Control Shift En-tier

About a week ago, Wrathblood asked me to take a more thorough look at the Tier 15 2-piece bonus. The earlier simulations I ran were based on a lower-ilvl gear set (496) which was more constrained in its itemization, and only considered the Control/haste gearing strategy. He suggested that the extra itemization afforded by higher-ilvl gear may have effects that we didn’t anticipate, as might the more extreme values of haste and mastery we’re pushing in the ilvl 522 gear sets because of the relaxed parry/dodge constraints. And of course, it’s not hard to imagine that a Control/Mastery build might interact more strongly with a block-based set bonus.

Even if the results don’t end up making the 2-piece an amazing bonus, it’s good to cover more of the parameter space. Better to have data and know what you’re dealing with than to guess and make blind assumptions, especially when the data isn’t hard to generate.

In addition, Mel and I have been discussing a few of the more significant limitations of our earlier simulations. One of them was our SotR model. It’s not clear how exceptional you need to be to effectively pull off the time-shifting hijinks that we suggested in the 5.2 Smoothness post. But it’s also probably not realistic to model it as if it were macro’ed to Crusader Strike. It would be tough to come up with a happy middle ground, but it’s not that hard to make some basic tweaks to improve the SotR usage model. So that’s exactly what I did.

I’ve introduced a new “Shift” finisher queue that tries to intelligently shift the usage of SotR around to react to spikes. The queue, which we’ll call “SH” will cast SotR when either of the following conditionals is satisfied:

  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 N boss attacks is greater than 80% throughput.

The first condition is the bleed valve, and taken in isolation gives identical results to the simple “spam SotR whenever you have 3 holy power” model we’ve been using in previous sims, also known as the “S” rotation. It’s only purpose is to make sure we don’t end up wasting holy power, and bleed it in the safest way possible.

The second condition is the one that acts as a spike filter. Maximum throughput in the simulation is a string of full hits, which looks like (1, 1, 1, 1, 1, 1, …). This filter just looks at the last N attacks, takes the average, and compares it to 0.8.

So for example, if I set N=1 (an “SH1″ queue) it filters based on only the last attack, and will use SotR (if possible) any time we take a full hit. It wouldn’t use SotR if we blocked the previous attack, because that only registers as 0.7 damage, nor if we mitigated it with an existing SotR or avoided it. It doesn’t look at any attacks beyond the previous one, however. You can probably already guess that this means it’ll be a short-spike filter.

If instead we used N=2 (“SH2″), then it would look at the mean of the last two attacks. So for example, it will try to cast SotR if we take two full hits (because $\frac{1+1}{2}=1$, which is >0.8), or one hit and one block ($\frac{1+0.7}{2}=0.85$, which is >0.8), but not two blocks (mean of 0.7) or any other combination.

We could, of course, set N higher than 2 if we wanted to defend only against extremely high spikes. However, by N=3 we’re already looking at 4.5 seconds, and it only takes 5.5-7.5 seconds to generate 3 holy power depending on your haste level. By N=3 or N=4 we’re already almost guaranteed to trigger the bleed valve condition during our filter period, which means we’re not very likely to see events that exceed 0.8 in the first place. As N increases, the filter gets engaged less frequently and the results should tend towards the “S” rotation.

We could also change the threshold level from 0.8 to something else, if we wanted to. The lower we set it, the more frequently it will engage, and if we set it too low it will just fire any time we don’t avoid N attacks. If we set it higher than 0.8, we limit the filter’s trigger condition to strings of full hits. I’ve chosen 0.8 specifically so that hit+block combinations trigger the conditional for N=2, but a single block won’t trigger it for N=1. We can explore the effects of changing this threshold condition in a later blog post.

Later this week, we’ll see how this shift queue performs with all of the gear sets we’ve compiled. Today, we’re just going to see how it performs with the Control/Haste and Control/Mastery gear sets, alongside the queues we tested in the last round of T15 2-piece simulations. Thus, a quick review of the T15 2-piece finisher queues is in order:

  • “T” (for “tier bonus”) will use WoG any time the tier bonus buff drops off, as well as if we hit 5 holy power so that we don’t waste anything.
  • “TS” tries t0 keep the tier bonus up at all times and uses SotR as a bleed mechanism. The logic is to cast WoG every time the tier buff drops off and only cast SotR when we’re at 5 holy power.
  • “ST” is a “first come, first serve” style rotation that tries to keep something up at all times, whether it’s SotR or the tier 15 set bonus. It will prioritize SotR over the tier bonus. The logic is roughly “if neither buff is up and HP>=3, cast SotR; if neither buff is up and HP<3, try to cast WoG.”

As a reminder, here are the stats for the Control/Haste and Control/Mastery gear sets:

|    Set: |  C/Ha |  C/Ma |
|     Str | 15000 | 15000 |
|   Parry |  1500 |  1500 |
|   Dodge |  1500 |  1500 |
| Mastery |  1500 | 13500 |
|     Hit |  2550 |  2550 |
|     Exp |  5100 |  5100 |
|   Haste | 12000 |     0 |
|   Armor | 65000 | 65000 |

Control/Mastery Gear Set

The mastery set runs a whopping 30.5% mastery, pushing SotR’s mitigation up to 60.5%; that’s as potent as a warrior’s critical block. Speaking of blocking, it also packs an impressive 41% block, though that gets knocked down to 36.5% against a boss. The downside is that we have 0% haste in this gear set, so our holy power generation is capped at around 0.41 HP/sec. There’s an inherent trade-off here – timing SotR will give you a strong effect, but the lower uptime of SotR means that we can’t cover as many attacks.

As before, we need to specify how much overheal we expect WoG to generate to properly evaluate the queues that include WoG. We’ll start with 100% overheal (meaning WoG itself does essentially nothing for you), and then look at the 50% and 0% overheal cases.

WoG overheal factor: 100%

| Queue: |       S |     SH1 |     SH2 |       T |      TS |      ST |
|     S% |  0.4101 |  0.4101 |  0.4105 |  0.0000 |  0.2126 |  0.2360 |
|   mean |  0.5417 |  0.5429 |  0.5423 |  0.6231 |  0.5433 |  0.5375 |
|    std |  0.1449 |  0.1325 |  0.1253 |  0.1455 |  0.1513 |  0.1475 |
| ------ |   --- 2 |  Attack |  Moving | Average |  ------ |  ------ |
|    60% | 42.3010 | 46.7672 | 42.9268 | 65.4025 | 48.0190 | 46.1590 |
|    70% | 21.9970 | 17.6668 | 19.5328 | 27.2403 | 18.5732 | 17.6793 |
|    80% | 21.9970 | 17.6668 | 19.5328 | 27.2403 | 18.5732 | 17.6793 |
|    90% | 10.2610 |  5.6763 |  8.4748 |  3.6463 |  2.4868 |  2.3977 |
| ------ |   --- 3 |  Attack |  Moving | Average |  ------ |  ------ |
|    60% | 32.7568 | 27.8945 | 33.7280 | 54.9875 | 36.7665 | 34.8075 |
|    70% | 18.2605 | 13.4463 | 13.9110 | 29.3353 | 17.6935 | 16.4103 |
|    80% |  7.4737 |  4.4705 |  1.7305 |  7.4543 |  4.3247 |  3.9830 |
|    90% |  2.7365 |  1.2675 |  0.6405 |  0.6740 |  0.4245 |  0.3802 |
| ------ |   --- 4 |  Attack |  Moving | Average |  ------ |  ------ |
|    60% | 33.9690 | 31.9660 | 31.7003 | 48.5408 | 30.4235 | 29.5403 |
|    70% | 12.5495 |  8.7740 |  4.0938 | 28.7690 | 15.3175 | 13.7128 |
|    80% |  5.2840 |  3.4635 |  1.2557 | 10.2150 |  5.0597 |  4.4720 |
|    90% |  1.2183 |  0.7810 |  0.2188 |  1.8105 |  0.9160 |  0.7803 |
| ------ |   --- 5 |  Attack |  Moving | Average |  ------ |  ------ |
|    60% | 39.1620 | 36.9288 | 35.6007 | 61.8642 | 36.6945 | 35.0003 |
|    70% | 10.6135 |  8.7673 |  6.9117 | 27.4010 | 12.7738 | 10.9168 |
|    80% |  1.4990 |  1.3630 |  0.3220 | 11.7240 |  4.7525 |  3.8885 |
|    90% |  0.0000 |  0.0375 |  0.0000 |  0.4228 |  0.1762 |  0.1425 |
| ------ |   --- 6 |  Attack |  Moving | Average |  ------ |  ------ |
|    60% | 35.8623 | 32.9168 | 32.5135 | 58.4110 | 35.6808 | 33.3798 |
|    70% | 11.1083 |  8.5260 |  7.2780 | 26.0475 | 10.6435 |  8.5690 |
|    80% |  0.0000 |  0.1315 |  0.0095 |  4.1433 |  1.3693 |  1.0128 |
|    90% |  0.0000 |  0.0000 |  0.0000 |  0.0948 |  0.0305 |  0.0248 |
| ------ |   --- 7 |  Attack |  Moving | Average |  ------ |  ------ |
|    60% | 32.3012 | 29.1515 | 29.4353 | 55.0732 | 31.1550 | 28.9513 |
|    70% |  9.7462 |  7.2493 |  4.8043 | 24.8360 |  8.8652 |  6.8185 |
|    80% |  0.8013 |  0.5012 |  0.2617 |  5.0090 |  1.2965 |  0.8367 |
|    90% |  0.0000 |  0.0000 |  0.0000 |  0.2248 |  0.0468 |  0.0355 |

