Fury Warrior Survey

[Zoopercat]

Hey all, hope it's ok to post this here. I'm researching an article for Blizzard Watch about how theorycraft affects our in-game choices. I have a 9-question survey that everyone can take, and I need to make sure I have enough Fury Warriors and Balance Druids in particular. There are more than 9 questions, but you are randomly given 9 of them. I want to make sure I cover the major fan sites so I get a representative selection of survey-takers.

I'll be posting my interpretation of the findings in my "Climbing the Ranks" column.

[Rikhart]

I highly doubt your

According to Simulations, Furious Strikes sometimes does 0.5% less and sometimes 0.5% more than the popular talent, Sudden Death

statement.

[Zoopercat]

It's true T18 4-pc warrior, standard rotation in simC. On a short fight, where the execute phase is short, Sudden Death wins out. On longer fights, Furious Strikes does.

[Bigbazz]

It's true T18 4-pc warrior, standard rotation in simC. On a short fight, where the execute phase is short, Sudden Death wins out. On longer fights, Furious Strikes does.

On single target my Fury sims were in the region of 3% in favour of Sudden Death (which has been fairly consistent throughout 6.2), with each additional target further increasing that difference. The thing is it's not really a make or break situation between those talents, but it's a clear difference and if the aim is to rank then Sudden Death is even more valuable with RNG allowing it to peak higher.. So it's better on an average pull and much better on a pull where RNG is favouring you.

Ask Mr Robot is a great tool but it has typically been very innacurate regarding trinkets and basic stat weights. For example your default for Arms has Crit almost as strong as Strength with Mastery a long way behind, it's impossible to think those weights could be the result of a simulation. Stat weights change a lot depending on gear, but since BRF times it has been fairly consistently Mastery that wins out for Arms and by a massive margin if you're running T18, by an astronomical margin if you also have the Archimonde trinket.

I did your questionaire, some questions need multiple answers to be ticked. Very few people go about things in just one way, it's usually a large combination of things that arrives at a popular conclusion.

[Zoopercat]

The stat weights are all based on simulations. We have a program that picks a real set of gear then runs simulations. We then use the stat weight results and put it in to the optimizer to pick a new set of gear (based on those weights). We then run the simulations again on that new set of gear. Rinse and repeat. We do this until there is a stable set of weights that picks the best set of gear, OR until we find out that weights dramatically change with something in particular.

If you look at our site there are 2 stat weight options for arms, one of them are weights specifically for the class trinket, which (should) favor Mastery. However, looking at it, the weights look the same as the default set. I've pinged our stat guy about it to investigate. It looks like those weights might not have been updated, in which case I apologize and they will be fixed soon. Until then, make sure you update them to your liking. In fact, if you have a set you'd like to share, I'd like to check them out.

As for trinkets, we simulate those across ST and MT fights and get the average value across all fights. We use that value when ranking them. That's a bit different than most theorycrafters, and the MT simulations in SimC aren't that great. Also, most people switch trinkets (if they have the choice to) between different types of fights. So it's hard to recommend just one. We're actually working on better ways to suggest trinkets for Legion, so stay tuned.

As always, I appreciate all feedback. It helps us direct our focus to future updates and features!

[Bigbazz]

The stat weights are all based on simulations. We have a program that picks a real set of gear then runs simulations. We then use the stat weight results and put it in to the optimizer to pick a new set of gear (based on those weights). We then run the simulations again on that new set of gear. Rinse and repeat. We do this until there is a stable set of weights that picks the best set of gear, OR until we find out that weights dramatically change with something in particular.

If you look at our site there are 2 stat weight options for arms, one of them are weights specifically for the class trinket, which (should) favor Mastery. However, looking at it, the weights look the same as the default set. I've pinged our stat guy about it to investigate. It looks like those weights might not have been updated, in which case I apologize and they will be fixed soon. Until then, make sure you update them to your liking. In fact, if you have a set you'd like to share, I'd like to check them out.

