Monte-Carlo simulation to show that my intuition was correct.

It’s also worth noting that all of the damage in this simulation is physical.  If we add any source of magical damage, stamina gets even better, because it doesn’t care what damage type you’re taking.  So if stamina wins in this best-case scenario for secondary stats, it will trounce them even more in a real encounter with a variety of damage sources.  And remember, we’re already undervaluing it by excluding the benefits to healer reaction time and mana.

Now that we’ve seen how stamina performs in 10-man normal raids, let’s up the ante a bit and look at slightly larger boss swings.

Data – 250k Swings

The 250k-swing category is a rough model of 25-man normal modes and 10-man heroic modes.  With harder hits, one might expect that more stamina is necessary to survive.  On the other hand, SotR mitigation automatically scales with boss hit size, so larger hits naturally make haste and mastery stronger too.  Stamina doesn’t scale with boss hit size, so one could even make an argument that stamina becomes less effective as a survivability stat as boss hit size increases.

In fact, both of these are true.  The relative strength of stamina vs. haste should decrease as boss hit size increases.  Each SotR mitigates more damage, making each point of haste stronger, while every point of stamina is still providing the same amount of hit points.  And that amount of stamina is now a smaller percentage of boss hit size.  In this case, for 250k attacks we’ve dropped to a little over 3 attacks worth of health for the baseline sets, but only 3.5 attacks for the C/StH set. Of course, when a boss hits harder, you need more stamina to survive a handful of boss attacks too.  So stamina is becoming more valuable indirectly at the same time that it’s becoming less efficient.

The question we want to answer is whether stamina drops off enough to make it less useful than haste once we can survive “enough” boss hits.  While still vague, we can roughly interpret that to mean “long enough for our healers to react and counter the spike.”  At any rate, what we’re looking for in the data is a case where the C/StH set trails the standard C/Ha set.

So let’s see if it’s there or not:

Finisher = SH1, Boss Attack = 250k, data set smooth-10000-42

| Set: |   C/Ha |  C/StH |   C/Ma |   C/Av |  C/Bal |   C/HM |  Avoid | Av/Mas | Mas/Av |
| mean |  0.336 |  0.349 |  0.330 |  0.307 |  0.349 |  0.323 |  0.282 |  0.304 |  0.309 |
|  std |  0.123 |  0.123 |  0.121 |  0.135 |  0.134 |  0.117 |  0.139 |  0.137 |  0.129 |
|   S% |  0.523 |  0.482 |  0.411 |  0.419 |  0.452 |  0.472 |  0.367 |  0.362 |  0.357 |
|   HP |   755k |   876k |   755k |   755k |   755k |   755k |   755k |   755k |   755k |
|  nHP |  3.021 |  3.506 |  3.021 |  3.021 |  3.021 |  3.021 |  3.021 |  3.021 |  3.021 |
| ---- | ------ |  --- 2 | Attack | Moving | Avg.-- | ------ | ------ | ------ | ------ |
|  20% | 46.595 | 46.810 | 54.570 | 41.292 | 51.317 | 48.799 | 36.159 | 45.405 | 50.331 |
|  30% | 35.926 | 14.938 | 35.570 | 32.936 | 41.260 | 34.056 | 29.073 | 34.938 | 32.673 |
|  40% |  6.860 |  8.010 |  7.179 |  8.109 | 12.581 |  8.060 |  7.851 | 10.003 |  7.135 |
|  50% |  0.020 |  0.000 |  0.789 |  0.074 |  0.645 |  0.080 |  0.054 |  0.381 |  0.580 |
|  60% |  0.000 |  0.000 |  0.053 |  0.006 |  0.129 |  0.000 |  0.008 |  0.088 |  0.002 |
| ---- | ------ |  --- 3 | Attack | Moving | Avg.-- | ------ | ------ | ------ | ------ |
|  30% | 63.104 | 49.406 | 58.880 | 55.756 | 66.220 | 57.338 | 49.341 | 56.215 | 53.846 |
|  40% | 34.331 | 31.648 | 23.055 | 28.416 | 37.490 | 32.754 | 24.270 | 28.690 | 22.703 |
|  50% | 19.006 |  5.030 | 10.638 | 13.249 |  8.533 |  8.266 | 10.753 |  7.580 | 10.333 |
|  60% |  4.156 |  0.588 |  5.542 |  3.510 |  4.563 |  3.245 |  3.169 |  4.345 |  5.443 |
|  70% |  0.278 |  0.035 |  2.592 |  1.006 |  2.252 |  1.019 |  1.104 |  2.180 |  2.458 |
|  80% |  0.005 |  0.000 |  0.401 |  0.225 |  0.619 |  0.112 |  0.341 |  0.577 |  0.428 |
| ---- | ------ |  --- 4 | Attack | Moving | Avg.-- | ------ | ------ | ------ | ------ |
|  40% | 61.965 | 57.398 | 54.526 | 51.666 | 59.961 | 58.657 | 44.476 | 49.014 | 49.824 |
|  50% | 45.051 | 29.046 | 35.889 | 36.134 | 37.860 | 32.099 | 29.877 | 28.250 | 31.758 |
|  60% | 19.806 |  2.373 | 20.050 | 17.852 | 27.129 | 14.528 | 14.422 | 17.228 | 16.721 |
|  70% |  1.338 |  0.450 |  8.129 |  4.063 | 12.113 |  3.891 |  3.871 |  8.218 |  7.622 |
|  80% |  0.198 |  0.024 |  1.685 |  1.602 |  4.868 |  0.804 |  1.742 |  2.523 |  1.904 |
|  90% |  0.005 |  0.000 |  0.144 |  0.138 |  0.477 |  0.082 |  0.251 |  0.474 |  0.229 |
| 100% |  0.000 |  0.000 |  0.020 |  0.001 |  0.050 |  0.000 |  0.002 |  0.037 |  0.010 |
| 110% |  0.000 |  0.000 |  0.003 |  0.000 |  0.018 |  0.000 |  0.001 |  0.009 |  0.000 |
| ---- | ------ |  --- 5 | Attack | Moving | Avg.-- | ------ | ------ | ------ | ------ |
|  50% | 64.146 | 51.156 | 59.567 | 54.779 | 60.858 | 56.683 | 46.843 | 47.754 | 52.800 |
|  60% | 43.273 | 27.848 | 42.980 | 35.392 | 49.374 | 37.181 | 28.778 | 34.183 | 36.046 |
|  70% | 23.415 | 13.826 | 22.950 | 17.813 | 30.001 | 19.425 | 13.991 | 21.325 | 20.049 |
|  80% | 11.898 |  3.263 |  7.558 |  9.860 | 17.352 |  8.310 |  7.859 | 10.840 |  7.380 |
|  90% |  4.309 |  0.160 |  1.947 |  3.546 |  5.669 |  2.519 |  2.817 |  3.769 |  2.270 |
| 100% |  1.457 |  0.013 |  0.865 |  1.180 |  0.958 |  0.248 |  0.977 |  0.871 |  1.119 |
| 110% |  0.004 |  0.000 |  0.394 |  0.120 |  0.385 |  0.072 |  0.202 |  0.441 |  0.538 |
| 120% |  0.000 |  0.000 |  0.111 |  0.030 |  0.093 |  0.016 |  0.073 |  0.132 |  0.163 |
| 130% |  0.000 |  0.000 |  0.009 |  0.005 |  0.010 |  0.001 |  0.020 |  0.023 |  0.022 |
| ---- | ------ |  --- 6 | Attack | Moving | Avg.-- | ------ | ------ | ------ | ------ |
|  60% | 63.811 | 52.319 | 62.497 | 52.253 | 67.569 | 58.613 | 43.626 | 51.026 | 53.647 |
|  70% | 46.505 | 32.327 | 42.700 | 35.896 | 50.816 | 39.697 | 28.347 | 36.676 | 36.427 |
|  80% | 29.339 | 15.025 | 24.101 | 23.699 | 34.918 | 23.105 | 18.530 | 22.820 | 20.898 |
|  90% | 13.879 |  3.923 | 12.694 | 12.101 | 18.525 | 11.309 |  9.379 | 11.710 | 10.804 |
| 100% |  6.391 |  1.083 |  6.103 |  5.949 |  8.187 |  3.932 |  4.787 |  5.364 |  5.383 |
| 110% |  1.299 |  0.011 |  2.167 |  1.636 |  4.122 |  0.974 |  1.558 |  2.576 |  2.176 |
| 120% |  0.004 |  0.000 |  0.468 |  0.189 |  1.554 |  0.094 |  0.306 |  0.802 |  0.552 |
| 130% |  0.000 |  0.000 |  0.038 |  0.027 |  0.377 |  0.012 |  0.096 |  0.166 |  0.109 |
| 140% |  0.000 |  0.000 |  0.000 |  0.003 |  0.011 |  0.001 |  0.013 |  0.029 |  0.009 |
| 150% |  0.000 |  0.000 |  0.000 |  0.000 |  0.001 |  0.000 |  0.000 |  0.003 |  0.001 |
| ---- | ------ |  --- 7 | Attack | Moving | Avg.-- | ------ | ------ | ------ | ------ |
|  70% | 64.218 | 51.512 | 61.659 | 52.033 | 66.623 | 58.797 | 42.529 | 51.330 | 53.747 |
|  80% | 48.342 | 33.314 | 45.131 | 39.160 | 52.671 | 42.564 | 31.209 | 37.377 | 38.036 |
|  90% | 31.252 | 16.770 | 29.761 | 25.488 | 36.930 | 26.821 | 19.665 | 24.054 | 24.415 |
| 100% | 19.183 |  6.731 | 17.233 | 15.273 | 24.039 | 13.899 | 11.816 | 14.130 | 14.360 |
| 110% |  8.843 |  1.348 |  7.812 |  6.777 | 14.794 |  5.296 |  5.282 |  7.881 |  7.043 |
| 120% |  3.057 |  0.204 |  2.515 |  2.432 |  6.646 |  1.718 |  2.029 |  3.386 |  2.588 |
| 130% |  0.823 |  0.013 |  0.477 |  0.944 |  2.526 |  0.439 |  0.890 |  1.359 |  0.713 |
| 140% |  0.076 |  0.000 |  0.091 |  0.271 |  0.516 |  0.056 |  0.285 |  0.410 |  0.224 |
| 150% |  0.002 |  0.000 |  0.037 |  0.048 |  0.083 |  0.001 |  0.074 |  0.074 |  0.100 |
| 160% |  0.000 |  0.000 |  0.012 |  0.002 |  0.028 |  0.000 |  0.013 |  0.026 |  0.039 |
| 170% |  0.000 |  0.000 |  0.002 |  0.000 |  0.001 |  0.000 |  0.004 |  0.005 |  0.008 |
| 180% |  0.000 |  0.000 |  0.000 |  0.000 |  0.000 |  0.000 |  0.001 |  0.000 |  0.001 |

It doesn’t look like stamina is struggling at all in this data set.  It’s still the dominant anti-spike stat at 250k boss swing sizes, again decreasing spike representation by factors of 10 or more for the largest spikes.  And in this data set, we’re seeing some truly deadly spikes (i.e. over 100% health).  Stamina always provides a very significant drop in those categories, which are ones we care quite a lot about.

It’s worth noting that the avoidance sets are starting to make a much better showing in this regime.  They still suffer from the “long tail” – i.e. they still have non-zero representations in the highest spike categories and see the largest absolute spikes.  However they quite frequently beat C/Ha if we consider “what percentage of attacks exceed my total health?”  This is an important distinction that we couldn’t effectively make with the old boss-normalized data table structure in previous posts.

An avoidance set might give you a nonzero chance to let through a sequence exceeding 150% of your health, which seems bad.  However, if the avoidance set gives you a lower chance of taking a spike exceeding 100% of your health overall, the distribution of those attacks isn’t that important.  In the rough sense, you’re dead either way, so why does it matter?  That glosses over some obvious nuances – for example, once you consider healers you might survive a 120% spike more easily than a 150% spike, so there’s still value in shifting the distribution to lower levels.  But it’s something to consider if we’re going to critically analyze what each gear set is really accomplishing.

