Legion Theorycrafting [7.3, All specs]

[Effinhunter]

I had some time during a lunch break today to look at MM AoE in the APL.

The following tests some action changes and lack of actions within MM AoE conditions over a 15s fixed window for 3 and 5 mobs. Among the things I'm varying here are:

1.) whether under ts (true shot) or agi potion (only applicable for a little while longer)
2.) not using Aimed Shot (noais)
3.) no barrage (nobarr)
4.) using marked shot immediately after gaining Hunters Mark debuffs

"norm" == "normal rotation in the current MM APLs". I used the default MM APL posted on the MM Simc discussion for 1-3, but #4 is definitely not included in the current MM APL. It is underutilizing MarkS during AoE periods and this results in undervaluing MM AoE by ~15%.

Three Target


Five Target


Simcraft source

Basically, this outlines some pretty fundamental things about AoE situations for MM:

1.) It's pretty damned important that you keep using AiS. You cannot just pool focus for MarkS and Barrage and use Sidewinders on CD. You have to be intelligent about each usage.
2.) The MM APL should probably be updated to include using MarkS when Hunters Mark is up in a much higher priority when sidewinders is enabled. For most hunters this is pretty obvious. The interaction between Marked Shot and Hunters Mark is extremely powerful, and parsing on fights with AoE is going to revolve around not only continuing to use AiS properly, but also taking advantage of MarkS with HM and obviously using Barrage at a reasonably high priority.

Barrage will be roughly 20% of your damage on 3 target aoe segments and ~22% on 5 targets, scaling up from there. Positioning for maximum Hunters Mark targets with your sidewinders is going to be a huge factor in your AoE dps. These sims assume you hit ALL targets with your sidewinders and apply HM to all. You should practice your SW so you understand the blind spots and how these SWs move. You also need to get used to clicking targets that are farthest away, but in range, or more importantly, positioning yourself in range of the farthest target as/before the adds come out, and adjust your angles so that you hit all adds with SW. Then use your MarkS appropriately. You want to make sure you are consuming those HM before the adds die.

- - - Updated - - -

About MM Artifact Relics:

The following tests different relic bumps from Storm, Blood and Life. This can help you with optimizing MM relics in Legion, but there are a few caveats about what this really means (and a bug in Simcraft).


Simcraft source (max_time=450s, 20% dev, 10k iterations)

The bug here is that increasing levels of Called Shot is not actually increasing Sidewinders damage. Should be a simple fix in Simcraft. So, cs4-6 are undervalued and performing at same level as not having any relics.

Here are some things to understand about relics (and this includes some information from sims that I haven't posted yet involving ilvl upgrades):
1.) Each raiding gem tier upgrade (e.g. heroic version to mythic version) will contribute 1.2k-1.5k dps by item level increases alone to your weapon. So, going from heroic relics to full mythic raiding relics will be ~3.6k to 4.5k dps. That's ANY relic, not just the specialized relic trait upgrades.
2.) The mythic dungeon gems are around ~3k dps under the heroic raiding gems.
3.) The best relic to get while leveling may be +1 to Quick Shot (reduces cd of Trueshot), simply because of quality-of-life improvements (getting a big AoE cd more quickly for trash between dungeon bosses, for instance) but according to sims, when you get the core traits done in MM, you appear to get more from MfD as it has good interplay with Deadly Aim (crit damage for AiS). Precision is pretty damn good too.

[3clipse]

Fascinating results on the talent sims, particularly the BM ones. Using Dire Stable and Killer Cobra together theoretically beats all other combos, but if you drop one then the other stops performing as well. Makes intuitive sense, but still a really interesting synergistic effect.

The MM results are really interesting as well, I think everyone (myself included!) assumed that Lone Wolf would be the go to there but that doesn't appear to be true!

[Zoopercat]

After our discussions in Discord, I do think that the planned AMR way of doing simulation, stat weights, etc. is going to be really invaluable. I probably need to talk to you about AMR and navigating around it. I'm still hoping to get this all investigated in Simulationcraft, though, so others can replicate the results without needing to pay anything.

To make sure it's clear: the AMR simulator is free. You won't have to pay

[Effinhunter]

The Called Shots bug with artifacts should be fixed in tonight's Simcraft build, btw. Nakauri helped us out! Also, he said he was going to be incorporating MS at elevated priority in the default MM APLs because it turns out that it will help even on single target.