As for trinkets, we simulate those across ST and MT fights and get the average value across all fights. We use that value when ranking them. That's a bit different than most theorycrafters, and the MT simulations in SimC aren't that great. Also, most people switch trinkets (if they have the choice to) between different types of fights. So it's hard to recommend just one. We're actually working on better ways to suggest trinkets for Legion, so stay tuned.

As always, I appreciate all feedback. It helps us direct our focus to future updates and features!

I'l give you an Example again relevant to Arms Warriors. AMR comparison against a 720 Archimonde Trinket With socket (default choice on all fights as Arms + bonus socket stats). Results were the same with the default weights (Crit best) and accurate weights (Mastery best).

AMR Suggests

Drinking Horn (730) > Drinking Horn (715) > Chorus (730) > Pebble (720) Worldbreaker (735) > Chipped Soul Prism (705) > Worldbreaker (720 + Socket) = Drinking Horn (700)


Drinking horn is a trinket that would not be used on any fights as Arms (let alone Chorus).. Arguably you could get away with using it on AOE fights but it's not desirable for Arms progress raiding nor ranking, maybe you could get away with it on an Iron Maidens fight if you removed the execute phase... The fight just doesn't exist where you would use it as Arms. On the other hand the Worldbreaker is arguably one of the most important items in an Arms Warrior toolset, we use it on every single fight.

Our choices at the top end of raiding, or even in the mid-tier players among enthusiastic Warriors are not just based on simulations but on ingame testing, on theorycrafting based on fight mechanics, different phases and based on pushing certain thresholds within a fight.. For example we need "X" ammount of dps at this specific point to push this phase in a fight.. So while the SC simulations are important they are only supporting theorycrafting that is formulated based on the actual content.

This trinket priority list just doesn't add up in any way, not in the sim and not in the game. So I definitely stand by what I say with the trinkets.. AMR itself is fantastic as a tool, but beginners should beware that it can mislead them.

[Archimtiros]

It's true T18 4-pc warrior, standard rotation in simC. On a short fight, where the execute phase is short, Sudden Death wins out. On longer fights, Furious Strikes does.

Actually that's not true. Furious Strikes isn't incredibly far behind, but isn't as good as Sudden Death, on either a long or short fight and pales in comparison when any amount of secondary targets are introduced. Off the top of my head, I would guess that it probably looks good because of the stated way you do simulations, using a rotating variety of gear. With the class trinket, Furious Strikes becomes incredibly wasteful, as we no longer have the GCDs to devote to it, and our added rage income causes Rage overflow and waste. Without the class trinket, it still isn't an amazing talent, but this does not happen, so it looks better in comparison.

As good as a tool as AMR can be, Bigbazz is right in that its default settings are often inaccurate specifically with regards to stat weights and trinkets rankings (for example, that blurb about ST and MT trinkets is simply inaccurate for Warriors), I believe we spoke of this during MoP as well. If you would like any help looking over your Warrior priorities or SimC profiles, let Collision, WarriorSarri or myself know.

Finished your questionnaire. Only a couple issues with it:

• The FS/SD scenario question doesn't specify level of progression making it difficult question to answer. I used the Why? box to enumerate.
• There was a Balance Druid trinket question after having selected that I do not play a Balance Druid. Not sure if intended.
• When choosing talents, etc, there was no option for "I write the guides and theorycrafting posts".

[Deleted]

As for Ask mr robot and trinkets. I can't imagine using something different than Worldbreaker's Resolve on any fight as Arms. When I play as Arms I stick to its "core" - ss/execute + amazing cleave. I cannot think of using EDH on Arms even on something like HFA. But I guess above posters explained it very well anyway. Just my 2 cents.

[Zoopercat]

I'll have our warrior expert post his simulations and results, as it does show to be very close in value. He'll be able to discuss the finer points with you, I don't want to get it wrong in translation.

I'll definitely hit you and Collision up for Legion. For WoD, would you like to add your weights to the theorycrafter-specific list? We add those as stat weight presets for each class. People have liked that a lot. It's hard to come up with 'default' weights that everyone likes, so we've been trying to expand and give users more options. And of course, they can always be edited. I could leave the weights blank and never suggest weights, but I think that would be worse as the average player probably wouldn't take the time to find the weights. So presenting them with more options feels like the right way to go

Again, thanks for taking the time to make a post I'm reading each survey response to 'why' and categorizing them. It's quite interesting to see how things are shaping up!

