# Is Nothing Sacred?

I’ve been hinting all week that there was some big news coming up regarding our gearing choices.  If you’ve followed the last two rounds of smoothness simulations, it looks an awful lot like Control/Mastery is the solid winner.  To put some numbers to that thought, let’s look at a side-by-side comparison of the queues:

         |       Control/Mastery       |        Control/Haste
| Queue: |       S |     SH1 |     SH2 |       S |     SH1 |     SH2 |
|     S% |  0.4108 |  0.4102 |  0.4105 |  0.5225 |  0.5222 |  0.5225 |
|   mean |  0.5412 |  0.5416 |  0.5422 |  0.6003 |  0.6013 |  0.6005 |
|    std |  0.1451 |  0.1328 |  0.1252 |  0.1481 |  0.1344 |  0.1425 |
| ------ |   --- 2 |  Attack |  Moving | Average |  ------ |  ------ |
|    60% | 42.2155 | 46.5763 | 42.8827 | 46.3508 | 47.4428 | 45.8948 |
|    70% | 21.9353 | 17.4630 | 19.5302 | 40.6535 | 43.3518 | 40.3810 |
|    80% | 21.9353 | 17.4630 | 19.5302 | 14.5957 | 10.0767 | 14.1655 |
|    90% | 10.2748 |  5.5675 |  8.4910 |  9.7813 |  4.4400 |  9.4358 |
| ------ |   --- 3 |  Attack |  Moving | Average |  ------ |  ------ |
|    60% | 32.6512 | 27.6415 | 33.6860 | 53.6503 | 53.9720 | 54.0338 |
|    70% | 18.2035 | 13.2302 | 13.8877 | 36.9320 | 36.7185 | 36.8463 |
|    80% |  7.4228 |  4.3803 |  1.7488 | 15.0675 |  7.4773 | 14.6283 |
|    90% |  2.7748 |  1.2550 |  0.6620 |  1.2695 |  0.4752 |  0.0010 |
| ------ |   --- 4 |  Attack |  Moving | Average |  ------ |  ------ |
|    60% | 33.8680 | 31.6200 | 31.6953 | 53.7005 | 52.9335 | 55.2110 |
|    70% | 12.4778 |  8.6483 |  4.0650 | 32.1685 | 26.9200 | 32.0120 |
|    80% |  5.2785 |  3.4120 |  1.3058 |  3.3473 |  2.0720 |  0.0050 |
|    90% |  1.2040 |  0.7582 |  0.2240 |  0.0000 |  0.0697 |  0.0000 |
| ------ |   --- 5 |  Attack |  Moving | Average |  ------ |  ------ |
|    60% | 39.2205 | 36.6190 | 35.4823 | 55.3028 | 55.8280 | 56.5597 |
|    70% | 10.5500 |  8.6173 |  6.9608 | 28.5143 | 25.6930 | 26.8380 |
|    80% |  1.4735 |  1.3350 |  0.3348 |  8.8370 |  5.0570 |  6.9102 |
|    90% |  0.0000 |  0.0425 |  0.0000 |  0.0000 |  0.0000 |  0.0000 |
| ------ |   --- 6 |  Attack |  Moving | Average |  ------ |  ------ |
|    60% | 35.7910 | 32.5177 | 32.3367 | 54.0813 | 56.5700 | 54.7773 |
|    70% | 11.0220 |  8.3745 |  7.3345 | 24.1095 | 23.5930 | 22.9925 |
|    80% |  0.0000 |  0.1320 |  0.0105 |  4.2360 |  2.4580 |  3.3653 |
|    90% |  0.0000 |  0.0000 |  0.0000 |  0.0000 |  0.0000 |  0.0000 |
| ------ |   --- 7 |  Attack |  Moving | Average |  ------ |  ------ |
|    60% | 32.2383 | 28.7430 | 29.3895 | 52.9763 | 54.9205 | 53.4872 |
|    70% |  9.6973 |  7.1182 |  4.8477 | 22.5190 | 20.2357 | 21.5467 |
|    80% |  0.8415 |  0.5030 |  0.2890 |  3.7915 |  2.0220 |  2.8915 |
|    90% |  0.0000 |  0.0000 |  0.0000 |  0.0000 |  0.0000 |  0.0000 |

Control/Haste wins the race to eliminating 90% spikes, and generally dominates that category.  But that’s mostly relevant for 2- to 4-attack spikes; beyond that point Control/Mastery catches up and both gear sets eliminate the category.  And C/Ha is miserably behind C/Ma in 70% and 80% spikes, and takes 10% more damage.  It’s such a significant difference that you can even see it in the damage distribution plots.  The first one below is for C/Ma with an SH2 rotation and a 5-attack moving average, the second is for C/Ha with SH1:

Control/Mastery gear set, SH2 rotation.

Control/Haste gear set, SH1 rotation.

Side by side like this, you can definitely see the difference.  C/Ha permits far more events above the 80% threshold, and has a larger skew towards the higher damage ranges.  C/Ma has much less skew, a lower mean, and chops off the high events much more effectively. Based just on the performance of the shifting queues, it’s not just the winner – C/Ma has taken C/Ha out back, beat it up a little, and kicked it while it’s down.

Talent Show

However, haste has one more trick up it’s sleeve.  When I wrote these simulations initially, I intentionally avoided adding any talent-based features.  The concept of talents was that they are supposed to be modular, and thus I couldn’t count on any given paladin taking any given talent.  This is also good for me, because it limits the number of configurations I needed to worry about – having to simulate different numbers for Execution Sentence healing vs. Light’s Hammer healing vs. Holy Prism healing, for example, would be a pain in the ass.

The L75 talents are a bit more relevant, as you could imagine Divine Purpose giving very different results than Holy Avenger.  I hope to get around to adding those talents some day so we can perform a realistic comparison of them, but I don’t think they will affect a player’s choice of gear very much.  Both should emphasize hit and expertise, both should favor haste and mastery, and your choice of haste vs. mastery is likely based more around play style than the small variations in smoothness each L75 talent will cause.

But there’s one tier of talents that hasn’t really panned out for protection.  Our L45 talent choices are heavily skewed – Selfless Healer and Eternal Flame simply cannot keep up with the sheer mitigative power of Sacred Shield.  I do know paladins that take the first two talents, but it’s a small minority.  And frankly, the rationalizations they give for making those choices don’t line up with my understanding of how tanks die or what prevents those deaths.

Since the vast majority of paladins are taking Sacred Shield, it seems reasonable to assume that it will be present, which we haven’t done in any of the simulations up to this point.  And that’s a big deal, because Sacred Shield scales. Oh does it scale.  It scales a lot, and not uniformly either.  Stacking mastery to the sky has almost no effect on Sacred Shield, and strictly zero effect if we’re only considering the spells characteristics.  But it scales very well (linearly, in fact) with haste, because we get more absorption bubbles and we get them faster.  It also scales with avoidance, because avoided attacks extend the average lifetime of the absorption bubbles, making each more effective.  We saw a similar effect with warriors and Shield Barrier usage, because it’s the same principle at work.

And this matters, because the haste variations between gears sets are huge.  The Control/Mastery set has 0% melee haste and the baseline 15.5% spell haste we get from Seal of Insight and the raid-wide spell haste buff.  The Control/Haste set has 28.24% melee haste and a whopping 48.11% spell haste.  In the C/Ma gear set we’re getting a Sacred Shield absorb bubble every 5.19 seconds.  The C/Ha gear set knocks that down to 4.05 seconds.

And these Sacred Shield absorbs aren’t anything to sneeze at.  To give you some idea of the magnitudes we’re talking about, the raw boss DPS I’m using is 310k, or 465k swings every 1.5 seconds.  Armor and other passive mitigation effects knock that down to a hair over 168k damage per swing as registered in the combat log.  In the steady state, that gives us 111.6k Vengeance attack power, or a total of about 156k AP.  With that much AP, each Sacred Shield bubble ends up absorbing about 67k, or ~40% of a boss swing.  That’s better than a block, and almost like getting a free SotR applied to one boss swing.  And the haste gear set lets you do that 20% more often.

So if we’re going to make a truly fair comparison between C/Ha and C/Ma, or any of the other gear sets for that matter, we should really be including Sacred Shield in the analysis.  At the same time that I added the shifting queues to the code, I finally implemented Sacred Shield as well.  And the results were fairly surprising.  Let’s take a look.

Gear and Simulation Details

First, the obligatory background information.  If you’ve been keeping up with these smoothness simulations, you can probably skip this section.  I forgot to include it earlier this week and got some questions about it, so I figured I should just keep including it.

The stats of the gear sets we’re using are listed in the table below.  Each set has 65k armor, 15k strength, and 24150 rating to distribute amongst the secondary stats.  This is roughly equivalent to an average equipped ilvl of 522.  For more details on the choices I’ve made, you can consult the original 5.2 Smoothness Simulations post.

|    Set: |  C/Ha |  C/Ma |  C/Av | C/Bal |    Ha | Avoid | Av/Mas | Mas/Av |
|     Str | 15000 | 15000 | 15000 | 15000 | 15000 | 15000 |  15000 |  15000 |
|   Parry |  1500 |  1500 |  7500 |  4125 |  1500 | 10825 |   7717 |   4000 |
|   Dodge |  1500 |  1500 |  7500 |  4125 |  1500 | 10825 |   7717 |   4000 |
| Mastery |  1500 | 13500 |  1500 |  4125 |  1500 |  1500 |   7716 |  15150 |
|     Hit |  2550 |  2550 |  2550 |  2550 |   500 |   500 |    500 |    500 |
|     Exp |  5100 |  5100 |  5100 |  5100 |   500 |   500 |    500 |    500 |
|   Haste | 12000 |     0 |     0 |  4125 | 18650 |     0 |      0 |      0 |

I’m using the standard Monte-Carlo code, updated with the 5.2 mechanics.  Just as last time, here’s the copy/paste summary of how the simulation works:

To better understand the data below, here’s a rough overview of how it’s generated: I run a Monte-Carlo sim that simulates 10k minutes of combat (think Simcraft, but paladin-specific and more limited in scope), making all combat rolls and logging all damage events.  I take the resulting string of attacks (something like “1, 0, 0.7, 1, 0, 0, …” where 0 is an avoid (no damage), 1 is a full hit, 0.7 is a block, and so on) and do some calculations on it.  I calculate the average damage intake normalized to 100% possible throughput (i.e. “1, 1, 1, 1, 1, …”), and report that in the “mean” row, representing mean damage intake (lower is better, represents better TDR).  “std” is just the standard deviation of that mean as averaged over 5 attacks.  “S%” is SotR uptime.

The rest of the rows are smoothing data for strings of N attacks.  For now, let’s just consider the first gear column (C/Ha).  I take the damage event sequence and perform a moving average on it (i.e. an N-attack moving average).  I then calculate how many of those N-attack averages exceed a certain threshold of maximum throughput.  So for example, if we look at a 3-attack moving average, the “70%” row tells me how many of those 3-attack averages exceed 70% of the maximum throughput.  Max throughput for 3 attacks is “1, 1, 1″ or 3 normalized damage, so the 70% row tells me how many exceed 2.1 damage.  And so on for 80% and 90%.  Note that they’re cumulative, so if 5% of attacks exceed 90% max throughput those attacks are also being counted in the 80% and 70% rows (thus, if 17% of attacks exceed 80% max throughput, the percentage between 80% and 90% is 17%-5%=12%).  I should add that the repeatability on these simulations is quite good thanks to the long integration time – results usually fluctuate by less than +/- 0.1% (absolute, i.e. 5% +/- 0.1%).

