Register

The true value of Haste(or Mastery) in T13.

Face-rippin fun.
Posts: 2
Joined: Tue Dec 20, 2011 2:27 am

The true value of Haste(or Mastery) in T13.

Postby Suphix » Tue Dec 20, 2011 3:00 am

Hello, all. First time poster and avid kitty DPS. I have read this and other threads with great detail and tried to collect as much data and math-work as possible when looking at the value of haste versus the value of mastery in several different situations. First, I will explain that like most, I run a very heavily haste stacked setup comming out of the back end of T12 and have done a lot of math and testing to vary how this works in various situations. I recently put aside my 4 piece T12 and rolled with 4 piece T13 to do some physical testing, both with various values of secondary stands and various gear setups.

A few night's back I was calculating secondary stat values in 4xT13 but with keeping a fully stacked haste setup and noticed something perculiar with the stat weights. Now, I know RSVs are not the be all and end all of gearing and execution but they can help, at the very least, to give some kind of indication as to what is going on while trying various setups. I am an engineer, so I of course intent to steer clear of the Tigers Fury glyph in 99.9% of cases, but I am having a bit of trouble with something here.

4xT13, haste stacked setup, Glyph of bloodletting - my haste value is as sky high as it has always been throughout T12, with crit some way behond and very closely followed by mastery. Hit and Expertise bringing up the rear, at the very bottom. This I suppose makes sense since it is a purposely stacked build - but haste is quite a way in front.

Now, I decided to run some simulations strictly in Mew without much of my own calculations, with tigers fury glyph even while being an engineer - same stat setup as above, mostly just to see what the results would be. I was quite suprised to see such a vast change in the stat weights. With no changes other than the glyph, my weights almost flipped 180 degrees. Hit, Mastery and Expertise were now way up the top, very closely followed by mastery, close enough to be called even I would say, with Haste then WAY behind. This has left me a little confused - I can understand how the value of Hit and Expertise can rise with 4xT13, after all there is a tasty Ravage to be used up and you don't particularly want to be swinging wildly at it. What I do not understand though, is how the value of haste depreciates so very much. I have been trying to work out the value of haste on the the bloodletting glyph instead, but the math I was putting out started to become very complex and just too messy. In the end I went back to basics - haste and extending rip for 6 seconds, per 1 duration.

As I said above, I know stat weights are not the colour and creed of kitty gearing, but I would like to know if any others are experiencing this too? I really don't understand how haste values are affected by a 6 seconds extension of rip, after all if doesn't matter how quickly those 6 seconds are stacked - other than once you reach the 6 second cap you are free to forget it and continue with shreds - but you would be using shred to stack those 6 seconds anyway (mostly).

Is this just an anomaly in Mew itself, or am I missing some vital mathwork regarding the value of haste?

Apologies for the long post, and thanks for reading. (If you made it this far)

Exalted
User avatar
Posts: 1526
Joined: Mon May 24, 2010 10:21 pm

Re: The true value of Haste(or Mastery) in T13.

Postby Tinderhoof » Tue Dec 20, 2011 1:49 pm

I could spend 30 minutes going through all of the recent posts by Yawning and Leafkiller about RSV's. But I at work and busy so I am just going to sum up.

Now, I know RSVs are not the be all and end all of gearing and execution but they can help, at the very least, to give some kind of indication as to what is going on while trying various setups.

This is an incorrect statment. Not only are they not the end all be all, they are misleading to anyone who does not understand how they work in Mew. It is clear by your statment here and the rest of the post you do not understand them.

Please do yourself a favor and turn the RSV's off. Mew is a fantastic tool that gives a lot of flexability in simulating not only Patchwork but a whole range of fight types. It is much better to try different gear sets like you were doing, but focus on what the simulated DPS output is.

With your current gear try and see which setups, be they Haste, Mastery, Crit or balanced give you the most bang for your buck and leave it there.

Exalted
User avatar
Posts: 1526
Joined: Mon May 24, 2010 10:21 pm

Re: The true value of Haste(or Mastery) in T13.

Postby Tinderhoof » Tue Dec 20, 2011 1:58 pm

Found this http://fluiddruid.net/forum/viewtopic.php?p=3129#p3129. Read it, learn it, love it. Then...TURN OFF RSV's.

