Demo secondary stat priority - mixed information
I'm getting Crit from Noxxic, Mastery from AskMrRobot, and Haste from Icy-Veins and MMO-Champ!
Becuasr all of them are godaweful sites for theory crafting. They're as bad as my spelling. the priority of Demo is is actually Crit=Mastery>haste, though crit aand mastery are usaully so close in stat weight you will need to sim to see which is better for you at any given time. and haste is not that far behind either.
http://code.google.com/p/simulationcraft/ Use it.
Now, how do we send this to 11 million in game mails?
Originally Posted by Tengenstein
Thanks for your informative post. I've been hesitant to really dig into simcraft, but at this early in the xpac it looks like I'll have to get my nose into it.
Originally Posted by Tengenstein
Originally Posted by sifuedition
The problem with all four of the previous sites mentioned is that they simply regurgitate info that was developed elsewhere. In doing so, they are subjet to several issues.
1. How well does the poster understand and regurgitate
2. How easy is the correct info to find so that they are repeating info that was right to begin with
3. Most places that have a static list are based on best in slot
3a. Is the best in slot the "actual" BiS or just the BiS in the opinion of the original source.
3b. What are the odds you are in the same gear as the BiS list used? You stat weights will not necessarily match those used to make that static list.
Simulationcraft fixes all that. If you import your toon properly and take the time to make sure the appropriate buffs and stats are analyzed, the result will be correct for YOUR toon, not just your class BiS list.
Thanks for the clarification. That's been what I've found over my time playing, it's just easy to reference to.
But, as the saying goes; "Reap what you sow".
Hey everyone, I'm one of the people from Ask Mr. Robot and wanted to clarify a few things.
1. Mastery > Haste. We only found a 0.1% DPS increase by swapping Haste and Mastery. Demonology is a spec that lends itself well to heavy AoE burns or single-target burns. And those situations favor Mastery more significantly since you almost always do them in demon form. That's why we chose Mastery over Haste in our default stat weights.
2. I can't speak for other sites, but we research all of our own stat weights, as well as work with a lot of the top theorycrafters. No regurgitation from us We actually wrote a program that interfaces with SimC to sim through a lot of different gear options. As Sifuedition said, a lot of weights that people quote are based of THE BiS list, which you normally don't have. We actually take that into account, getting weights that work across most high-end gear levels, knowing you won't have BiS. And when you want to change the weights on your own, you can do that in AMR
3. We have an option that lets you export your gear to SimC. You can then sim your gear and put new stat weights into AMR if you want to be advanced It's a very flexible tool.
4. Our BiS sets are actually dynamic - all based on the stat weights, selected options, as well as racials, professions, etc. Any time you change one of those options, Mr. Robot recalculates your BiS list, as well as the rankings for all of the other gear in the lists.
That all being said, if anything thinks we aren't handling your characters right. Tell me your character's name and realm and I can dig into it.
Wow, I had a huge wall 'o text and accidently hit back when I grabbed the mouse. /rage
I will try to re-create the spirit but I doubt I redo the depth of my comments, lol.
It's nice to have you here for feedback! I apologize in advance for previous or forthcoming bad assumptions or generalizations. I am also not a theorycrafter myself, so I apologize if any of this comes across as a double standard or anything. I do recognize and appreciate the massive amount of time and effort that must be invested in AMR.
I kind of categorize AMR, Noxxic and Icy Veins as a do this site. I have a problem with the concept of that approach. I need the explanation and discussion that comes from other sources. I feel that this is the major reason my dps off spec almost always tops the meters over our main spec dps. I care more about the why. AMR just doesn't have enough of that for me.
I think the concept of a static listing for stat weights is invalid to begin with. For the casual player, I'm sure that's all they really want is a "close enough" simple answer. I understand that taking the step do be dynamic with that information on your site would be a huge time and effort investment for a tool that is a free resource for us. I don't feel that it would be fair to expect that kind of functionality to the site. The other side of that coin, however, is that if I do care about that level of detail, SimC provides it. Different tools for different approaches. Since the stat weights are changable on AMR, it can be used in conjunction with SimC as you mentioned. However, when I compare AMR with wowreforge, wowreforge is generally 0.0x or occassionally even 0.x more exact. If I have gone this far to optimize, I may as well get that last 0.0x. Unfortunately, I'm not the math guy who might be able to help with optimizing your calculations there or I would actually gladly offer =/
As to the BiS, I do like the way that your site handles them from the view of it changing based on professions, crafting options, lfr or heroic level gear, etc. It does tend to lead to user error as I have guild members ask me why their list is different every time and I'm sure it's their own fault for not being consistent with settings, lol. I imagine that takes a TON of work for your people and I like the concept of the function. What I miss is the why. Help me understand why you went with the non-tier shoulders instead of the crafted legs or w/e. I have used your BiS quite a bit but always with a grain of salt and my own lazy critique, lol. If I know I want haste and the crafted legs have it but the non-tier shoulders don't, I may go against your list. You may be right, but I lose confidence without the why being available.