First, let’s consider the shifting queues. Both show a distinct improvement over the S queue, suggesting that timing SotR casts really does make a difference in spike representation. SH1 does the best job at eliminating 2-attack spikes, which makes sense since it’s a short-spike filter. While it outperforms S across the board, it falls behind SH2 for 3+ attacks, suggesting that SH2 is probably the stronger of the two queues overall for a mastery gear set. It’s also worth noting the magnitude of the improvement: compared to S, we’re getting a 40%-80% reduction in some of the spike categories, especially in the important 4- to 5-attack regime. That’s not a trivial gain for a reasonably simple time-shifting algorithm.

As in the last round, the T queue is pretty much garbage. The set bonus isn’t strong enough to stand on its own, so it’s not worth spamming WoG in an attempt to keep the buff up 100% of the time.

The TS queue doesn’t perform all that poorly compared to S, even showing a pretty significant improvement for 2-3 attacks. But it only manages to match the performance of S for 4 attacks, and lags a bit in the 5+ attack categories. When compared to SH2, it falls significantly behind in nearly every metric.

A similar story could be told about ST, which performed reasonably competitively in the last round of testing. It can hold its own against S, but falls behind SH2 by a significant margin. At 100% overheal on WoG, there’s just no situation where the set bonus seems to give a clear survivability advantage.

Let’s see what happens when WoG’s overheal is cut to 50%.

WoG overheal factor: 50%

| Queue: |       S |     SH1 |     SH2 |       T |      TS |      ST |
|     S% |  0.4104 |  0.4105 |  0.4104 |  0.0000 |  0.2122 |  0.2356 |
|   mean |  0.5420 |  0.5408 |  0.5413 |  0.5551 |  0.4898 |  0.4989 |
|    std |  0.1449 |  0.1333 |  0.1256 |  0.1373 |  0.1484 |  0.1373 |
| ------ |   --- 2 |  Attack |  Moving | Average |  ------ |  ------ |
|    60% | 42.3070 | 46.4395 | 42.8245 | 49.6615 | 36.9973 | 39.7920 |
|    70% | 22.0145 | 17.4175 | 19.5012 | 17.5070 | 14.1575 | 16.2463 |
|    80% | 22.0145 | 17.4175 | 19.5012 | 16.0098 | 13.2680 |  8.5230 |
|    90% | 10.2410 |  5.6247 |  8.3945 |  2.1725 |  1.7857 |  1.7715 |
| ------ |   --- 3 |  Attack |  Moving | Average |  ------ |  ------ |
|    60% | 32.7285 | 27.4853 | 33.6465 | 39.5823 | 29.2448 | 31.4512 |
|    70% | 18.2455 | 13.2448 | 13.8183 | 15.8858 | 12.2385 | 14.6877 |
|    80% |  7.5020 |  4.3853 |  1.7138 |  3.3298 |  2.6967 |  3.4908 |
|    90% |  2.7553 |  1.2767 |  0.6318 |  0.2680 |  0.2367 |  0.3100 |
| ------ |   --- 4 |  Attack |  Moving | Average |  ------ |  ------ |
|    60% | 34.1212 | 31.4797 | 31.5030 | 33.3507 | 23.0412 | 26.3468 |
|    70% | 12.5640 |  8.6035 |  4.0767 | 13.8833 |  9.7470 | 10.6955 |
|    80% |  5.2990 |  3.4830 |  1.2605 |  3.5175 |  2.4328 |  2.1993 |
|    90% |  1.2255 |  0.7878 |  0.2120 |  0.3693 |  0.3953 |  0.0762 |
| ------ |   --- 5 |  Attack |  Moving | Average |  ------ |  ------ |
|    60% | 39.3628 | 36.3190 | 35.3318 | 40.4225 | 23.0955 | 23.1992 |
|    70% | 10.6638 |  8.7133 |  6.8473 | 12.2817 |  7.4853 |  7.0390 |
|    80% |  1.5003 |  1.3500 |  0.3285 |  1.0110 |  1.5378 |  0.7823 |
|    90% |  0.0000 |  0.0498 |  0.0000 |  0.0462 |  0.0590 |  0.0290 |
| ------ |   --- 6 |  Attack |  Moving | Average |  ------ |  ------ |
|    60% | 35.9303 | 32.3367 | 32.2488 | 40.4633 | 21.5243 | 21.2708 |
|    70% | 11.0507 |  8.3502 |  7.1970 | 11.1772 |  5.5295 |  4.5475 |
|    80% |  0.0000 |  0.1467 |  0.0110 |  0.6997 |  0.4557 |  0.3740 |
|    90% |  0.0000 |  0.0000 |  0.0000 |  0.0037 |  0.0077 |  0.0067 |
| ------ |   --- 7 |  Attack |  Moving | Average |  ------ |  ------ |
|    60% | 32.3418 | 28.5257 | 29.1668 | 35.7612 | 18.7482 | 18.8182 |
|    70% |  9.7500 |  7.1750 |  4.7520 | 10.0488 |  3.8830 |  2.9875 |
|    80% |  0.8268 |  0.4840 |  0.2512 |  0.7990 |  0.2135 |  0.1840 |
|    90% |  0.0000 |  0.0000 |  0.0000 |  0.0005 |  0.0028 |  0.0008 |

The S, SH1, and SH2 data doesn’t change here outside statistical fluctuations. The T queue is still a lot less garbage-y now, arguably matching the performance of S for shorter strings. It still loses ground for 6- to 7-attacks, however.

TS and ST show a fairly significant improvement over S across the board. But neither is clearly superior to our new gold standard, the SH2 queue. They generally outperform the shift queues for 2 attacks, but it’s questionable how much we really worry about 2-attack strings, since most bosses don’t melee you for 400k or more. SH2 tends to beat them in the 3- to 5-attack categories, but falls behind again at 7 attacks. All in all I think it’s pretty much a three-way tie between SH2, TS, and ST at this point.

So if WoG’s giving you 50% overheal, gaming the new set bonus buy syou a small improvement over just spamming SotR (“S”), but it’s not a clear improvement over just sticking with SotR and being careful with your timing (“SH”).

