[TC] 2T18 Temporal Heroes - Damage and Haste Details

[Komma]

This is partly migrated from a thread I made on NGA Mage forums (Chinese), explaining the 2T18 temporal hero mechanic, with slight updates. The original Chinese thread can be found at http://bbs.ngacn.cc/read.php?tid=8295006&_ff=182&_fp=5" target="_blank

Temporal Heroes
Arcane's 2T18 bonus (http://www.wowhead.com/spell=186166/ite ... e-2p-bonus" target="_blank) reads as follow:

Arcane Missiles has a 35% chance to create a Time Anomaly, summoning a random hero through time to assist you in battle for 10 sec.

This thread aims to explain the exact nature of how this set bonus works.

Temporal Heroes are summoned upon casting Arcane Missiles. Only the parent cast has a chance of summoning them, instead of individual missiles. They last for 10 seconds after being summoned, and come in 4 flavors: Jaina, Arthas, Sylvanas, Tyrande. The order in which they can appear is cyclic, but varies a tiny bit: it goes either J->A->S->T or J->A->T->S. However, since Sylvanas and Tyrande do exactly the same things, for theorycrafting purposes we can assume the order is fixed.

The heroes underwent at least a dozen tuning changes on the PTR, going up and down repeated times. The final change was made in a hotfix several hours before patch 6.2 was released, where all hero damage was nerfed by 40%. We tested most of these changes as they happened on the PTR, and recorded them in the PTR testing result thread (http://altered-time.com/forum/viewtopic.php?f=2&t=1614" target="_blank). I think it's safe to assume that at this point, Blizzard considers their damage balanced, and further tuning changes are unlikely to happen. This post has taken all hotfixes and changes into account.

Attacks and Damage
The damage done by Temporal Heroes is different from most spells. They have a high amount of base damage (which our spells have none of). Their attacks can crit, multistrike, scale with haste and versatility, but they do not scale with mastery. The strength of their attacks are further increased by their spell power and attack power. Their crit/multistrike/haste/versatility ratings are inherited directly from the player, while their spell power and attack power are dependent from player spellpower:

Code: Select all

pet_spell_power = 500% * player_spell_power pet_attack_power = 1140.8% * player_spell_power

With 4T18, the pets not only cast the "Temporal Power" 12% damage buff on the player, but also cast one on themselves. Each pet is only affected by their own buff, so they only get 12% increased damage even when multiple pets are up. All other player damage buffs (such as Arcane Power or Rune of Power) do not affect them.

The pets have 3 attacks. Arthas melees, Jaina casts Frostbolt (http://www.wowhead.com/spell=191764/frostbolt" target="_blank), while Sylvanas and Tyrande are archers that use shoot arrows (http://www.wowhead.com/spell=191799/shoot" target="_blank). To determine exactly how much damage the pets do, a number of us went on PTR to test them. The details of the attacks are as follow:

Code: Select all

Melee: spell ID 0, 2.0 second base swing timer melee_damage = 0.1963 * ( ( ( pet_attack_power * 2.0 ) / 3.5 ) + 7262.57 ) Frostbolt: spell ID 191764, 3.0 second base cast time, coefficient = 0.5 * SP frostbolt_damage = 0.6 * ( ( pet_spell_power * 0.5 ) + 2750 ) Shoot: spell ID 191799, 2.0 second base swing/cast time, coefficient = 0.266 * AP shoot_damage = 0.5285 * ( ( pet_attack_power * 0.266 ) + 3255.19 )

Melee and Shoot are physical attacks, which means they get reduced by armor, and do higher damage against lower level targets.