- - - Updated - - -

To make sure it's clear: the AMR simulator is free. You won't have to pay

Sorry, I'm really not all that familiar with the AMR system. I logged in for the first time in a long time yesterday or the day before, and I noticed most of the other features like Bag Optimizer and the planners seemed to be paid. Is the simulator always going to be free? Are you guys actually going to be running some of the more intensive sims for people for free? I mean, I wouldn't blame you for charging for that as these can be processor intensive and if lots of people are trying to run free, massive sims on your hardware, that takes up a lot of paid hardware time (some of my sims takes me half a day to run at this point on a high-end computer, so I can only imagine what would happen to your servers if lots of us were doing these). I know you also offer the sim for download, too, according to your wiki, but I haven't had a chance to look into the videos you have posted on how to use it.

[keygy]

Fascinating results on the talent sims, particularly the BM ones. Using Dire Stable and Killer Cobra together theoretically beats all other combos, but if you drop one then the other stops performing as well. Makes intuitive sense, but still a really interesting synergistic effect.

The MM results are really interesting as well, I think everyone (myself included!) assumed that Lone Wolf would be the go to there but that doesn't appear to be true!

you think gonna be the optimal choices? i tried AMR with way of the cobra/stampede and youre right the simulator show killer cobra/dire stable gives more dps than the other option.

[Azortharion]

You would be a fool to try and make talent comparison sims for Legion now, and even more so to try and draw conclusions from them.

[3clipse]

you think gonna be the optimal choices? i tried AMR with way of the cobra/stampede and youre right the simulator show killer cobra/dire stable gives more dps than the other option.

Well, for right now, it seems like Dire Stable/Killer Cobra is the optimal choice. Things are going to be fairly different at 110 - we'll have passive damage bonuses from Titanstrike, and a second pet, and another 1 minute cooldown. It wouldn't surprise me to see Way of the Cobra displace Dire Stable since having Hati will be an additional permanent +8% damage for Cobra Shot, which in turn would probably mean Stampede is the best level 100 choice (unless Hati's Kill Commands also benefit from Aspect of the Beast, anyone who's played beta know the answer there?).

Basically who knows what things will be at 110, but I sincerely doubt the best talent choices will be the same.

[silverstarzs]

To make sure it's clear: the AMR simulator is free. You won't have to pay

still curious - how accurate is the AMR simulator atm?

[Azortharion]

Dire Stable + Killer Cobra is very demonstrably a worse choice for live BM.



Base APL used is on http://pastebin.com/u/Azortharion

[3clipse]

Going by the OP and his posts in this thread, those BM talents are from simming level 100 in T18 Mythic gear (please correct me if I'm wrong Effin). The 110 sims hover around 200k DPS, not 60K, and he hasn't done in depth talent breakdowns for 110 yet.

That's an enormous disparity in both absolute DPS and talent setups between Azor and Effin's sims. I have absolutely zero idea which one is correct.

[Kwaai]

You would be a fool to try and make talent comparison sims for Legion now, and even more so to try and draw conclusions from them.

I think "fool" is a bit strong, don't you? What if people enjoy theory crafting for the sake of theory crafting? What if Blizz decides not to change things? What if Blizz decides not to change things that impact talent choices? It's not as if anyone's presenting this information as if it's gospel.

[Effinhunter]

Not really sure. I can try rerunning BM talent sims in a bit. With 89k, what is the max_time you are using here and what's the gear?

P.S. Haven't had much chance to tweak the APL for BM from hunter lodge forums

[Azortharion]

Correct 110 sims hover around 280-300k. That's for Emerald Nightmare raid content, I do not know what others are running.

My APL/base profile with gear and all is public.

I cannot speak for what Effin or anyone else is running, only that I believe my own way to be the best way (120s with 15% variation to get a decent split of 1 and 2-ring encounters while staying close to something realistic for as many people as possible), with fixed_time enabled since SimCraft is treating these short kills oddly for some reason (I actually don't know, mage nerds are working on that), and of course a very slightly adjusted APL, as well as running trinkets that are easily accessible, but also very good, and also ones that do not have a disproportionate effect on anything else (bit dumb to run sims with a Skull of War for example).

- - - Updated - - -

I think "fool" is a bit strong, don't you? What if people enjoy theory crafting for the sake of theory crafting? What if Blizz decides not to change things? What if Blizz decides not to change things that impact talent choices? It's not as if anyone's presenting this information as if it's gospel.

It's a foolish endeavour in the sense that it is a waste of time, time that could be better spent theorycrafting other things, or working towards future theorycraft/sims on these subjects being more valuable such as relevant APL craft.

It's purely a question of spending time efficiently. Obviously theorycrafting results now such as talents will rely on the APL, so it is foolish to work on those things in depth before working the APL.