I do this for a bunch of different gear sets, i.e. “C/Ha” for Control/Haste, etc.  The first table lists all of the gear configurations so you get a rough idea of what they look like.  They’re roughly equivalent to stats in ilvl 496 gear.

The code can be found in the matlabadin repository, as usual.  The two files in particular are pally_mc.m and pally_mc_smooth.m.

Results – Simple Shield of the Righteous Spam

First, let’s look at the data for the simple “S” queue and with Sacred Shield disabled.   This won’t include any fancy shifting tricks, so it’s a better estimate for a novice raider who’s still getting comfortable tanking and hasn’t mastered their active mitigation yet.  This isn’t new data – you’ve seen this before, but we need it to establish the baseline so that we can tell what sort of difference Sacred Shield makes.

| Set: |    C/Ha |    C/Ma |    C/Av |   C/Bal |      Ha |   Avoid |  Av/Mas |  Mas/Av |
|   S% |  0.5224 |  0.4106 |  0.4191 |  0.4528 |  0.4990 |  0.3664 |  0.3632 |  0.3577 |
| mean |  0.6013 |  0.5410 |  0.5333 |  0.5559 |  0.6082 |  0.4946 |  0.5036 |  0.5158 |
|  std |  0.1478 |  0.1453 |  0.1759 |  0.1611 |  0.1523 |  0.1879 |  0.1771 |  0.1650 |
| ---- |  ------ |   --- 2 |  Attack |  Moving | Average |  ------ |  ------ |  ------ |
|  60% | 46.4365 | 42.2245 | 36.0532 | 42.0223 | 46.8495 | 30.6360 | 32.6615 | 38.9607 |
|  70% | 40.7703 | 21.9045 | 32.3510 | 32.0562 | 41.5132 | 27.8290 | 27.5790 | 22.6745 |
|  80% | 14.6537 | 21.9045 | 16.8478 | 17.5670 | 17.2325 | 16.7553 | 19.2880 | 22.6745 |
|  90% |  9.8918 | 10.2063 | 11.3503 | 10.7805 | 11.6783 | 11.2970 | 10.6098 | 10.0570 |
| ---- |  ------ |   --- 3 |  Attack |  Moving | Average |  ------ |  ------ |  ------ |
|  60% | 53.8772 | 32.5935 | 39.5635 | 45.3188 | 54.4480 | 34.0340 | 35.6205 | 31.5645 |
|  70% | 37.0875 | 18.1815 | 24.6943 | 28.7108 | 37.0938 | 18.9820 | 17.2312 | 17.7622 |
|  80% | 15.1958 |  7.4453 | 13.2320 | 12.3843 | 16.6545 | 11.3433 | 11.0848 |  8.8108 |
|  90% |  1.2962 |  2.7432 |  3.2145 |  2.6515 |  3.1935 |  3.6222 |  3.2912 |  3.0520 |
| ---- |  ------ |   --- 4 |  Attack |  Moving | Average |  ------ |  ------ |  ------ |
|  60% | 53.9435 | 33.9070 | 39.4125 | 42.8023 | 54.8250 | 32.6215 | 33.1535 | 30.3133 |
|  70% | 32.4230 | 12.4927 | 21.3595 | 25.8540 | 32.6262 | 17.0912 | 17.2385 | 13.6712 |
|  80% |  3.3830 |  5.2843 |  6.2967 |  6.5102 |  6.9973 |  5.6180 |  5.2748 |  6.7035 |
|  90% |  0.0000 |  1.2263 |  0.9175 |  0.6537 |  1.4930 |  1.9408 |  2.3090 |  2.7103 |
| ---- |  ------ |   --- 5 |  Attack |  Moving | Average |  ------ |  ------ |  ------ |
|  60% | 55.5823 | 39.0590 | 40.1595 | 42.7240 | 56.8680 | 33.0050 | 30.1425 | 34.5108 |
|  70% | 28.7618 | 10.5665 | 19.3762 | 22.4620 | 30.4220 | 15.4128 | 13.3293 | 11.5990 |
|  80% |  8.9097 |  1.4918 |  7.0112 |  5.3835 | 10.8432 |  4.9352 |  2.7422 |  3.4403 |
|  90% |  0.0000 |  0.0000 |  0.5200 |  0.3295 |  1.0175 |  1.0540 |  0.6695 |  0.7688 |
| ---- |  ------ |   --- 6 |  Attack |  Moving | Average |  ------ |  ------ |  ------ |
|  60% | 54.3558 | 35.7295 | 34.1108 | 41.2287 | 55.5080 | 25.9360 | 28.3017 | 31.3457 |
|  70% | 24.3050 | 11.0603 | 13.4900 | 16.6937 | 26.1385 | 10.4652 | 10.5652 | 11.1263 |
|  80% |  4.2673 |  0.0000 |  4.3717 |  4.8828 |  6.8168 |  3.4403 |  2.7655 |  1.4698 |
|  90% |  0.0000 |  0.0000 |  0.0000 |  0.0000 |  0.2012 |  0.3240 |  0.2855 |  0.2027 |
| ---- |  ------ |   --- 7 |  Attack |  Moving | Average |  ------ |  ------ |  ------ |
|  60% | 53.2705 | 32.1395 | 32.5230 | 38.7373 | 54.7385 | 24.7587 | 25.7575 | 28.3533 |
|  70% | 22.7233 |  9.7105 | 12.6490 | 15.3603 | 24.6157 |  9.4672 |  9.3260 |  9.6970 |
|  80% |  3.8538 |  0.7848 |  3.0023 |  2.9965 |  5.5295 |  2.0965 |  1.9265 |  1.4485 |
|  90% |  0.0000 |  0.0000 |  0.0000 |  0.0000 |  0.0445 |  0.1020 |  0.1087 |  0.1403 |

As we saw before, Control/Mastery has a pretty strong advantage.  Control/Haste makes a good showing in certain categories, but lags significantly in most of them.

However, if we turn Sacred Shield on, we find get very different results:

| Set: |    C/Ha |    C/Ma |    C/Av |   C/Bal |      Ha |   Avoid |  Av/Mas |  Mas/Av |
|   S% |  0.5229 |  0.4106 |  0.4195 |  0.4520 |  0.4980 |  0.3667 |  0.3621 |  0.3575 |
| mean |  0.4111 |  0.3898 |  0.3663 |  0.3893 |  0.4175 |  0.3378 |  0.3448 |  0.3644 |
|  std |  0.1362 |  0.1357 |  0.1521 |  0.1440 |  0.1424 |  0.1597 |  0.1553 |  0.1495 |
| ---- |  ------ |   --- 2 |  Attack |  Moving | Average |  ------ |  ------ |  ------ |
|  40% | 47.8083 | 48.5945 | 39.4050 | 45.5303 | 48.5420 | 34.5910 | 36.9545 | 45.0358 |
|  50% | 38.6730 | 29.1845 | 31.8520 | 32.9937 | 39.4498 | 27.3310 | 28.9110 | 28.6592 |
|  60% | 14.2155 | 25.0162 | 16.4483 | 18.5405 | 16.8133 | 16.3825 | 19.6868 | 24.3062 |
|  70% |  9.0985 | 12.2625 | 10.7745 | 11.6047 | 10.6237 | 10.6597 | 10.8950 | 11.3018 |
|  80% |  9.0983 | 11.1268 | 10.7743 | 10.6713 | 10.6235 | 10.6595 | 10.2533 |  9.9063 |
|  90% |  0.0005 |  1.8365 |  0.0005 |  0.5215 |  0.0003 |  0.0003 |  0.3465 |  1.2070 |
| ---- |  ------ |   --- 3 |  Attack |  Moving | Average |  ------ |  ------ |  ------ |
|  40% | 53.2233 | 47.9372 | 41.6563 | 46.2660 | 54.7275 | 37.3810 | 38.3425 | 44.8230 |
|  50% | 33.1910 | 29.1695 | 28.1168 | 31.0420 | 34.3940 | 25.1808 | 25.1827 | 27.6882 |
|  60% | 15.8120 | 15.3685 | 13.5373 | 12.3215 | 17.3197 | 11.6468 | 10.7345 | 14.6328 |
|  70% |  6.0195 |  6.2542 |  6.4393 |  7.1407 |  7.7778 |  6.0847 |  5.1463 |  6.4992 |
|  80% |  0.5590 |  2.0340 |  1.2630 |  1.3333 |  1.3225 |  1.3603 |  1.4642 |  1.8325 |
|  90% |  0.0003 |  0.0003 |  0.0003 |  0.0000 |  0.0000 |  0.0000 |  0.0000 |  0.0000 |
| ---- |  ------ |   --- 4 |  Attack |  Moving | Average |  ------ |  ------ |  ------ |
|  40% | 55.6395 | 50.5580 | 45.6000 | 50.7708 | 56.6330 | 39.9558 | 41.5223 | 44.1125 |
|  50% | 32.0505 | 22.0170 | 23.8438 | 29.0312 | 32.5883 | 19.4885 | 20.2703 | 19.7258 |
|  60% |  3.3115 |  7.8463 |  6.6013 |  7.8057 |  7.0792 |  6.3295 |  6.5520 |  8.6195 |
|  70% |  1.5733 |  2.3697 |  3.3905 |  1.0685 |  3.8997 |  3.5480 |  2.4675 |  3.3725 |
|  80% |  0.0000 |  0.5643 |  0.4318 |  0.3043 |  0.6975 |  0.9630 |  0.8832 |  0.8835 |
|  90% |  0.0000 |  0.0860 |  0.0000 |  0.0000 |  0.0000 |  0.0000 |  0.0003 |  0.0000 |
| ---- |  ------ |   --- 5 |  Attack |  Moving | Average |  ------ |  ------ |  ------ |
|  40% | 56.0972 | 47.6848 | 43.1835 | 49.6810 | 57.1600 | 35.3995 | 37.6767 | 39.3170 |
|  50% | 28.0432 | 23.6345 | 20.8675 | 23.0615 | 30.4812 | 17.0093 | 16.7255 | 20.8955 |
|  60% |  9.0510 |  2.8620 |  7.2145 |  6.7045 | 11.3188 |  5.9373 |  5.1080 |  4.2838 |
|  70% |  0.0000 |  0.3158 |  0.4495 |  0.2433 |  1.0105 |  1.0470 |  1.1665 |  1.4310 |
|  80% |  0.0000 |  0.0000 |  0.0000 |  0.0000 |  0.0562 |  0.1053 |  0.1070 |  0.1305 |
|  90% |  0.0000 |  0.0000 |  0.0000 |  0.0000 |  0.0000 |  0.0000 |  0.0000 |  0.0000 |
| ---- |  ------ |   --- 6 |  Attack |  Moving | Average |  ------ |  ------ |  ------ |
|  40% | 57.0305 | 47.5482 | 42.5625 | 47.8055 | 58.5522 | 34.9025 | 34.1675 | 40.9050 |
|  50% | 28.6013 | 19.9410 | 20.1412 | 21.2048 | 30.8865 | 15.7623 | 15.3158 | 15.7263 |
|  60% |  3.9783 |  3.8653 |  4.6835 |  5.9700 |  6.6668 |  3.9325 |  3.5230 |  3.5655 |
|  70% |  0.0000 |  0.0000 |  0.0000 |  0.0747 |  0.2290 |  0.4517 |  0.5048 |  0.6350 |
|  80% |  0.0000 |  0.0000 |  0.0000 |  0.0000 |  0.0210 |  0.0635 |  0.0578 |  0.0615 |
|  90% |  0.0000 |  0.0000 |  0.0000 |  0.0000 |  0.0000 |  0.0000 |  0.0000 |  0.0000 |
| ---- |  ------ |   --- 7 |  Attack |  Moving | Average |  ------ |  ------ |  ------ |
|  40% | 55.1845 | 47.9093 | 42.2420 | 47.8852 | 56.8610 | 34.7955 | 33.5052 | 41.3620 |
|  50% | 23.3448 | 18.5660 | 16.4868 | 18.9205 | 26.0792 | 12.7240 | 11.9865 | 16.0440 |
|  60% |  3.0648 |  3.1160 |  3.1250 |  3.7430 |  5.1485 |  2.6605 |  2.4692 |  2.8385 |
|  70% |  0.1460 |  0.0473 |  0.2498 |  0.2550 |  0.4335 |  0.2563 |  0.1417 |  0.1662 |
|  80% |  0.0000 |  0.0000 |  0.0000 |  0.0000 |  0.0017 |  0.0055 |  0.0043 |  0.0067 |
|  90% |  0.0000 |  0.0000 |  0.0000 |  0.0000 |  0.0000 |  0.0000 |  0.0000 |  0.0000 |