Haste and Breakpoints
Due to the fixed 10 second lifespan of Temporal Heroes, this leads to some breakpointing issues in theory. Arthas' melee and Tyrande/Sylvanas' Shoot ability both have 2 second base times, while Frostbolt has a 3 second base cast. In theory, with 20% haste, the heroes would be able to perform 10 * (1 + 20%) = 12 seconds' worth of actions within their 10 second life span, which is equal to 6 melees/arrows (ie. 6 * 2 = 12), or 4 Frostbolts (4 * 3 = 12). The amount of haste rating needed to reach this level is fairly straight forward to compute:

Code: Select all

Normal: ( (1.2 / 1.05) - 1 ) * 9000 = 1285.7142857142857142857142857143 ~= 1286 rating Gnome/Goblin: ( (1.2 / 1.05 / 1.01) - 1 ) * 9000 = 1183.8755304101838755304101838755 ~= 1184 rating

This breakpointing behavior is really obvious when we look at simulations:

Credits to Frosted for providing this reforge chart

The problem however, is that these breakpoints are not accurate in practice.
In game, all 3 heroes have a tiny delay, which mean they don't start attacking the very millisecond they are summoned. Jaina, Tyrande and Sylvanas all have delays between their casts, during which they do nothing. Here are some excerpts from Deathdefier, during a PTR dummy test with ~21% haste (https://www.warcraftlogs.com/reports/jC ... iew=events" target="_blank):

Code: Select all

00:02:05.556 Lady Jaina Proudmoore 3 (Deathdefier) casts Frostbolt on Weak Damage Dummy 2 00:02:06.103 Lady Jaina Proudmoore 3 (Deathdefier) begins casting Frostbolt

Code: Select all

00:02:50.973 Lady Jaina Proudmoore 5 (Deathdefier) casts Frostbolt on Weak Damage Dummy 2 00:02:51.574 Lady Jaina Proudmoore 5 (Deathdefier) begins casting Frostbolt

As the above logs show, the ladies seems to enjoy her half second tea breaks in between casting Frostbolts and shooting arrows.

This delay is affected by haste, but not in a predictable fashion. Performing another test at 25.49% haste, Deathdefier sees little to no delay (https://www.warcraftlogs.com/reports/jR ... iew=events" target="_blank):

Code: Select all

00:01:54.281 Lady Jaina Proudmoore 3 (Deathdefier) casts Frostbolt on Weak Damage Dummy 2 00:01:54.281 Lady Jaina Proudmoore 3 (Deathdefier) begins casting Frostbolt 00:01:56.662 Lady Jaina Proudmoore 3 (Deathdefier) casts Frostbolt on Weak Damage Dummy 2 00:01:56.776 Lady Jaina Proudmoore 3 (Deathdefier) begins casting Frostbolt

But when I tested it at around 34.06% haste, the delay came back, although slightly shorter than at low haste (https://www.warcraftlogs.com/reports/nH ... &options=0" target="_blank:

Code: Select all

00:02:51.631 Lady Jaina Proudmoore 3 (Tilde) casts Frostbolt on Weak Damage Dummy 5 00:02:52.205 Lady Jaina Proudmoore 3 (Tilde) begins casting Frostbolt 00:02:54.493 Lady Jaina Proudmoore 3 (Tilde) casts Frostbolt on Weak Damage Dummy 5 00:02:54.619 Lady Jaina Proudmoore 3 (Tilde) begins casting Frostbolt

These delays are not modelled in SimulationCraft, because we have no details on how exactly they work. This also means that the breakpoints seen in sims are not reproducible in game. For the same reason, Sylvanas and Tyrande seem to do noticeably more damage in the sim than in game. This is because they suffer heavily from delays, sometimes standing around for over a second.

The key takeaway here is that haste has weird effects on their attacks, and we cannot assume that 21%, or 25%, or any specific amount of haste is going to give them that extra attack. The only way to tell whether you have reached the breakpoint, is to look at logs and count the number of attacks gotten off by the hero. If Tyrande/Sylvanas are firing 6 arrows and jaina is firing 4 Frostbolts, you've reached it. If not, you probably need more.


FAQ
Q: Does all this mean that I shouldn't stack haste if I can't reach the breakpoint?
A: No, haste is still our strongest stat with T18 set bonuses. This is because mastery does not increase pet damage, while haste no only increases temporal hero damage, but also increases the number of hero summons, and the benefit we get from 4T18's Temporal Power damage buff. Even beyond whatever breakpoint there is, haste is still our strongest stat.
Haste being strong has many reasons, and whether you reach the fictional breakpoint matters very little for stat priorities.