Posts: 2
Joined: Tue Dec 20, 2011 2:27 am

Re: The true value of Haste(or Mastery) in T13.

Postby Suphix » Tue Dec 20, 2011 4:30 pm

Hello and thanks for the reply. However, it is abundantly clear from your post that you either A) didn't bother to read the post or B) read it and changed it to read completely different, or C) Have absolutely zero to contribute, don't really understand the post and just reposted something that someone else has already said (many, many times) and I have said myself. I suspect it is a mix of B and C. Above all else, I probably understand more about general wow math and specific RSV's work with specific calculations than you ever could. I was hoping for at least an educated reply, or something a little more substantial to my findings, not some half-assed prattle. That said, thank you for the effort (or lack thereof) that you made when responding, though in hindsight you haven't really contributed a single thing to this post.

I will take this finding some place else, best of luck.

Exalted
User avatar
Posts: 568
Joined: Sat May 28, 2011 3:43 pm

Re: The true value of Haste(or Mastery) in T13.

Postby Konungr » Tue Dec 20, 2011 5:31 pm

inb4 Humiliating SMACKDOWN by Yawning/Leafkiller/Qbear/Tinder/etc.

Tinderhoof read this the exact way I did, you were doing multiple tests within Mew with RSVs enabled and changing certain variables you managed to get vastly different results which caused confusion, leading you to post here, which is why he said what he said and linked you to YAWNING's post about RSVs and why you should not be using them since you clearly do not understand them. I can go back to your main post and quote you directly on how you "do not understand". I do love your self-glorifying attitude, its a very nice touch after the "not understanding", it made me squirm with excitement at someone that is utterly clueless.

"Above all else, I probably understand more about general wow math and specific RSV's work with specific calculations than you ever could."

I'm sorry, I just can't stop laughing.

On a side note, if you are going to take your "findings" someplace else, please take them to Elitist Jerks, they are already an unreliable source of Feral information, so it will fit right in. Try to avoid the ODF, MMO-C, FluidDruid, Team Waffle, and other decent places.

Posts: 3
Joined: Sun Oct 09, 2011 5:54 pm

Re: The true value of Haste(or Mastery) in T13.

Postby noctuary » Tue Dec 20, 2011 7:37 pm

Suphix wrote:Above all else, I probably understand more about general wow math and specific RSV's work with specific calculations than you ever could.


I had to stop reading here, as your ego has clearly defined your own value of haste. I award you no points, and may God have mercy on your soul.

Exalted
User avatar
Posts: 1145
Joined: Fri Jun 04, 2010 4:28 am

Re: The true value of Haste(or Mastery) in T13.

Postby Leafkiller » Tue Dec 20, 2011 8:10 pm

Suphix wrote:As I said above, I know stat weights are not the colour and creed of kitty gearing, but I would like to know if any others are experiencing this too? I really don't understand how haste values are affected by a 6 seconds extension of rip, after all if doesn't matter how quickly those 6 seconds are stacked - other than once you reach the 6 second cap you are free to forget it and continue with shreds - but you would be using shred to stack those 6 seconds anyway (mostly).


If you read through what Yawning wrote, then you should realize that RSV calculation in Mew does not apply to the value of your current stats, but only to what are the best places to add the next 100 or so stat points. In other words, they are "relative" to what you already have and not absolute measurements. This is why we see the stat values moving around and also see certain "peaks" in stat values with different reforging schemes.

Now your observation about haste being particularly volatile is not a new one. Several people have observed this and commented on it. While there may be some notion of haste breakpoints, they are quite dynamic and affected by fight mechanics, fight length etc. This is because haste simply gives us more energy rather than extra ticks as casters get, so the extra energy may or may not be useful. If you are interested in the value of increasing your haste by a small amount vs. increasing other stats, you should make sure that the "Encounter Duration Randomization" is non-zero and also consider using the "Randomize RNG Seed". This will help smooth out the large swings in haste RSV value that Mew sees with a fixed length fight and a fixed RNG seed.

So now that we have discussed how it is possible for haste to take large swings in value, the next question you ask is how having Rip last 6 seconds longer can contribute to a "breakpoint." The answer is obvious if you think about it. With Rip lasting longer, there is less need for combo points and with the fixed length/fixed seed Mew, you may or may not see a breakpoint popping up. In other words, it simply changes things and the RNG may or may not be favorable for haste with what has been changed.