That said, even though the avoidance sets perform pretty well, there’s one set that always beats them in every category.  Yep, you guessed it: stamina.  Good ol’ stamina. Nothin’ beats stamina.

Let’s crank the boss attacks up once more and see what happens.

Data – 350k Swings

This is our 25-man heroic boss model.  350k is a pretty large chunk, enough that even 3-attack sequences become very dangerous.  This sort of boss necessitates smart use of active mitigation, and in theory that’s what Control/Haste excels at.  We could make the same argument about stamina becoming less efficient because each point is a smaller portion of a boss attack.  Let’s see if it’s enough to allow C/Ha to regain the lead:

Finisher = SH1, Boss Attack = 350k, data set smooth-10000-43

| Set: |   C/Ha |  C/StH |   C/Ma |   C/Av |  C/Bal |   C/HM |  Avoid | Av/Mas | Mas/Av |
| mean |  0.346 |  0.360 |  0.340 |  0.317 |  0.331 |  0.332 |  0.291 |  0.311 |  0.317 |
|  std |  0.124 |  0.124 |  0.121 |  0.136 |  0.128 |  0.118 |  0.141 |  0.139 |  0.130 |
|   S% |  0.523 |  0.482 |  0.410 |  0.419 |  0.452 |  0.472 |  0.366 |  0.363 |  0.357 |
|   HP |   755k |   876k |   755k |   755k |   755k |   755k |   755k |   755k |   755k |
|  nHP |  2.158 |  2.504 |  2.158 |  2.158 |  2.158 |  2.158 |  2.158 |  2.158 |  2.158 |
| ---- | ------ |  --- 2 | Attack | Moving | Avg.-- | ------ | ------ | ------ | ------ |
|  30% | 46.449 | 46.837 | 54.767 | 41.383 | 47.114 | 49.855 | 36.321 | 45.444 | 50.230 |
|  40% | 40.377 | 32.098 | 41.131 | 37.230 | 41.885 | 38.653 | 32.574 | 37.420 | 37.343 |
|  50% | 12.406 |  8.091 | 22.042 | 14.356 | 15.733 | 16.736 | 13.682 | 18.980 | 19.010 |
|  60% |  6.761 |  7.795 |  6.542 |  7.880 |  7.963 |  7.232 |  7.696 |  8.511 |  6.607 |
|  70% |  6.484 |  0.000 |  5.900 |  6.925 |  6.582 |  6.299 |  7.038 |  6.416 |  5.956 |
|  80% |  0.001 |  0.000 |  0.396 |  0.006 |  0.006 |  0.000 |  0.013 |  0.121 |  0.302 |
|  90% |  0.001 |  0.000 |  0.001 |  0.000 |  0.001 |  0.000 |  0.000 |  0.076 |  0.001 |
| ---- | ------ |  --- 3 | Attack | Moving | Avg.-- | ------ | ------ | ------ | ------ |
|  50% | 46.395 | 37.558 | 48.541 | 38.766 | 39.831 | 44.063 | 33.432 | 40.787 | 37.438 |
|  60% | 30.619 | 27.007 | 22.089 | 26.063 | 29.254 | 26.503 | 22.502 | 23.973 | 20.894 |
|  70% | 23.180 |  7.297 | 17.133 | 20.279 | 21.650 | 17.964 | 18.485 | 16.027 | 16.953 |
|  80% |  7.098 |  3.306 |  7.984 |  5.695 |  5.468 |  5.758 |  5.031 |  6.279 |  8.072 |
|  90% |  3.284 |  0.037 |  3.567 |  2.743 |  2.863 |  2.496 |  2.614 |  2.986 |  3.587 |
| 100% |  0.292 |  0.036 |  2.570 |  1.016 |  1.121 |  1.022 |  1.127 |  2.230 |  2.482 |
| 110% |  0.003 |  0.000 |  0.507 |  0.280 |  0.229 |  0.112 |  0.393 |  0.606 |  0.443 |
| ---- | ------ |  --- 4 | Attack | Moving | Avg.-- | ------ | ------ | ------ | ------ |
|  60% | 56.754 | 55.440 | 53.365 | 47.329 | 53.298 | 54.568 | 40.695 | 46.216 | 45.382 |
|  70% | 48.492 | 31.055 | 40.975 | 41.379 | 46.096 | 45.224 | 35.998 | 36.290 | 37.011 |
|  80% | 28.212 | 15.474 | 26.515 | 24.304 | 28.395 | 23.862 | 19.953 | 23.568 | 24.066 |
|  90% | 19.434 |  1.977 | 15.791 | 16.555 | 14.740 | 13.293 | 13.364 | 13.572 | 14.036 |
| 100% |  1.448 |  0.436 |  8.149 |  3.891 |  4.698 |  4.167 |  3.800 |  8.338 |  6.983 |
| 110% |  0.347 |  0.029 |  3.556 |  2.075 |  1.714 |  1.851 |  2.280 |  4.340 |  3.597 |
| 120% |  0.167 |  0.026 |  1.111 |  0.613 |  0.535 |  0.587 |  0.812 |  1.504 |  1.280 |
| 130% |  0.003 |  0.000 |  0.114 |  0.125 |  0.105 |  0.084 |  0.245 |  0.396 |  0.224 |
| 140% |  0.003 |  0.000 |  0.080 |  0.081 |  0.056 |  0.048 |  0.204 |  0.225 |  0.173 |
| 150% |  0.000 |  0.000 |  0.007 |  0.000 |  0.000 |  0.000 |  0.001 |  0.015 |  0.008 |
| 160% |  0.000 |  0.000 |  0.000 |  0.000 |  0.000 |  0.000 |  0.000 |  0.011 |  0.000 |
| ---- | ------ |  --- 5 | Attack | Moving | Avg.-- | ------ | ------ | ------ | ------ |
|  80% | 50.558 | 40.979 | 49.978 | 43.146 | 50.703 | 49.095 | 36.166 | 42.514 | 43.889 |
|  90% | 40.570 | 21.871 | 39.236 | 34.371 | 34.183 | 34.659 | 27.939 | 29.174 | 33.165 |
| 100% | 24.584 | 13.959 | 23.434 | 18.242 | 20.572 | 21.305 | 14.372 | 21.539 | 19.833 |
| 110% | 13.469 |  5.539 | 13.016 | 11.449 | 14.122 | 11.795 |  9.490 | 13.317 |  9.707 |
| 120% |  8.269 |  2.054 |  6.181 |  6.446 |  5.516 |  5.438 |  5.171 |  7.400 |  5.010 |
| 130% |  4.094 |  0.145 |  1.742 |  3.241 |  2.821 |  2.553 |  2.561 |  2.814 |  2.091 |
| 140% |  1.621 |  0.013 |  0.968 |  1.426 |  1.495 |  0.418 |  1.371 |  1.259 |  1.348 |
| 150% |  0.068 |  0.000 |  0.488 |  0.245 |  0.150 |  0.110 |  0.354 |  0.609 |  0.704 |
| 160% |  0.002 |  0.000 |  0.145 |  0.068 |  0.057 |  0.037 |  0.154 |  0.245 |  0.246 |
| 170% |  0.001 |  0.000 |  0.107 |  0.026 |  0.029 |  0.019 |  0.080 |  0.189 |  0.166 |
| 180% |  0.001 |  0.000 |  0.014 |  0.005 |  0.004 |  0.001 |  0.026 |  0.040 |  0.030 |
| ---- | ------ |  --- 6 | Attack | Moving | Avg.-- | ------ | ------ | ------ | ------ |
|  90% | 60.244 | 45.574 | 58.716 | 51.118 | 54.595 | 56.448 | 42.744 | 46.347 | 50.065 |
| 100% | 47.642 | 32.876 | 43.999 | 37.268 | 40.903 | 42.223 | 29.307 | 36.870 | 36.846 |
| 110% | 34.189 | 17.807 | 30.691 | 26.780 | 32.397 | 29.817 | 21.137 | 26.363 | 24.299 |
| 120% | 22.629 | 11.938 | 20.383 | 18.446 | 18.497 | 18.729 | 14.231 | 17.686 | 15.958 |
| 130% | 13.016 |  3.907 | 11.913 | 10.922 | 11.498 | 11.429 |  8.231 | 10.460 |  9.792 |
| 140% |  9.130 |  1.108 |  6.652 |  7.685 |  7.466 |  6.293 |  5.926 |  6.281 |  6.142 |
| 150% |  2.797 |  0.014 |  3.915 |  3.266 |  3.477 |  1.961 |  2.772 |  3.750 |  3.645 |
| 160% |  1.283 |  0.003 |  1.247 |  1.494 |  1.111 |  0.912 |  1.421 |  2.000 |  1.487 |
| 170% |  0.007 |  0.000 |  0.489 |  0.180 |  0.194 |  0.100 |  0.325 |  0.946 |  0.662 |
| 180% |  0.002 |  0.000 |  0.148 |  0.086 |  0.106 |  0.049 |  0.185 |  0.424 |  0.301 |
| 190% |  0.001 |  0.000 |  0.022 |  0.012 |  0.015 |  0.011 |  0.054 |  0.117 |  0.074 |
| 200% |  0.000 |  0.000 |  0.001 |  0.002 |  0.002 |  0.001 |  0.014 |  0.026 |  0.013 |
| ---- | ------ |  --- 7 | Attack | Moving | Avg.-- | ------ | ------ | ------ | ------ |
| 100% | 64.870 | 52.838 | 63.598 | 53.849 | 58.715 | 61.342 | 43.943 | 51.614 | 54.406 |
| 110% | 53.934 | 38.221 | 51.906 | 43.436 | 50.520 | 49.201 | 34.627 | 41.911 | 42.450 |
| 120% | 42.200 | 28.308 | 39.911 | 33.993 | 37.324 | 37.621 | 26.469 | 31.884 | 31.229 |
| 130% | 30.305 | 15.648 | 29.074 | 23.774 | 26.315 | 27.016 | 17.815 | 22.849 | 22.592 |
| 140% | 23.316 |  8.907 | 18.838 | 18.630 | 19.281 | 18.177 | 13.964 | 15.654 | 15.438 |
| 150% | 14.243 |  2.723 | 12.724 | 10.602 | 12.109 |  9.844 |  8.126 | 10.609 | 10.133 |
| 160% |  7.416 |  1.308 |  6.446 |  6.207 |  5.722 |  4.886 |  4.994 |  6.330 |  6.054 |
| 170% |  3.724 |  0.201 |  2.651 |  2.830 |  2.438 |  2.056 |  2.287 |  3.956 |  2.771 |
| 180% |  1.335 |  0.016 |  1.130 |  1.268 |  1.332 |  0.737 |  1.213 |  1.940 |  1.394 |
| 190% |  0.416 |  0.001 |  0.400 |  0.609 |  0.559 |  0.279 |  0.605 |  0.960 |  0.504 |
| 200% |  0.078 |  0.000 |  0.109 |  0.272 |  0.108 |  0.066 |  0.283 |  0.320 |  0.231 |

Well, that’s a resounding “nope.”  Stamina still dominates even in this regime.  In many cases, your chance of taking a spike exceeding your total health is reduced by 10% or more (and that’s in absolute terms – in relative terms it can be as large as 40% or more).  There’s just no doubt about it – stamina is the king smoothness stat for large boss hits too.

Conclusions

I don’t think there’s any question about these results.  They provide a fairly convincing argument that stamina is your best survivability stat, hands-down, no matter what level of content you’re raiding.  And, more to the point, prove that the “common sense approach” of stacking stamina until you can survive a few boss hits and then turning to other smoothness stats is misguided, at least for survivability.

Now, I want to make one thing abundantly clear.  This data does not mean that it’s wrong to stop stacking stamina and start stacking haste at some point.  In fact, many of the best tankadins in the world do exactly that.  And they aren’t misguided about anything, they know exactly what they’re doing.

Those tanks aren’t increasing their survivability by stacking haste.  They’re increasing their DPS by sacrificing some of their survivability (i.e. stamina).  Because unlike DPS, survivability isn’t a stat that you should stack blindly without end.  Stamina will always give you more survivability, but that added survivability isn’t always the best practical benefit for your raid.