I’ve had to include a lot more data here, because Sacred Shield absorbs a LOT of damage.  We’re getting about a ~30% reduction in damage taken across the board by turning on Sacred Shield.  As we get into sequences of many attacks we can zero out almost all of the 60%-90% categories.  When that happens, we need to consider the lower sizes to get a clear idea of what each gear set is doing.  It makes the tables a bit more unwieldly, which is annoying, but most people just read my summary of the table anyway.  Either way, I’d rather have all of the data on the table so that inquiring minds can look for things I may have missed.

The first thing you note on this table is that Control/Haste fares significantly better than it did without Sacred Shield.  It is competing with or beating Control/Mastery in pretty much every category.  It isn’t until we get to 7 attacks that it starts to trail C/Mastery by any measurable amount, and even that’s not very much.  The small advantage C/Ma has in the  7-attack category doesn’t make up for the significant advantage C/Ha gives you in the 2- to 4- attack categories.  And while C/Ma maintains it’s lead in damage taken, that lead has dropped from 10% to 5%.

The other surprising thing about this data set is that the avoidance gear sets don’t perform all that badly.  They’re rarely winning in a category, but they often rival C/Ma.  Their usual weakness in the 80% and 90% thresholds is still there, but it’s not as pronounced as before.  And it still takes about 18% less damage than C/Ha.

This is likely due to the inherent synergy between avoidance and absorption effects.  Avoiding an attack during SotR means that there’s one fewer attack to which SotR’s mitigation is applied, effectively undermining the use of SotR – that’s the negative scaling we’ve consistently seen between avoidance and control stats in the past.

But avoiding an attack doesn’t waste an absorption bubble, allowing that bubble to apply to the next unavoided attack.  In that way, absorption bubbles automatically time-shift to make themselves useful. The only time they’re wasted is if they expire before you take an attack, which means you didn’t take damage during the duration – an automatically safe situation to begin with.

The other gear sets all perform pretty well too, though nothing stands out as particularly interesting.  None of them trail by much, so the “use whatever gear you can find” strategy isn’t going to be at that much of a disadvantage.

In any event, it seems Control/Haste has won this round.  Control/Mastery isn’t that far behind, but it also doesn’t provide the raw DPS increase that the haste build does.  When one build is better for DPS and for survivability with the “macro SotR to Crusader Strike” queue, it’s a bit of a no-brainer.

However, we also know that shifting seemed to give mastery a bigger benefit than haste.  So maybe enabling the shifting algorithm will allow mastery to regain some of that ground.  Let’s see if that changes anything.

Results – SH1 Finisher Queue

If you remember from last time, the SH1 queue tries to cast SotR if:

1. We have 5 holy power and a generator is available within the next second
2. We have 3+ holy power and the mean of the last boss attack exceeded 80% throughput.

This means that we’ll use SotR early if the last boss attack wasn’t blocked, avoided, or mitigated with SotR.  Otherwise we’ll just bank holy power to 5 and bleed it as safely as possible.

First, let’s look at the data with Sacred Shield disabled.  This is the same data we saw earlier this week for C/Ha and C/Ma, but this table includes all of the other gear sets.

| Set: |    C/Ha |    C/Ma |    C/Av |   C/Bal |      Ha |   Avoid |  Av/Mas |  Mas/Av |
|   S% |  0.5224 |  0.4103 |  0.4197 |  0.4523 |  0.4980 |  0.3672 |  0.3625 |  0.3577 |
| mean |  0.6004 |  0.5416 |  0.5328 |  0.5562 |  0.6082 |  0.4936 |  0.5043 |  0.5164 |
|  std |  0.1348 |  0.1332 |  0.1643 |  0.1480 |  0.1397 |  0.1766 |  0.1638 |  0.1503 |
| ---- |  ------ |   --- 2 |  Attack |  Moving | Average |  ------ |  ------ |  ------ |
|  40% | 79.9800 | 65.3280 | 69.9758 | 74.0790 | 80.4850 | 65.0438 | 67.4257 | 63.1565 |
|  50% | 64.7512 | 50.0140 | 46.1452 | 51.0942 | 64.8940 | 37.6743 | 37.4632 | 45.0725 |
|  60% | 47.3285 | 46.5570 | 36.3417 | 42.8345 | 48.1180 | 30.6615 | 34.7815 | 42.3635 |
|  70% | 43.2215 | 17.4820 | 33.6598 | 32.6245 | 44.1107 | 28.5703 | 27.5072 | 18.3808 |
|  80% | 10.0245 | 17.4820 | 12.2670 | 12.9318 | 12.6908 | 12.3615 | 15.1310 | 18.3808 |
|  90% |  4.4145 |  5.6712 |  6.5693 |  5.8338 |  6.3378 |  7.0395 |  6.7092 |  6.1442 |
| ---- |  ------ |   --- 3 |  Attack |  Moving | Average |  ------ |  ------ |  ------ |
|  40% | 85.6148 | 83.1100 | 71.9463 | 79.9485 | 85.7280 | 63.5208 | 68.6168 | 77.3255 |
|  50% | 77.8120 | 58.6900 | 63.8260 | 70.0968 | 78.2438 | 56.0808 | 45.6707 | 53.3563 |
|  60% | 53.8410 | 27.6640 | 37.3355 | 43.9935 | 54.8933 | 31.0248 | 34.3588 | 26.9438 |
|  70% | 36.5553 | 13.3447 | 23.9787 | 27.9680 | 37.4950 | 18.2287 | 13.7600 | 13.5083 |
|  80% |  7.4112 |  4.4365 |  8.3697 |  7.3390 |  9.7087 |  7.5145 |  7.5018 |  5.6735 |
|  90% |  0.4773 |  1.3013 |  1.5867 |  1.2285 |  1.6435 |  2.1285 |  1.9637 |  1.7528 |
| ---- |  ------ |   --- 4 |  Attack |  Moving | Average |  ------ |  ------ |  ------ |
|  40% | 86.0850 | 83.9205 | 70.7567 | 79.5945 | 86.5430 | 62.6885 | 71.0963 | 78.4237 |
|  50% | 79.6950 | 61.9600 | 59.6745 | 64.4532 | 79.7443 | 48.4910 | 46.9840 | 54.0313 |
|  60% | 52.7840 | 31.5972 | 37.1283 | 41.4445 | 54.3593 | 30.1280 | 30.9165 | 27.8300 |
|  70% | 26.7898 |  8.7100 | 18.4998 | 22.4168 | 28.8195 | 14.6303 | 14.7073 |  9.9785 |
|  80% |  2.0325 |  3.5068 |  3.7713 |  3.8595 |  4.5088 |  3.6730 |  3.6110 |  4.6870 |
|  90% |  0.0653 |  0.7993 |  0.5550 |  0.4383 |  0.8952 |  1.2337 |  1.5215 |  1.7453 |
| ---- |  ------ |   --- 5 |  Attack |  Moving | Average |  ------ |  ------ |  ------ |
|  40% | 93.0055 | 84.9635 | 79.4570 | 84.4562 | 93.0613 | 69.4290 | 71.0673 | 77.7092 |
|  50% | 79.3107 | 60.1430 | 60.8635 | 67.3745 | 79.8648 | 51.1845 | 49.2092 | 53.2810 |
|  60% | 55.6362 | 36.5780 | 38.8565 | 41.5920 | 57.4337 | 31.2155 | 27.9410 | 31.6680 |
|  70% | 25.5045 |  8.8130 | 16.8290 | 18.7098 | 28.0448 | 13.0993 | 11.0980 |  9.4880 |
|  80% |  5.0088 |  1.3685 |  5.0305 |  3.7135 |  7.2525 |  3.8527 |  1.9560 |  2.6338 |
|  90% |  0.0000 |  0.0487 |  0.2490 |  0.1515 |  0.5242 |  0.6455 |  0.4367 |  0.4740 |
| ---- |  ------ |   --- 6 |  Attack |  Moving | Average |  ------ |  ------ |  ------ |
|  40% | 93.6545 | 85.7085 | 80.9963 | 86.8032 | 93.8197 | 72.0932 | 75.3320 | 78.7903 |
|  50% | 81.1658 | 65.3350 | 61.0710 | 67.5583 | 81.6785 | 49.4887 | 49.8540 | 55.8625 |
|  60% | 56.3747 | 32.6410 | 34.1103 | 40.5217 | 57.6178 | 25.0623 | 26.3197 | 28.7055 |
|  70% | 23.3570 |  8.5092 | 12.1910 | 14.3090 | 25.4800 |  9.0840 |  8.7905 |  9.1058 |
|  80% |  2.4442 |  0.1440 |  2.9768 |  3.2437 |  4.6220 |  2.6045 |  2.0735 |  1.0507 |
|  90% |  0.0000 |  0.0000 |  0.0000 |  0.0000 |  0.1085 |  0.2122 |  0.1905 |  0.1110 |
| ---- |  ------ |   --- 7 |  Attack |  Moving | Average |  ------ |  ------ |  ------ |
|  40% | 94.5883 | 87.7143 | 82.3010 | 89.3173 | 94.7705 | 72.6070 | 77.9573 | 81.1767 |
|  50% | 81.7600 | 62.9725 | 61.0677 | 68.3070 | 82.3575 | 49.7385 | 50.8487 | 54.5155 |
|  60% | 54.7430 | 28.8410 | 31.6180 | 37.7540 | 56.2775 | 23.1782 | 23.9253 | 25.5347 |
|  70% | 20.0370 |  7.2920 | 10.7345 | 12.8502 | 22.5915 |  8.0165 |  7.6148 |  7.8110 |
|  80% |  1.9885 |  0.5120 |  2.0800 |  1.9633 |  3.6658 |  1.6352 |  1.4348 |  1.0933 |
|  90% |  0.0000 |  0.0000 |  0.0000 |  0.0000 |  0.0225 |  0.0697 |  0.0850 |  0.0852 |

We see a reprisal of the Control/Mastery superiority here, but we talked about that last time.  It’s clear that the shifting queue benefits Control/Avoidance less than the mastery and haste variants, and Control/Balance tends to fall somewhere in the middle.  The avoidance builds generally perform a little worse than the Control builds again, much like they did with the simple S queue.