Q: Why do pets behave this weirdly? Is it a bug, and will it be fixed?
A: Pets have always had cast time delay issues. For the longest time, Water Elemental had the same problem, and it was only rectified during late MoP. There is no saying whether Blizzard will fix this, but it does seem unlikely.

Q: Why are the pet damage values so weird? Are you sure they are accurate?
A: They are as accurate as we've been able to tell. Here is a screenshot of the spreadsheet we used to analyze the data gathered from PTR dummy testing:

[Frosted]

I think that saying the breakpoints are not accurate is not portraying the situation honestly.

The breakpoint(s) absolutely exist, and they exist at the values stated for the duration of time the pets spend attacking. What is in question is the magnitude of the effect in game vs. what is the full value of the breakpoint seen in the sim.

[Darthy]

https://www.warcraftlogs.com/reports/7v ... rt=6300718" target="_blank
After scrolling through those logs with my 24,63% haste it seems like i havent hit the breakpoint yet. For some reason jaina always casts 4 frost bolts while the 3 physical attackers only attack 5 times unless my haste enchant procced. Then again one time jaina spawned and only casted 2 frostbolts before dissapearing which i dont understand either (2:03)
I'll try out some haste food next raid and compare if theyll get one more attack in.

[Lak]

This is indeed very interesting and thank you both for very good work.

But what this means in practice, is that we cannot trust our sims and that we therefore can not know how to gear.
We don't even know if Arcane is in fact stronger than frost on single target?

* From this post I read Haste>Multi=Crit=Vers>Mastery
* Previous experience tells me I should probably put > instead of = above.
* My own (not trustworthy) sims tell me Haste=Mastery>Multi>Crit>Vers

So my assumption would be to go Int>Haste>Multi=Mastery>Crit>Vers, but thats pretty much based on educated guessing and I have no idea of relations. Is a worforged item with say multi+crit secondaries better than a normal with Haste+Multi?

Am I missing something or are we left in the dark until the pet damage can properly simulated?

[Frosted]

https://www.warcraftlogs.com/reports/7v ... rt=6300718
After scrolling through those logs with my 24,63% haste it seems like i havent hit the breakpoint yet. For some reason jaina always casts 4 frost bolts while the 3 physical attackers only attack 5 times unless my haste enchant procced. Then again one time jaina spawned and only casted 2 frostbolts before dissapearing which i dont understand either (2:03)
I'll try out some haste food next raid and compare if theyll get one more attack in.

This is exactly what komma is trying to clarify here - pet delay effecting the number of attacks the pets get.

This is indeed very interesting and thank you both for very good work.

But what this means in practice, is that we cannot trust our sims and that we therefore can not know how to gear.
We don't even know if Arcane is in fact stronger than frost on single target?

* From this post I read Haste>Multi=Crit=Vers>Mastery
* Previous experience tells me I should probably put > instead of = above.
* My own (not trustworthy) sims tell me Haste=Mastery>Multi>Crit>Vers

So my assumption would be to go Int>Haste>Multi=Mastery>Crit>Vers, but thats pretty much based on educated guessing and I have no idea of relations. Is a worforged item with say multi+crit secondaries better than a normal with Haste+Multi?

Am I missing something or are we left in the dark until the pet damage can properly simulated?

It means that the value of a breakpoint is questionable. The pet delay effectively will distribute the DPS gains from the breakpoint into the haste values beyond it, instead of acting as one large gain at a single rating. This is what the sim wont see. The sim is still fine for figuring out the average gains in haste value with things like more pets and more pet damage auras - which is where the majority of the non-1300 related strength for haste is coming from.

[Komma]

But what this means in practice, is that we cannot trust our sims and that we therefore can not know how to gear.
We don't even know if Arcane is in fact stronger than frost on single target?

There's a big difference between "pet damage is slightly inaccurate" and "sims cannot be trusted". For example, Tyrande and Sylvanas are each around ~1.5% of overall damage in a single target scenario. It's closer to ~2.0% of overall damage in simulations, due to not modelling the pet delay effect. Arthas uses melee autoattacks, which don't suffer from a delay. Whether an additional 1% of your overall damage scales with mastery, has a very small effect on the overall scaling strength of mastery.