I know a majority of the visitors to your site don't care about the things I am mentioning here as room for improvement. Maybe there could be a way to provide some of this in an optional form. An explanation button or a toggle for simple/advanced or something similar. An explanation like you were able to provide in your response about why you chose mastery is great and part of what I feel the site lacks. Same for BiS as I mentioned. If there was a button or revealable section to explain that This gear set achieves x more of stat y while still meeting condition z that goes a long way toward giving me confidence in the conclusions you've reached. I think it's important to be transparent about how you came to the conclusions you did. I probably will never personally challenge your numbers on any math you post but it's important to me to see that the math has been done and is open for discussion/peer review.
All-in-all, of the do this sites, I feel AMR is easily the best. When I have a guild member have their head spin 360 when I start talking about SimC, I will usually send them to AMR instead. It's just that when someone comes to a theorycrafting site to ask a question, I assume they want more than the "simple, close-enough" answer or they would have went to MMO-C. Even though AMR is often right or at least quite close, I have at times felt it was wrong and without the transparency and discussion available at sites like TS or EJ or the actual sims of my toon with my gear from SimC, I lose confidence.
In cata, I tended to find that if i put my main in and pressed optimize the little gear rating thing askmr.robot churns out would go down.
Originally Posted by Zoopercat
sifuedition, thanks for the well thought out feedback
You're right that we don't offer a lot of the discussion and theorycraft. A lot of other people already do that really well, so we want to send people to those discussions. With 5.0.4 we added a new feature that is linking people to guides over at Icy-Veins. We can target these messages based on class and spec, which is pretty awesome. So we are also looking for good discussions to put in that spot as well. That way, the people who want to read up on the theory behind everything, can.
As for SimC: SimC will give you weights that are very specific to your current gear setup, and aren't necessarily useful for moving you towards an optimal setup. Does that distinction make sense? Now, you seem like you know a lot about SimC, and I figure you already knew that, since you talk about reoptimizing at each gear level.
As for WoWReforge: we see the comparison between our tools a lot. But there's a big difference: WoWReforge is a calculator, AMR is an optimizer. Both are good at what they do. Let me explain: If you want to get to an exact hit cap or some other number, WoWReforge will get you there. If you want to do the most DPS, even if that means you might be 15 Hit Rating under the cap, Mr. Robot will do that for you. You see, it's all in the numbers for Mr. Robot: he takes you stat * your stat rating to get a value. He tries to get you the highest possible value, which turns into the highest theoretical DPS. You can actually do a test: Put both AMR's results and WoWReforge results into SimC and see which on has higher DPS. That all being said - both tools will get you close enough, that even if you sacrifice a little DPS to get exactly to that hit cap, you won't notice in a raid. The DPS you would lose is so small, one extra lucky crit or two would make up the difference.
As for the transparency issue: We don't try to be transparent It just takes time to explain everything that is going on behind the scenes. What have been some of your top questions when using it? Maybe I'll spend some time posting some 'documentation' in our forums.
Again, thanks for the in-depth comments It's hard to get people to engage in well-thought out discussions about tools!
PS: Glad you like the dynamic BiS lists - a lot of people have no idea why that's cool! Fun fact: There are billions of BiS combinations with just the settings on our site. We cache about 25 million of the most popular ones on our servers so they load fast. Crazy, right?!
Well, several questions come to mind regarding your response.
1. I'm not sure I would use Noxxic for theorycrafting reference. I apologize if that offends or you have professional or personal ties to them, but I am less than impressed with the break-downs I have seen there for classes/specs. I have seen some that were spot on and it seems like just as many that were clearly off from the conclusions of sites like TS and EJ but they did not show any of their work for critique/review as TS and EJ do.