Remember, the goal of WoW is not to max out the “Survivability” stat and e-peen around in Ironforge.  It’s to kill bosses, and survivability is a means to that end.  But if you have enough survivability to regularly make it through the boss encounter without faceplanting, then more survivability doesn’t necessarily help that much.  And given our affinity for haste and tank DPS mattering a lot more than it did in previous expansions, it’s often a good call to trade some survivability for haste to help meet enrage timers.

The difference here is that these tanks know that they’re making a survivability-for-DPS trade-off when doing this.  They don’t claim that they’re doing it for improved smoothness or better survivability.  They’re very up-front about the fact that it’s for increased DPS and better overall raid performance.  And they’re also skilled enough to make that decision in an intelligent fashion.

I think in part, that’s what’s led to the zombie-like recurrence of the “get enough stamina and then stack haste for survivability” mantra.  People look up the armory of a “good” tank they know and see them gearing for haste instead of stamina, and then assume that it’s being done for the purposes of survivability.  Then they log on to forums and repeat that opinion, or even berate others for having an affinity for stamina.

But in fact, stamina really is a higher priority for survivability than anything else you could gem or enchant for.  That’s why I suggest such a high stat weight value for it in AskMrRobot and recommended that Icy Veins add a comment about its worth beyond what’s found innately on gear.

Because a beginner tankadin does not play perfectly.  Hell, I don’t play remotely perfectly and I consider myself a pretty good tankadin.  A beginner does not have that experience, and can’t be expected to intuitively know what “enough” stamina is after a handful of boss pulls.  At best they’ll be able to clearly decide they don’t have “enough” if they faceplant on every pull.   And for both of these reasons, the beginner shouldn’t be trying to dance around the razor’s edge of having barely enough stamina under perfect conditions.  They are far better served by having a little bit of a buffer – some more wiggle room for when they or their healers make a mistake.

It’s the advanced tanks that should be shedding stamina for haste, because they’re the ones that have the experience and know-how to be able to intelligently assess their situation and determine whether they have “enough” stamina for an encounter.  They can then weigh the options and decide whether sacrificing some survivability for more DPS is a net benefit to their raid.  But I think it’s hopelessly optimistic to assume that the average AskMrRobot user has the experience to accurately make those judgment calls.  And if so, that’s when they can transition from a “novice tank” to an “intermediate tank” and learn how to adjust the custom weights to suit their needs.

While the simulation and data in this post are specific to tankadins, I think that the conclusions generalize across all of the tanking classes at least somewhat.  Other classes may have different damage smoothing mechanisms and mechanics, but the effect of stamina on damage smoothing is very much the same for a Paladin as it is for a Warrior or Death Knight.  It may be a little stronger or weaker depending on their exact stamina multiplier (paladins get +25% from Guarded by the Light, but not all classes do).

But the benefits we’re seeing here are an order of magnitude stronger than those of the secondary stats, so it’s not likely to change the results significantly enough to dethrone stamina.  While I can’t speak conclusively for any of the other classes, I highly encourage other theorycrafters to consider these results in the context of their own classes and see if they don’t come to the same conclusions I have.

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

94 Responses to Stamina: Keeping You Up Longer

  1. Weebey says:

    Wow. Admit it Theck: even you were a little surprised at HOW much better stamina is than secondary stats! We are, as you say, literally talking about an order of magnitude difference in the relevant categories.

    I am trying to think of what someone who disagrees might say in response. It is, of course, true that while all these spikes are a smaller percentage of your health, they are larger in absolute terms, hence at least in theory (although probably not in practice) require greater healer attention and mana usage.

    This just doesn’t work, though. There are a number of reasons why–many of which Theck has already discussed in the posts linked above–but perhaps the simplest is an ad hominem: if this argument were valid, it would apply, mutatis mutandis, to show that avoidance is a stronger survivability stat than haste. After all, the anti-spike benefits of haste over avoidance are much, much smaller than the anti-spike benefits of stamina compared to any other stat. And I don’t think I have seen a single critic actually advocate avoidance stacking.

    You could also make the boring argument that this is is still just a sim, that it doesn’t capture all the aspects of real tanking, etc. Even if one finds that critique convincing in general, we are still talking about an order of magnitude gap. That is so large, it is just hard to see how any further complication is going to come close to reversing the result.

    • Theck says:

      It is, however, worth noting that the margins will shrink some for gemming. Arielle pointed out on twitter that I traded 4k haste for 6k stamina, using the usual 1:1.5 ratio for itemization. On gems it’s actually 2:1.5 though, so that 4k haste is really 3k Stamina. I was thinking in terms of swapping trinkets, but we won’t be able to make such a sizable trade until T16 gear is available (right now we could probably do at best ~3k haste for 4.5k Stamina).

      So for gemming, the margins will drop by a factor of 2 or so. However, given that we’re looking at an order-of-magnitude gap, that still leaves Stamina far, far ahead. I’ll have an addendum post up later this week with gem-biased results.

      What I’m most curious about from these results, however, is not how they apply to paladins. It’s how well they apply to other classes. I have taken part in some debates about why AMR values stamina so highly for, say, Warriors, DKs, and Monks. Each time I presented the same hand-waving argument that I’ve been using up until this post, which quantifies that argument. And in most of those cases, the other commenters disagreed with me, suggesting that the smoothing mechanisms available to other classes simply dwarf Stamina’s value (Warriors/DK’s and mastery are ones I especially remember, as it’s definitely a survivability-to-survivability trade rather than a survivability-to-DPS trade).

      But it really seems to me like these results generalize fairly well. While it’s possible that the active mitigation or smoothing mechanics of another class are slightly stronger than a Paladin’s, I find it hard to believe they’re 5x-10x stronger. It seems to me that this post is going to make it harder for other classes to hand-wave away Stamina, which will hopefully spur some more critical analysis of their own class mechanics.

      (And, of course, once I get around to updating the Warrior simulations, we’ll have some quantitative results for them too).

  2. Astur says:

    It’s great to see a comparison between stamina and secondary stats finally, espescially as it’s so difficult to model. It’s also nice to see stamina doing so well, because as you stated in the conclusion – stamina offers nothing in the way of output to the raid (apart from possibly keeping you alive longer).
    Though how are you trading 4k haste rating for 6k stamina? – as gems would give a 4 haste to 3 stamina ratio, whereas you have a 2 haste to 3 stamina ratio which is closest to only switching trinkets from secondary stats to stamina.
    I personally would like to see a model where you assume that the tankadin is using stamina trinkets already and switching between haste gems to stamina gems or an intermediate value between the two, just to compare those gearing strategies.

    • Theck says:

      Yeah, I was thinking about trinket swaps initially. I’m running another quick sim as we speak to generate results for a gem-biased swap (i.e. 4k haste for 3k stamina). It won’t change the order of the results, it will just slightly narrow the huge gap stamina has over haste.

      • Wrathblood says:

        Not to, um, be difficult, but I think even shifting to a straight 4:3 ratio doesn’t quite completely cover it, because that assumes you’ll always match socket bonuses. With results like these, I wouldn’t be surprised at all if, assuming you’re gearing for pure survival, skipping socket bonuses and gearing pure Stamina is the way to go. However, to be sure, you’d need to go as far as 475:300.

        • anafielle says:

          Well, maybe I’m missing something here, but people geared haste are skipping most socket bonuses as well. Edit: So it would be a straight trade, skipped socket bonuses to skipped.

        • Theck says:

          4:3 assumes the socket bonus is irrelevant. Adding the socket bonus will depend on what color the socket is – for a blue socket, it means the cost going from haste/stam to haste is steeper than going from haste/stam to stam.

          Generally, single-socket bonuses are 60 stats, but we could imagine extreme cases where a single blue socket determines the result for a 120-haste or 180 stamina bonus. That would mean going from situations where you’re trading 120 stamina for 40 haste (obviously a bad trade) to 120 stamina for 280 haste (160 base + 120 socket).

          Either way, none of those trades are going to give you 10 haste for 1 stamina, which is about the level at which they’d be equaling out.

  3. anafielle says:

    I already gave this feedback, but: this is a great post. I think it’s one of the most important blog posts we’ve ever had here.

    It really hurts to see low level tanks malign stamina, complain that they feel wrong or uncomfortable stacking stamina, or brag about how control & haste are “the right” stats “according to Theck” because they blindly follow someone’s 2 sentence summary of some of the higher level posting on this blog. I see people wonder why Absalom or Trekkie are gemmed stam, and it just makes my head hurt. And then casual tanks sit there and call Ask Mr Robot incorrect for recommending stamina by default when you, yourself, gave them those default stat weights. It drives me nuts.

    Although the blog is your location for high level theorycrafting posts, I feel like so many tanks of varying levels of skill read this, that we have a responsibility to put all the information on the blog into context. You are great at doing that in the conclusions, but sometimes a bigger whack on the head is necessary. This is one of the most important posts you have made here, because – if nothing else – it will trickle down (I hope) to the greater prot community as “Theck says stamina is important too.” And hopefully it will do something to fight the pervasive (and completely incorrect) opinion that control-haste is just the right way to gear for all paladins specced protection. That is simply not true.

    We can’t expect every prot paladin that hears 7th hand “what Theck says” to understand every bit of the information here – that’s not our purpose – but I’m really glad you’ve made this point in its own post with, of course, the best & most memorable title. :)

    • Bere says:

      I for one am shocked at how the tankadin (and by extension the entire tank community) undervalues stamina. I have to include myself in that realm, as you could group me with the “stam haters” up until reading this blog. Very insightful Theck.

      Only thing I wonder now is the tradeoff regarding total boss damage in a stamina versus haste setup. You’ve already said that the haste build is there for the increase in DPS, which we can all agree on. What I’m wondering now is how much of an impact that would have, for example turning a 10 minute boss fight into a 9 minute boss fight. Would the damage you save yourself by finishing the encounter sooner be worth more or less than the extra stamina you’ve had during the fight?

      • Theck says:

        I think the “damage saved” metric is more or less worthless for the same reason that TDR metrics are worthless. Total damage taken in the fight is meaningless *unless* that extra damage causes your healer to run OOM and create a higher potential spike danger.

        However, you could make a reasonable argument based on probability. If you make the encounter 10% shorter, then that’s 10% fewer potential spikes you would be subject to in the first place. That sort of argument would bump haste’s value up quite a bit (and increase its lead over mastery and avoidance strategies).

        That said, I think that a 10% decrease in encounter duration isn’t realistic. Even in a 10-man, where we could assume you contribute 1/8 (12.5%) of the team’s damage, increasing your own DPS by 10% is still only a 1.25% increase to overall raid DPS, and thus only a 1.25% decrease in spike representation. Nothing to sneeze at, but not the sort of margins we’d need to give haste a real chance at catching stamina.

  4. Kihra says:

    As Theck stated, the reason you stack Haste is not for survivability. You do it for DPS. Especially this tier where tanks are often top DPS in 25H, stacking Haste on a paladin (who benefits so very greatly from it) is a big deal. If I moved away from Haste and back to Stamina, I can think of at least two of our Heroic kills that would have been a wipe instead of a kill as a result. It’s that big a deal to some guilds, and it’s a relatively easy adjustment to make to get a lot more raid DPS.

    Really it comes down to whether you feel like tank DPS is needed to meet the DPS checks of a fight. My experience so far in Temerity, where we’re pushing right up against the enrages on some of these fights when we first get to them, is that the additional tank DPS has been far more beneficial than the occasional death I might suffer as a result.

    This has also been a tier of predictable danger periods and telegraphed spikes, and that means a skilled tank can avoid being bursted during those periods with smart use of cooldowns and active mitigation. This has also been a tier where many fights on 25H can add a third tank and that gives you more cooldown coverage for known danger periods.

    • Theck says:

      Just to embarrass Kihra, he’s one of several tankadins I was thinking of in the conclusion (“In fact, many of the best tankadins in the world do exactly that”).

      • anafielle says:

        I still don’t know whether to call haste gemming “Kihra style gemming” or “the Slootbag build.” We’ll see…

        • Kerriodos says:

          Slootbag is pretty aggressive with his haste stacking, might have to advocate for him on this one. Sorry Kihra.

          • Zothor says:

            Also, all else being equal, “The Slootbag Build” is awesome for the same reasons the title of this post are awesome. I totally want to be able to say tell people I’m rocking the Slootbag build, and it’s objectively badass.