[Archimtiros]

As for Ask mr robot and trinkets. I can't imagine using something different than Worldbreaker's Resolve on any fight as Arms. When I play as Arms I stick to its "core" - ss/execute + amazing cleave. I cannot think of using EDH on Arms even on something like HFA. But I guess above posters explained it very well anyway. Just my 2 cents.

This is what I meant by the idea of swapping trinkets. While it's absolutely true for some classes, for Warriors, we pretty much exclusively use Worldbreakers and Empty Drinking Horn (Fury) or Unending Hunger (Arms). While a pure 5+ target simulation with 100% uptime might show the Cleave trinket as being much more valuable than Worldbreakers, that type of fight doesn't actually exist in the game, so it's practical value decreases significantly.

- - - Updated - - -

I'll have our warrior expert post his simulations and results, as it does show to be very close in value. He'll be able to discuss the finer points with you, I don't want to get it wrong in translation.

I'll definitely hit you and Collision up for Legion. For WoD, would you like to add your weights to the theorycrafter-specific list? We add those as stat weight presets for each class. People have liked that a lot. It's hard to come up with 'default' weights that everyone likes, so we've been trying to expand and give users more options. And of course, they can always be edited. I could leave the weights blank and never suggest weights, but I think that would be worse as the average player probably wouldn't take the time to find the weights. So presenting them with more options feels like the right way to go

Again, thanks for taking the time to make a post I'm reading each survey response to 'why' and categorizing them. It's quite interesting to see how things are shaping up!

Feel free to, that's what we're here for.

You're completely right that default weights are useful, and for some classes, those weights never change dramatically. Warriors have been in a weird spot for the last couple of expansions though because their weights do shift depending on where our individual stats are at or what trinkets we are using. This is why we create generalized "rules" for our guides, such as "Once you exceed 30% Crit, Mastery becomes better than Critical Strike". I'm not sure how your scripts are set up, but it might be possible to set up "evolving" stat weights that automatically update based on the input profile.

Example: Player uses default Arms profile stating Crit > Mastery, etc. Player selects Tier 18 gear and Worldbreaker, profile automatically updates stats to Mastery > Crit.

That might be too much work, but I'm just brainstorming here.

[Zoopercat]

Archimtiros, I love brainstorming! We have some ideas queued up for Legion that will make weights more dynamic. We love solving hard problems - take the optimization algorithm when reforging was around, for example. Most characters had a few quintillion combinations of relevant gems, enchants and reforges, and we were able to find the single best optimization (based on the given weights), in about 1/10th of a second. That was a pretty huge project.

What I'm getting at, is hard problems are fun to solve. But we tend to group them into big site redesigns with expansions. Since we have to re-do a bunch of code to support new things (like artifact weapons), we use that opportunity to solve other hard problems. So stay tuned on having dynamic weights. We hope to be able to solve that.

As for swapping trinkets (or not swapping them in the case of warriors), you might very well be correct. I'm NOT the warrior expert, and was just thinking more generally about how trinkets are hard to recommend overall. But I see what you are saying - if Warriors tend to not swap them, could we adjust things to suggest different trinkets? Yes, we can. But I will leave that up to Yellow to discuss with you (our warrior expert). He knows much more about that than I do.

[Yellowfive]

Hey -- I did a lot of the simulation for warriors, and I play a fury warrior as well. This was meant to be an example more than anything, one that we used in a talk at blizzcon. Here are a few of the simulation results (using 622-01 version of simc... web page doesn't have the 6.2.3 build yet I guess, but won't change things).

For starters, I was working with a heroic-geared SMF fury warrior (that is what I play). If I run a simulation using roughly BiS gear at that level, empty drinking horn + worldbreaker's, on a typical 450 sec single-target fight in simc, using sudden death:

81857 DPS

Using furious strikes, all else the same:

81419 DPS

That is a 0.5% difference in favor of Sudden Death, not very much.