2. SimC is indeed most directly a tool to analyze your current gear and the most effective reforges of that gear. However, it does have the capability to load gear that you do not have equiped and test upgrades/combinations as well. Admittedly, that takes a higher level of understanding of the tools (which I have NOT mastered, lol) but the functionality is there. Clearly, that is more work than finding a list where that has been done for you in effect, but you are returning to the issue of a static list. For instance, if I am clearing 3/6 hm, with SimC, I can look for upgrades on the first four (hoping for a kill) without using a list of full heroic or look at full heroic if I want. Basically, you can subset the list of possible gear that way. I understand that with AMR, you can also lock individual slots to achieve a similar effect, but you don't get the feedback of an actual simulated dps number. If the difference between that painful rep grind for best and the drop I can get this week is ~1%, maybe I'll skip the grind.
3. Feral dps used to basically ignore expertise and largely hit as well. Since they were a fast attack speed that was limited by energy which most was recovered on a miss/dodge/parry, it was not the same level of dps influence for them as most other dps. I understand that being perfectly capped will not always be top dps.
3a. However, since the SimC numbers are giving the weights, shouldn't that already be accounted for before you start the calculate/optimize? If hit/exp are not the top weighted stats, then capping them will not come first. I could be misunderstanding something and just need to be put on the right scent here.
3b. Since the weights indicate the effective dps increase of each stat, how does AMR determine that the last x rating to cap is not as much increased dps as x rating being lost to the reforge? Your response almost sounds like you are putting your own multipliers or weights on top of the stat weights (without giving up any trade secrets of your work if there are any, obviously). I definitely will test with SimC and the reforge recommendations between AMR results and wowreforge results. Should be cool to experiment with.
And for me, transparency is just that important if I'm going to have any confidence in a tool/source over one where that is available. When one source can "show their work" vs another just saying "trust us"...I will go with the transparent one every time. As it relates to reforges, wowreforge seems to be going more strictly off of SimC weights vs a system that is not explained on AMR. As it relates to spec/glyphs etc, Noxxic does not document methods/math like TS or EJ. As it relates to BiS, well, nobody else seems to usually have any better explanations than anyone else. Even EJ BiS lists rarely break down the differences between the recommended sets and the alternatives but given the nature of the site, there is enough peer review that I'm sure someone SimC'd them to verify at the least. The big drawback to EJ BiS lists is they are totally static and lose the ability to plan and subset totally, vs AMRs quick methods to do so and SimCs tedious but (IMO until AMR shows math) more accurate ways to do so.
This thread has been a trove of information, I'm glad that I subscribed to it. Also, thanks to all the contributors to this thread!
sifuedition - good questions. You're my new favorite person this week - most people don't engage in a geek-out session like this about optimizing gear!
1. We link to Icy-Veins (I assume you meant that) - they are actually currently working on updating everything for MoP We like linking there because it's a nice starting point for a lot of users, with guides for everything from their class to bosses. We'd also like to link to in-depth theorycrafting sources, but need to pick good posts with good summaries. We're working on gathering those sources
2. I totally get what you're saying. And you're more of an expert on SimC than most it appears. Not much to add here, it's a great tool - keep using it
3. Very good question, here's an example of why getting exactly to hit cap might not be optimal. Let's say your stat priority is Hit > Crit > Mastery. There is a limited number of combinations to get you exactly hit capped. Let's say to get exactly hit capped, you have to reforge all of your Crit to Hit. Now, if Mr. Robot looks at it, he might reforge all of your Mastery to Hit, and get you close to the hit cap, but 15 hit rating below it. In this scenario, you'd have higher DPS because Mr. Robot preserved your 2nd best stat (Crit) and sacrificed your 3rd best stat (Mastery) to get as much hit as possible. Does that make sense?
3.a. Mr. Robot does everything based on pure math. No secrets - really just pure math. He does this: stat weight * stat = score. So if Hit has a weight of 2, Crit 1.5 and mastery 1... he just multiplies everything to get you the highest score. He puts a few smart checks in for haste breakpoints and what not, and some clever things to speed it up, but it's all based on getting the highest score.
3.c. A side note about Mr. Robot - we optimize gems, enchants and reforges. If you lock in one thing, Mr. Robot reoptimizes everything else around it, so it's not uncommon that gems will change as a result. If you use wowreforge for reforges, but AMR for everything else, you should take an extra step: Go back to AMR and lock in every reforge you got from wowreforge. Then let Mr. Robot reoptimize the gems to make sure you don't go over the hit cap, etc. This will also give you a new score - you can compare that score against Mr. Robot's original optimization and see which one is higher. The easiest way to compare is by saving a profile for each one, so you can go back and forth. Again, that score is purely stat weight * stat.
Transparency: We haven't shown our work because it is purely simple math. Sure we have an algorithm that figures out what to check and how to do it, but Mr. Robot is just trying to get the highest score in a short amount of time. I should spend some time explaining a few of these things and linking to it for people who want to know... that would be a very worthwhile project.