Now let’s turn Sacred Shield back on and see what happens:

| Set: |    C/Ha |    C/Ma |    C/Av |   C/Bal |      Ha |   Avoid |  Av/Mas |  Mas/Av |
|   S% |  0.5227 |  0.4103 |  0.4192 |  0.4522 |  0.4988 |  0.3669 |  0.3620 |  0.3572 |
| mean |  0.4112 |  0.3923 |  0.3670 |  0.3897 |  0.4178 |  0.3370 |  0.3453 |  0.3671 |
|  std |  0.1297 |  0.1272 |  0.1420 |  0.1334 |  0.1321 |  0.1479 |  0.1412 |  0.1372 |
| ---- |  ------ |   --- 2 |  Attack |  Moving | Average |  ------ |  ------ |  ------ |
|  40% | 47.6090 | 53.3802 | 39.9920 | 46.3927 | 48.5255 | 34.8960 | 38.7608 | 48.7848 |
|  50% | 39.0460 | 30.6213 | 32.9847 | 33.9475 | 40.0908 | 28.0093 | 29.3300 | 28.8960 |
|  60% | 13.0910 | 24.1870 | 14.1325 | 16.8777 | 15.2587 | 14.0968 | 17.5545 | 23.0013 |
|  70% |  7.4175 |  7.3740 |  8.1310 |  8.9853 |  8.4995 |  8.1120 |  8.5627 |  7.6457 |
|  80% |  7.4175 |  6.9742 |  8.1295 |  7.7037 |  8.4987 |  8.1117 |  7.6287 |  6.9807 |
|  90% |  0.0003 |  0.1588 |  0.0000 |  0.0037 |  0.0000 |  0.0003 |  0.0135 |  0.1363 |
| ---- |  ------ |   --- 3 |  Attack |  Moving | Average |  ------ |  ------ |  ------ |
|  40% | 54.6732 | 53.5018 | 43.6617 | 49.0445 | 56.5392 | 38.5643 | 40.1080 | 48.6035 |
|  50% | 33.3215 | 22.9238 | 27.4898 | 31.5487 | 35.2305 | 24.1510 | 23.0220 | 22.8960 |
|  60% | 13.8798 |  9.2035 | 10.5063 |  7.7275 | 15.5575 |  8.9612 |  6.3230 |  9.3345 |
|  70% |  3.9423 |  3.9482 |  3.2218 |  3.6018 |  4.8708 |  2.9163 |  2.5517 |  4.0790 |
|  80% |  0.0017 |  0.8727 |  0.2373 |  0.2875 |  0.2022 |  0.3850 |  0.5318 |  0.8918 |
|  90% |  0.0000 |  0.0003 |  0.0000 |  0.0003 |  0.0000 |  0.0000 |  0.0000 |  0.0000 |
| ---- |  ------ |   --- 4 |  Attack |  Moving | Average |  ------ |  ------ |  ------ |
|  40% | 57.7482 | 50.3973 | 47.5403 | 53.0330 | 59.5875 | 41.2575 | 42.6945 | 45.2568 |
|  50% | 29.6090 | 21.8163 | 21.9488 | 27.0672 | 31.3215 | 17.1445 | 17.8478 | 18.7295 |
|  60% |  1.1785 |  5.7568 |  2.4093 |  4.0728 |  2.7228 |  2.7700 |  3.6998 |  6.0033 |
|  70% |  0.1500 |  1.4965 |  0.7970 |  0.7388 |  0.8792 |  1.1280 |  1.2070 |  1.9040 |
|  80% |  0.0013 |  0.1310 |  0.1045 |  0.0710 |  0.1643 |  0.2465 |  0.2515 |  0.2935 |
|  90% |  0.0000 |  0.0105 |  0.0000 |  0.0000 |  0.0000 |  0.0000 |  0.0000 |  0.0005 |
| ---- |  ------ |   --- 5 |  Attack |  Moving | Average |  ------ |  ------ |  ------ |
|  40% | 57.0225 | 47.6938 | 43.5565 | 50.3275 | 59.0108 | 35.0115 | 37.2285 | 39.8558 |
|  50% | 26.1370 | 21.6223 | 17.4095 | 19.3090 | 28.3583 | 13.7083 | 13.7485 | 19.2475 |
|  60% |  7.0750 |  2.4827 |  5.0105 |  5.1677 |  8.5480 |  3.8860 |  3.2785 |  2.7635 |
|  70% |  0.0843 |  0.6472 |  0.2590 |  0.1657 |  0.4012 |  0.3640 |  0.4855 |  0.9257 |
|  80% |  0.0000 |  0.0233 |  0.0037 |  0.0020 |  0.0060 |  0.0248 |  0.0333 |  0.0643 |
|  90% |  0.0000 |  0.0000 |  0.0000 |  0.0000 |  0.0000 |  0.0000 |  0.0000 |  0.0003 |
| ---- |  ------ |   --- 6 |  Attack |  Moving | Average |  ------ |  ------ |  ------ |
|  40% | 58.3780 | 47.0712 | 42.9440 | 48.4048 | 60.3563 | 34.2907 | 32.9735 | 40.4083 |
|  50% | 27.7150 | 16.1460 | 17.9885 | 18.7237 | 30.0160 | 13.3348 | 12.4510 | 13.1953 |
|  60% |  3.1587 |  3.8188 |  2.8743 |  3.9238 |  4.5448 |  2.4330 |  2.2203 |  3.2333 |
|  70% |  0.0003 |  0.2053 |  0.0308 |  0.0473 |  0.0583 |  0.1335 |  0.2328 |  0.4163 |
|  80% |  0.0000 |  0.0017 |  0.0000 |  0.0000 |  0.0040 |  0.0145 |  0.0130 |  0.0165 |
|  90% |  0.0000 |  0.0000 |  0.0000 |  0.0000 |  0.0000 |  0.0000 |  0.0000 |  0.0000 |
| ---- |  ------ |   --- 7 |  Attack |  Moving | Average |  ------ |  ------ |  ------ |
|  40% | 56.0545 | 48.6200 | 43.1815 | 49.0210 | 58.3612 | 34.3670 | 32.6210 | 41.2830 |
|  50% | 22.4383 | 16.2710 | 14.7237 | 16.5923 | 24.9438 | 10.7795 |  9.4425 | 14.2403 |
|  60% |  2.0663 |  1.8787 |  1.5637 |  1.8995 |  3.1532 |  1.3958 |  1.4743 |  2.0515 |
|  70% |  0.0005 |  0.0730 |  0.0762 |  0.0818 |  0.1150 |  0.1095 |  0.0670 |  0.1415 |
|  80% |  0.0000 |  0.0000 |  0.0000 |  0.0000 |  0.0003 |  0.0010 |  0.0013 |  0.0035 |
|  90% |  0.0000 |  0.0000 |  0.0000 |  0.0000 |  0.0000 |  0.0000 |  0.0000 |  0.0000 |

Whoa. The shifting algorithm was giving C/Ma a significant edge before, but turning on Sacred Shield completely reverses that result.  Control/Haste completely blows everything else out of the water here for shorter strings, and not by any small margin.  In the 3- to 4-attack categories, it’s almost a full threshold level ahead.  It’s showing fewer spikes in the 70% category than C/Ma is showing in the 80% category!  It isn’t until we get up to 5- attacks that mastery manages to catch up by virtue of its stronger probabilistic footing.  The more attacks to which you can apply block chance, the better C/Ma perform.

I think that what’s happening here is the following: C/Ha is getting Sacred Shield bubbles every ~4 seconds, which means it’s covering roughly every third boss attack.  Those attacks are guaranteed to not trigger the shifting algorithm, and C/Haste generates holy power fast enough that it can basically always cover the other two with SotR if it needs to, even if the bleed valve engaged recently.  That makes it incredibly strong for short-attack strings.

It falls behind a little in the longer strings because it can only keep this up for so long before running out of gas and needing another ~5 seconds to recharge.  So for the long strings, it loses a little ground to the more stochasting C/Ma set.

C/Ma can’t muster up that same level of HPG, and it’s only covering one in four boss swings with the guaranteed absorb, so it’ll end up struggling for short strings.  But through sheer probability it’s going to be able to brute-force long strings and claw its way back into contention.

This is rather eye-opening, because it suggests that there’s very little point in running Control/Mastery if you’re an advanced tank (using Sacred Shield, of course).  There doesn’t seem to be a category in which C/Ma gives a distinct advantage, and you’d be giving up a lot of DPS as well.  In one fell swoop, it’s upended the paradigm and put us right back where we were in 5.0.

But we have one final simulation to run, this time with the SH2 queue.  We know that favors C/Ma a little, so let’s see if it can pull out a last-minute hail mary.

Results – SH2 Finisher Queue

Again, first we’ll look at the data with Sacred Shield disabled to set our baseline.