* From this post I read Haste>Multi=Crit=Vers>Mastery

This post specifically discusses damage done by temporal heroes. Temporal Heroes do roughly ~6% of player overall damage in single target fights. Don't confuse that for general secondary stat strength for Arcane Mages overall.

Am I missing something or are we left in the dark until the pet damage can properly simulated?

We're not in the dark. What we need to do is take into account known differences between simulations and in game, and use that to adjust our interpretations of simulation results. That is why I remind players not to take the 1286 haste breakpoint literally, even if the sim suggests otherwise.

[Neculsa]

From what I have learned in this thread and other discussions, haste has beat over mastery and becomes the first stat we should build, largely due to the benefit from T18 bonus. However, one of my concerns is that whether this is true for all item levels of 4P T18, and whether Simc takes such advantages into account in its rotation. By looking at the sample rotation of my own sim result (4p T18, no mage trinket, ilvl 698), I didn't see the specific usage of AM as well as the pet bonus before the burning phase, which makes me really doubt on the simulation . It looks to me there are not too much changes in terms of rotation priority comparing to patch 6.1. Would spamming all AM collected before dropping the crystal be a better idea? If all Arcane missiles are used more efficiently, the benefit from haste might not be that greater over mastery. I do believe haste becomes the first priority stat after we strongly geared and get the mage trinket. However, thats mainly because you have already gained pretty much mastery from your gear. Therefore, is it a good idea to maintain some basal level of mastery at low item levels?

[Komma]

Would spamming all AM collected before dropping the crystal be a better idea? If all Arcane missiles are used more efficiently, the benefit from haste might not be that greater over mastery.

Why do you think spamming AM before Crystal/Burn is a better idea? How do you avoid wasting mana regen? How are you sure that this doesn't lose more than whatever gains you have in mind? There's a lot of leaps of logic in your post, and it's not anywhere close to leading to any haste/mastery conclusions.

What do you mean by more efficient AM usage? The simulation already does its best at avoiding losing Arcane Missiles due to 3 charge capping. 1.5 AMs out of 72.3 are lost to overflow in a simulation, and that is largely unavoidable due to 4 consecutive Arcane Blasts proccing AMs, or SN proccing double AM.

The general belief is that T18 set bonuses don't really alter our rotation at all, and mostly act as a passive benefit. You can maybe capitalize on Temporal Power with Supernova, but I do not see much reason

[Neculsa]

Why do you think spamming AM before Crystal/Burn is a better idea? How do you avoid wasting mana regen? How are you sure that this doesn't lose more than whatever gains you have in mind?

Even by "spamming" the AM, you only have 3 charges, and you might not need to cast the AM in a row. An AB could be introduced in between to avoid mana overflow. The key point is to obtain at least one stack of T18 damage buff before burning phase. This may be particularly true when the tome is equipped since you wanna spam AB during AP phase. What I saw from simc is that it directly drops the crystal even with 2 charges of AM. Would it be better to cast those 2 charges of AM and see whether we are lucky enough to get the bonus?

By "efficient usage of AM", I'm not saying about the capitalization of charges, again, its all about the banking of AM before the crystal. Whenever we see PC cool-down is gonna over, reduce the cast numbers of AM and get some stacks, just like what we did for SN. This may allow us to have a buffered damage during the crystal phase.

[Komma]

What I saw from simc is that it directly drop the crystal even with 2 charges of AM. Would it be better to cast those 2 charges of AM and see whether we are lucky enough to get the bonus?

You are welcome to experiment and crunch the numbers for this, but there is no established theory for this being a gain. I encourage you to publish any findings you might get. However, this doesn't come anywhere near a haste/mastery discussion. To jump from "maybe the sim isn't being optimal with AM usage" to "therefore the sim is giving us the wrong stat priorities" is a big leap of logic. It's also unlikely, because you're discussing whether two to three actions should get their order of usage switched up, once every 90 seconds or so. It requires much more than that to change stat priorities.

[supervixen]

Hey quick question for 2pc - if we don't have the PoF and Tome trinkets does our rot change at all? I'm assuming we're still using PC and SN. Again, I only have my 2pc tier 18.

Cheers