When I am testing rotation changes I prefer to test with fixed length/seed tries to get an apples to apples look at the data. But for the purposes of what to reforge the next piece of gear to, some randomization is important, especially when dealing with haste.

Honored
Posts: 76
Joined: Sat Jun 04, 2011 4:21 am

Re: The true value of Haste(or Mastery) in T13.

Postby Yawning » Tue Dec 20, 2011 11:40 pm

Suphix wrote:Above all else, I probably understand more about general wow math and specific RSV's work with specific calculations than you ever could. I was hoping for at least an educated reply, or something a little more substantial to my findings, not some half-assed prattle.


People beat me to this but I'm failing to see how "Don't use RSVs in the first place" qualifies as "half-assed prattle" in response to "findings" that can be summed up as "Generating RSVs without using options that the author put in place for smoothing volatility produce results that are volatile".

If you truly were the depth less fountain of knowledge that you claim to be, you would have done the right thing in the first place (not using RSVs and instead using a baseline gear set + different reforges).

Leafkiller wrote:If you read through what Yawning wrote, then you should realize that RSV calculation in Mew does not apply to the value of your current stats, but only to what are the best places to add the next 100 or so stat points. In other words, they are "relative" to what you already have and not absolute measurements. This is why we see the stat values moving around and also see certain "peaks" in stat values with different reforging schemes.


It is also entirely possible to change the 100 to something else (and hit/exp are handled differently regardless so it will do the right thing wrt to caps) to further reduce volatility. If it was not for the fact that the returned values are actually somewhat useful for the tank specs that I chose to model, I would have ripped out the functionality a long time ago because it causes more headaches for me than it's worth.

Leafkiller wrote:If you are interested in the value of increasing your haste by a small amount vs. increasing other stats, you should make sure that the "Encounter Duration Randomization" is non-zero and also consider using the "Randomize RNG Seed". This will help smooth out the large swings in haste RSV value that Mew sees with a fixed length fight and a fixed RNG seed.


For the most part that information is fairly worthless because you (almost) never increase stats in a vacuum. If I cared enough about enhancing this functionality I would generated a NxN matrix showing deltaDPS/statscale when accounting for trading off another stat (Which is what reforging does), but there's better ways to optimize gearing (That will probably appear in Mewtwo).

As a side note, it shouldn't be strictly necessary to change the PRNG seed since the underlying code doesn't use the same PRNG seed per iteration (The way things work is that each iteration starts off the PRNG seeded with "base seed + iterationNumber". Base seed being 0 or something time based depending on what the user selects). Yes, it's entirely possible that the pool of random numbers that it generates from number-of-iteration seeds will tend to favor one stat or the other, but it's extremely improbable assuming sufficient iterations are used since all three possible PRNG algorithms are quite good (Out of all of them WELL and IBAA are more unlikely to cause PRNG related headaches since Mersenne Twister has certain edge cases where it produces fairly terrible output depending on initial state[0].).

[0]: For those that are curious about this, see "Improved Long-Period Generators Based on Linear Recurrences Modulo 2" http://www.iro.umontreal.ca/~lecuyer/my ... lfsr04.pdf (TLDR: The author of MT and some French Canadian dudes were like "Hey, we can do better than Mersenne Twister" so they wrote WELL)
"Understanding is a three-edged sword."

Mew: http://code.google.com/p/mew-wow-druid-model/

Honored
Posts: 116
Joined: Thu May 19, 2011 7:34 am

Re: The true value of Haste(or Mastery) in T13.

Postby Floofles » Wed Dec 21, 2011 7:29 am

In reference to the original post and the response by the same chap later in the thread;

You mention that you were doing your own calculations and yet neglect to show your working. You obviously do not understand how Mew calculates the RSVs,

I would refer you to the previous threads on the forum where we discussed RSVs and stat priorities in detail, often the most informative threads start with arrogant/ignorant posts such as yours, but the previous threads, I feel, have fully covered the subject.

