Page 2 of 2 FirstFirst
1
2
  1. #21
    That's one of many local mins, where the values are so close that they flipflop. Watch the EP values change slightly, but you'll see like:

    You are reforged for haste, and have a lot of haste. Your mastery EP is X, and your haste EP is .95X. If you add one mastery, you'll get more dps than adding one haste. When you reforge, it uses those values.

    But, that CHANGES the values. Now you have a lot of mastery. Your mastery EP is Y, and your haste EP is 1.06Y. If you add one mastery, you'll get LESS dps than adding one haste.

    You can, of course, just pick whichever one predicts the highest dps.

    You can also go and force a priority, reforge, write down the dps, force another, reforge, write down the dps, etc. Personally if the difference is small enough, I will take the easier to play one, or the one that is better at aoe, etc, as those will result in more damage in practice. But if the difference is substantial I'll aim for whatever that needs.

  2. #22
    Quote Originally Posted by Verain View Post
    You can also go and force a priority, reforge, write down the dps, force another, reforge, write down the dps, etc.
    Or force a priority, reforge, read the DPS, force another, reforge, compare the new value to the old int the DPS chart. ShC rocks!
    A witty saying proves nothing.
    -Voltaire
    winning
    plus ça change, plus c'est la même chose

  3. #23
    I have downloaded the simDPS last night and doing my own tests... cant believe the changes, and at the same time, I understand, the nerfs at the trinkets of ToT just fucked up the reforges, doing like the trinket of valor points getting better than Renataki's Sou Charm Heroic... I think the only way of trusting those things are testing

    - - - Updated - - -

    I'm almost going insane, was now, just kidding at the ShC, when I do the reforge like Crit > Mastery > Haste the values of DPS go down, when I go back for Haste > Mastery > Crit the values get increased, I'm a little bit confused

  4. #24
    Just as a point for the people here wondering about crit, it will not be a dps-gain unless you've got t16-2p as far as I can remember, personally I've gone Mastery>Crit=Haste and it's producing fairly good results, doing quite a bit more today than I did yesterday, although this will ofc be a little from the extra stats from the new pieces of loot I got, but not that high.

    Currently running 12.5k Mastery, 7.2k Crit and 7.1k Haste.

    What really makes me feel underpowered is the trinket duration only being 10 seconds so limited how much I can squeeze in there, like, it's alot harder to pool up during Renataki's to get out 2-3 Envenoms during the last few seconds

  5. #25
    Quote Originally Posted by zerynax View Post
    What really makes me feel underpowered is the trinket duration only being 10 seconds so limited how much I can squeeze in there, like, it's alot harder to pool up during Renataki's to get out 2-3 Envenoms during the last few seconds
    I suffer with that too, but the ExtraCD fool me over and I keep using Renataki's Soul Charm for like 2 days after the launch of 5.4, if you have enough valor points or the trinket of valor points of ToT (Vicious Talisman of the Shado-Pan Assault) or you have it on your bank, use it, still 20 seconds and its just like... 9.4k at 2 upgrades, and how we're not depending of the haste anymore, as ShC and SimDPS shows you have no doubt of replace it

    english isnt my main language, just a warn haha

  6. #26
    Deleted
    Quote Originally Posted by intrinsc View Post
    Simcraft gave me these values:

    http://i.imgur.com/kUn1wfz.jpg

    I think I want to the reforge plot for all 3 secondary stats, though.
    Is this with T16 2p?


    I think I'll stay with more mastery for aoeing, in SoO there are not enough tanknspank encounters

  7. #27
    For now I'm using 12k mastery > 6k crit ~= 6.9k haste and I think its pretty balanced since i'm gearing up for Mastery and Haste, probbably it will change at the middle of the patch, but for now, thats me

    - - - Updated - - -

    12k mastery > 6.4k crit ~= 6.5k haste

    just an update, I thought I couldnt get more balanced, but I was wrong and I love myself for use both ShC and AskMrRobot

  8. #28
    Quote Originally Posted by Verain View Post
    It's not entirely uncommon for a new tier to be unsupported, or for some important effect not to be modeled.
    Just a quick post,
    When I was actively developing it (cata), we had plenty of those:
    -we never got around to model the 3rd stage of the legendary proc (it would have been a nightmare at that time).
    -had the stacking of KsP and BG wrong for most of the xpack.
    -the munching of the FL tier was half-assed.
    -matrix trinket was able to crash the back-end at some point.
    -subtlety modeling was always behind: we launched the tool without a sub model; hemo weaving was more of a hack; eviscerate clumping during FW was not even considered.