Let’s crank WoG up to maximum effectiveness and look at the 0% overheal data:

WoG overheal factor: 0%

| Queue: |       S |     SH1 |     SH2 |       T |      TS |      ST |
|     S% |  0.4108 |  0.4102 |  0.4105 |  0.0000 |  0.2121 |  0.2360 |
|   mean |  0.5412 |  0.5416 |  0.5422 |  0.5022 |  0.4465 |  0.4597 |
|    std |  0.1451 |  0.1328 |  0.1252 |  0.1373 |  0.1490 |  0.1337 |
| ------ |   --- 2 |  Attack |  Moving | Average |  ------ |  ------ |
|    60% | 42.2155 | 46.5763 | 42.8827 | 38.5640 | 33.1002 | 27.4115 |
|    70% | 21.9353 | 17.4630 | 19.5302 | 15.8348 | 12.1708 |  9.5460 |
|    80% | 21.9353 | 17.4630 | 19.5302 | 15.8348 | 11.4650 |  7.9205 |
|    90% | 10.2748 |  5.5675 |  8.4910 |  2.1077 |  1.6105 |  0.9662 |
| ------ |   --- 3 |  Attack |  Moving | Average |  ------ |  ------ |
|    60% | 32.6512 | 27.6415 | 33.6860 | 26.2135 | 21.5578 | 20.9620 |
|    70% | 18.2035 | 13.2302 | 13.8877 | 11.5970 |  9.2747 |  6.9993 |
|    80% |  7.4228 |  4.3803 |  1.7488 |  2.8165 |  2.1900 |  1.2845 |
|    90% |  2.7748 |  1.2550 |  0.6620 |  0.2760 |  0.2053 |  0.1045 |
| ------ |   --- 4 |  Attack |  Moving | Average |  ------ |  ------ |
|    60% | 33.8680 | 31.6200 | 31.6953 | 18.7090 | 14.8930 | 17.8592 |
|    70% | 12.4778 |  8.6483 |  4.0650 |  7.2133 |  6.1665 |  4.8243 |
|    80% |  5.2785 |  3.4120 |  1.3058 |  2.0125 |  1.8793 |  0.4870 |
|    90% |  1.2040 |  0.7582 |  0.2240 |  0.3785 |  0.2710 |  0.0468 |
| ------ |   --- 5 |  Attack |  Moving | Average |  ------ |  ------ |
|    60% | 39.2205 | 36.6190 | 35.4823 | 28.2785 | 14.6325 | 14.4177 |
|    70% | 10.5500 |  8.6173 |  6.9608 |  4.3925 |  3.8047 |  2.8465 |
|    80% |  1.4735 |  1.3350 |  0.3348 |  0.6145 |  0.8670 |  0.2318 |
|    90% |  0.0000 |  0.0425 |  0.0000 |  0.0210 |  0.0413 |  0.0017 |
| ------ |   --- 6 |  Attack |  Moving | Average |  ------ |  ------ |
|    60% | 35.7910 | 32.5177 | 32.3367 | 23.0938 | 13.8513 | 11.4323 |
|    70% | 11.0220 |  8.3745 |  7.3345 |  3.6457 |  2.1958 |  1.4570 |
|    80% |  0.0000 |  0.1320 |  0.0105 |  0.0703 |  0.1385 |  0.0610 |
|    90% |  0.0000 |  0.0000 |  0.0000 |  0.0000 |  0.0025 |  0.0003 |
| ------ |   --- 7 |  Attack |  Moving | Average |  ------ |  ------ |
|    60% | 32.2383 | 28.7430 | 29.3895 | 19.1578 |  9.9872 | 10.3063 |
|    70% |  9.6973 |  7.1182 |  4.8477 |  3.2942 |  1.3065 |  0.7248 |
|    80% |  0.8415 |  0.5030 |  0.2890 |  0.0478 |  0.0342 |  0.0128 |
|    90% |  0.0000 |  0.0000 |  0.0000 |  0.0000 |  0.0000 |  0.0000 |

Here we finally start to have a real competition. The raw healing power of WoG is enough to tip the scales a bit, such that even the T rotation is competitive. It flat-out beats the simple S queue, and does a reasonable job of competing with SH2. It edges out SH2 for longer (5+ attacks) and shorter (2 attacks) strings, though it trails ever so slightly in the 4–attack category.

TS and ST fare about the same, outperforming SH2 in a number of categories but trailing in isolated spots, especially in the 4-attack range. ST seems to have pulled ahead of TS enough to be strictly superior in all categories. Overall one could make the argument that either TS or ST is competitive enough to be pereferable to even SH2 at this point. the fact that they beat out S by a pretty handy margin is notable for tanks that aren’t trying to micromanage SotR – mixing some WoGs in to maintain the buff may be a net gain for those players.

It’s also worth noting that at 50% and 0% overheal, the WoG queues beat out the S and SH queues in total damage reduction (TDR) by a large amount. TS and ST offer a whopping 17% reduction in total damage taken. Remember though, most of that effectiveness is coming from WoG, not the set bonus.

So, much as we saw last time, queues that leverage the power of the T15 2-piece set bonus don’t become truly competitive until WoG is operating near maximum efficiency. We’ll talk more about what this means shortly, but first let’s look at how things change for a Control/Haste gear set.

Control/Haste Gear Set

The haste gear set shifts to the opposite extreme. We only have a meager 10.5% mastery, giving 19.5% block chance against a boss and 40.5% SotR mitigation. However, we’ve piled on the haste – 28.24% of it, to be exact, and we’ll reap the benefit of that by having significantly higher holy power generation. In this gear set, HPG caps out around 0.52 per second. If you recall the last smoothing post, we posited that this extra 28% uptime on SotR would make a bigger impact on the shifting queues than the extra 20% mastery would. Let’s see if we were right.

Again, we start with the 100% overheal data:

WoG overheal factor: 100%

| Queue: |       S |     SH1 |     SH2 |       T |      TS |      ST |
|     S% |  0.5228 |  0.5225 |  0.5226 |  0.0000 |  0.3242 |  0.3530 |
|   mean |  0.6010 |  0.6015 |  0.6016 |  0.6652 |  0.5773 |  0.5749 |
|    std |  0.1479 |  0.1342 |  0.1418 |  0.1562 |  0.1496 |  0.1470 |
| ------ |   --- 2 |  Attack |  Moving | Average |  ------ |  ------ |
|    60% | 46.3575 | 47.4338 | 46.0037 | 65.5367 | 47.2860 | 46.8760 |
|    70% | 40.7343 | 43.3862 | 40.4728 | 42.3603 | 30.2638 | 29.2868 |
|    80% | 14.6405 | 10.0218 | 14.2113 | 42.3603 | 22.0120 | 20.4937 |
|    90% |  9.9078 |  4.4075 |  9.4393 | 10.8740 |  5.6080 |  5.2313 |
| ------ |   --- 3 |  Attack |  Moving | Average |  ------ |  ------ |
|    60% | 53.7650 | 54.0125 | 54.2755 | 59.2830 | 46.6840 | 46.6097 |
|    70% | 37.0855 | 36.8943 | 37.0173 | 41.9015 | 26.6883 | 26.2925 |
|    80% | 15.1855 |  7.4750 | 14.6333 | 19.1935 |  9.7335 |  9.0783 |
|    90% |  1.2980 |  0.4585 |  0.0008 |  3.6005 |  1.3678 |  1.2080 |
| ------ |   --- 4 |  Attack |  Moving | Average |  ------ |  ------ |
|    60% | 53.8647 | 52.9630 | 55.3687 | 57.4893 | 46.3350 | 46.0938 |
|    70% | 32.2542 | 26.9342 | 32.2070 | 40.2890 | 23.4513 | 23.4288 |
|    80% |  3.3598 |  2.0602 |  0.0017 | 23.0570 |  8.0365 |  7.2527 |
|    90% |  0.0000 |  0.0730 |  0.0000 |  8.0308 |  2.1803 |  1.6980 |
| ------ |   --- 5 |  Attack |  Moving | Average |  ------ |  ------ |
|    60% | 55.5468 | 55.8940 | 56.7255 | 71.8725 | 46.1725 | 45.1663 |
|    70% | 28.6008 | 25.7090 | 26.9973 | 39.7900 | 21.4207 | 21.0795 |
|    80% |  8.9258 |  5.0655 |  7.0110 | 23.4135 |  6.1088 |  4.8078 |
|    90% |  0.0000 |  0.0000 |  0.0000 |  3.2043 |  0.6520 |  0.4323 |
| ------ |   --- 6 |  Attack |  Moving | Average |  ------ |  ------ |
|    60% | 54.1748 | 56.6675 | 55.0505 | 69.3145 | 44.5780 | 44.0457 |
|    70% | 24.2147 | 23.5965 | 23.1848 | 39.8963 | 19.3180 | 18.5075 |
|    80% |  4.2663 |  2.4810 |  3.3905 | 13.6180 |  3.6472 |  3.0048 |
|    90% |  0.0000 |  0.0000 |  0.0000 |  1.2370 |  0.1580 |  0.0800 |
| ------ |   --- 7 |  Attack |  Moving | Average |  ------ |  ------ |
|    60% | 53.0793 | 55.0057 | 53.8088 | 67.3342 | 45.6855 | 44.9018 |
|    70% | 22.6185 | 20.2733 | 21.7108 | 40.0643 | 16.8960 | 15.8635 |
|    80% |  3.8282 |  2.0460 |  2.8815 | 15.1383 |  2.0480 |  1.4508 |
|    90% |  0.0000 |  0.0000 |  0.0000 |  2.3283 |  0.0700 |  0.0250 |