  5. Sunnier says:

    This is a great post, and I’d expect the results would be very similar for monk tanks. Brewmasters seem to have an extra helping of stamina hatred, and yet so many complain about dying to burst. It can be very difficult to encourage some stamina here or there when the bulk of the community can’t see anything beyond the red herring that is TDR. I’ve long wanted to replicate the math that often appears here for Brewmasters, and I love this stamina model so much I might actually be motivated to do it.

    • yellowfive says:

      Yes, monks have been the biggest stamina haters in general, to the point that we (the ask mr. Robot team) have had to really lower our stamina recommendation for the very practical reason: we literally could not keep up with the customer service required to back up our high-stamina weights for monks. It’s nice to see this very informative simulation, that I think will make all tank classes give stamina a second look.

      Monks have been the hardest for me to decide on stamina due to the stagger mechanic… and the fact that it is a guaranteed smoothing mechanic. For all other classes though, I would be surprised if stamina did not show similar results.

      The most interesting point to me in this article is that “good” paladins are trading stamina for dps. That is the true and meaningful motivation. Warriors (the class I play personally at the moment) can only trade for hit/exp to gain significant dps, so I would not be surprised if the warrior simulations show that crazy stamina stacking turns out to be best. Looking forward to it!

  6. Awyndel says:

    I haven’t played this game in a year. But this still makes my day. After all this time I still don’t get tired of seeing stamina win.

    Thank you so much Theck.

  7. Felade says:

    I wish we could see something like this for healers an the throughput/spirit debate. Next project, Theck? :P

    • Theck says:

      Don’t get your hopes up, I’m busy enough so as it is. I’d have to give up sleeping if I wanted to start work on healer models.

      • Zothor says:

        Particularly since the first thing you’d have to do is throw out any premise of rational decision making….

  8. Wrathblood says:

    Excellent post. Very interesting.

    • Wrathblood says:

      Ok, thought about it for a few minutes and I’ve got two slightly more substantive things to say. First, a question. Does Stam scale negatively with itself as you stack more? To get 10% more health you need to add X Stam. But to increase your health by another 10%, you need to add 1.1*X. Whereas something like Haste probably scales positively with itself, getting better as you add more. I don’t think it’ll be enough to affect anything, I’m just wondering if I’m seeing it properly.

      Second, it suddenly hit me how right you are about the value of Stam and how intuitively obvious it should have been. To start, two points. First, hypothetically, lets say an overall 10% reduction in damage taken by stacking more Haste would also reduce your spikes by 10%. Not exactly true, but probably vaguely in the right neighborhood. Next, since we don’t really care about TDR, lets assume that the concept of Damage Sponges as a bad thing died in a fire long ago.

      So, with these two ideas, it becomes very simple. How much Haste does it take to reduce your damage taken, and thus your spikes by 10%? I don’t exactly know, and it varies depending on different things, but I bet its a significant amount. 8k? More? Now, how much Stam does it take to increase your health by 10% and thus reduce your spikes by 10% (well, not exactly correct but close enough)? A lot less than 8k. Probably more like 3.5-4k. Thus, Stamina is ~2x as valuable as Haste per actual stat point for survivability.

      • Bram says:

        @Wrathblood: I am not sure I fully follow you. I add a replay to your posting with the hope Theck will be able to answer your question together with mine as it is connected. :)

        As my understanding goes, the paladin would be on the most safe side, as having the most smoothed intake, if he stacks primarily stamina.

        The question is what to do with the rest of his stats on the gear? Simulations show that Control strategy delivers the best results. The interesting part is that Haste, which is designed as purely damage stat, helps paladin to smooth in some degree his damage intake. But Haste itself is not perfect in it, as shows C/HM set.

        Theck and Khira then explicitly stated, that once they see they are “safe”, they start to trade some of survivability/smoothing towards more DPS, shifting their set either to C/HSta (Theck) or even more aggressively to C/Ha (Khira). But in both cases they consider Haste a damage stat and sacrifice the survivability.

        I do see where your question heads. But it feels to me like you want to compare apple with oranges.

        @Theck: Could you in your simulation introduce C/StaHM (34k Sta, 4750 Haste, 4750 Mastery), please?

        Do you still plan follow-up on the C/HM set in regard of the optimal Haste-to-Master balance?

        • Theck says:

          I do plan on following up on the haste/mastery question. I actually have the simulation written (and even generated results), but decided this post was more important. Also, I wanted to look into implementing Seal of Insight first, figuring that it would put the finishing touches on the sim (it’s the last significant self-provided mechanic we’re ignoring). It’s really a matter of finding the time to write the blog post, really – I’m always busy, and we’re nearing the end of a semester which always makes things more hectic (so. much. grading. /wrists)

          • Wrathblood says:

            Have you seen any of the Kurt Vonnegut assignment letters from when he was teaching? I recall one where he was teaching, I think, a writing survey class. He notes that since there are 80 students in the class, papers should be no longer than 20 pages or he might go blind or be forced to kill someone while grading them.

      • Theck says:

        Yes, stamina scales negatively – that’s what I mentioned in the post about the natural drop-off in effectiveness as you stack more of it. Haste doesn’t scale as positively as you’d think because it starts doing funky things at 50% haste, and the nonlinearity is pretty slow since we’re at relatively low values.

        Regarding the “how much haste does it take to reduce damage/spikes by 10%” – it’s not a simple calculation. 10% more haste isn’t the same as 10% fewer spikes, which is why we need the Monte-Carlo. On the other hand, 10% more stamina really does decrease all spikes by 10% compared to player health. The sim shows pretty clearly that it’s much easier to boost stamina by 10% than it is to reduce spikes by 10% through haste. I’d say the 2x estimate is even too low, because if it were that close they’d come out even in the gemming simulation (spoiler alert: they don’t, stam is still ahead by a factor of 5).

        • Wrathblood says:

          While I don’t want to disagree with you, I’m confounded by something I noticed while pondering data on just this point. I was thinking about how to figure out how much Haste would be needed to cut damage taken and/or spikes by 10% and I thought a decent starting place might be comparing the work you did last Tier on the different gear-sets with the stuff you’ve just done.

          Leaving out the impact of SS and using an SH1 rotation (because they obviously weren’t in the old one), I checked out the difference in spikes and damage taken between the old and new C/Ha builds. In a nutshell, the difference between the two in itemization is +8K haste, but minus 500 Dodge and Parry.

          Anyway, overall damage dropped from roughly 0.65 to 0.60 (which isn’t terribly far off that 10%, considering the loss of Avoidance stats), but the spikes just get destroyed. 70% spikes drop by 30-35%, 80% spikes drop by 60-65% and 90% spikes go from non-zero to zero. 8K haste is a lot of haste, but still.

          I’m sure this’ll be ugly, so I apologize in advance.
          4 swings
          Old New
          70% 45.01 32.17
          80% 12.95 3.35
          90% 1.46 0

          5 swings
          70% 43.13 28.51
          80% 18.83 8.84
          90% 1.03 0

          6 swings
          70% 36.76 24.11
          80% 13.97 4.24
          90% 0 0

          • Theck says:

            While a good chunk of that is the +8k haste, keep in mind that SS alone makes a pretty huge difference. I don’t think it’s fair to credit haste with the bulk of that change.

          • Wrathblood says:

            No, I used the data from the non-SS chart.

            However, its just occurred to me that frequency =/= size. These numbers are spike frequency. Increasing Stam 10% doesn’t reduce the number of spikes you take by 10%, it reduces the *size* of your spikes by 10%, essentially turning 90% spikes into 80% spikes and so forth.

            So, on 6 swings, its nice that another 8k haste shrinks 80% spikes from 13.97 to 4.24, but 10% more Stam essentially would reduce them by 10%, giving them the frequency of 90% (well, not quite but close enough) spikes, so roughly in the zero-ish range. 8k Haste lets you move your results “over” a column to the right, while perhaps half that much stamina allows you to more your results “down” a row. Stamina wins in a landslide.

  9. Kaarn says:

    Something I wonder now is if baseline tank survivability (due to class mechanics + appropriate ilvl on gear) is overpowered right now.
    Many tanks now neglect survivability stats like Stamina or TDR-stats like Parry and Dodge in favour of DPS stats and they are successful with it. Some may do it for the wrong reasons but they still don’t hurt their raids noticeably, else you would read more whines about tank squishiness I suppose.
    Is that good design that promotes different tanking styles, an oversight, or something that should be adjusted so that tank gearing follows traditional gearing schemes again?

    • anafielle says:

      It’s a very fine line. I wish that tanks were squishier outside AM, but I don’t know if that would be best for everyone. That is incredibly un fun for the people who play with the tanks who make errors; imagine LFR with tanks whose baseline survivability was much worse. I understand why AM simply can’t be so punishing at every level of tank, that using it wrong causes death — except obviously, on heroic 25, where requiring that level of play is acceptable.

      There is a level of punishment for error which is fun in normals and LFR and 5 mans and the vast majority of the game… Wiping the group because the tank didn’t hit their button in the right split second is over the line. Except, again, for the pieces of the game where we explicitly enjoy that challenge: heroic 25, challenge modes, etc.

      But as a result, yes, I think tank survivability is probably too high overall. If you look at it from the perspective of the level of tank which Theck speaks to and which Theck simulates. Baseline tank survivability has to be at a certain level to prevent the vast majority of people who play with a tank from hating on them… sure, it could be tuned down overall, and the high level players would be less likely to make the DPS-for-survivability trade, but would that be best for the entire game? I care about my part of the game, but I know my part is a small part of the whole.

      I think one of the problems with tank gearing is simply a matter of perception. Most tanks really should gear “traditional gearing schemes” – and they don’t – but they don’t make that like an informed decision. They just blindly do what everyone else is doing. That’s, I bet, one of the reasons why the devs have hesitated to change how tanks work. There’s probably not much wrong with how they work now, people just don’t in general understand the implications of their decision to gear a certain way.

      And I think the devs also care about our fun – if the devs broke haste gearing, they would upset a huge number of people who truly enjoy having the option available, who gear stam when they must and gear haste when they can. It’s important that we have fun too :)

  10. Huh says:

    I feel like the 1:1.5 stamina:stat ratio throws of this entire blog post. Wouldn’t a 2:1.5 ratio be much more relevant seeing as 99% of the question sta vs stat is about gems?

    • Theck says:

      See my response to Weeby’s comment. I was thinking about trading trinkets (which, admittedly, right now is at best a <~3k stat for <~4.5k stamina trade). I’ll have an addendum post up later this week with gem ratios (2:1.5). It doesn’t change the conclusions at all, it just narrows the gap from a factor of 10 to a factor of 5.

  11. SucellusAP says:

    Really interesting post, I just hope that instead of “Theck says don’t stack stam” people don’t now blindly say “Theck says just stack stam”!

    You have it just right that it’s a tradeoff that each individual tank needs to make for themselves. My guild is normal progression with some heroics when pretty well geared for them (as opposed to being in heroics in week 2 of a new raid) and generally speaking tank death is not an issue for us. 99% of our wipes are people failing to a mechanic or non-tanks getting themselves killed. So I don’t think I’ll be stacking stamina just yet as I feel the dps contribution (and minor TDR contribution) is worth more, but I’m glad to see these results as I’d feel much more comfortable moving to stamina now if we did start to have a problem.

    I love to see these results for the other classes, particularly guardian druids as I feel that right now they are very squishy when they are unlucky enough to get a damage spike which is extra noticeable because of their relatively low health.

    • anafielle says:

      I’d rather people erred on the side of “Theck says stack stam!” I would be ok with the majority blindly following that, if they have to blindly follow something.

      I realize that not everyone in the game enjoys knowing every single intricate detail of how their class works. Some people want to hear “Theck says _____!” and then do it. Those people need to be filling the blank in with stam, not haste.

      • SucellusAP says:

        Good point, if that’s the default then those who aren’t automatons can make the decision for themselves to trade stamina out for something else.