Note: We get 99% of people within 0.25% of THE most optimal score. And for the 1% outside of that range, they are within 1% of the optimal score. We want to hear about those cases though, those cases help us improve the optimizer. Why not 100% you ask? To do that, we'd have to build an optimizer that runs on your computer since it would take up too much processor time. There are about 500,000 viable combinations of gems/enchants/reforges, and running through every combination on demand, on a web server, would choke! We do about 100-200 optimizations a minute, so it needs to be fast.
My apologies, I did mean Icy Veins. Unfortunately, I've seen the same issues with that site as I have with Noxxic.
If your time here was scouting sources for theorycrafting links, I would highly recommend any guide you can find here by any of these authors:
Theckhd (Theck from maintankadin)
Satorri (althogh I think Satorri retired from WoW and no longer does guides)
There are other very solid authors and contributors here and I don't mean to slight them in the least. These are just the ones that came to mind immediately probably because they have been some of the more active threads lately.
I can see where I misunderstood earlier. I'll try to summarize my understanding just in case I'm still missing it.
If SimC gives stat weights of:
You are simply then multiplying the amount of that stat/rating by the weight? If that is the case, I better see where you are coming from. I do seem to remember some optimization problems from math and programming classes that worked exactly that way, I just don't remember if the application of that solution fits this model. Given that you are associated with AMR, I am sure you have given this a lot more time and thought than I have. While it is true that wowreforge may reforge haste to reach an exact hit cap instead of crit even though haste is better than crit; I am uncertain that determining the theoretical best dps is as simple as multiplying the total of each stat/rating by the weight for a score. Since you stated earlier that SimC will show a higher dps with AMR reforges than wowreforge, I am assuming you have tested this already enough to convince you it is valid. Forgive me if I feel the need to check that myself a bit. I wish I could apply all that college math I took well enough to simply pop out a calulator and verify it, but alas, I don't LIKE math that well even if I have done it before =). I'll just have to brute force test some and see how consistent the results are.
As to gems, I haven't seen any math to say that things have changed from Cata. Primary stats are (were) so far and away better than any other stat that you should be doing nearly anything to increase them. As such, gems are one of the only ways to increase your primary stats. I doubt there will be many situations were anything but a hybrid will have a stat other than primary. Since secondary stats on gems now have twice as much rating as primary stats, I am aware that this may have change. I simply haven't seen any math on it yet.
I know for warrior DPS specs Hit and to a lesser extent Expertise are far more valuable than strength
sifuedition, I made a blog post with some details. It's the same type of info in our discussion here, but I wanted to let you know I made a blog post as a result of your questions I'll get it linked up to our UI soon so everyone can find it. Thanks again for all of the questions and feedback. http://blog.askmrrobot.com/2012/10/h...ptimizer-work/
I liked the example with the hybrid gem. Assuming that multiplying rating * weight is valid, and I haven't thought of a reason why it can't be, then that example works very well. I think the catch would only be that usually primary stats as well as hit and expertise are more than 0.2 higher in weighting than secondary stats...but not always. As an example of the math, it does the job very well.
The part about the BiS is something that I played with on my toon until I understood it pretty well. I can't say how well the "average" user gets it, however, based on how many times I have had to explain it to guild members. I hope your explanation will be relatively easy to see for visitors using that list.
I have frequently seen four main problems from guild members trying to use your BiS.
1. They don't pay enough attention to the settings to ensure they are right (they leave crafted in when they have no intention of paying for those expensive pants, etc)
2. They are not consistent when they go back. They get the settings different and then wonder why their list changed.
3. They are not realistic with the levels they can clear. They leave it on Heroic tier for example and then try to build the same gear set but with normal modes. However, the dynamic nature of it may mean that the x difference in hit rating from normal to heroic changes the set. Or maybe that trinket is BiS on heroic but the crafting one is better than the normal, etc.
4. They don't understand locking in pieces to evaluate their best options for a slot. For instance, if they feel they have no shot at upgrading their weapon and trinkets even though they do not have BiS, they do not lock them in to analyze the best chest and legs, etc. They just continue to try to collect the BiS based on having the Sinestra or Archeology trinket, for example, even though they have never even pulled that boss and probably never will or never gotten past lvl 50 arch.
Almost all of those are user error that I'm not too sure can be corrected from a design standpoint, lol. A profile for getting the gearing options the same each time they look is about the only site-side change I can see that could effect any of those and I'm not sure it's worth the overhead or the need to login to protect people from themselves.