The behavior of SH1 and SH2 is a little different here. With the mastery gear set, we saw SH2 dominate in all categories except for the 2-attack range. However, with the haste gear set SH1 seems to be doing the dominating. This is mostly due to having faster holy power generation – 0.52 HP/sec compared to 0.41 HP/sec for the mastery set. That drops your mean SotR time from 7.3 seconds to 5.8 seconds, and as I mentioned in the introduction, faster holy power generation makes higher-N queues less efficient. We now have enough HPG to be a little riskier with our reaction filter, because we’re generating holy power fast enough that the downtime is only going to be about 2 boss swings.

This is also the cause of the curious data in the 4-attack category, which is the one place that SH1 clearly falls behind SH2 (at least, for 80 and 90% spikes). Because the downtime of SotR is 2 boss swings, the SH1 becomes a little more vulnerable to 4-attack strings thanks to the “2 on, 2 off” pattern of SotR. SH2 is a little more conservative with its SotR usage, and thus covers that particular case a little more effectively.

How does the improvement from S to SH compare to what we saw in the mastery set? Well, it’s… about the same, actually. We see a lot of the same 30%-50% drops going from S to SH, with larger reductions in a few categories. But it’s not a strong victory for haste over mastery like we predicted.

Part of the problem is that haste’s strength comes from shaving off the top category – it’s good at combating 80% and 90% spikes, but it doesn’t do as much for 60% or 70% spikes. But the default S queue already eliminates most of the 90% categories for the haste gear set, so there’s no more room for improvement there. The 80% categories still show large gains, but not larger than the mastery data showed. And in the 60% and 70% categories, the time-shifted queues tend to show less improvement with the haste gear than the mastery gear. So it’s not clear that there’s a large untapped well of survivability that one can draw from with this simple shifting model.

Moving on to the WoG queues, you could probably have guessed this but… yes, the T queue is still garbage at 100% overheal.

TS and ST both perform fairly similarly, doing a better job of suppressing 60%-70% spikes than the S or SH queues, but lagging in the 80%-90% categories. Just as in the last round of simulations, this has to do with the damage profile of each distribution. TS and ST have a broader profile with a longer tail, while S (and SH) suppress the long tail at the expense of shifting the mean of the distribution to higher spike levels. I won’t reprise the plots here, because they look pretty much the same as last time, so you can take a look at the last post to see example graphs.

Short version: at 100% overheal, the set bonus still doesn’t seem to be worth prioritizing for a haste gear set.

Onwards and upwards (for WoG anyway); the 50% overheal data beckons:

WoG overheal factor: 50%

| Queue: |       S |     SH1 |     SH2 |       T |      TS |      ST |
|     S% |  0.5226 |  0.5224 |  0.5232 |  0.0000 |  0.3242 |  0.3532 |
|   mean |  0.6008 |  0.6006 |  0.6005 |  0.5745 |  0.5339 |  0.5391 |
|    std |  0.1481 |  0.1348 |  0.1423 |  0.1489 |  0.1422 |  0.1393 |
| ------ |   --- 2 |  Attack |  Moving | Average |  ------ |  ------ |
|    60% | 46.3222 | 47.3290 | 45.7760 | 53.1528 | 39.7963 | 41.6880 |
|    70% | 40.7070 | 43.2488 | 40.2820 | 25.5638 | 23.9830 | 23.3118 |
|    80% | 14.6335 | 10.0162 | 14.0865 | 19.8077 | 15.3640 | 11.3565 |
|    90% |  9.8972 |  4.3823 |  9.3970 |  5.0638 |  3.8420 |  2.9267 |
| ------ |   --- 3 |  Attack |  Moving | Average |  ------ |  ------ |
|    60% | 53.7365 | 53.8780 | 54.0440 | 46.7885 | 39.3038 | 40.0625 |
|    70% | 37.0468 | 36.7013 | 36.7800 | 24.0000 | 19.6335 | 18.7733 |
|    80% | 15.1913 |  7.3915 | 14.5518 |  7.0505 |  6.0220 |  5.0628 |
|    90% |  1.3120 |  0.4600 |  0.0023 |  0.7963 |  0.6673 |  0.5020 |
| ------ |   --- 4 |  Attack |  Moving | Average |  ------ |  ------ |
|    60% | 53.8390 | 52.8623 | 55.0905 | 41.6155 | 36.7965 | 38.4763 |
|    70% | 32.2512 | 26.8493 | 31.8983 | 22.9363 | 15.6793 | 14.8535 |
|    80% |  3.4008 |  2.0172 |  0.0030 |  8.0065 |  3.7687 |  2.8995 |
|    90% |  0.0000 |  0.0720 |  0.0000 |  0.3538 |  0.6822 |  0.2405 |
| ------ |   --- 5 |  Attack |  Moving | Average |  ------ |  ------ |
|    60% | 55.5238 | 55.7700 | 56.4808 | 40.8570 | 34.2160 | 35.9790 |
|    70% | 28.6000 | 25.5805 | 26.7625 | 21.4312 | 12.6945 | 12.0057 |
|    80% |  8.8878 |  4.9570 |  6.9828 |  2.4095 |  1.9552 |  1.7050 |
|    90% |  0.0000 |  0.0000 |  0.0000 |  0.2692 |  0.1355 |  0.0532 |
| ------ |   --- 6 |  Attack |  Moving | Average |  ------ |  ------ |
|    60% | 54.2253 | 56.4615 | 54.6968 | 45.6250 | 32.6095 | 34.4695 |
|    70% | 24.2168 | 23.4888 | 23.0660 | 19.7000 |  9.8537 |  9.8712 |
|    80% |  4.2467 |  2.4075 |  3.3208 |  3.1580 |  1.0467 |  0.6258 |
|    90% |  0.0000 |  0.0000 |  0.0000 |  0.0540 |  0.0168 |  0.0067 |
| ------ |   --- 7 |  Attack |  Moving | Average |  ------ |  ------ |
|    60% | 53.1063 | 54.8103 | 53.4983 | 45.6503 | 31.3818 | 32.3642 |
|    70% | 22.6195 | 20.1075 | 21.6425 | 17.4415 |  7.6185 |  8.1792 |
|    80% |  3.8020 |  1.9550 |  2.9023 |  1.9150 |  0.4410 |  0.2603 |
|    90% |  0.0000 |  0.0000 |  0.0000 |  0.0075 |  0.0013 |  0.0008 |

In this data set, we’re starting to see a clear advantage towards the TS and ST rotations. They’re solidly competing with SH1, and even though they permit more 90% spikes, they more than make up for it with suppression in the 60%-80% categories. T is also in the running now, though there’s no situation where it’s clearly preferable to either of the mixed queues.