  12. Psykewne says:

    I have an additional question here,
    Is it reasonable to consider the effect that stamina stacking would have on our 2 piece tier 15 bonus usage.
    Previously you Theck analysed it and made the assumption that when overhealing does not occur it was very good however obviously that was not going to be common, but by raising our stamina levels that means we are more likely to benefit from the healing from word of glory (or even find use of it in general as healers take a little longer to top us off).
    Also with less haste we will see less shield of the righteous availability perhaps making us sit on it a little more for the predictable larger hits etc… on many bosses. In that case the 2 pieces gives us a much longer relatively reliable reduction on hits between using our main active mitigation on the stronger strikes.
    I could be very wrong in assuming that stamina would have any impact on our 2 piece effectiveness, but i thought it worth asking.
    I suppose in actual fact now that I think about it, does a shift from haste to stamina impact on our play style at all (aside from the obvious less holy power), does it actual affect our choices?, obviously with less haste, holy avenger and unbreakable spirit both lose some effectiveness.
    To finish, you opened my eyes in this post, for something that should have stood out as obvious you made a compelling read and a though provoking series of possibilities, It may have even resolved an issue in our raids.

    • Zothor says:

      My gut is that the problem with this isn’t going to be the 2-Piece itself, or gearing strategy: it’s going to be Thunderforged offset pieces. To start, only two pieces of our tier set can be considered well itemized: the Exp/Mastery Helm (which has an unfortunate 180 parry socket bonus), and the Hit/Mastery shoulders (same problem with the socket). The gloves are exp/avoidance and the legs are hit/avoidance so they’re not terrible (although both are heavily weighted towards the avoidance), and the chest is straight up awful given that (a) it’s double avoidance, and (B) the ret chest is perfectly itemized for control-haste tanking. For now, since you’re only talking 2 piece and not 4 piece (the 4 piece, in my opinion, is unjustifiable from a survivability standpoint unless you’re trading that survivability for the DPS boost of the extra HP), let’s focus on the Helm and the Shoulders, since they’re the two best pieces to use.

      While the tier helm itemization is excellent and the Puncture Proof Greathelm is poor, the Puncture-proof has a huge 270 stam socket bonus on it—and it’s on a blue socket. Even assuming that you don’t match the socket bonus on the tier helm (and gem it straight stam), the Thunderforged Puncture Proof is going to have a 404 stam advantage on the Tier helm (as well as an 87 armor advantage, for whatever that’s worth). That’s a no-joke difference. The Doomed Crown of Lei Shen is itemized effectively as well as the Tier helm, and the 528 Doomed Crown has a 134 stam and 87 armor advantage on the tier helm. Given the evidence in this post of how strong stamina is, leaving that stam on the table seems questionable for a buff that we’ve established (a) isn’t a sure thing, since block is still a chance, and (b) probably is only up after we’ve WoGged ourself out of the danger zone.

      The shoulders tell the same story: the poorly itemized Thunderforged tanking shoulders have an ~171 stam bonus because of their better socket, and an 80 armor bonus. The well itemized “ret” trash drop shoulders (admittedly rare as hell and hard to assume you’ll get) have a 101 stam bonus and the same 80 armor bonus.

      If our 2 piece were objectively valuable, we could argue that it was worth taking. But as of now, I’ve got my Tier helm on until I get a Thunderforged Punctureproof, but I’m doing it for the DPS boost—even the normal Puncture Proof has a buttload of extra stamina on it. My tier gloves and pants? They’re sitting in my bag behind thunderforged offset pieces (although i’ll probably put the gloves on for the 2 piece and the better itemization since my TF gloves are the awful double avoidance ones, again, for the DPS; if I had the dodge/exp or hit/haste gloves I’d absolutely be wearing them). If you want the two piece, maybe the way to roll isn’t with the best itemized pieces (shoulders/helm) given the strength of the alternatives on the helm; maybe you roll with the shoulders/gloves. The ret tier chest and ret-offset 528 legs are just way too good to seriously consider the 4-piece, and I personally think the offset 528 helms also fall into that category.

      But really, when I can have 528 gear with hit/haste/exp/mastery, you are going to have a very hard time convincing me the questionable 2 piece is worth wearing avoidance. The change to Grand Crusader just wasn’t that big a deal.

    • Theck says:

      Replying to both: More stamina may not have a very large an effect on the 2-piece bonus. Having more stamina does not necessarily mean that you’re regularly sitting at a lower relative health deficit – your healers are still going to top you off about as quickly as before, they just have more time to react during a big spike (and in those situations, may be able to top you off a little more slowly without risking your death).

      That still pretty much means that SotR is the default choice though – it just does a better job of smoothing than the 2-piece bonus does alone (assuming a wasted WoG). *During* a spike, the WoG is useful, and the 2-piece is a nice benefit which acts as a recovery mechanism. So I think optimal WoG/2pc usage doesn’t really change with high stamina.

      I share Zothor’s concern about thunderforged off-set pieces. I wish tier could be thunderforged via an item (or valor upgrade – I’d totally pay two 4-ilvl upgrades worth valor to bump a tier piece by 6 ilvls just to keep the tier bonus perks). The raw stamina advantage of thunderforged off-set pieces and more optimal stat configurations make them very attractive.

      That said, I like the 4-piece much more than Zothor does. In 25H raiding, with bosses hitting for 350k a clip, that 4-piece means you’ll be picking up ~1 Holy Power every boss attack (give or take, rough estimate based on SotR mitigation, ignoring avoidance). It’s probably going to be enough to make Divine Protection a mini-Holy Avenger, which is pretty strong. Especially when you consider the possibility of chaining them (DP first, then HA following it to get ~40 seconds of almost continuous SotR uptime).

      The problem is that I think the 4-piece is far less attractive for 25N or 10N/H, because damage intake is much smaller. If I were running 10-mans, I’d probably be skipping 4-piece entirely. Even in 25H, it may not end up materializing to the degree I’m predicting, so we’ll have to see.

      That said, I have a trick up my sleeve for that. When I updated the sim to include stamina, I also put in the 4-piece bonus. So it’s already in there, just waiting to be simmed. :) All I need to do is figure out an approximate trade value. For example: I can toggle on and off the 4-piece to see exactly how effective it is, but that’s not really what we want to know. What we want to know is how 4-piece compares to, say, 2 pieces of thunderforged off-set gear that are itemized well (more haste and stam and etc.).

      In fact, since Zothor seems to have been looking into this far more closely than I have, maybe he would offer to put together ideal 2-tier to 2-thunderforged-off-set and 4-tier to 4-TF-OS trades to nail down the amount of stats I need to be comparing? That will save me some time, which means the sim and blog post will happen faster!

      • Zothor says:

        For reference, I used the official Theckweights from AMR for Control/Haste, which, based on the latest numbers, might even under value stamina. But those weights are:
        Stamina 1.5; Physical Hit 1.1; Expertise 1.09; Haste 1; Armor 0.99; Mastery 0.8; Strength 0.6; Dodge 0.5; Parry 0.5

        Bear in mind that like Theck I raid 25 man content, so accessing all the Thunderforged pieces off normal bosses isn’t at all unrealistic. For a baseline progression wise though, I’ve stuck to non-heroic pieces, since once you have access to the Heroic Lei Shen Thunderforged loot, progression gearing is a bit of a moot point. I limited consideration to the two best non-prot-set pieces available, and only if there were two that were demonstrably better than the Tier piece.
        I’ve ignored enchants and metagems since they’ll be universal by slot; I’ve matched socket bonuses only when the weighted return was greater than the value of the stamina given up; I have used reforged versions of items where their stats weren’t optimal to begin with, but I have not reforged into hit/exp if there’s already hit or expertise on a piece. I just don’t think that’s realistic, since most of us are going to have enough (if not more than enough) hit from our gear in this set and be reforging out of it, and be stacking lots of expertise pieces to come close to where we need without reforging. But if you were allowed to reforge all your pieces into hit, sure, they’d look better on the stat weights. That’s not realistic. Many of the ret pieces are so good precisely because they’re hit/haste or expertise/haste.

        I have no idea if I can embed a table in here, so I’ve uploaded my sheet to Google Docs:
        https://docs.google.com/spreadsheet/ccc?key=0Au3QRwGu1OsPdG1Fd1llbkt5Wl90MDFJbFMyLW8yV2c#gid=0

        The conclusions:
        The Tier shoulders are really good. They are all but best-in-slot, even considering thunderforged, because of the rarity of the only shoulders that are demonstrably better. The better shoulders are barely better overall, and are a Thunderforged trash/random drop. That makes them significantly harder to come by that a Conq token. Just use the tier shoulders and enjoy their sparkly lightningness.

        The rest of the slots are, unfortunately, basically a wash. I had expected the Tier chest to grossly underperform relative to slots that had at least one non-avoidance stat (it didn’t), and the tier helm to perform more favorably given its solid itemization (it’s getting tanked by its crappy socket). I had even figured gloves would be a slot to swap simply because of the lower budget on the slot, but the Pathogenic Gauntlets are really good, even in the non Thunderforged variety.

        That leads to what feels to me to be a counterintuitive conclusion: if you can get them, the Pathogenic Thunderforged gauntlets are your off piece if you want the 4-piece.
        Swapping from ideal Thunderforged offset to 4-piece tier in the other slots costs:

        579 Stam, 261 Armor, (1115) Hit, 1089 Expertise, 1212 Haste, (637) Mastery, 686 Strength, (114) Dodge, (923) Parry. The tradeoff is a weighted itemization loss (based on the AMR weights listed above) of 2141.54 Theckpoints.

        If you’re only taking the 2 piece, the only piece you should absolutely use are the shoulders; the only “better” shoulders are better by 48.18 Theckpoints, an ultimately meaningless difference across an entire gearset (effectively 32ish stamina). Since we know you should never use the T15 gloves compared to the Thunderforged Pathogenics, that leaves the Helm, Legs, and Chest to swap to complete the 2 set. I recommend the legs.

        The tradeoffs, vs the best Thunderforged alternatives, are: Helm 714.88 TP, Chest 754.06 TP, Legs 624.42 TP. That alone says swap the legs because you’re losing less weighted value, but I will also add that the better options for the helm and chest are demonstrably easier to get.
        The best helm option isn’t even the hit/haste piece you’ll be competing with DPS for; it’s the Thunderforged Puncture-Proof Greathelm, which just has a metric butt-load of Stamina on it that pushes it ahead despite its avoidance itemization. Only another plate tank is going to even want this. When it eventually drops, it’s probably yours. As for the best chest option? It’s the Ret Tier 15 chest—a 100% “drop” once you have the tier token, of which you get 3 per kill on 25 man. Eventually, statistically, you will get the Ret T15 chest with much greater certainty than you will get the Thunderforged Ret/DK legs that they are going to fight you for.

        So, if you’re wearing the T15 Shoulders and legs, you are trading away:
        55 Stam, 174 Armor, (1140) hit, 724 Expertise, 1022 Haste, 26 Mastery, 416 Strength, 0 Dodge, (782) Parry, for a total weighted tradeoff of 672.6 Theckpoints.

        • Wrathblood says:

          I was just looking at that same comparison yesterday and came to the same conclusion. The Tier chest’s hidden saving grace, and the doom of the seemingly better itemized tier gloves is gem sockets. The chest has terrible itemization but an extra socket vs its best competitor, so while its more overall itemization points to use badly, gem sockets are currently quite OP in terms of value and the Pathogenic gloves are not only better itemized than the tier gloves, they also have an extra socket.

          • Zothor says:

            I don’t know if he was reading these comments (he’s a former tank playing Ret right now) but when I logged on yesterday after work, my raid leader handed me the Thunderforged Pathogenics he’d picked up the night before, because he still had time to trade them. Suffice it to say… I was pretty stoked.

  13. Zothor says:

    This post is making my consider Colossus again, which is almost-stamina, whereas I’ve been rocking Dancing Steel because it feels like so much more pure itemization value. I am not happy about this potential trade back. Ugh.

    • Theck says:

      If you’re already trading survivability for DPS, then Dancing Steel isn’t a bad choice. Colossus is “almost-stamina,” but it’s not a lot. 8k absorb bubbles if I recall correctly, so about 396 stamina when it’s active, which is only for one hit. 6 RPPM, so ~6 hits every minute out of a potential 60/1.5=40 attacks, or 15% of the attacks you take. So it’s about 15% of 396 stamina, or about 59 stam.