My warrior is roughly the same item level, but I have been unlucky with trinkets, so I have a ilvl700 drinking horn and a discordant chorus. This is pretty representative -- not everyone is going to have the best stuff. So for my character the sims look like:

65372 DPS -- sudden death
65853 DPS -- furious strikes

So in my case, furious strikes comes out to 0.7% better than Sudden Death.


Pretty close in both cases, but you can see how depending on your particular gear, it can go up or down.


For AoE... Furious Strikes is probably the worst choice, but still, not by a large margin. Sudden Death or Unquenchable Thirst both make more sense. It's really hard to get a good idea about AoE in simc though, and the default rotation makes particularly bad use of Unquenchable Thirst. I don't have time at the moment to dig up my custom rotations... but I'd be glad to post more info if people are interested.

Bottom line: Sudden Death is not always the best, and when it is, it's not hugely ahead of the others. So the basis of the example in the survey is sound.

[Darkfriend]

First off, you shouldn't be using a 450s fight length. If you are trying to model an actual fight you'll be using a time significantly shorter. Results that don't mirror any actual in game reality, while interesting, aren't relevant.

Oh and don't use DR for ST.

[Archimtiros]

You're right, it's definitely not a large margin, but the topic of discussion was what is best. For that we have to decide what we are looking at, a simulated vacuum using a specific setup and situation, or a realistic situation based on the tier.

To give you an example: AoE works fine in SimC, you just need to have it set up right. Simply setting X number of targets is unrealistic, since there are few actual encounters with fulltime cleave, and there are ways to set up custom fights, but that requires intimate knowledge with the script. The best way to do it is to change the Fight Style to HecticAddCleave. This was designed in MoP to reflect a Horridon style encounter, which consists of a single priority boss target, intermittent waves of adds, and moderate movement.

The default profile doesn't make great use of UqT simply because we couldn't find a good way to make it work. It's quirky and requires an inordinate amount of extra scripting due to the way it fits into the rotation, and the results were always too low to justify it.

You're logged out in tank gear, so I can't really say why you got Furious Strikes ahead of Sudden Death, though I would reckon part of it is due to being SMF, and part due to not having WB, or using Dragon Roar instead of Storm Bolt for Single Target, thereby freeing up an extra GCD for Wild Strikes, or maybe using your stronger weapon in the offhand to reinforce the damage of Wild Strike. All of these are common, suboptimal mistakes that players use, and will cause results to vary.

I'm not trying to say there isn't a way to make Furious Strikes better, you can always find a situation in which a talent shines when it normally wouldn't (ex: Shockwave, Arms Bladestorm, Fury UqT), but they are generally unique, and thus not normally given as advice. In 99% of situations, Sudden Death is the better talent. It scales better with increasing Crit and Haste, it uses less GCD's, and there are only two purely single target fights in the tier to even consider using Furious Strikes on.

So all things considered, Sudden Death really is the best choice, though I agree not always by a huge margin. Whether or not a particular player cares to optimize to that degree is an entirely personal choice.

- - - Updated - - -

Still, I feel like this has gotten a bit off topic. I don't mean to start a debate quibbling over a <1% optimization. I only meant to point out some of the inconstistency with stat weights and how they could better represent the shifting priorities that Warriors often have to deal with. The soft cap function introduced in SoO (I believe?) was a great step in mitigating that problem, but there are still some inconsistencies, such as looking at stat weights for my Arms profile defaulting to Crit > Mastery as Arms, when that should be reversed. Crit can be better without the T18 set bonus, but the system doesn't seem to take into account that I do have it.

- - - Updated - - -

Regarding Bigbazz's trinket concerns,

I'l give you an Example again relevant to Arms Warriors. AMR comparison against a 720 Archimonde Trinket With socket (default choice on all fights as Arms + bonus socket stats). Results were the same with the default weights (Crit best) and accurate weights (Mastery best).