It seems that the extra holy power fueling WoG in this gear set is having a significant effect. With this much HPG, we’re able to get a respectable SotR uptime while also taking advantage of the set bonus, which makes it more valuable than it was in the mastery gear set. I think it’s fair to say that ST has a slight lead over all of the competitors this round. Let’s see what’s in store when we crank WoG up to full power:

WoG overheal factor: 0%

| Queue: |       S |     SH1 |     SH2 |       T |      TS |      ST |
|     S% |  0.5225 |  0.5222 |  0.5225 |  0.0000 |  0.3247 |  0.3534 |
|   mean |  0.6003 |  0.6013 |  0.6005 |  0.4974 |  0.5072 |  0.5027 |
|    std |  0.1481 |  0.1344 |  0.1425 |  0.1503 |  0.1445 |  0.1370 |
| ------ |   --- 2 |  Attack |  Moving | Average |  ------ |  ------ |
|    60% | 46.3508 | 47.4428 | 45.8948 | 30.6310 | 34.8588 | 32.9618 |
|    70% | 40.6535 | 43.3518 | 40.3810 | 19.8600 | 22.1415 | 17.7995 |
|    80% | 14.5957 | 10.0767 | 14.1655 | 18.9675 | 14.8998 |  8.9260 |
|    90% |  9.7813 |  4.4400 |  9.4358 |  5.0583 |  3.7705 |  2.0580 |
| ------ |   --- 3 |  Attack |  Moving | Average |  ------ |  ------ |
|    60% | 53.6503 | 53.9720 | 54.0338 | 29.0413 | 31.9298 | 30.9422 |
|    70% | 36.9320 | 36.7185 | 36.8463 | 12.1288 | 15.7963 | 11.7378 |
|    80% | 15.0675 |  7.4773 | 14.6283 |  4.3130 |  5.3172 |  2.4430 |
|    90% |  1.2695 |  0.4752 |  0.0010 |  0.7910 |  0.6440 |  0.1398 |
| ------ |   --- 4 |  Attack |  Moving | Average |  ------ |  ------ |
|    60% | 53.7005 | 52.9335 | 55.2110 | 24.9007 | 30.3105 | 30.3485 |
|    70% | 32.1685 | 26.9200 | 32.0120 |  9.0545 | 11.2550 |  8.7277 |
|    80% |  3.3473 |  2.0720 |  0.0050 |  1.0317 |  2.6930 |  0.8783 |
|    90% |  0.0000 |  0.0697 |  0.0000 |  0.3053 |  0.5988 |  0.0318 |
| ------ |   --- 5 |  Attack |  Moving | Average |  ------ |  ------ |
|    60% | 55.3028 | 55.8280 | 56.5597 | 27.1045 | 28.0080 | 26.2122 |
|    70% | 28.5143 | 25.6930 | 26.8380 |  8.6128 |  8.5400 |  6.5027 |
|    80% |  8.8370 |  5.0570 |  6.9102 |  0.2565 |  1.2795 |  0.5853 |
|    90% |  0.0000 |  0.0000 |  0.0000 |  0.0000 |  0.0813 |  0.0020 |
| ------ |   --- 6 |  Attack |  Moving | Average |  ------ |  ------ |
|    60% | 54.0813 | 56.5700 | 54.7773 | 21.6617 | 24.0912 | 22.7147 |
|    70% | 24.1095 | 23.5930 | 22.9925 |  6.6400 |  6.1165 |  4.4040 |
|    80% |  4.2360 |  2.4580 |  3.3653 |  0.5295 |  0.4777 |  0.1800 |
|    90% |  0.0000 |  0.0000 |  0.0000 |  0.0000 |  0.0047 |  0.0000 |
| ------ |   --- 7 |  Attack |  Moving | Average |  ------ |  ------ |
|    60% | 52.9763 | 54.9205 | 53.4872 | 21.2062 | 23.0325 | 20.7325 |
|    70% | 22.5190 | 20.2357 | 21.5467 |  3.8108 |  4.1210 |  2.7390 |
|    80% |  3.7915 |  2.0220 |  2.8915 |  0.2178 |  0.2122 |  0.0222 |
|    90% |  0.0000 |  0.0000 |  0.0000 |  0.0000 |  0.0000 |  0.0000 |

Again, leveraging the full power of WoG pushes the tier set queues to the forefront. ST in particular has a commanding lead in every category, but TS and T aren’t far behind. Even T is eclipsing the S and SH queues by a fair margin in almost all categories. There’s no question that at 0% overheal, these WoG queues take the lead.

Conclusions

So we’ve waded through a lot of data, but we still haven’t addressed what it means in a practical sense. The shifting queues are obviously a success, and we’ll be seeing more of those later this week to try to get a more complete view of how they impact our gearing. The improvement from S to SH is fairly significant, reinforcing the idea that smart use of SotR will always trump “lazy” usage (i.e. macroing it to Crusader Strike).

Though unfortunately, it doesn’t seem that the haste gear set gives a larger benefit than the mastery gear set in this regard. Mel and I had both expected a more significant drop in the haste gear set, but it never materialized. This could just be a product of how high the uptime on SotR is already; having more uptime to shift around could be more valuable when you have less of it to work with in the first place. Maybe by the time you’re at 40%-50% uptime on SotR, there just may not be that much room for shifting that uptime around.

Or maybe we just under-estimated the interaction of the passive benefit of mastery with the shifting algorithm. A higher block chance may just cause the filter to apply that shifting algorithm more efficiently since it isn’t engaging as often. Or maybe it’s something else I haven’t thought of yet. It’s not entirely clear what interaction is the limiting factor here, but something is certainly doing it.

From this data you might conclude that Control/Mastery has cemented its claim as the dominant survivability gearing strategy for this tier of content. However, haste has one final ace up it’s sleeve, and later this week we’ll see exactly how that hand plays out.

Regarding the Tier 15 set bonus, I’ll echo my sentiment from last time – while the 0% overheal data makes the 2-piece look very impressive, we also need to keep in mind that it’s the least realistic of the simulations. If you’re mindlessly spamming WoG on yourself, you’re guaranteed to waste some of that with overhealing.

And it’s not just our overhealing that counts here. If you WoG yourself to full health right before your healer drops a large heal on you, all of that heal just turned into overheal. And we have to count that against us, because all the healing that WoG performed was irrelevant in that case. We’d have been returned to full health whether we used WoG or not, so we would have gotten more mileage out of a Shield of the Righteous (not to mention the loss of Bastion of Glory stacks).

So what is a realistic estimate of our overhealing with WoG? If we’re following the ST or TS rotations, I’d wager that it’s 50% or more. 20%-30% of it from your own overhealing, and another 20%-30% from overhealing you indirectly cause for your healers. You could probably do better than that if you’re running light on healers for an encounter, but I think that creates more risks than it eliminates. That situation leaves you with fewer backup sources of healing for when things go awry, and increases the time it takes before someone else can react and step in to help.

So under most conditions, I don’t think it’s worth weaving WoGs into your rotation for the sake of the set bonus. The situations where it does make sense to slip WoG in are the ones where it’s nearly guaranteed to not cause overheal – in other words, when we’re already fairly low on health. However, that’s generally a case where we would already be using WoG (often with banked Bastion of Glory stacks) even without the set bonus. In a bizarre situation where you are your only healer, such as soloing old raid content, I think the set bonus might become very attractive because you can guarantee you’ll be in the no-overhealing regime. But I don’t think it’s likely we’ll see that situation in regular old “bring-a-healer-along” raiding.

In short, I still don’t find the 2-piece compelling as a maintenance buff. The optimum usage is dictated more by WoG’s effectiveness than the value added by the extra block chance. The most effective way to make use of the set bonus seems to be to use WoG exactly like we do now: not as a proactive tool (like SotR), but as a reactive tool, an emergency heal for when things get dicey.

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

31 Responses to Control Shift En-tier

  1. blizzhoof says:

    I want to preface my post by saying, that I know I’m not nearly as well-versed as you or others on mathematical models and number-crunching.

    That said, does it not matter that you’ll always have an extra 3000 Mastery? You don’t include the buff in your sets, but I’d imagine it would make a difference on how SotR is used (considering more blocked hits) especially with the SH2.

    Also, since higher Dodge/Parry chances effectively make block worse (due to the 2-roll system), would it not be advantageous to have more realistic avoidance chances on the C/Ha and C/Ma gearsets?

    Thanks for doing what you do by the way.