      Compare to Dancing Steel: 1650 STR, 2.3 PPM, 12-sec duration, so about 27.6 seconds uptime per minute, or 46%. 46% of 1650 is 759 STR.

      So you’re trading about 59 stamina for 759 STR, or over 10x as much. Colossus is a bubble, so it adapts to damage intake (much like Sacred Shield), but it’s small enough that it’s not a huge effect (unlike SS). The disparity may even be enough to make DS comparable to Colossus in term of smoothing. Probably not quite, since avoidance lags in that department, but it’s probably close enough that you shouldn’t feel bad for using it.

      Hell, after doing that math, I’m tempted to re-enchant too. That’s a hefty DPS trade for very little survivability cost. I didn’t consider haste, which should scale both of them multiplicatively, further increasing the raw stat advantage of DS.

  14. Lakh says:

    Very interesting stuff. An old topic covered in a great new way.

    Would very much like to see you touch base & update the other side of this again – maximising our DPS. For example, Vengeance abuse and stam gear vs haste gear and perfect rotation. It just seems to me like this is an authoritative end of the story (for now) on how-to-survive, and there’s not a lot else to say.

    • blizzhoof says:

      Stamina doesn’t affect Vengeance in MoP. You’ll have the same Vengeance with 500k hp or 800k hp. Also, it’s calculated pre-mitigation (and pre-avoidance I do believe) so you get the same with GoAK up or down, with or without Sacred Shield. So it’s pretty intuitive that Haste beats nothing (stamina offers no dps benefit), but I do wonder if crit overtakes haste at certain haste levels.

      • Theck says:

        Crit won’t overtake haste until you start getting into the >50% haste region, where funky things start happening to the rotation (see “Haste Breakpoints for Tankadins” blog post from November).

        The maintankadin MATLAB DPS thread covers most of the DPS questions you could possibly ask:
        http://maintankadin.failsafedesign.com/forum/viewtopic.php?p=718673#p718673

        Crit is strictly inferior to haste for us. It starts out weaker, and haste scales faster than crit because of mechanical synergy (more haste -> more CS/HotR -> More GC procs; no such synergy exists for us with crit).

      • Lakh says:

        If I run 800k+ hp and a survival build I can deliberately take more incoming damage more safely, whereas deliberately fishing for higher incoming damage in a haste build feels significantly more chancy. Especially when the sort of things you can deliberately stand in tend to be magic-based raid damage (e.g. Crashing Thunder, the melee range effect, on Lei Shen).

        That’s the approach I’m thinking of with stam+vengeance vs haste.

        • Huh says:

          I noticed that the ground effect that crashing thunder leaves on the ground do not grant vengeance.

          • Zothor says:

            This appears to be woefully inconsistent; like some of you, I’ve tried (particularly while offtanking during Bloodlust) to use ground effective to get or maintain SOME vengeance. From what I can tell, things that DPS are supposed to avoid do not grant vengeance; things that tanks MIGHT be subject to do. I’ve been taking the first double swipe on Horridon on purpose for the 40k or so Vengeance it grants and it’s made my add pickup on the first door significantly easier. But stacking bleeds and whatnot don’t seem to generally grant vengeance because you’re not supposed to gain an advantage by letting them stack up (meanwhile, on the pre-Horridon snakes yesterday, with 40+ stacks of poison I had 300k+ vengeance).

            There’s got to be a flag in the particular damage type. Like all other things vengeancey, the frustration is in the inconsistency.

  15. Zao says:

    I could be wrong, but have you ever attempted to factor in the dual gain of Battle Insight+Seal of insight with higher haste levels? Your model makes stamina seem like the “god” stat when these two BIG benefits aren’t included in the simulation. Pulling up a random log from last night I have the following data over the course of a 7 minute fight:
    Seal of insight 5.66M(13,476HPS)
    Battle Insight 4.07M (Around 9,690HPS)

    Now people weren’t very stacked up for this fight, so Insight would have gotten far more heals off(at times there aren’t people who are not topped off that are in range of me), but you can already see that these two numbers are huge instant heals that occur at a higher frequency with more haste.

    Our healers have the following numbers over the duration of the fight.
    Healer 1: 44,454HPS
    Paladin Tank: 42,240HPS
    Healer 2: 39,249HPS

    The argument would basically be that the trade off is a marginally lower health pool(while still being high enough to survive a reasonable amount of boss swings) allows a paladin tank to free up more overall raid healing over the duration of the fight, ie, less people that healers need to waste mana on if you were ever to try and create a simulation that incorporated that element.

    • Theck says:

      Seal of Insight is something I want to add to the sim, and with any luck will have time to do so in the coming weeks. That said, it is a small enough effect that it won’t change the results. Remember that you get SoI procs without haste – adding 10% haste just gives you 10% more of them. But they aren’t large enough to account for an order of magnitude difference in survivability.

      Battle Insight only heals others, so from the perspective of these simulations it’s irrelevant. That said, it’s certainly a non-trivial amount of raid healing. But I don’t think there’s any good way to sim that effect, because there are so many variables involved – e.g. how many people are in melee range? how many of them are not topped off? Do they have HoT’s ticking that will top them off anyway? The results would vary wildly from one raid to another based on composition (ex: less effective with 2 tankadins, less effective with 1+ druid healers, less effective for a caster-heavy composition, more effective for a melee-heavy DPS composition).

      I guess the best we can say about Battle insight is that yes, you can trade your own survivability for more raid healing by converting stamina to haste. How effective that trade will be depends a long on your own raid situation, and how much you and your healers value incidental healing. I think in practice your own DPS boost from that haste is probably the more important factor to consider.

  16. Zothor says:

    The only thing I feel bad about is when I have to double-salv myself just to not pull aggro back on a tank swap…

  17. Weebey says:

    Perhaps unsurprisingly, early fallout is not encouraging: http://us.battle.net/wow/en/forum/topic/7591979125?page=17

    Not a single person even considers the possibility that they should rethink their assumptions; the whole thrust of the discussion, insofar as anyone tries to address the points raised here at all, is to hand wave vigorously (“we have so many anti-burst tools” “stamina is useless once you can’t be three shot” “our mastery provides EH” and so on ad nauseum) and insult the poster who tries to argue that some of the results may transfer. No one even seems to appreciate that for the results here to not apply to Monks their AM stats must be literally 5-10 times more effective at preventing spikes than haste is for Paladins–perhaps true, but demonstrating it is a tall order indeed.

    Granted, this is the official forums, but still sad.

    • Raed says:

      So I use the control/mastery build. Looking at regemming to get more stam is 60k more HP going to out weight 7% less mastery? I was at 31% fully buffed and now I’ll be at 24%. Is this an effective trade off?

    • Theck says:

      I had a good laugh at “Good job, you found a statement by Theck that says EH helps in Time To Live.” Given that that’s exactly the opposite of what the data and blog post are discussing (I don’t think I even mentioned TTL).

  18. Kort says:

    I spent much of T14 (10 man) worrying about how low my health was compared to my co-tank. I had many discussions with my healers, read a lot of sites, tried different gearing strategies now and then etc.

    In the end, my healers & I concluded that I was not “hard” to heal, I was not dropping dead randomly, and our wipes were very rarely because of a tank dying. I made the decision to just have the base stamina on my gear, pouring everything else into hit/exp/haste. Again, this is 10 man.

    Now we’re ploughing through T15, and I haven’t heard any complaints. I’m always chatting to my healers, making sure they are happy with my tanking and my ability to survive bosses. I think the one time I dropped dead this tier was a Horridon attempt during his enrage, where I didn’t have ShotR up (or any cooldown!) for an unmitigated duo of Triple Puncture/melee swing within a few milliseconds. If I had stacked some stamina, I probably would have survived that, but I would have also survived it if I didn’t have a bit of an insane moment in not pressing my ShotR button.

    In the end, I do agree with you. I do purposefully gear haste for more DPS (and more healing) with little thought for my survivability. I will therefore say the taboo phrase of “I have enough stamina”, because that is what my healers & I have decided. If, someday after a wipe, I sit back and think “damn, if only I had survived that hit” (not like the Horridon kill, because as I said, that was a 100% mistiming of ShotR) then I may consider more stamina to increase my survivability. Until then, DPS ftw!

    As you say in your post, I think tanks that understand why they gear how they do should decide for themselves. Hopefully I can justify why I gear the way I do, and with good reasons. I don’t want to be a zombie paladin!

  19. bryjoered says:

    I got in a huge verbal battle with some arrogant paladin tank in LFR last night. “haha what a newb for stam stacking” God I wish he would read this post. I’ve even done personal tests by swapping out trinkets to see which helps more, the stam trinkets do of course. I’m a warrior tank, can’t wait to see your warrior sims! I just came on here because I remember a post you had a couple weeks back speaking of how these people act like elitists snobs and that they know better than you when they haven’t backed it up with anything, especially math. They really should L2R.

    • SucellusAP says:

      Happens the other way around as well, I have been kicked from LFR twice (once after 2 clean kills and once before we’d even pulled the first trash) for being “in DPS gear” because I have haste on my gear as a paladin tank. :(

      • bryjoered says:

        yeah, he said “all the paladins that are 12/13 heroic haste stack” I said yeah after stamina. Look at Blood Legions MT he has 749k UNBUFFED stamina, and stamina on every single gem.

  20. tailias says:

    As the guy who brought your name up in that Monk thread on the official forums (in retrospect, I definitely should not have replied to that guy at all, but bleh), I do feel obligated to weigh in on what bugs me about this post and your overall views on Stamina gemming.

    I do agree with the idea that people who draw a gigantic line between (Damage Reduction) and (Health) are extremely wrong. They both serve to further our survival, that is totally correct. The issue is that you can’t simply say that the two are one and the same either – if we took every point of damage from every hit but had an unlimited health pool, and instead healers had to heal every point of it back up, would that be ideal? We could never die! (Of course that’s a ridiculous analogy and nobody would agree, just sayin’.)

    One thing that’s become clear as the expansion goes on, ESPECIALLY to a Holy Paladin-turned-Prot Paladin like myself, is that ‘tank healing’ is more of a part-time job with the new AM models. If a tank is constantly getting trucked for massive amounts of their health pool, it’s either a gimmick fight like Heroic Empress (bring 8 healers!) or Lei Shi (the one fight where all your physical damage reduction buttons are entirely worthless against the boss) or there is something substantially wrong.

    And if you’ll recall from past expansions, one of the primary reasons that Stamina was so strong was because we basically just had healers sitting on the tanks no matter what. We might as well stack more Stamina to be able to survive the spikes we DO take, rather than just stack more… parry… and make them bored. Right? Am I bringing back Festergut memories?

    That isn’t really the case anymore, at least not that I’ve seen – not doing normals or heroics last tier as a healer, and not tanking the normal content or the first one (and working on the second) heroic(s) in ToT. It’s not even necessarily the ‘mana sponge’ concept nowadays, not with 520+ ilvls and legendary meta procs with 10% uptimes for free spells and such (might as well be magical mana unicorns!). It’s time! That healing could go to someone else – that smartheal could go to someone else. Other people take damage, so mitigating a lot of damage yourself while losing very little if any potential survivability is a huge bonus in my eyes.

    So, I am not arguing that we should be caring about TDR (lord no), nor that we should be ignoring Stamina. I’m saying that if you’re given a choice where you can stack an extremely potent “all-around” stat that gives you reliable damage reduction (via both ShoR uptime AND Sacred Shield procs) the way Haste does, it tends to be worth using over Stamina. At least when looking at gem budgets and such. If we were still hampered with a RNG defensive stat option like Parry… well, I’d be gemming Stam too.

    AND all that being said

    I have two stamina trinkets in my bags. One of them (the 2/2 upgraded Shado-Pan VP trinket) has such a damn good on-use with Holy Avenger that I’m basically waiting for any slight whiff of an opportunity to need to use it. I also will never hesitate to regem if I feel that I need to, in fact, regem for ‘survival’ rather than the all-around damage reduction+survival that Haste affords.

    • tailias says:

      There’s no edit button, so I have to add this in another reply:

      Not related to my previous post, but your Sacred Shield WeakAura is basically the most amazing thing ever. Even if you overvalue Stamina (or more appropriately, undervalue Haste), you earned a spot on my bookmarks for that a few weeks ago.