| Set: |    C/Ha |    C/Ma |    C/Av |   C/Bal |      Ha |   Avoid |  Av/Mas |  Mas/Av |
|   S% |  0.5223 |  0.4100 |  0.4194 |  0.4523 |  0.4993 |  0.3666 |  0.3620 |  0.3577 |
| mean |  0.6000 |  0.5428 |  0.5319 |  0.5565 |  0.6084 |  0.4942 |  0.5050 |  0.5171 |
|  std |  0.1424 |  0.1250 |  0.1652 |  0.1486 |  0.1410 |  0.1747 |  0.1594 |  0.1415 |
| ---- |  ------ |   --- 2 |  Attack |  Moving | Average |  ------ |  ------ |  ------ |
|  40% | 77.8200 | 65.0427 | 68.2365 | 72.6840 | 79.0070 | 63.8262 | 66.6148 | 62.8997 |
|  50% | 64.6552 | 48.8390 | 45.9310 | 51.0615 | 64.8900 | 37.7113 | 37.6560 | 43.9873 |
|  60% | 45.7122 | 43.0228 | 35.8450 | 42.0607 | 46.6923 | 30.4370 | 32.7540 | 39.3973 |
|  70% | 40.2465 | 19.5655 | 32.0827 | 31.5863 | 41.3385 | 27.4450 | 27.0597 | 20.1478 |
|  80% | 14.1430 | 19.5655 | 14.5213 | 15.6145 | 15.8750 | 14.4053 | 17.0558 | 20.1478 |
|  90% |  9.3812 |  8.5258 |  9.7062 |  9.4170 | 10.5368 |  9.5890 |  9.0388 |  8.4430 |
| ---- |  ------ |   --- 3 |  Attack |  Moving | Average |  ------ |  ------ |  ------ |
|  40% | 81.8700 | 80.9910 | 70.0565 | 77.7767 | 82.9153 | 62.2112 | 66.2630 | 75.3718 |
|  50% | 74.4885 | 59.3250 | 62.3030 | 68.4200 | 75.9533 | 55.2432 | 49.8645 | 54.8380 |
|  60% | 53.8223 | 33.8610 | 41.0710 | 47.0465 | 55.7272 | 35.3563 | 37.0870 | 32.0658 |
|  70% | 36.7032 | 13.9912 | 24.1533 | 28.1390 | 37.4335 | 18.3825 | 15.5112 | 13.3758 |
|  80% | 14.5350 |  1.7413 | 10.9438 |  9.5242 | 15.4828 |  8.9765 |  7.6713 |  3.4228 |
|  90% |  0.0010 |  0.6500 |  0.3152 |  0.1653 |  0.4120 |  0.8832 |  1.0208 |  1.1808 |
| ---- |  ------ |   --- 4 |  Attack |  Moving | Average |  ------ |  ------ |  ------ |
|  40% | 85.7677 | 85.2933 | 72.4490 | 80.5705 | 87.1720 | 66.1072 | 73.5293 | 80.4767 |
|  50% | 76.3633 | 66.6900 | 58.2595 | 65.1142 | 77.4215 | 47.6770 | 51.2200 | 58.7230 |
|  60% | 54.9747 | 31.7698 | 40.8983 | 44.3940 | 57.2597 | 33.2178 | 33.4380 | 27.1323 |
|  70% | 31.8188 |  4.1273 | 19.5360 | 24.5765 | 33.1850 | 14.8280 | 13.6393 |  6.6363 |
|  80% |  0.0035 |  1.2910 |  0.6425 |  0.4238 |  0.9350 |  1.4013 |  1.6883 |  2.7095 |
|  90% |  0.0000 |  0.2180 |  0.0508 |  0.0217 |  0.1497 |  0.4260 |  0.6865 |  1.0378 |
| ---- |  ------ |   --- 5 |  Attack |  Moving | Average |  ------ |  ------ |  ------ |
|  40% | 91.1868 | 87.1790 | 78.0393 | 84.1690 | 91.9443 | 68.5353 | 73.5373 | 80.9253 |
|  50% | 77.6485 | 62.6120 | 62.1093 | 69.7057 | 79.4778 | 53.1682 | 51.7235 | 55.2273 |
|  60% | 56.3043 | 35.5973 | 40.1310 | 42.6025 | 58.6827 | 32.3425 | 28.2125 | 31.0635 |
|  70% | 26.7188 |  6.9745 | 16.5240 | 18.3535 | 28.6548 | 12.6120 |  9.4627 |  7.5395 |
|  80% |  6.8837 |  0.3365 |  4.5732 |  3.4560 |  8.0095 |  3.3715 |  0.8610 |  1.4175 |
|  90% |  0.0000 |  0.0000 |  0.0305 |  0.0113 |  0.1123 |  0.2295 |  0.1958 |  0.2898 |
| ---- |  ------ |   --- 6 |  Attack |  Moving | Average |  ------ |  ------ |  ------ |
|  40% | 92.4952 | 86.9205 | 80.7193 | 86.2888 | 93.3290 | 72.9005 | 76.6795 | 80.3848 |
|  50% | 79.2488 | 66.7635 | 60.8193 | 67.7832 | 80.9382 | 49.7330 | 51.7130 | 57.6740 |
|  60% | 54.5883 | 32.6580 | 33.2920 | 40.7420 | 56.3898 | 24.5072 | 26.2225 | 28.1565 |
|  70% | 22.8665 |  7.3368 | 11.1203 | 13.7167 | 24.2715 |  7.9290 |  7.6668 |  7.7412 |
|  80% |  3.3348 |  0.0128 |  2.4295 |  2.9755 |  4.4640 |  2.0282 |  1.5145 |  0.5995 |
|  90% |  0.0000 |  0.0000 |  0.0000 |  0.0000 |  0.0245 |  0.0643 |  0.0847 |  0.0762 |
| ---- |  ------ |   --- 7 |  Attack |  Moving | Average |  ------ |  ------ |  ------ |
|  40% | 93.3610 | 88.5393 | 81.7302 | 88.7318 | 94.1352 | 73.3020 | 78.4955 | 82.5045 |
|  50% | 79.4857 | 65.5380 | 61.3417 | 69.2993 | 81.4112 | 50.9712 | 52.6560 | 56.6260 |
|  60% | 53.2993 | 29.5740 | 31.6163 | 38.8610 | 55.4382 | 23.1475 | 24.0225 | 25.2393 |
|  70% | 21.3562 |  4.7965 | 10.4565 | 12.9678 | 23.2783 |  7.3540 |  6.5375 |  5.8625 |
|  80% |  2.8672 |  0.2820 |  1.6265 |  1.1868 |  3.7050 |  1.1700 |  0.8225 |  0.6787 |
|  90% |  0.0000 |  0.0000 |  0.0000 |  0.0000 |  0.0047 |  0.0178 |  0.0320 |  0.0480 |

There isn’t a lot new here – this is much the same as the SH1 data.  Control/Mastery dominates, the other Control sets lag by varying degrees, and avoidance tends to lag even further until we look at very long strings of attacks.  Business as usual.

However, turning on Sacred Shield one last time…

| Set: |    C/Ha |    C/Ma |    C/Av |   C/Bal |      Ha |   Avoid |  Av/Mas |  Mas/Av |
|   S% |  0.5229 |  0.4109 |  0.4191 |  0.4524 |  0.4981 |  0.3660 |  0.3617 |  0.3566 |
| mean |  0.4120 |  0.3902 |  0.3673 |  0.3890 |  0.4176 |  0.3380 |  0.3455 |  0.3661 |
|  std |  0.1324 |  0.1276 |  0.1449 |  0.1374 |  0.1337 |  0.1492 |  0.1434 |  0.1368 |
| ---- |  ------ |   --- 2 |  Attack |  Moving | Average |  ------ |  ------ |  ------ |
|  40% | 47.6120 | 50.2428 | 39.8375 | 45.7452 | 48.3047 | 34.8185 | 38.0930 | 46.7488 |
|  50% | 38.9460 | 28.5045 | 32.7732 | 33.3143 | 39.7985 | 27.9272 | 29.1158 | 28.0058 |
|  60% | 14.2005 | 23.6138 | 15.2243 | 17.4722 | 16.0515 | 14.8583 | 18.1448 | 23.1213 |
|  70% |  8.6380 | 10.5908 |  9.2225 | 10.2643 |  9.4262 |  8.9523 |  9.4345 |  9.8748 |
|  80% |  8.6378 |  9.4778 |  9.2225 |  9.2380 |  9.4262 |  8.9520 |  8.6443 |  8.3620 |
|  90% |  0.0005 |  1.7873 |  0.0003 |  0.4998 |  0.0005 |  0.0003 |  0.3010 |  1.0542 |
| ---- |  ------ |   --- 3 |  Attack |  Moving | Average |  ------ |  ------ |  ------ |
|  40% | 54.1732 | 49.7465 | 43.3820 | 47.6243 | 55.8525 | 38.5485 | 39.8533 | 46.8705 |
|  50% | 33.9008 | 29.2475 | 29.0235 | 32.0695 | 35.4320 | 25.4790 | 25.5342 | 27.7323 |
|  60% | 16.0977 | 14.1055 | 12.8713 | 10.5247 | 17.4093 | 10.6113 |  8.2012 | 12.5555 |
|  70% |  5.2683 |  2.6925 |  4.0322 |  5.1090 |  5.8770 |  3.3680 |  2.0065 |  3.3333 |
|  80% |  0.0005 |  0.2215 |  0.0328 |  0.1455 |  0.0620 |  0.1715 |  0.3105 |  0.5557 |
|  90% |  0.0003 |  0.0000 |  0.0000 |  0.0003 |  0.0003 |  0.0000 |  0.0003 |  0.0000 |
| ---- |  ------ |   --- 4 |  Attack |  Moving | Average |  ------ |  ------ |  ------ |
|  40% | 56.8932 | 54.5630 | 48.4860 | 52.9208 | 58.8742 | 42.7167 | 44.7313 | 48.1942 |
|  50% | 32.3667 | 19.4830 | 23.6827 | 29.0388 | 33.4400 | 18.0577 | 19.5975 | 17.5515 |
|  60% |  1.4580 |  3.5452 |  2.5282 |  4.3430 |  2.9897 |  2.8442 |  3.2468 |  4.9950 |
|  70% |  0.0535 |  0.6055 |  0.3357 |  0.3340 |  0.4828 |  0.7567 |  0.7258 |  1.0982 |
|  80% |  0.0003 |  0.0170 |  0.0053 |  0.0010 |  0.0298 |  0.1063 |  0.0865 |  0.0828 |
|  90% |  0.0000 |  0.0013 |  0.0000 |  0.0000 |  0.0003 |  0.0000 |  0.0000 |  0.0000 |
| ---- |  ------ |   --- 5 |  Attack |  Moving | Average |  ------ |  ------ |  ------ |
|  40% | 57.3653 | 47.7255 | 44.5812 | 50.7285 | 59.1245 | 35.9823 | 38.9498 | 40.3257 |
|  50% | 27.4577 | 20.2110 | 19.1030 | 20.9293 | 29.4162 | 14.7647 | 14.4468 | 18.8630 |
|  60% |  7.4005 |  2.2590 |  5.0818 |  5.0638 |  8.5432 |  3.7573 |  3.0122 |  2.1902 |
|  70% |  0.0325 |  0.1300 |  0.1363 |  0.0567 |  0.2250 |  0.2560 |  0.2508 |  0.4215 |
|  80% |  0.0000 |  0.0000 |  0.0000 |  0.0000 |  0.0028 |  0.0095 |  0.0055 |  0.0100 |
|  90% |  0.0000 |  0.0000 |  0.0000 |  0.0000 |  0.0000 |  0.0000 |  0.0000 |  0.0000 |
| ---- |  ------ |   --- 6 |  Attack |  Moving | Average |  ------ |  ------ |  ------ |
|  40% | 58.3780 | 47.0772 | 43.6273 | 48.4960 | 60.4503 | 35.2037 | 34.0155 | 41.4363 |
|  50% | 28.5205 | 17.8475 | 19.2605 | 19.8965 | 30.5623 | 14.0432 | 13.2497 | 13.7620 |
|  60% |  3.3697 |  2.8927 |  2.9888 |  4.3565 |  4.4585 |  2.3605 |  2.0293 |  2.2940 |
|  70% |  0.0005 |  0.0253 |  0.0043 |  0.0600 |  0.0212 |  0.0692 |  0.0895 |  0.1422 |
|  80% |  0.0000 |  0.0000 |  0.0000 |  0.0000 |  0.0015 |  0.0053 |  0.0030 |  0.0045 |
|  90% |  0.0000 |  0.0000 |  0.0000 |  0.0000 |  0.0000 |  0.0000 |  0.0000 |  0.0000 |
| ---- |  ------ |   --- 7 |  Attack |  Moving | Average |  ------ |  ------ |  ------ |
|  40% | 56.1492 | 49.2988 | 43.4217 | 48.6828 | 58.2107 | 34.9405 | 33.4910 | 42.2753 |
|  50% | 23.0362 | 16.5573 | 15.8103 | 17.8190 | 25.2770 | 11.2423 | 10.0868 | 14.1870 |
|  60% |  2.5048 |  1.5065 |  1.6330 |  2.2555 |  3.3623 |  1.3305 |  1.1208 |  1.4793 |
|  70% |  0.0005 |  0.0107 |  0.0135 |  0.0200 |  0.0460 |  0.0570 |  0.0275 |  0.0548 |
|  80% |  0.0000 |  0.0000 |  0.0000 |  0.0000 |  0.0003 |  0.0005 |  0.0000 |  0.0005 |
|  90% |  0.0000 |  0.0000 |  0.0000 |  0.0000 |  0.0000 |  0.0000 |  0.0000 |  0.0000 |

Again, we see Control/Haste dart out to an early lead in the short-string categories. But this time starts falling slightly behind Control/Mastery in the 5-attack category, and never quite regains the lead. This is still not a banner showing for Control/Mastery, mind you, but at least it didn’t lose in every category this time. I guess we can give it a medal for participation?