[Effinhunter]

It would be similarly wayward to think we can only work on the APL without changing talents. We have to look at our class from several angles if we want to get this right.

[Azortharion]

Yes, but it takes APL work to get the most out of different talents. You won't see the best of every talent by not tweaking the APL's for them, and a lot of that remains to be done. Killer Cobra has a pretty decent implementation right now, and that's the only talent that requires in-depth APL tweaking for BM, fortunately. With MM, it's obviously way different.

[Kwaai]

It's a foolish endeavour in the sense that it is a waste of time, time that could be better spent theorycrafting other things, or working towards future theorycraft/sims on these subjects being more valuable such as relevant APL craft.

It's purely a question of spending time efficiently. Obviously theorycrafting results now such as talents will rely on the APL, so it is foolish to work on those things in depth before working the APL.

And the APL will at least partially depend on talents. You could just as easily turn around and ask what if they buff Way of the Mok'Nathal, and it becomes optimal to *never* let it drop, even at the expense of an X stack of Mongoose Bite? With talents and numbers in flux, there's nothing guaranteed to be not-potentially-wasted.

[Zoopercat]

@Effinhunter - yes the simulator will always be free, good question. The web sims let you go down to a 0.25% margin of error. If you want to go lower, we let you do that for free too, but you need the client to run it on your computer (so you don't put too much strain on our servers, like you guessed).

The main uses for the site, like the gear explorer, optimizing, playing with stat weights, Best in Slot, those are all free. The premium features ($1 a month when paying annually), are convenience features. These are things you can do on your own with spreadsheets, but Mr. Robot will save you time and do them automatically. Like finding the best gear from your bags or calculating the best bosses to do bonus rolls on.

@silverstarzs The simulator is actually very accurate. But let me explain what that does and doesn't mean:
--The spell mechanics are all very solid with no outstanding bugs. We've combed for bugs twice per spec now, and are going for a 3rd time soon. That doesn't mean there aren't bugs. But we are super thorough which is why there are so few bugs left. Part of the reason it's so bug free is all mechanics are verifiable on the wiki and a lot of people have been checking it out to verify the correct implementation (and report bugs).
--If a bug is found, we fix it within 24 hours, but usually within a couple of hours of reporting it. Bugs don't stay around longer than that.
--We verify spell data and changes with patches against actual in-game testing, to be super sure everything is accurate. We take a lot of pride in this project and want it to be in tip-top shape.

What it doesn't mean:
--APLs (rotations) are good, but not perfect. And as APLs change, so will simulator results. So be sure to look over each APL that you use in case you want to change any of the logic.
--Trinkets - while we coded in all of the trinkets, some of them have guessed values. For example, if something says it spawns a fun little helper dude that spams fireballs to do damage with you for 20 seconds, the rate of firing these fireballs is unknown until we get our hands on that trinket and do actual tests.

If you want to play with it, here's a page with help videos. Get started here: http://beta.askmrrobot.com/wow/simulator/run, and don't forget to change any gear or artifacts that you want (it's all super simple to change).

Hope that helps!

[Azortharion]

And the APL will at least partially depend on talents. You could just as easily turn around and ask what if they buff Way of the Mok'Nathal, and it becomes optimal to *never* let it drop, even at the expense of an X stack of Mongoose Bite? With talents and numbers in flux, there's nothing guaranteed to be not-potentially-wasted.

No one is talking about what Blizzard may or may not do. What I am saying counts regardless if they do anything or not.

The APL does not "depend" on talents, I am not sure what you mean by that. You build the APL around what the talents do, in order to extract the most from those talents.

[Effinhunter]

Dire Stable + Killer Cobra is very demonstrably a worse choice for live BM.



Base APL used is on http://pastebin.com/u/Azortharion

Are you just selecting two talents here? Instead of a full combination of dps talents? If the latter, what other talents were you taking?

[Kwaai]

No one is talking about what Blizzard may or may not do. What I am saying counts regardless if they do anything or not.

The APL does not "depend" on talents, I am not sure what you mean by that. You build the APL around what the talents do, in order to extract the most from those talents.

When I said the APL depends on the talents, I was saying that how play changes based on talent balance. If talents are likely to wildly change (such that simming one vs another isn't worth the time), why bother coming up with a priority list that might be wrong when the numbers pass hits?

You would be a fool to try and make talent comparison sims for Legion now, and even more so to try and draw conclusions from them.

Or did I misinterpret this? I assumed you were saying that because talents and numbers are in flux, it's not a good use of time to sim them.