    • Weebey says:

      Theck is obviously more than capable of defending himself, but I must say, that is some mighty thin gruel being served. You are really going to go all-in on the healer attention argument? Really? There is an effect there, I suppose, but it is very, very weak–while hard to formally quantify, I would guess around two orders of magnitude less important than the effects modeled in this post. There are a lot of effects not included in this post that one would obviously want to include in a fully specified model–a detailed healer reaction function being perhaps the most important–and the diversion of raid healing is going to be pretty far down the list.

      That being said, you do make an interesting point, but I think it cuts against your position. It is true that, even in 25H raiding, tank healing is somewhat of a part-time job. The most cursory perusal of logs reveals that the overwhelming majority of the heals a Paladin tank receives–typically in excess of 95%–are either self-healing, hots, aoe, smart heals, or cheap, cooldown limited mana-efficient heals (rapture boosted PW:S, Holy Shock, etc). It’s only in the last 5% that you find your Greater Heals and Divine Lights. Which isn’t of course to say that those heals are unimportant–without them the frequency of tank death would increase dramatically, as (assuming competent healers) they will be overwhelmingly concentrated in the few moments during the fight when the tank is in danger of actually dying.

      What it does imply is that the level of through-put damage in the tank is, by itself, even less relevant than people tend to assume. The average HPS from coming from background, extremely efficient healing is more than sufficient to keep up with the average damage taken by the tank; if DTPS were constant and at the same level (or even substantially higher), you could tank forever without receiving a single direct, expensive heal. The only thing–literally the only thing–that matters for tank survivability are those rare but definitely existent occasions when DTPS spikes far above its mean, and background healing is suddenly insufficient. And I think we all know the best way to deal with those.

      • Theck says:

        I think I echoed much of what you said (I started writing the reply this morning, but ran out of time before I had to go teach my engineering math methods class), but didn’t think to include a detailed log analysis of healing sources. Thanks for bringing those numbers up, I think they’re a strong piece of evidence in favor of spike analysis.

      • tailias says:

        I do appreciate Weebey phrasing it as “defending Theck” rather than merely replying to me, haha. Obviously I’m on the offensive against Theck, not merely arguing against his ideas because I think he’s incorrect ^^ THIS IS PERSONAL RAWR

        As far as this:

        “didn’t think to include a detailed log analysis of healing sources. Thanks for bringing those numbers up, I think they’re a strong piece of evidence in favor of spike analysis.”

        That would be quite interesting, actually. I’d be interested in seeing an analysis of healer throughput from a lifelong tank perspective.

    • Theck says:

      I disagree with your assertion about healers freeing up uptime. Keep in mind that we don’t have the option to transfer all of our stamina to haste or vice versa – a lot of it is just what comes on gear. So the amount we can shift is limited, which is why I only shifted 4000 haste around in the simulation.

      Your argument has a hidden assumption, namely that by shifting to stamina we’re getting hit like a truck all of the sudden and thus require a lot more healer attention. But that’s simply not the case. Look at the mean damage taken for the two sets: it’s ~0.345 for C/Ha and ~0.36 for C/StH. That’s a measly increase of about 4%. I don’t honestly believe that ANY healer is going to notice a difference of 4% time-averaged damage taken per second, nor that it will cause them to heal differently. In fact, I’d argue that they won’t notice it *at all* because it will just soak up some of the 30% overheal they and the other healers are *already doing*. Hell, Seal of Insight’s overheal alone probably absorbs that small of a difference.

      What does cause them to heal differently is exactly what you described – getting hit like a truck over and over. But you’re errantly equating that to overall damage taken per second, when really what you’re describing is spike damage. If trading 4k haste for 3k-6k stamina caused you to get hit by a truck and raised your DTPS by 50%, then it would be a real concern, and I might agree with your hypothesis. But it doesn’t have that large of an effect, and thus isn’t a concern at all.

      Now consider that by taking 4% more damage, which in my view doesn’t change your healer’s behaviors or reactions at all because of overhealing, cross-healing, and self-healing, you reduce the chance of taking a really big spike by a factor of 10 or more. You trade something your healers literally do not notice and do not care about to help eliminate something they absolutely DO notice and DO care about, because the spikes are the events that scare them. That’s when they go into emergency panic mode and tunnel their big life-savers onto you, taking their attention away from everyone else.

      So in fact, I think you could argue the exact opposite – that stamina actually *frees up* GCDs for them to spend on the rest of the raid, because you reduce the number of events that cause them to ignore the rest of the raid and “tunnel-vision” on you.

      In fact, as another counterpoint – your hypothesis is that when you don’t take damage (or take less damage), the healers can spend time healing someone else. That argument leads directly to TDR – that by reducing damage taken, you’re easier to heal and less of a strain on your healers. However, haste is a TERRIBLE stat for total damage reduction – it’s literally our worst secondary stat for that. We stack it because of its smoothness contribution and its DPS. The damage reduction it gives is incredibly small (1k rating for ~1% relative reduction in damage), and really not worth considering. If that hypothesis were correct, we should all be gearing for avoidance and taking crazy, spiky damage but freeing up many more GCDs for our healers (ex: when we dodge two attacks in a row).

      In practice though, we don’t see that. In highly-progressed guilds, tanks are stacking haste *to increase DPS* on progression bosses. That means that they’re already comfortable with their survivability, and clearly their healers are able to keep up with it without significantly impacting their ability to crossheal. My experiences in my own guild (also 1/12 25H, probably 2/12 tonight) and the other heroic-progressing tanks I talk to regularly match up with that as well.

      I will note that it may vary significantly based on your guild’s healer core. I’ve seen guilds that use cross-healing extensively and to great effect – basically any spare GCD gets spent throwing a HoT on someone who isn’t your responsibility (usually the tank), which leads to loads of incidental healing to pad against damage spikes. I’ve also seen guilds that are terrible at cross-healing, and use one healer “tunneling” on the tank and ignoring everyone else. Both versions work, though I think the latter is far less effective in general. However, your hypothesis may apply very well to the second type of guild, and maybe that’s how your healer core is used to operating. In the first type of guild, though, that extra damage taken just gets soaked up by the “donated” HoTs from the raid healers.

      • tailias says:

        You do make very good points, but again you’re confusing my argument of “we should care about reducing our damage taken” with “Guys let’s gear around TDR!”. I think smoothing damage is what I want to say.

        Our guild definitely has the former kind of healer core, we do assign tank healers but like I said before it’s a part time job – between all the HoTs, smart heals, absorbs, bouncing Cascades, ground effects, etc. there is enough cross-healing and such to do much of the work already. It’s not that your healer ignores you entirely – it’s that they don’t have to sit on you as frequently and can help with other things.

        You argued that healers wouldn’t notice shifting 4000 Haste around, and speaking as somebody who actually went up 5.5k Haste in one night early in the tier (I was following your “stack stam in ALL THE SLOTS” advice and then realized it was not working) that is absolutely not remotely true. Your sim says that the Haste value would be a trivial difference that impacts next to nothing – my experience says otherwise. Given the limitations of simming out tank damage, I’m tempted to go with my own eyes and ears.

        • Zothor says:

          Without passing judgment on your personal play, this kind of post exposes the fundamental flaws with the “eyes-and-ears” argument compared to objective analysis. It’s not unique to WoW; this was the argument for decades against advanced metrics and objective analysis in Baseball, and even a casual sports fan knows who lost that one (spoiler alert: the eyes and ears lie). There are just too many uncontrolled variables in your eyes and ears evaluation for anyone to truly and clearly respond to an argument founded on “well that’s what I felt.”

          Like, for example, I’m sure you know by now that the heavy haste build requires MUCH more frequent SoTR use. Consequently, it’s also far more sensitive to any mistake in your rotation, such as holy power generator pushback. There’s no way for anyone to respond to your “it worked better for me, so there” argument without, at the very least, logs of the “one night early in the tier” (was it this tier? last tier?) you’re referencing. Your healer comp certainly matters as well; are you carrying an extra healer than normal (3 v. 2 in 10 man? 6-7 v. 5 in 25?)? We don’t know. Not to mention, as always, the single biggest caveat that gets forgotten in this kind of discussion: the content you’re raiding makes a TREMENDOUS difference to tank damage mechanics. 10 Man normal vs. 25 Man Heroic use the same number of tanks for an enormous difference in incoming damage.

          The entire purpose of Theck’s model is to give an objective baseline to evaluate around, which takes away the noise that confuses your eyes and ears. It’s how it became clear that haste was so good in the first place early in Mists, something you don’t seem at all unhappy about.

          But Math? Math is what it is. It’s clean, it’s dry, and unless you’re challenging the actual mechanical assumptions underlying the model (which would be a good thing, since that’s how models improve), eyes and ears rarely add much to objective analysis.

          • tailias says:

            Well, I should’ve been more clear – I was also referencing the fact that he was quoting damage reduction numbers from a sim that is specifically designed to measure spike damage and ignores Sacred Shield (which he proved in a previous post has a tremendous reliable damage reduction aspect, and scaled extremely favorably with Haste). So the numbers he quoted are irrelevant and he knows it.

        • Zothor says:

          I think you should probably hit ctrl-F for “Sacred Shield,” since he specifically says the model DOES include Sacred Shield. More importantly, even if it didn’t include SS the numbers in this model would be far from irrelevant.

          Look back at the post you’re referencing, where the immense importance of Sacred Shield is demonstrated. Compare the SH1 queue, for example, with and without Sacred Shield. Yes, the primary conclusion of that comparison is that SS makes Haste freaking awesome (without SS, Mastery was clearly the dominant strategy). But you’re skipping the secondary conclusion: SS made EVERY single one of those gearing strategies more awesome than the BEST non-SS strategy alone. It’s impact is massive, and almost gear irrelevant. With Sacred Shield on, not ONE set saw more than one spike per sample in the 90 or 80% range, until you got down to the (questionably dangerous) 2 attack range.

          So your point that Sacred Shield massively improves the value of haste is well taken—but it’s poorly taken in the vacuum where you want us to assume that Sacred Shield doesn’t also greatly increase the value of Stamina. What the SS post showed us it that SS alone basically shifts us up an entire row or two on the chart; what were 90% spikes became 80-70% spikes, and 70% spikes became 60-50% spikes, etc. That’s before your gearing, and almost gear irrelevant. Haste’s effect on it matters and is strong, but not by an order of magnitude. Sacred Shield massively increased the smoothness of Control Haste, Avoidance, and balance sets, and their non-control variants. It improves everything, yet you want to assume that somehow the Stamina build won’t see the same result?

          • tailias says:

            Hmm. My apologies, I missed the part about SS being included in this model. (derp)

            However, I still am going to be filtering the conclusions provided by simulations through my own experience and what I’ve done/seen. Maybe I’m playing wrong, maybe my raid is playing wrong, maybe we’re doing something different. Guess what – that’s all totally relevant. Or maybe the sim is wrong. I honestly don’t know, and frankly I don’t have the math acumen to break it all down and tell you.

            What I can flat out tell you is that I tanked the first 11 bosses of Throne of Thunder 25 with a heavy Stam-oriented gear setup, and then after a night of Lei Shen I bit the bullet and regemmed/oriented for Haste. I’m sure there were other circumstances (I got a tier helm and a new… uh, ring? that night), but going up 5.5 Haste reduced my damage taken in a huge way. Not in a “TDR” way but in a smooth, reliable way that impacts the way my healers looked at me as a tank.

            That all being said… you guys are absolutely making an impact on me, and you’re right. I have been subconsciously thinking of it as a “increases damage taken” stat even though I know damn well that’s only superficially true. I am flat-out not going to regem primarily for stam the way I was previously, but when it comes to relatively minor tweaks like subbing in trinkets, swapping out the occasional gem, etc… I will definitely be less hesitant to do so in the future.

          • Meloree says:

            It’s funny that you phrase it that way, because it sounds exactly like some of the background reading that Theck linked at the beginning of the post.

            Tank theorycraft isn’t absolute, especially when it doesn’t consider context: Your raid, and your healing team, and your content. And your playstyle, and your mistake-recovery, and your co-tanks, and any number of other factors. If you’re thinking things through and communicating with your healers and evaluating options… then you’re doing it right.