It’s worth commenting on the other sets here as well. The other Control sets still lag a little, but again aren’t that bad. The avoidance gear sets actually compete pretty well at lower threshold levels in most of the categories, often winning by a pretty wide margin. But like usual they can’t suppress the largest spikes the way a Control set can. And the mantra we’ve been repeating is “Spikes kill tanks,” so we have to consider those high-damage categories as a little more important. All in all, we’re still not very impressed with the other offerings.

Conclusions

It looks like Control/Haste has a pretty commanding lead when Sacred Shield is turned on, even though it lags significantly when Sacred Shield is absent.  Unless we’re really concerned with 6- and 7-attack strings, there’s a decent bit of survivability to be gained by sticking with haste.  And of course, there’s the DPS angle to consider, which is nontrivial for raiders in a variety of categories; from 10-mans raiders where tank DPS is a large percentage of raid DPS to heroic raiders looking for that last little bit of juice to meet the enrage.

What surprised me most about these results was just how powerful Sacred Shield is.  I’ve been touting it as the obvious choice in that talent tier since release, but these results really hammer that point home.  Sacred Shield is essentially a floating “smart SotR” every six seconds that automatically adapts to your damage intake pattern.  It’s steady, can’t overheal, won’t be wasted when you get a lucky avoid, and best of all free because it only costs a GCD.  It’s the damage smoother’s dream talent, and there’s literally nothing else in the tier that can compare.

Once you recognize the power of Sacred Shield, the dominance of haste in these simulations becomes more understandable.  Getting more of these “smart SotRs” is a huge benefit for a tank that adapts their actual SotR usage to intake patterns, because they can often guarantee coverage of every melee attack.

That said, we see that even with the simple S queue, Control/Haste pulls ahead.  Not only does it provide better active mitigation when you’re shifting, it provides better passive mitigation through Sacred Shield and “lazy” SotR usage.  Sacred Shield single-handedly turns Control/Haste from an expert’s gearing strategy to one that a novice will be able to use effectively.

And that’s… well, to be honest, it’s a little overpowered.  Broken, even.  As much as I like the haste gearing strategy, it does seem like it’s too good to be true.  Maximum DPS in conjunction with maximum survivability through a stat that is not traditionally desired by tanks is a bit lopsided, and part of the reason we saw changes to Grand Crusader in 5.2.  Were the developers blinded the same way we have been by not including Sacred Shield in our models? I don’t know, obviously, but it seems strong enough that I could believe it wasn’t intended.

And of course, there’s an even more important question from our perspective:  is this interaction is something that, now that the cat is out of the bag, will end up drawing their attention and call down the nerf bat?

And the corollary: can we stack enough haste to outrun it?

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

### 60 Responses to Is Nothing Sacred?

1. Newsom says:

Shhh! Don’t tell Blizzard!

As someone who does 10s where I can afford to really stack haste (closing in on 30% melee haste unbuffed now), I have definitely seen this in action for a while now. In heroic ToT my sacred shield alone usually accounts for well over 20k AbsPS.

I’m often close to healers in HPS output with Battle Healer – so much that our holy paladin complains loudly when i try to tank with Truth. Feels a bit overpowered, yes.

• Daishan says:

^^I’ll send you cookies if you hide the blog from Blizz somehow

• Zao says:

I’m sorry, but if you tank with truth on you deserve to be yelled at.

2. Weebey says:

I was smacking my forehead throughout reading this post. We all knew, from simply looking at logs or doing the most basic theorycrafting, that Sacred Shield accounted for an enormous amount of healing, and yet it didn’t occur to me until reading this that this could have an impact on our gearing strategies. D’oh. In any case, great work as usual.

Another, similar but subtly different effect that could be worth considering is Seal of Insight. The non-overheal healing it provides is, if I recall correctly, actually greater on average than Sacred Shield (per Theck’s Matlab thread, it seems to be almost double). Of course, you would need to take overhealing into account; logs seem to indicate a figure of around 50%, but it’s harder to know what the “effective” overhealing is (e.g. how much of this was simply turning e.g. healer hots into overhealing.) Perhaps 60% is a reasonable number for these purposes, to take a wild guess?

SoI should further improve Haste sets, since SoI healing scales almost perfectly linearly with haste. Unlike Sacred Shield, there is no synergy with avoidance; indeed, higher avoidance is actually likely to INCREASE the overhealing %, although that effect will be difficult to detect in this model, since there is no health.

In any case, thanks as always for the great work. I had actually shifted over to mastery reforging, and am logging on as soon as this sentence is finished to shift over.

• Theck says:

SoI is a weird case. Haste certainly increases SoI and Battle Healer healing. But it’s also a random proc (20 PPM), so you aren’t guaranteed a SoI heal on each melee attack. I could model it the same way that I model WoG, as a short-duration absorb bubble (iirc I used 3 seconds for WoG, which was probably being too generous). But I really don’t like doing that, because I think it vastly overestimates the value of healing. One point of absorption is generally worth a lot more than one point of healing once overheal and healer reaction is factored in.

The other issue is that the way we handle overhealing is pretty lazy – if we estimate 50% overheal, I’ve just been halving the amount of absorption you get from WoG. If I were being more rigorous (and this is something I may consider changing in the simulation), I would make it more like avoidance. You have a 50% chance to get the full amount, and a 50% chance to get zero. That would introduce some of the spikiness that’s characteristic of accidental overheal, which would probably knock the T15 2-piece bonus down even further.

An even more rigorous way to handle it would be to use a continuum of absorb sizes and a probability density function that governs how much you get. So a simple uniform PDF would give you an equal chance of getting a heal bubble that’s 28% the maximum size or one that’s 64% of the maximum size, and so on (assuming 1% bins here, probably finer than is necessary). In practice I think that a uniform distribution isn’t going to be accurate though – it’s probably highly skewed towards the high and low ends, such that a function like P(x)=12*(x-0.5)^2 (which is normalized on 0<x<1) would probably be a better approximation.

3. Zoroth says:

Which begs the question with all haste scaling of periodic effects: are the breakpoints for extra ticks (30% and 50% haste, if I recall correctly from your previous post) significant targets? Or, does the gain in faster ticks make more of the difference?

• Meloree says:

The breakpoints are only significant for DPS (and even then, it’s a minor effect) due to using less GCDs on SS. They don’t pay an additional premium in survivability.

• Newsom says:

My shield seems to reapply every ~4.6 seconds (tested with a weakaura just now, I have around 28% haste) so there shouldn’t be any hard breakpoints. Correct me if I’m wrong.

• Theck says:

There *are* hard breakpoints – 10%, 30%, 50% etc. That’s where you get new ticks of the buff (i.e. going from 5 bubbles to 6 bubbles @ 10% haste). However, they’re all but meaningless for us in practice.

If you’re refreshing SS early (i.e. any empty GCDs go towards refreshing it), then you have nearly 100% uptime and those extra ticks doesn’t make a difference whatsoever. In practice, it will have a small effect because it means you need to refresh SS slightly less often, so in the rare situation where it fell off because you were too busy pewpew-ing it will lead to slightly less downtime.

But all in all, it’s a very small effect, and not worth trying to adjust your haste rating around. More is always better.

4. Daishan says:

Brilliant work as always Theck
I’ve been feeling for quite a while that Sacred Shield has a large impact on our survival and haste benefits it a lot, very nice to see the numbers back that up, having an even bigger impact than i suspected.
Would a fast swinging boss or many adds reduce the haste bias that SS brings to your simulations?

• Theck says:

Good question! I can think of two effects that faster-swinging bosses would have. The first is that mastery should perform better over a fixed period of time, as it has more chances to trigger a block. Note, however, that this would not manifest itself in the tables above, which are naturally normalized to the boss attack timer (since we’re counting consecutive events).

It should also affect how SotR is perceived though. That high-powered SotR from Mastery can now cover more than two boss attacks, which might give it a decent edge. On the other hand, haste’s higher uptime will also cover more swings. I think which one of those two effects ends up dominating will depend on exactly how the SotR timing is being performed – if we time the cast to be milliseconds before the boss swing, Mastery will see a bigger benefit than if we don’t time it that tightly (and thus only cover 2 boss attacks instead of 3).

Many adds will also tend to fall in favor of Mastery, i would think. The absorb bubbles would get used up quickly, so you’d end up in a cycle of high and low damage periods based on SotR uptime. Mastery’s higher block chance should tend to smooth those a lot more effectively than the haste set will.

5. BentBlyant says:

Great post, thanks! Do you feel like the value of haste should be increased slightly in the default weights for Control/Haste?

• Theck says:

If you’re talking about AMR, not particularly. Those stat weights are sort of arbitrary – they’re chosen to encourage certain gearing patterns, so they’re not rigorously based on a numerical relationship between the stat and, say, damage prevented or smoothness. The C/Ha weights on AMR are set so that it prefers to gem haste/stam in yellow slots, and will generally prefer haste/mastery gear to dodge/parry gear unless it’s a significant ilvl upgrade. It still accomplishes that fairly well as-is.

6. Kihra says:

Great post as usual. Before you count Control/Mastery out, though, I really think you need to theorycraft the concept of the “big special” and work it into your model. Especially this tier, it has become a theme, and it is almost always these “big specials” that occur at the same time as melee attacks that lead to spike deaths.

I’ve worked on four Heroic bosses so far, and all of them have a special attack that combines with melee to create potential spikes, with the incidental melee attacks that don’t occur around the special being pretty inconsequential (or being coverable by cooldowns, e.g., Horridon last phase, big angry green head on Megaera).

Jin’rokh – Static Burst magic damage can happen at the same time as melee.
Horridon – Triple Puncture. Unavoidable but mitigatable by SotR.
Tortos – Snapping Bite. Avoidable, not armor-mitigated, but mitigatable by SotR.
Megaera – Breaths. Magic damage.

While I embrace Control/Haste for the additional damage it provides, I feel like Control/Mastery is not getting a fair shake as long as you don’t theorycraft different types of specials. If DPS checks were not a concern, I’m fairly certain Control/Mastery would be a superior choice for the first half of this tier given that the spike damage is telegraphed.

I even think the 2pc bonus gains in strength with telegraphed spikes, and have found it to be much more powerful than I was expecting on these first few bosses.