    • Theck says:

      You’re right, I don’t include the buffs in my sets. However, it only adds another ~3.4% block and mitigation on SotR. It may make a small difference, but it shouldn’t skew the results very heavily in favor of either gear set. Remember that both gear sets would see the benefit, and that the synergy between mastery and haste would probably cause the haste set to see a larger benefit than the mastery set would. That said, it’s a good question, and seems like a reasonable (and simple) thing to add to the simulation, so I’ll do that for future work.

      As to dodge/parry, it’s interesting you should say that. In the last round, i put much more stringent minimums on dodge and parry (2500 of each, I believe) for exactly that reason. And I got some strong criticism from players who really were dropping down to nearly zero dodge/parry rating, suggesting that my choices were unrealistic. So you’re telling me exactly the opposite!

      I think your suggestion is more common, in truth. Most players won’t have the gear availability to push dodge and parry below 1500. I’m regularly making do with dodge or parry items because they drop, they’re a large stamina/armor upgrade, and nobody else is going to use them. The juicy hit/haste/expertise items that want generally make a bigger difference in the hands of a plate DPS player, so I make sure to give Retribution paladins (and Frost DKs, I think?) first dibs on them. Luckily, our two DPS warriors prefer crit to haste, so they don’t generally want the same gear I do.

      The solution I came up with for this simulation conundrum, however, was to just run more gear sets. In the 5.2 Smoothness sims and the post you’ll see later this week, I run both of these “extreme” gear sets along with a “Control/Balanced” gear set that splits the itemization roughly equally between haste, mastery, dodge, and parry. That set gives a better glimpse of how things work for a regular tank. I haven’t run that gear set for these tier simulations yet, but it’s something I can do. I probably won’t bother posting it unless it shows something out of the ordinary though.

  2. Lakh says:

    I’m not panicking & spamming WoG while I hover at 25% HP… I’m cunningly optimising my play, really. Nice to have a set bonus that syncs like that – becoming increasingly effective in the situations you’re most likely to use it.

  3. Joe rosen says:

    I am much of a similar situation as Blizzhoof. I am not nearly as versed in the mathematics and simulation as you are. I did however have a needlessly complicated idea. I know the EF/SS debate was settled a long time ago with SS on top by a lot. However, I think it worth revisiting at least mentally because the T queue seems to be the strongest if we can push our overhealing down to close to zero. Or, either the ST or TS queue if our overhealing is in between 0-50%.

    EF provides the same amount healing from the HOT than SS does when there are 10 EF in the raid. This of course assume that only 1 holy power is used per EF. In addition, because of the buff to EF it actually passes SS at 9 targets (if the paladin is one of them).

    I think it might be worth considering a situation in which we have EF rolling on everyone in the raid. It is not unreasonable to think we would be able to do this. I think that we would have an extra 10 seconds per 30 where no one in the raid needs to have EF refreshed. While this certainly wouldn’t push overhealing down to zero it might very well push it below the 50% margin where the TS queue becomes viable.

    In order to fully maximize a strat like this I think Divine Purpose would be mandatory. This would mean that we would basically be using the a modification of the T queue where we only SoTR when we have 3 HP which would be during DP procs. This should ensure that when we refresh our own EF we have a 5 stack of Bastion buffing the HoT.

    I think that this is worth considering because your data seems to indicate that overhealing is the largest difference between the success of the set bonus queues and the regular queues. By driving down our overhealing the set bonus queues become more and more viable. EF might be a way to do that.

    • Meloree says:

      Damage prevented is generally more useful than damage healed, and healing on the raid doesn’t translate well to healing on the tank. Your scheme would, I think, give up a great deal of personal survivability for not much benefit.

      • Joe says:

        I actually have to disagree with you on healing on the raid does not translate to healing on the tank. If (and I admit it is a big if) there is enough raid healing being put out by the tank then the raid can free up a healer to focus on tank healing. I do agree that damage prevented is generally more useful than damage healed which is the premise of my scheme, I am trying to find a way to get as close to the 0% overhealing set bonus model which according to the tables above seems to be far and above the best.

        I also disagree with the notion that I would be giving up a “great deal of personal survivability for not much benefit.” I would certainly give up some personal survivability but I think it would be for not such a small benefit. Assume each raid HoT ticks for 2.5k. That would 22500 raid healing. Plus 7500 for the HoT ticking on the tank (assuming no bastion stacks). We are at 30k raid healing. If the tank used a 5 stack Bastion its a 37500 HoT on the tank. So now we 60k of raid healing which becomes 20k HPS. Even if that is 30% overhealing we still have a large chunk of healing helping the healers.

        In my raid I usually do 25-30k HPS on a boss kill. Usually 20% of that is sacred shield (which is probably low due to my lack of noticing when it has fallen off). But if I could put out 40k HPS to help my healers by smoothing damage on everyone else I would not say that is insignificant benefit.

    • Theck says:

      There are a number of reasons that this strategy is not going to result in any improvement.

      First, the premise is flawed. T is not the strongest queue at 100% overhealing, ST takes that title. If you’re throwing out 1-HP EFs on your raid, you’re necessarily casting it every time you get one holy power, which means you’ll never get to cast SotR.

      More importantly, we are solely interested in healing received by the tank. An EF on a raid member is equivalent to 100% overhealing from the perspective of keeping the tank alive. While one could (and you did) make the argument that healing applied to the raid could free up another healer to focus on the tank, I think that is optimistic and unlikely at best, and flat-out incorrect at worst. One-HP Eternal Flames are not going to be sufficient to replace a dedicated raid healer. Even if we’re generous and say that you free up a few GCDs for that raid healer to spend on the tank, it’s not going to be a steady stream of healing, and it’s a poor replacement for the raw mitigation power of SotR.

      It’s worth noting that HoTs generally have a much larger percentage of overhealing than normal heals. Unless you are literally the only person healing the raid, which seems unlikely, EF is likely to be even more significant overhealing than had you used that HP to WoG yourself. In general, I don’t think it’s at all reasonable to assume a 1:1 conversion of tank->raid healing to healer->tank healing.

      Third, I think you’re overlooking the opportunity cost of EF. EF may produce more healing than SS does if you’re blanketing the raid, but EF also costs Holy Power. Thus, the comparison you’re making isn’t fair. You need to compare the healing power of EF to the combined value of SS and all of the SotRs you would get from an equivalent holy power expenditure. That tips the scales strongly in favor of SS+SotR.

      Finally, keep in mind that we’re approximating healing here. We’re not making a distinction between points of damage healed or points of damage absorbed/mitigated. In practice, there’s a very large distinction between those two. Mitigating is almost always preferable because your health fluctuations will be steadier. Trying to work in the T-rotation regime will lead to your health ping-ponging up and down due to large unmitigated boss attacks and large self-heals. That’s naturally more unstable than the smoother damage intake profiles that we see in even the SH2 queues. So even if you could approach the ideal 0%-overheal limit you’re aiming for, I’m not convinced it would actually be advantageous to use a T queue instead of SH.

      • Joe rosen says:

        Thank you for your very thorough reply. It makes a lot of sense of why it wouldn’t work out the way I envisioned. I have a tendency to get an idea in my head and not let go until a logical argument is made to prove it wrong.

      • Geodew says:

        “I think that is optimistic and unlikely at best, and flat-out incorrect at worst.”

        As a long-time raid healer, I personally agree with Theck here, and have a bit more to say on the topic.

        First, more HPS is not always better, even on healer end. AOE heals inherently do more HPS and are more mana efficient at 0% overheal (otherwise we’d rarely use them). However, if there is a DPS at 10% HP because he stood in fire, and I know it’s possible for them to take a fair amount of damage soon, it is rarely a good move to e.g. Circle of Healing, even though the smartheal will hit them, especially if the rest of the raid is in fairly good or decent health. Your job is not to put out crazy HPS; your job is to keep the raid alive. These two things are similar and go mostly hand-in-hand but in that situation, the first does not yield the second. Instead, I would use a fast single-target heal (even if my only available tool is Flash Heal or equivalent) on the dying raid member before returning to AOE. The only reason a good healer might do otherwise is if they could be sure another healer will bring them up instead, and they’d burn a lot of mana bursting them up, which can happen if you heal with the same people for a long time or have pre-coordinated this.