I would suggest you repeat your tests by taking the advice in this thread - base gear set, reforging several ways with RSVs disabled with an amount of encounter duration jitter and RNG seed enabled -or at least return to us with more details of your "findings", surely someone with your impressive mathematical and analytical background understands the need for showing your working.

Honored
Posts: 92
Joined: Sun May 29, 2011 2:58 am

Re: The true value of Haste(or Mastery) in T13.

Postby Terias » Wed Dec 21, 2011 8:39 am

Out of all the threads you recommend someone roll a warrior, you overlook the one that deserved it most.

Honored
Posts: 116
Joined: Thu May 19, 2011 7:34 am

Re: The true value of Haste(or Mastery) in T13.

Postby Floofles » Wed Dec 21, 2011 9:13 am

A good point, sir, I do point you to the warrior forums wherein you will find a hive of great minds.

Posts: 45
Joined: Fri Jul 29, 2011 5:04 pm

Re: The true value of Haste(or Mastery) in T13.

Postby BoldTM » Wed Dec 21, 2011 6:46 pm

So the conclusion is pretty much.

RSV is not anything usefull, Mew can not and will not provide any usefull information about stats comparision what so ever and if you want to get an understanding of how different stats interact go somewhere else.

To be honest to the OP when the Mew are showing RSV then you could somewhat expect it to actually do something usefull which it pretty much is not doing?

Its like, "look i have a very nice shiny button you can press and it will show you something that you think is useful", and when someone comments on them here then the reply is, " WTF u stupid? Why are u pressing that shiny button and think it actually will tell you something".

Revered
User avatar
Posts: 358
Joined: Tue Mar 08, 2011 4:12 am

Re: The true value of Haste(or Mastery) in T13.

Postby Qbear » Wed Dec 21, 2011 7:47 pm

BoldTM wrote: and when someone comments on them here then the reply is, " WTF u stupid? Why are u pressing that shiny button and think it actually will tell you something".


Ha, well the main issue here is that all this has been cover quite a lot even since before I started helping with the site.

Honestly at first I didn't understand it either since the actual technique of theorycrafting is just mind blowing to me, I've always prided my ability to preform the class well on being able to filter all the data that's out there into what is useful and what is fluff.

Well the admins and regulars to the forums have been saying over and over again that the rsv's generate by mew fall into the fluff catagory when you look at just your toon and just your gear.

I love and think that most people that come here love and respect this site for the fact that you can come in and generally ask or comment about anything and not get the rule book of the elitist jerks community thrown down at you; however, there's a point where you can only answer the same question so many times before you just laugh it off when its brought up.

I don't think a week has gone by where someone hasn't asked the same question as the OP and points for him for at least wording it in a more fancy way; however, it really is the same question everyone else has already asked and gotten answered for months.

So nah Bold it's not WTF u stupid, it's WTF not this again.

Exalted
User avatar
Posts: 1526
Joined: Mon May 24, 2010 10:21 pm

Re: The true value of Haste(or Mastery) in T13.

Postby Tinderhoof » Wed Dec 21, 2011 8:02 pm

BoldTM wrote:So the conclusion is pretty much.

RSV is not anything usefull, Mew can not and will not provide any usefull information about stats comparision what so ever and if you want to get an understanding of how different stats interact go somewhere else.

To be honest to the OP when the Mew are showing RSV then you could somewhat expect it to actually do something usefull which it pretty much is not doing?

Its like, "look i have a very nice shiny button you can press and it will show you something that you think is useful", and when someone comments on them here then the reply is, " WTF u stupid? Why are u pressing that shiny button and think it actually will tell you something".

To be fair the big shiny button in Mew gives exactly what we suggested the OP do. A simulation of the current gear set based on the fight mechanics given by the user. As was suggested by several posters making gear sets and testing in MEW over and over will give a very clear idea of what will give you a dps up or down. The RSV's that the OP was refering too are turned off by default. This was because there have been MANY threads just like this. In fact only a week or so ago we had this very same issue come up (I linked that thread here). In order to see the RSV's the user has to find the option and enable it. Like Yawning said he would have removed RSV's altogether by now if they wern't useful for bears.

I may have been blunt in my reply, but I did not insult the OP and I did give solid advise that Yawning and Leaf had given several times before. If you read the link to Yawnings post at the start of the thread he in great detail broke down why RSV's are misleading. The rude reply was what triggered the backlash.