• Theck says:

As I said on twitter, I think that Control/Mastery will pull ahead for smoothness in that situation (big predictable spikes). However, there are a couple points to consider about that. First, if the spike is infrequent, we may just cooldown it, making it a non-issue. If the spike is frequent (like most of the bosses you’re mentioning), such that we’re expected to use SotR to mitigate it, the question becomes “how much SotR is enough SotR?”

Mel probably could (and probably would, if you ask him to) give you his thoughts about cooldown planning, but in short – what matters is that you have a sufficient cooldown, not necessarily the largest cooldown. If a 40% SotR is more than sufficient to make the telegraphed attack survivable, then it may still be better to have that extra uptime on either end of the boss attack to smooth out the entire damage event.

Finally, it’s worth noting that healers matter. That’s an obvious statement, but we don’t include healers in this sim. And healers react differently to large, predictable spikes of damage than they do to random fluctuations in the throughput damage from boss attacks. If your healer is properly pre-casting or pre-shielding for the big burst event, then that extra mastery might just be creating extra overheal on their part. That’s partly why I haven’t put it in the sim (yet – random spikes and casts are something that I do want to add eventually, but as always, it’s a matter of having the time). Is it truly realistic to model a predictable, telegraphed spike without modeling your healer’s reaction to that spike? Probably not, and it’s a valid criticism of the sim as-is, but I think the discrepancy will becomes more significant as the magnitude of that predictable spike gets larger.

7. Wrathblood says:

• Wrathblood says:

Ok, had a minute to think about it and the only thing I can really add is that, if Haste is arguably OP now (without, as a poster above noted, counting SoI which is big for Haste) imagine how good it was *before GC was changed*. If anything, we’ve been underestimating how good Haste has been for us all along.

For context, I don’t think Prot Pallies are particularly overperforming in terms of survivability or raid dps, so we don’t necessarily need Haste outright nerfed. However, if their intent was to make the other stats more competitive with Haste, they were successful in only the most marginal of ways.

I’m not personally a fan of DP as a talent, but someone who did would be getting even more value out of Haste because you’d be getting 25% more procs than with Mastery. With a 50%+ ShoR uptime, it does occur to me that maybe I should give DP some more thought.

• Wrathblood says:

Avoidance is pretty powerful, isn’t it?

• Theck says:

Can you elaborate? It performs pretty well in certain categories, but not well in others.

• Wrathblood says:

Oh, just survivability. For both the shift rotations, Avoidance is 10-30% better on almost all of the spike categories than either C/H or C/M while also taking 13-18% less damage. There are a few areas where its comparable to or worse on spikes, but they tend to be few and far between.

Its hard to think of a single target fight where you get hit hard enough for Avoidance to be the way to go for survivability given the enormous loss in tank dps. However, given the interaction with GC and the smoothing effect of tons of melee blows, I’d feel comfortable saying that Avoidance is the defensive stat to go with for tanking a challenging AoE fight. My main set will be C/Haste, but I’m going to have an Avoidance set on the side in case the trash from Mount Hyjal ever comes back.

• Theck says:

Eh, I’m not sure I agree about avoidance. It generally does better in the bulky categories (40%-60%) where you’re guaranteed to have a lot of events anyway. But at least in the Sacred Shield data, it tends to permit more 70%-90% spikes, often by a factor of 10 or 100. For example, SH2 5-attacks, Avoidance halves the 50%-60% numbers but permits 10x as many 70% events and a nonzero number of 80% events.

That said, for a challenging AoE fight I completely agree with you – Avoidance will be very strong.

• Wrathblood says:

Well, Haste wins some of those categories by an order of magnitude but the numbers are so small as to be irrelevant. Like 80% on 5 swings. Haste is zero while Avoid is 0.0095. Sure thats a win for Haste but its the difference between an event happening never and it happening once per month. Neither is really relevant.

The only real wins for Haste that matter are 4 swings
60% and 70% and 5a swings 60%. Everything else is a win for Avoidance, a tie or so small as to be irrelevant.

• Theck says:

Yeah, DP has some interesting effects in the high-haste regime.

8. Scoutyou says:

There is 1 department that haste lacks in, and that is bosses that have these predictable nukes.

Horrdion tripple puncture
Tortos Quake/Snap
Megaera Breaths (BoG is a beast there)
Ji-Kun Talon RAke
Durumu Whatever that thing is called
Primordius front cleave thing
Lei-Shen nukes
Qons spear

Though I still agree that haste is the strongest stat, especially in 10 man, just that a lot of the power in mastery is being overlooked.

• Theck says:

See my response to Khira above – I do think that mastery has the advantage (at least numerically) for mitigating that predictable spike. But it’s also not clear that having a bigger SotR for the predictable spike is always strictly the better scenario. Mitigating 60% of the big spike but 0% of the 3-4 melees around it may not be as good as mitigating 40% of the spike and 2-3 melees, leaving only one potentially un-mitigated attack go through.

• bryjoered says:

I’m not sure, not math behind this, but I was tanking Tortos last night and let me tell you if you don’t block it it hits you for about 400-500k. I’d rather take some melees in between than take that spike unmitigated. I’m sure you’ve tanked Tortos and understand, just saying though.

9. Zao says:

I think that you will find in a majority of cases that DP is a better talent overall. Higher dps, significantly more periods of mitigation, while random, can be timed with predictable attacks like horridon’s TP. Tried this last night on heroic using the strategy to tank him the entire fight and even at high stacks(14+) the incoming damage was laughable; I almost always covered his tp and the next melee attacks, allowing insight and raid healers enough time to bring me up from wherever his attack brought me to. Coupled with SS absorbs his special and melee swings are all huge joke.

• Kihra says:

Divine Purpose is certainly fun, but Holy Avenger gives you more survivability control. It’s an extra tanking cooldown that has 100% reliability. For example, on the last phase of Heroic Horridon, when he’s hitting for 500k melee swings, HA is pretty amazing.

I’d be curious to see the math regarding the DPS difference between DP and HA. I’m not convinced it’s large enough to be worth giving up on the control that HA provides, unless you’re doing easy content or parse chasing. Given that HA use also tends to coincide with higher Vengeance periods, I’m not really sure DP is much of a DPS gain.

10. David says:

@Zao: You do know that with Divine Shield and Hand of Protection you can (and must) reset your stacks on all gates except for the last since it does not come out of CD, so you say you tank him with 14 stacks+ and it’s laughable! I laughed that you stress your healers with no purpose and unnecessary… great use of your class….

• Zao says:

@David I’m just sorry that you haven’t been playing your class to the highest potentials. With a disc priest sitting on Horridon there is no “stress” added on the healers, that’s funny that you think there is. Also, take a look at logs before you assume someone is doing something wrong. The majority of people logging in the top percent on that fight are following this strategy. I’m not sure what difficultly or raid size you are used to, but tank dps does matter, and anyway you can safely squeeze more out should be attempted.

• blizzhoof says:

I’m doing H Horridon next week on 25 and if I can go this high with Triple Puncture, then I’d like to. Are you doing 25s or is the damage you’re referring to from 10s? Also, do you mind linking the log so I can have a look?

• Daishan says:

I’m sure it’s possible to go that high on 10hm, I went to 12 or something myself when I used LoH by mistake and had forbearance, but I don’t see the need if you spec clemency you can reset at 6-7 stacks every time.

• Zao says:

You do it purely for the raid dps/hps.

• Zao says:

I was doing 10 man(we only did 3 pulls because we wanted to work on Ji kun on heroic, but you can factor your theoretical maximum by doing the following:

Triple Puncture*stack size*armor reduction*Shield of the Righteous mitigation

The highest I got before the raid wiped to the 2nd door’s adds was 14 stacks, but even at that level I only lost about half my health pool.

On our normal kill I think I reset at 21 stacks, but i’m not sure(I only reset because I didn’t want to have a high number going into the soft enrage, the damage was really easy to deal with as long as I had 5 HP with a generator up as a TP was coming in): http://worldoflogs.com/reports/rt-my8j0dlt9dydv0nd/spell/136767/?s=7660&e=8168

• Theck says:

I’m not sure what data you’re using to draw that conclusion. Holy Avenger will put out more DPS for most players – DP should pull ahead at very high haste levels, but I’m not sure exactly where the breakpoint is.

• Zao says:

I’d hate to pull the blizzard defense, but that may be because HA has been touted as the best talent choice for so long. The nature of the talent can work for and against you, but having higher haste (which we’d need to determine the break point for) allows you a chance to proc DP more frequently than was probably intended. Not to mention that periods of proc can and do coincide with high vengeance.

You have to also look at it from this side: not all people are going to use HA when it’s off cool down. Most wait to time it correctly with some type of mechanic that they want to mitigate more heavily. In those cases it is obvious that DP would result in more smoothed out damage than a banked HA would.

• I don’t really think this is that compelling an argument as you think it is. While higher haste gives you more DP procs, higher haste also allows you to fit more holy power generators into Holy Avenger’s duration, so it’s not as if you’re getting zero benefit for HA from haste.

While DP might pull ahead at very high haste levels (I’m not really sure why it would, so maybe Theck can explain why he thinks that) I expect that the difference between the two would mostly be determined on a fight to fight basis. The bigger the Vengeance differential between when you pop HA and when you don’t, the better HA looks; the more your Vengeance is uniform throughout the fight, the better DP will look.

Heck, if you’re tanking a huge damage phase for 30 seconds every 3 minutes, Sanctified Wrath looks pretty good.

• Theck says:

DP will definitely pull ahead in terms of DPS. It scales very well with haste, while HA doesn’t. HA is a fixed time window, so while you’ll get a few more 30%-buffed HPGs during the fixed duration, you will get a lot more DPS from the extra SotR’s over the course of 2 minutes at high haste levels.

The key is that you need high haste levels, though. At 18% haste, HA still holds a DPS lead over DP in steady state (see http://maintankadin.failsafedesign.com/forum/viewtopic.php?p=718671#p718671). The break point where DP pulls ahead is probably in the 22%-25% range.

Also note that this ignores interactions with AW. Combining AW and DP together likely pushes the effective breakpoint much higher, since you can line up every other AW with DP for a multiplicative boost.

• blizzhoof says:

I don’t really think HA gets banked much by good tanks. I personally plan my cooldowns and externals around HA because it’s probably the best defensive CD we have (duration wise). The only time I can think of not using it when it comes off CD is when I’m not tanking and DP is useless then too due to a lack of Vengeance.

The only situation that I can see DP being better in is when tanking a boss the whole time by yourself. So, I can see it being stronger on Tortos and then one-tank strats for Horridon and Iron Qon. Even if you one-tank Durumu, there is no way it’s good enough on him due to Disintegration Beam giving you quite a long break. Actually, there’s probably a good argument against using it on Iron Qon for the same reason. For two-tank fights, it can’t even be close. Don’t forget the 30% damage increase HA gives you.

• Zao says:

HA was banked in a few kill videos by the top raiding guilds for the majority of the fight (as in it would have been off cool down by the point they decided to use it). Hadn’t thought of one tanking Durumu, but I was doing something similar with the stacks already. Thanks for the idea!

• Theck says:

Yes and no. If you’re banking HA, then you’re implicitly saying you care more about mitigation during a specific section of the fight than smoothness overall. Having smoother damage intake during the entire fight (including irrelevant portions) isn’t necessarily as good as having smooth damage intake during the sections that do matter.