        Second, a tank would have to do a lot of HPS (on the order of 5.3 Glyph of Battle Healer or something) to allow healers to focus their attention more on the tank enough to make a difference in health stability. In order for a tank to do enough healing that I’d want to spend my Chi on Enveloping Mist instead of Uplift, they’d have to do enough healing that it reduces the number of Uplift casts I feel I need to stabilize raid health. If I were a Priest, it would have to be enough that I’d want to spend my GCDs casting Greater Heal instead of Prayer of Healing. Even to match a portion of a dedicated healer’s HPS like that would be difficult for a tank to do, as Theck points out. Otherwise, you’re just adding to overheal, and occasionally changing a death into a 99% health spike at best, which happens even more rarely than a tank hitting exactly 0-1%, for reference.

  4. Mavalda says:

    Thanks again for the work you do! I have a side question I don’t seem to be able to find an answer to, maybe you have a feel for the answer.

    When does it become worthwhile to break our T14 4-pcs set bonus?

    I just got a huge upgrade (483 to 522!) which I would equip in a heartbeat if it didn’t mean breaking the set bonus.

    Thoughts?

    • Scoutyou says:

      Break it.

    • Theck says:

      I really like the T14 4-piece, such that I haven’t given it up despite having 3 509->522 upgrades for those slots. I think once I get another 522 piece I’ll probably make the jump.

      In your case, the upgrade is more significant – you could probably feel safe making the upgrade right now, definitely if you can grab a 2nd upgrade.

  5. Thels says:

    What effect would the set bonus have on Sha like fights where we get a single strong hit with long intervals? Since we generate enough HP in between these intervals, it seems quite likely that we could enable both SotR and the set bonus for these attacks.

    I do realize that Sha was an exception for T14 with it’s Trash attack, but we seem to be getting more and more of these abilities in T15, such as Horridon’s Triple Puncture, and Tortos’ Bite attack.

    • Theck says:

      It obviously depends on whether the attack is blockable, of course, but it seems obvious that it would be worth saving up 4 Holy Power to use SotR and a 1-point preemptive WoG to put up the block buff. Since SotR and WoG don’t share a cooldown or GCD, it should be possible to do this ~2 seconds before the attack.

  6. Wrathblood says:

    Theck, this is magnificent. I had to go clean my shorts after reading it.

    However, I modestly disagree with a couple of your conclusions. First, I’m not surprised that Mastery > Haste in terms of the “shift” rotations (I’m going to dispense with the whole “Control/” part because it just kinda goes without saying at this point, other than possibly for a Pure Avoidance build). I had this discussion on the WoW Forums the other day, and someone pointed out that while you get more opportunities to play with your ShoR timing with Haste, you get more time in which to time each ShoR with Mastery and of course its exactly proportional. The issue then becomes that, since you have more time to work with it, the odds are good that you’re likely to be able to get more value out of *each* ShoR with Mastery than you are with Haste. Over the course of a boss fight, this could add up to a significant advantage for Mastery.

    I’m *VERY* curious to see what’s up for late this week. I’m pretty sure I know, but I’m a little surprised as I wouldn’t have expected it to be a significant reduction in spikes which is the real issue here. I mean, from the data above, Haste vs Mastery isn’t even close.

    Second, perhaps I’m biased, but I’m not sure that I agree with 2-piece not being worth maintaining for Survivability. To some degree it depends on your gear choices and obvious WoG OH%, but the numbers are quite strong. If you’re going Haste, even at 100% WoG overheal the ST and TS rotations both are clearly to superior to the regular rotation and both Shift rotations in supressing spikes while also taking 5% less damage. Of course, the TS and ST rotations will have much lower DPS than the “normal” rotations and if you’re stacking Haste higher DPS is probably one of your objectives. However, if you’re in a situation in which you’re taking a lot of damage (and the incoming physical damage you’re taking is blockable), maintaining the 2-piece looks like a no-brainer.

    The 2-piece isn’t quite as attractive for Mastery but at 50% OH% (which I personally agree is probably slightly low but closer to correct than 100% based on nothing but my gut) it takes 10% less damage than the best Shift rotations and is superior at spike suppression for 4+ swings which is the part we care about more.

    • Theck says:

      Regarding haste vs. mastery: it’s not exactly proportional. Remember that:
      1) Haste has a much better conversion than mastery. You get 1% haste per 425 haste rating, but it takes 600 mastery rating to give you 1% mastery.
      2) You start off with 38% mitigation on SotR baseline, which should multiplicatively help haste (and is irrelevant for mastery).

      The argument that Mel and I batted back and forth is that while mastery makes every SotR stronger, haste gives you more of them. The logic being put forth was that the difference between a 40% and a 50% SotR isn’t that big a deal to a healer, either one blunts the attack enough to make it “safe.” So having more attacks mitigated by 40% may be a better deal than having fewer attacks mitigated by 50%.

      Now, I don’t think the premise itself is flawed. But the numbers didn’t seem to work out that way, at least in terms of raw damage intake. There may be an argument to be made that healer reaction still favors haste, but that gets into wishy-washy territory that I prefer not to tread on.

      I think that what is likely the deciding factor is the passive bonus of block. We get about 20% more block chance from the mastery gear set, which does wonders for spikes because you’re randomly shaving 30% off of attacks here and there, which is almost like getting free (albeit weaker) SotRs. Especially when coupled with an intelligent shifting queue that sees a block as “safe” (threshold of 0.8), that allows you to apply the more powerful SotRs very effectively.

      Regarding your opinion on the 2-piece: I don’t agree that ST/TS are strictly superior in haste gear at 100% overheal. In the critical 3-5 attack category, they just permit too many 90% spikes, and don’t universally win in the 80% category. They’re more or less tied for 6 attacks, and ST/TS hold an advantage in the 2- and 7-attack categories, but those last two are also the least relevant.

      • Wrathblood says:

        No, no. I meant proportional in the sense that if you have twice as many ShoRs to potentially time, you’ll only have half as much time with which to time each individual ShoR. Twice as much time means you’ll almost certainly do better with each individual ShoR, giving Mastery an advantage.

        Which makes sense looking at it by GCD. Looking at the SH1 and SH2 methodologies, Mastery generates HoPo slower, meaning there will be, on average, more GCDs per ShoR during which it is at 3 or 4 HoPo and thus potentially able to jump in and disrupt a streak. With Haste, there are simply less opportunities to do it per ShoR. From the results above there might be less opportunities to do it overall which would explain why the Shift rotations added so much less value than they did for Mastery. Much less opportunity to take advantage of superior timing.

        As for a Haste build with 2-piece tier and 100%OH on the WoGs… um, I must have been reading the wrong chart because you’re entirely right. Hmm. Perhaps a 65-75% WoG OH% chart would worth considering down the road.

        • Meloree says:

          With Haste, the GCDs are shorter. You have a nearly uniform number of GCDs per ShoR in any control scheme, what changes is the length of GCD. It absolutely doesn’t make sense to look at it by GCD.

          There will be more wall-clock time per ShoR, or more boss-swings-per ShoR, or almost any other metric that doesn’t depend on the paladin’s personal haste, though, and I think your argument, if I understand it, depends on wall-clock time anyway, so I’m really just being pedantic.

          I’m not sure I buy the argument, though. The haste scheme will almost certainly use more “bleed valve” ShoRs by virtue of having a higher overall uptime. Since there’s less wall-clock time uncovered by ShoR, there are just fewer opportunities to make a difference with timing. I suspect, then, that that’s part of why mastery shows better.

          The other half, of course, is break-points. C/HA’s ShoR is at about 40% mitigation, and C/MA is at about 60%. H/H/ShoR is an 80% spike for haste, and a 70% spike for mastery – as expected. But timing, rather than blind usage, prevents mastery from wasting those ShoRs on “dodge/dodge” strings – and the opportunity cost of a wasted ShoR is much higher for mastery.

          • Wrathblood says:

            Mmm, yeah. Really, I should have said “boss swings” rather than GCDs since I was thinking in 1.5 second increments. GCDs muddles things unnecessarily.