Posts: 25
Joined: Wed Dec 07, 2011 4:08 pm

Re: The true value of Haste(or Mastery) in T13.

Postby peki » Wed Dec 21, 2011 8:03 pm

why, in almost every post you do floofles, you tell someone to go play a warrior?
Image

Exalted
User avatar
Posts: 568
Joined: Sat May 28, 2011 3:43 pm

Re: The true value of Haste(or Mastery) in T13.

Postby Konungr » Wed Dec 21, 2011 8:10 pm

peki wrote:why, in almost every post you do floofles, you tell someone to go play a warrior?


Because he only replies to those unworthy of playing Feral.

Honored
Posts: 76
Joined: Sat Jun 04, 2011 4:21 am

Re: The true value of Haste(or Mastery) in T13.

Postby Yawning » Wed Dec 21, 2011 8:51 pm

BoldTM wrote:RSV is not anything usefull, Mew can not and will not provide any usefull information about stats comparision what so ever and if you want to get an understanding of how different stats interact go somewhere else.


Depends on what model you're using. RSVs with sufficient iterations is useful for bears and prot warriors because stats in general aren't close in value to each other and tend to be non-volatile. For kittys not so much, because stats are close and are volatile (Though it can tell you some things).

BoldTM wrote:To be honest to the OP when the Mew are showing RSV then you could somewhat expect it to actually do something usefull which it pretty much is not doing?

Its like, "look i have a very nice shiny button you can press and it will show you something that you think is useful", and when someone comments on them here then the reply is, " WTF u stupid? Why are u pressing that shiny button and think it actually will tell you something".


It provides information, the usefulness is just somewhat selective. Assuming you use jitter and sufficient iterations, it can tell you (quite definitively) how to gem and enchant because those are cases of increasing a stat in a vacuum. It will not tell you anything that's all that useful for reforging.

There really isn't anything else that's correct either AFAIK:
  • SimulationCraft's Stats Scaling - Works near identically to Mew. Values may be somewhat smoother by default and there's a few more nobs to tweak (in particular different stat shifts per stat), but IMO the results will be equally worthless.
  • Rawr - Not a simulator. There's nothing fundamentally wrong with using a closed form model to calculate DPS if the closed form model is correct. Last I checked it wasn't.
  • AskMrRobot - Can't comment because source code isn't public. I'd assume dubious at best.

If people care about reforging, SimulationCraft can generate CSV formatted data by simulating DPS as you shift X stats from one secondary to another, but I don't particularly think that's all that useful on it's own either.

It would be possible to do this with Mew as well ("If I cared enough about enhancing this functionality I would generated a NxN matrix showing deltaDPS/statscale when accounting for trading off another stat" is basically this) but:
  1. I think this approach is sort of naff and the usefulness of the results would be questionable at best at least for Feral (It would examine single reforge tradeoffs, but the values obtained will not be worth much when trying to optimize all of your reforges). I've always tried to take a more holistic view[0] when examining this problem.
  2. Coming up with a UI for it would be annoying.
  3. I don't care enough. The source code to the project is available. If other people care more than I do, they can do it themselves and optionally contribute their changes (As long as the code isn't totally awful, I will be more than happy to include third party contributions).
  4. It will be really really slow since it will require N - 1 times more simulation passes (assuming I only use a single fixed stat scale).
  5. People would misinterpret those results as well, so what's the point, really.

[0]: To me, the value of a certain stat for a certain gear set seems non-interesting except as an academic question (or as a minute facet of the interesting question). What matters the most is solving "given a certain gear pool/set, how do I maximize my DPS?". RSVs/Stat Scales/AEP systems can provide limited insight (The quality of said "limited insight" being dependent on the spec and encounter) into the answer, but the only way (provably so) to find the correct answer is an exhaustive search (You can get "not totally correct but reasonably close" with a number of algorithms but all of them require considerably more sophistication and run time than spitting out RSVs do).

Edit: Added stuff.
"Understanding is a three-edged sword."

Mew: http://code.google.com/p/mew-wow-druid-model/

Return to Kitty DPS

Who is online

Users browsing this forum: Bing [Bot], Google Adsense [Bot] and 13 guests