Also note that almost every fight makes use of two tanks nowadays, so you will spend about 50% of your time not tanking anything (making DP and SotR coverage irrelevant). That weights the dice heavily in favor of HA, because you can control it to make sure the 18 seconds of coverage occurs during the ~60 seconds of the cooldown you’ll be tanking instead of the ~60 seconds you won’t be. Given that you actually get about 25-30 seconds of SotR coverage from a single HA cast, you can have almost 50% effective uptime on SotR if considering only the parts of the fight you’re actively tanking.

You may very well be right about the “blizzard defense,” but I think there are also a good number of compelling reasons to take Holy Avenger instead of DP. Both have their pros and cons, but I think HA has more pros than DP does.

11. Two questions:

1) About the nuts and bolts of your simulation: are you taking absorb “munching” due to avoidance into account? For example, if your Sacred Shield drops to ticking once every 4.05 seconds, two avoided boss swings in a row means a wasted absorb, and if I’m not mistaken even an avoided hit + a hit that is reduced by both blocking+SotR will result in a partial munch.

I wouldn’t necessarily expect that to alter the results really, though (maybe it would knock avoidance down a bit?) Mostly curious.

2) How does this change at different Vengeance values? Is it basically proportional (with Sacred Shield pretty much always translating to ~40% of the value of an incoming swing at 1.5s swings?

If so, that strikes me as basically insane, and probably illustrates that this is a problem with the talent and not necessarily with haste. That’s a huge percentage of your incoming damage stopped by an easily-maintained buff even WITHOUT stacking haste, and it’s hard to imagine that Selfless Healer or Eternal Glory could ever really compete with it no matter how much healing you have them doing.

• Theck says:

1) Yes, it does. It creates an absorb bubble with a duration equal to the tick interval, and eliminates the bubble when the duration expires.

2) It does vary by a fair bit, but probably not the way you expected. It actually gets a little weaker with higher Vengeance. That makes more sense if you think about the limits – if the boss hits for 1 damage and gives you 0 Vengeance, you’d fully absorb the attack.

That also gives us the interesting corollary: SS is even *more* powerful in 10-man, where you’re likely to have less Vengeance. Each absorb is smaller, obviously, because you have less Vengeance. But those smaller absorbs are a larger percentage of a boss swing (at ~60k vengeance, it’s 50% of a boss swing).

It’s pretty simple to find the mathematical relationship between the SS absorb (in percentages of boss swing) and vengeance. Here’s the short version:

${\rm SS~Absorb} = 5879 + 0.39 \times AP$
$AP = AP_0 + V$
$V = 0.36\times \frac{\rm boss~raw~swing~damage}{1.5}$
${\rm boss~swing~damage} = {\rm boss~raw~swing~damage}\times M$

Here $M=0.362$ is the mitigation factor for Weakened Blows, armor mitigation (65k armor), and spec mitigation (Sanctuary).

Solving those equations for ${\rm SS~Absorb}/{\rm boss~swing~damage}$ gives you:

${\rm \% Absorb} = \frac{5879 + 0.39\times AP_0 + 0.39\times V}{4.1667 M V}$

When plotted, that looks something like this:

As you can see, it generally hovers between ~45% and ~35% of a boss attack, leveling off at around 30%. It eventually saturates to 25.86% in the infinite-Vengeance limit.

• Thanks for the math on scaling, that’s interesting.

• Weebey says:

Sorry to be a bother, but where is the documentation for the SS absorb formula? The WoW Db has 343+1.17*SP=343+.585*AP, see http://www.wowdb.com/spells/20925-sacred-shield. Some quick in game testing in a few different gear sets seems to confirm that formula is at least close. I tried looking through the Call to Arms thread, but I didn’t see it there.

• Theck says:

I’m sure it’s buried in the CtA thread somewhere, but it’s possible that it has changed since the last time we tested it. I don’t trust WoWDB tooltips in general, because they’re frequently wrong. But this might be a case where the spell was modified to match the tooltip instead of the other way around (i.e. tooltip was intended, spell was accidentally using a different formula at the time of our testing, and was subsequently corrected server-side).

What I do trust is data – if you have the combat log files you can e-mail them to me or host them somewhere and I’ll upload them. otherwise I’ll try to take some data later today to confirm.

That said, if your formula is correct, it’s going to be a fairly significant *buff* to SS….58.5% AP scaling instead of 39% is going to be huge when you apply 100k Vengeance (almost 20k more absorption per tick, which more than makes up for the nerf in base absorbtion). It will also flatten out the decay curve considerably, making it more even across AP levels.

• Theck says:

Seems that you’re correct, we’ve been undervaluing SS. The exact formula (down to the last point of mitigation) seems to be round(342.5+0.585*AP).

Corrected formula for % absorb should be:
${\rm \%~Absorb} = \frac{342.5 + 0.585\times(1.1\times(AP_0 + V))}{4.1667MV}$

Where I’ve also included the 1.1 modifier for the raid AP buff. Now it looks like this:

Which means we’re absorbing upwards of 50% of a boss hit at a time.

• Weebey says:

I don’t have a WoL account, so I can’t upload to show other people, but recount at least agrees with the tooltip/buff numbers for Sacred Shield (which in turn agree with the WoW DB formula.)

As a sanity check, I went to Hyjal and took off all my gear. That left me with 301 SP, and a tooltip SS absorb of 695 (which agrees with 343+1.17*301). I then fought one of the rock elementals. Melee attacks without SS were ~1100 — 1400 damage. When I put SS up, a single attack broke it, and seemed to do around 700 less damage. The formula Theck posted above would imply a SS of 6131 at this SP, which would be impossible for the elementals to break in a single hit.

Soit certainly appears that this is not simply a display error, and that the WoW DB tooltip is actually correct.

12. Zao says:

While this has absolutely nothing to do with damage mitigation, I would like to know why we are able to pop ret kings on the pull of a boss. It seems almost broken as it can add upwards of 40k dps in the first 20 seconds of a fight.

• Jackinthegreen says:

Do you mean Ret’s Guardian of Ancient Kings? In practice it’s probably going to be less than 20 seconds because of having to put up SoI and RF too not to mention a bit of running for the pull where melee isn’t happening.

As for why it’s possible, Blizz simply hasn’t gone in to make swapping specs automatically stop all buffs and such. If they think it’s a problem then chances are they’ll fix it. Perhaps you should post on the official forums and provide a log of you doing it to get their attention?

13. Bram says:

What is a bit sad how C/Bal set is under-performing both C/Ma and C/Ha. Not good news for those tanks who would like to change gearing strategy.

14. Paendamonium says:

So we are left with a question…. What are the possible actions Blizzard could take to try to bring the different stats into more of an equilibrium?

Now this is just pure conjecture on my part, but I think that Sacred Shield aside, Blizzard is probably pretty happy with how the different stats benefit us. Haste is something that they wanted tanks to benefit from, and as a dps stat it makes sense that it provides a larger dps boost but a smaller survivability boost compared to mastery and avoidance. I think your modeling sans Sacred Shield, which showed mastery (and even avoidance to a lesser extent) catching or overtaking haste for survivability purposes was what Blizzard was hoping for….. But they overlooked Sacred Shield.

So what would bring Sacred Shield in line with Blizzard’s stated goal of having mastery and avoidance stats still be at least equally desirable for survivability purposes alongside haste? The solution that immediately comes to mind is to make Sacred Shield unaffected by haste, but I don’t know if they would want to also change it for ret and holy paladins…. Theck I really appreciated last patch when you laid out different solutions for making haste less desirable (which Blizzard followed!) so I’m very interested to see your thoughts. And everyone else’s too!

15. Wrathblood says:

Theck, I have a question and I’m not sure how difficult it would be to model. Haste, we as know, is more dependent on skill of play than other gearing options. If you mash your keys slower, all gearing options suffer, but Haste suffers the most. Folks reading here likely aren’t among the folks who’ve got ShoR macro’ed to CS, but I’d argue that almost none of us (unless some posters are among those fighting for world firsts) are getting 95-100% of our potential. In fact, I realized that my guild generally only averages about 85% of Simcraft results. I’m probably in the same boat.

How badly hurt is Haste if a person is only playing at 85% instead of 100%?

• Theck says:

I think it depends on the failure mode you’re considering. “85% of potential” is a little too vague to make specific conclusions. For example, if you play with 200ms latency (network or reaction), then both sets might suffer equally, and there wouldn’t be much difference. On the other hand, if you cast on a ~1.5-second GCD rather than adjusting for haste (i.e., instead of pre-casting or spell queuing, you’re waiting and creating dead time between casts) then haste’s benefit will plummet compared to mastery.

Trying to code the sim to play poorly is a bit of a pain because you need tho choose specifics like that. At the worst, I think playing at 85% could be approximated by just giving the gear set 85% as much haste rating as we should. But I think that would be a gross over-estimate, because poor play would affect mastery as well. So the proper way to do it would probably be to include latency-like effects or a RNG “oops” clause that sometimes forces you to use spells out of order or delay by half of a GCD.

However, I’d argue that 95% efficiency is not that hard to hit for a hard-mode raider, so I don’t know that it’s worth simulating it in that much detail.

• Wrathblood says:

Oh, sure, and that’s why I was pondering how you’d even do it. I mean, sure, no one is hitting the theoretical levels for ShoR uptime for most fights, but its not because they’re bad. Its because they’re pooling and possibly even wasting HoPo while they’re OTing so they can maximize ShoR uptime while they’re on the boss rather than maximizing it globally.

However, the varying impact of different stats was kind of the heart of my question. Playing suboptimally might give you only 85% of the value of your haste (and I agree that its probably low. While I may be only 85% efficient overall, I’m probably better than at hitting ShoR during the phases when I’m on the boss), but you’re likely to get a higher value than that of your Mastery and I’m curious by how much it varies.

16. queldan says:

Hey Theck! Quick confirmation – you mention using SS for every empty GCD. So do you imply that SS is only cast when no offensive abilities are available, even when SS is down? Or is such an occurence rare enough to be a non-event?

In short – with SS looking so strong, are we safe to assume it’s *really* important to keep as close to 100% uptime as we can afford?

• vehlin says:

You REFRESH it on an empty GCD, you’ll get an empty GCD between first cast and the 30s buff dropping off. The only exception to this is if your vengeance does a massive spike up or down (up you refresh sooner, down you let it run its course). Aim is to maintain as close to 100% uptime as is practicable.

So in practice: Pop with 2 secs on the pull timer, refresh in your first empty GCD and each subsequent one after that if it makes sense based on your vengeance.

• Theck says:

In this simulation, I’m only modeling four GCD-bound spells: CS, J, AS, and SS. So in a sense, it’s following the priority CS>J>AS>SS>(everything else). That ensures 100% uptime on Sacred Shield, but isn’t strictly the same as “use SS when all of your offensive abilities are on cooldown” because of HW and Cons.

However, it’s powerful enough that yes, I recommend you do something similar to this in practice. You certainly don’t have to use SS every time it’s off cooldown – it’s silly to use it 6-7 seconds after your last refresh. But I would make sure to prioritize it ahead of fillers when there’s less than 6 seconds left on the buff. You often won’t even get to that point, as we do have an occasional empty GCD that we can use to refresh it, but if you get lucky with Grand Crusader procs that’s not guaranteed.

17. Pingback: Sacréd bleu! | Sacred Duty