            On the one hand, Haste will certainly have more ShoRs with which one can extra extra value by cleverly timing. However, Mastery is more likely to get extra value for each extra ShoR because you’ve got more boss swings over which to decide whether or not to use each ShoR. While it seems like this might be a win for Haste by sheer volume, the problem is that this is *all* Haste brings to the table while Mastery gets the bigger ShoRs (which technically will help stomp on spikes better, though honestly I don’t think it changes things much) plus still has Block% in its back pocket which gives Mastery a big advantage.

            Of course, I see Theck’s just posted something new so my argument might be out the window already.

  7. Booi says:

    Not sure if you’re going to bounce warrior simulations through again – But if you do, do you think you could modify the barrier model towards intelligent-use like you have with SotR?

    Essentially, if the next ability will rage cap you, or if you take > x% of incoming damage then use barrier. If you allowed buff clipping, you’d need another parameter for remaining shield size vs boss swing size (or similar).

    Either way – excellent reading as always,
    Booi

  8. Pingback: [Prot] 5.2 - Its Time (to Begin... not hating avoidance) - Page 2 - Elitist Jerks

  9. Greywolfamakir says:

    Hello

    I read your great entry called “Control Shift En-tier” and I have one question for You (REALLY EASY QUESTION).

    What mean the marked zone?
    [IMG]http://i.imgur.com/lrKTTQV.jpg[/IMG]

    What mean the marked zone?
    [IMG]http://i.imgur.com/WsIejoH.jpg[/IMG]

    Thank You very much.

    • Thels says:

      The 2 attack, 3 attack, 4 attack, etc… looks at chains of attacks. For example, the 4 attack considers attacks 1-4, 2-5, 3-6, 4-7, etc…

      The 60%, 70%, 80% and 90% list the number of attack strings that dealt more than 60%/70%/80%/90% of full damage.

      So for example, say a boss hits us for 10000 damage per swing. A string of 4 attacks that wouldn’t be avoided/blocked/mitigated with SotR would then deal 40000 damage. The line after 60% indicates how often a string of 4 attacks in a row deals 24000 (60% of 40000) damage in total.

  10. Bassm says:

    Well this is more for Theck or Wrathblood.

    Is it possible to contact either of you privately? I kind of have some things to input on my own with the whole Mastery vs. Haste, but of course it’s nothing mathematical like everyone else has. I guess I’ll shed what I mean.

    The other day I decided to “try” the Mastery set over Haste. Of course my Haste went from a whopping 15% to a measly 4%, so my HotR / CS went from under 4 sec to about 4.31 sec for a cd. Anyway, I’ve been using Skada’s Avoidance and MItigation to at least eye out what’s going on. We sadly didn’t down Thunder King (though whether this is because I’m 484 iLvl gear or not is debatable, though I’ve had crappy luck with the Shield / weapon drops sadly even in LFR), but I also feel that Haste has a better sense of “control” behind it than even the Mastery build does. Again, I realize that I don’t have numbers and, sadly, I’ve been only doing 10m Normal content (it’s… been slow progress and even my computer is on its last life I think), I guess this is what I’m trying to say:

    I felt I had more control with what was going on in the Haste build than the Mastery.

    With the Haste build, I had a better time aligning SotR – especially if I needed it for an extended period of time or if I accidentally shotgunned it (whether due to lag or simply poor timing). Faster HoPo regeneration also meant that if I did hiccup, with Bastion of Glory charges, I could always heal myself back up. I have no clue why but we were using 3 healers on ToT’s first boss last night, and they stated that they didn’t really realize much difference with my damage intake switching from Haste to Mastery. I know when I was doing Vault content (the group I’m in has mediocre graphics cards atm… so HoF is… kind of hard), and I was in 480ish gear (which is arguably high even for Vaults when entry level is 460).

    Now the Skada addon I was using did show something different of course, but I think it was because there was more passive Blocks going on (due to Mastery’s benefit of increasing Block) as well as the minor addition of absorption of attacks.

    I don’t know, because I realize I’m no expert and I really don’t have mathematics on my side (I wish I delved further into it much like you, but alas a computer degree was what I wanted); however, I think the Mastery vs. Haste debate really hinges on a lot of what is going on behind the player more than anything – which you noted nicely on your recent blogs as well.

    Sorry for the bloated post. I’m trying to update the guide I wrote on the WoW forums, and I don’t really want to sound inaccurate. It’s not as great as people like Wrathblood’s, but I guess “one can try, right?”

    • Theck says:

      Sure, you can e-mail me at theckhd @ gmail dot com.

      I think what you’re observing is a combination of a real effect and observer bias. Play feels much faster and fluid at higher haste, and it does give you a better sense of control. For one thing, it’s easier to recover from mistakes, because your holy power generation rate is much higher. If you blow your SotR early in the Mastery configuration, you’re often just screwed for 7 seconds. Do the same in the haste set and you might be able to recover.

      Mel and I both hypothesized exactly what you’re experiencing – that haste *should* be better at smoothing because it affords more control, and covering more attacks with a smaller SotR should be smoother than covering fewer attacks with a larger SotR. The data didn’t seem to agree in this post, partly because I think the algorithm might be *too* robust. Having a really strong SotR via mastery makes a big difference when you can time it flawlessly, like the sim can, especially given the extra blocks. But maybe for a player, who will occasionally make mistakes that a simulation won’t, the reverse is true?

      It’s hard to say. Writing a simulation that accurately models player mistakes is tough, and probably not worthwhile either since different players “Screw up” differently.

      That said, you were probably running Sacred Shield, yes? If so, a new blog post just went live which should help explain some of what you experienced. I suggest you read today’s post and see if that syncs up with your impressions.

  11. Pingback: Is Nothing Sacred? | Sacred Duty

  12. Geodew says:

    So if my understanding is correct,

    “S” is spam SotR at 3(+) HoPo
    “SHx” is reactive SotR usage with a specified cutoff (0.8)
    “T” (for “tier bonus”) will use WoG any time the tier bonus buff drops off, as well as if we hit 5 holy power so that we don’t waste anything.
    “TS” tries t0 keep the tier bonus up at all times and uses SotR as a bleed mechanism. The logic is to cast WoG every time the tier buff drops off and only cast SotR when we’re at 5 holy power.
    “ST” is a “first come, first serve” style rotation that tries to keep something up at all times, whether it’s SotR or the tier 15 set bonus. It will prioritize SotR over the tier bonus. The logic is roughly “if neither buff is up and HP>=3, cast SotR; if neither buff is up and HP<3, try to cast WoG.”

    Several questions:

    1. What's the "bleed" for ST? What does it do at 5 HoPo when tier bonus is up?
    2. Won't tanks dynamically change their rotation (WoG at very low hp) in reality, not just hit the button blind of current HP, resulting in ~50% overheal? Is this sim useful considering in that case we'd be more likely to have 5 Bastion of Glory stacks in that case?
    3. What if you use WoG in reaction to a spike, like "SHx" but with WoG? Should we use it only with a certain amount of stacks? Or since we've discovered "SHx" is way better than "S" shouldn't we apply that logic to TS and ST somehow?

    • Geodew says:

      Oh, and I’m aware that considering Sacred Shield makes Haste much stronger, for purposes of discussion. That is, I read “Is nothing sacred?” also.

    • Theck says:

      1. SotR
      2. Maybe? Not all tanks do, and not in all circumstances. Also, there’s no way to use a health-threshold conditional in a sim that doesn’t model current health. In any event, a good portion of this sim was intended to address the suggested idea of keeping the tier bonus up as a maintenance buff, which necessarily requires casting WoG independently of health.
      3. I haven’t tried this, but it’s certainly something we could consider doing in SimC, which has vastly more options for conditionalization.

      • Geodew says:

        Oh, you *did* reply; I guess I didn’t “follow” the thread. Oops 😛

        By a health conditional, I really mean we could work with “after taking a certain amount of damage from the boss,” our Theck-sim equivalent. I guess what I mean for #2 then would be “maybe it’s only worth it to grab the block buff if you’re not going to overheal, in which case you’d be more likely to have 5 BoG stacks. For the sims here, we were essentially spamming WoG before the stacks of BoG got very high, so it may not be realistic.”

Leave a Reply