            The worst thing you can do – almost the only truly bad thing you can do – is close your mind to other options and refuse to evaluate the impact it might have on you in certain situations.

            Stamina gets a bad rap on a regular basis. The term “mana-sponge” gets thrown around a lot. It has a place, and it has value, and it always has. It isn’t The One Stat to Rule Them All anymore. And it shouldn’t be. But neither is it a trash stat, or an indication that someone is “doing it wrong”.

            So, shockingly enough – I don’t think you actually do disagree with Theck. Because Theck hasn’t really told you how to gear – he’s just demonstrated how to evaluate certain statistics in a certain context. The rest is, and always has been, up to you – to fit into your own personal context, to mold and adapt it to your raid situation.

        • Zothor says:

          I’m curious if your raid has World of Logs parses I could look over for your “stamina wipes”, in no small part because I am in the exact same place you were—11/12 25 man the last two weeks. The thing is, I’m stacking stamina almost exclusively and having absolutely no trouble with tank survivability (nor is my Warrior partner); at least as far as we’ve gotten in the fight, tank damage has been an almost nonexistent factor. Yeah, Lei Shen hits hard and fast, but not in an unmanageable fashion by any means.

          I pulled up what I believe to be your guild’s logs for your first kill of Lei Shen, after 6 wipes on 3/28/13. What I can’t find is any indication that, after a night of attempts, you regemmed to get your first kill, since it looks like it was that same night, after consecutive attempts, only minutes apart. Even if you did regem haste in the middle of that, you appear to have stumbled onto a sheer coincidence, because most of your deaths in that first-kill night involved getting zapped by the Diffused Lightnings & Unharnessed Powers in bunches because your raid wasn’t spread out properly (this isn’t surprising; you guys were obviously learning the fight). Besides, that’s magic damage that haste doesn’t help you with significantly (but which stamina can save you from). Take a look at your damage intake on your kill (very clean—most of your damage is directly from Lei Shein, and you’ve got almost none of the avoidable damage from diffused lightning, etc) vs. your wipes. But the things that killed you? Not mitigable by haste. And most of those deaths were after 5-10 people in your raid were already dead… they were midwipe.

          Is there a parse missing? Because if so, I’d love to see what changed. I can fathom that, especially given the vengeance levels in this fight, switching from a full stamina build to a full haste build might have gotten you over the hump—but my gut is telling me that it would have had to be because of the DPS value (your kill DPS jumped significantly compared to most of your other parses), assuming nothing changed significantly in your Shield of the Righteous strategy. Because you’re absolutely correct about at least one thing: if you’re playing wrong, or your raid is playing wrong, or you’re doing something different, that IS relevant. But it’s impossible to plan around, learn from, or build on if we can’t identify where the difference is. So if you have other parses, I’d be very curious to see them. But what your eyes told you is doesn’t seem to be visible in the data your guild uploaded to World of Logs.

  21. Taser says:

    I know that my answer micht not be 100% what you are looking for, but in relation to this topic, you might want to have a look at my EH-calcs in Pisshands Thread at EJ, Page 11 and following

    http://elitistjerks.com/f99/t131791-like_water_brewmasters_resource_4_2_13_a/p11/

    Stamina is indeed a good stat for Monks. Calculating TTL there is no better stat than stamina. And, according to my calcs, Monks can hit the breakeven, when Mastery provides more EH than Stamina, somewhere during T15. And Mastery is a fantastic stat for damage smoothing for Monks. Especially if you are doing 25H.

    • Theck says:

      Thanks, I’ll take a look. It doesn’t surprise me that mastery can catch up, as it should scale with boss attack size and the combination of Stagger + Purifying Brew turns it into essentially a flat “reduces damage taken by X%” stat, which is excellent for both TDR and smoothing. It’s also good to hear that stamina is competitive, and not completely irrelevant.

    • Yellowfive says:

      That’s interesting Taser, as is some of the other discussion in that thread on EJ. We’ll definitely take a closer look, I feel like our default brewmaster strategy could use some love.

      • Taser says:

        Problem with default stat priorities is, that there are none. It all depends on your raid, playstyle and several other parameters.

  22. Taser says:

    Sry for double post, but my post was supposed to be as answer to Sunnier and Yellowfive

  23. Parq says:

    Please don’t take this as criticism, there is just something I can’t figure out and would really like to get an understanding of.

    The simulation model was altered and now shows an increase in the value of Stamina whereas it didn’t before. The catalyst for this alteration was “Stamina has a much larger effect on healer reaction than either of the other stats though, so by throwing out any pretense of modeling healers, we are ignoring a significant portion of stamina’s benefit.”

    This argument must go both ways. Assuming a low-Stamina tank receives a damage spike large enough to cause a healer to “become concerned”, in the same respect the heals or HoTs on that tank will appear larger, percentage-wise, offsetting this concern.

    In earlier days where health bars did not show incoming heals I would agree the healer reaction would be quite negative if Stamina were lower. But now, I throw out a heal on low-health tank and, boom, I can see they’ll be up to good health in one cast even before that cast is complete.

    So, I guess my question is, why would a model be altered on the basis that losing a bigger percentage of health will have a “large effect on healer reaction” but not consider that it means all heals will appear larger, percentage-wise?

    • Theck says:

      While the size of the heal will be a bigger percentage of the health bar, it’s time, not heal size, that is ultimately important. If the tank dips sharply, the healer is forced to act more quickly than before because the tanks overall life expectancy has suddenly become shorter. So they have to reach into their toolkit and break out the fast, mana-inefficient heals and emergency cooldowns to keep you up. They simply don’t have time for a slow Holy Light or Greater Heal.

  24. tailias says:

    Some of the initial criticism I got about suggesting healers and healing had anything to do with Stamina or our stat priority really prompted me to think further, and analyze this with a little more of a critical eye.

    Where’s the healing? After all, Stamina is a boost to our *maximum* health, and our health bar is refilled by other people, not ourselves. This post merely measures how much abuse we can take before we die without any kind of healing whatsoever, which I suppose is kind of useful if we run into a boss where we have a debuff that reduces healing recieved by 100% and our death is inevitable yet we will need to survive for as long as possible anyway. If our only source of health at all is what we have at the very beginning then naturally, our starting health is going to be inflated much higher than anything else we could possibly care about.

    Are the Prot Paladins in these models WoGing themselves at low health or merely spamming Shield of the Righteous?

    I actually found this to be a pretty interesting analysis model/tool for comparing both the TDR and smoothing values of our secondary stats, because those are completely relative to one another. Mind you, we still look at your previous posts with a keen eye – I looked at your initial post where Mastery was a fair bit “better” than Haste and said to myself, well Mastery doesn’t smooth my damage anywhere near as much, so if there isn’t a tremendous reason to go there then I will stay with Haste. (Math only goes so far, logic and common sense trump it at times.)

    Stamina is much, much more subjective and simply put, affixing arbitrary damage numbers to one end and a Stamina figure somewhere in between and letting the crash test dummy run is not going to tell us anything that relates to actual raid tanking.

    • queldan says:

      Do remember Theck’s basic premise, which is to wit: What is currently killing tanks is not the slow trickle-death that comes over 15+ seconds, but the instagib or near instagib – something so fast that healers can’t meaningfully react to it.

      What does that imply? Basically, that we can assume that healers have more than enough mana and time to bring us back up between each of those “spike” events, either via direct or incidental healing. So Theck assumes healers will be able to bring us back up in between spikes. Incidentally, you’ll notice that Theck tends to cut on his sims at the 8-10 second marks, because he assumes that if a spike is THAT long, cooldowns would have been applied somewhere along the line.

      How do we observe it? First, some personal experience. If I’m not mistaken, my last serious death was an instagib on Horridon, where I got caught without CDs at the far end of the War-God phase, where a nasty melee + shout dealt around 750k in around 0.4 seconds, whereas every “trickle-down” death I suffered this patch was already telegraphed by multiple players dying to avoidable mechanics of all kind.

      Further, pulling some WoL parses, you’ll notice that direct heals form a notable minority of our healing intake (I’m at around 25% myself), meaning healers don’t need to focus us in particular, merely buff us up with HoTs and strong incidental heals (Beacon and the like), implying we’re not THAT hard to bring back up.

      Lastly, check your healer’s Overheal numbers and you’ll notice they’re crazy high. Part of the reason is that the threat of instagibs is currently bigger than the threat of ressource depletion. In other words – healers are capable of throwing a metric fuckton of healing and not be bothered by it.

      Now, your point has validity (actually, it was *the* critical element in Burning Crusade after Crushing Blows/instagibs) – if healers start complaining about mana issues, and you can’t track that back to failures in the fight mechanics, it’s time to consider a tradeoff between damage smoothness and TDR. But even then, Theck did debate that TDR might not pull ahead because of the new healing triage model and the efficiencty penalty big and fast heals suffer

      • bryjoered says:

        Yeah, I can’t believe than even when this post clearly proves Stamina’s superiority by a vast margin (like you should have it on every gem, PERIOD) People still come back with “What about when I switched to haste build and we were able to down this boss?” That’s merely simply a coincidence or that you or your healers/ raid as a whole executing the mechanics better, not that haste is more effective at survivability. With adequate healers, you could afford to not stack stamina at all and just pile on haste for a huge DPS and active mitigation increase. This requires you to be better at your rotation and means one or two mistakes could often result in your death, rather than something your healers could recover from, if you had the hit points…While Stamina is the optimal route from a survivability and smoothness standpoint, stacking haste is very beneficial to tanks that want more dps and tanks with great healers, while at the same time being a decent mitigation stat.

        You’ll notice that most tanks pushing heroic content undergeared aka hardest situation requiring most optimal play, that they stack stamina to the roof. (With the exception of monks) You can’t go wrong with stacking stamina, and it will make your time as a tank easier, but stacking secondary stats still works.

    • bryjoered says:

      Why are you such a Stamina hater? Math is literally how the stat priorities are created for every class in WoW, why should protection paladin’s or any tank be different? The “smoothness” metrics are based on being easier to heal because less spike means giving the healers more time to react. Stamina is smoother than haste and therefore is easier to heal, granted you are using an optimal rotation. Haste is still a very effective stat, but in terms of pure survivability, stamina is better in all regards, this post and the subsequent post prove this. You can continue to stack Haste and be just fine and probably will be able to clear the current content, that still doesn’t mean that your way of gearing is the most optimal path for survivability, stamina is the blatant reality of math is there is no “gray” area. It’s just what is right and what is wrong, Stam is the best damage smoother, therefore stam is the best survivability stat, end of story.

    • Theck says:

      @tailias: If you are receiving steady healing, you will almost never die. Healer throughput is more than enough to keep a tank up in current content. Tanks don’t die when the healer is able to stand still and spam their efficient heal on the tank. They die when the healer is distracted, impaired, or otherwise unable to stand and cast. In other words, a period of low healer throughput and higher-than-average damage intake – a spike.

      In any event, your discussion of “where’s the healing” and “if our only source of health is what we have at the beginning…” suggests you don’t understand what the simulation is doing. You’ll have to go back through the archives a bit to find the earlier posts where I explain the methodology – in short, there are no healers because there *is no health.* I don’t track tank health at all, just overall damage intake.

      The rationale is that, if spike death is the major concern, what we care about most is minimizing the number of high-damage periods we take. In short, we want to make our damage as smooth as possible so that the healers *can* get distracted without as much risk.

      It has nothing to do with stacking stamina so that we “don’t run out of health.”

  25. Djellibeybi says:

    A very nice post, thank you for sharing those thoughts and calculations.

    Can i sum up your argument by stating: (this maybe a bit over simplified)
    Your healers will likely never run out of mana. (in this expansion that is) They will be able to heal up almost all spikes given enough time.Therefore giving them more buffer (=time) to heal you up after some big damage spike is more beneficial than anything else you can do or provide yourselves in terms of survivability.

    I like the way you proved that healers are better at keeping tanks alive than tanks are. :D As a bear i am naturally convinced it is ‘the fur’ that keeps me alive. But incoming heals are always nice.

    Cheers.

  26. Pingback: Blood Tanking - Actively Mitigating since before it was cool - Page 16 - Elitist Jerks

Leave a Reply