AMR Suggests

Drinking Horn (730) > Drinking Horn (715) > Chorus (730) > Pebble (720) Worldbreaker (735) > Chipped Soul Prism (705) > Worldbreaker (720 + Socket) = Drinking Horn (700)

Here is the correct list of trinkets for ST:
*Note 1: I used T18M profile for this, which will show higher base numbers, though the order of results should be about the same regardless.
*Note 2: Individual trinket rankings compares one trinket to another without a second trinket. Trinket combinations are entirely different but I don't think AMR processes those anyway. I can post if interested, but for Warriors you'll get the results you'd expect, as our trinkets don't have a large impact on one another.



Here is the list of trinkets for HAC:



You'll notice the second column (I hope! ). The first is taking adds into account, the second is only accounting for the damage that actually hits the boss, which is very similar to single target, but has some overlap with AoE. For Arms this overlap isn't really apparent, but for classes with exploding effects, dot spreading, and the like, you can see a large influence. The largest impact it has for Arms is that you will have a lot of Rend damage on the adds translating into a lot of Mortal Strike damage on the priority target, but that really has nothing to do with trinkets.

You'll notice EDH is high, this is because Arms isn't especially proficient at dealing with large groups of adds, so they'll generally live longer and take less damage from primary attacks, allowing the DoT to do more damage in comparison. In an actual raid setting (M Mannoroth), Chorus would likely swap places, though the difference between them is very minor to being with (550 DPS).

[Yellowfive]

For the record, I always put my best weapon in my main hand, and always use storm bolt, avatar, and siegebreaker for single target fights: I did so in the above simulations. I have found it to be the best combo in simulation and in practice for single target fights. For my particular setup, Furious Strikes is simulating just slightly higher, though the difference diminishes as fight length shortens, breaking even around a 200 second fight. In better gear than mine, Sudden Death stays slightly ahead of Furious Strikes.

But like you say... this is really digressing from the main point that I wanted to help reinforce: the example in the survey is "good enough" that the survey serves its purpose. Zoopercat is trying to get at how people react to different ways that information is presented to the end user, which will hopefully be illuminating to some degree. But who knows, the point of doing a survey is to test a theory, might not work out.


I can take a closer look at our Arms warrior weights and rankings, but that's a separate topic. I haven't looked at them in a while, one of our other members ran the simulations that generated the weights we are currently using back closer to the release of hellfire citadel. We're happy to take feedback and re-evaluate a spec as time goes by and theory shifts.

[Archimtiros]

For the record, I always put my best weapon in my main hand, and always use storm bolt, avatar, and siegebreaker for single target fights: I did so in the above simulations. I have found it to be the best combo in simulation and in practice for single target fights. For my particular setup, Furious Strikes is simulating just slightly higher, though the difference diminishes as fight length shortens, breaking even around a 200 second fight. In better gear than mine, Sudden Death stays slightly ahead of Furious Strikes.

But like you say... this is really digressing from the main point that I wanted to help reinforce: the example in the survey is "good enough" that the survey serves its purpose. Zoopercat is trying to get at how people react to different ways that information is presented to the end user, which will hopefully be illuminating to some degree. But who knows, the point of doing a survey is to test a theory, might not work out.


I can take a closer look at our Arms warrior weights and rankings, but that's a separate topic. I haven't looked at them in a while, one of our other members ran the simulations that generated the weights we are currently using back closer to the release of hellfire citadel. We're happy to take feedback and re-evaluate a spec as time goes by and theory shifts.

Well we weren't even talking about the survey, I don't think anyone had an issue with that. The topic of FS only came up because Zooper asserted it was better, and asked you in to expand upon the claim.

The real issue brought up was regarding trinkets and stat weights. Trinkets I expanded on above, but the stat weights for Arms were obviously set up pre-T18, which is fine as it can be expected that many players won't get T18 quickly, but I would also argue that the expectation of T18 is to obtain T18. This is why a single set of stat weights isn't accurate, and I suggested some type of automatic update based on selected talents or gear.

[Darkfriend]

Archi I've been asking AMR to do something like that since SoO but apparently they can't/won't do it. A single set of stat weights doesn't work outside of as a guideline when the gear can range from NT18 to M18 upgraded+10.