    We tried to cover our deficiencies through feedback on the forums. I think we did a pretty decent job. When something was known to not be sufficiently modeled, we developed several hack-ish engines to figure stuff in our personal machines. Those would never see the light of day (they usually rely on huge inefficient loops, and were prone to timeouts). I was never as vocal as Pathal is, but I think we were very honest with our shortcomings, and tried to give a measure of how much the tool could be off and in which directions.

    Ultimately, I think that our tool relies heavily on feedback. And it's wonderful that we have posters publicly fiddling with the tool in a wide range (Fiery). There are always things that can be better modeled (most of the time we know what those are). And there's always the option to download the source and toy with it on your own: I know some people that personally developed an attack_counts for Spine in DS. After all, it is a collaborative modular project in its core: every bit of feedback counts.

    So yes, caution is advised, of course. Skepticism just for the sake of it, not so much: we usually know where and by how much we're erring. It's only a matter of communicating the issues.

    Lastly, I always wonder if we should actively encourage people to download and look through the code: I think it's impressively easy to follow for such a huge task. We built an amazingly cohesive-looking tool for something so complex in nature. Its modular nature allows for anyone to plug their own model on anything: maybe you'd like to have a simulation-oriented model, or a different way to handle procs, or support, say, boomkins. While it currently follows on the trend set by Aldriana, it really is enough of a strong framework to support other kinds of theorycrafting.

  9. #29
    Deleted
    Reforge haste regem mastery.

    Confusing.

  10. #30
    Why do all the fun conversations happen while I’m asleep?

    Quote Originally Posted by Kulestu View Post
    It's not so much that I think they're incorrect as it's that I think that Sims are built for on-paper fights.
    Aldriana made an interesting post on the “endless dps” assumption, its not a long post and very worth reading but the short version is endless dps is a good assumption most of time. Within the range of typical raid fights there is an upper bound on how much the fight can deviate from strict single target and that upper bound is surprisingly low. Pick pretty much any raid encounter since wrath and the amount of the fight will be a single target encounter is probably not much lower than 90%.

    There are some interactions between mechanics and procs and cooldown timing that can be hugely relevant but those are also very hard to model. You can already do that in SimCraft, if someone wanted to they could sit down and define Lei Shen in SimCraft but most people don’t do that because that is a huge amount of work. Endless dps works because for the most part if works. We could probably do a better job of quantifying how some common cases impact EP values (time off target, multitarget) as nextormento was talking about but for the most part endless dps should be good enough for most things. This is why I like to emphasize the similarities between SimCraft and ShadowCraft results, ShadowCraft clearly doesn’t implement things, proc synergies for instance but as we can see in practice that doesn’t matter all that much for rogues.

    More broadly I suspect the biggest thing that hurts confidence in ShadowCraft is the reforge “ringing” because that just looks wrong to people. When an “optimize me” button yields lower dps it definitely seems wrong. Once you know whats going on it is totally logical but I’ve seen a decent number of people who see that funny result and basically give up on ShadowCraft because if its optimizations yield a dps loss then clearly this tool is totally wrong. I have a couple ideas of how to hide that but haven’t had the time to experiment with them yet.

    Quote Originally Posted by nextormento View Post
    Lastly, I always wonder if we should actively encourage people to download and look through the code: I think it's impressively easy to follow for such a huge task. We built an amazingly cohesive-looking tool for something so complex in nature. Its modular nature allows for anyone to plug their own model on anything: maybe you'd like to have a simulation-oriented model, or a different way to handle procs, or support, say, boomkins. While it currently follows on the trend set by Aldriana, it really is enough of a strong framework to support other kinds of theorycrafting.
    That was always the big motivation for using Python but I suspect that may be a bit optimistic. Thinking back to the initial discussions of ShadowCraft it was always a somewhat aspirational, a tool that other classes could use and anyone could contribute to. I’ve encouraged people to dig around in the code from time to time but I don’t think many (any?) people actually take me up on it. ShadowCraft is pretty easy to work your way through but it is still a pretty daunting task.
    Fierydemise-ShaowCraft Engine Guy
    Rogue Chat-Blogging about Rogue PvE and Theorycrafting (Twitter)
    Rogue IRC: #Ravenholdt on Quakenet

  11. #31
    Hail has looked into it. He pointed out an issue or two while I was fixing Sub, which was pretty nice actually.

    Really though, the largest reason why it was done in Python 2.6 was because Ald was comfortable in it. There were a lot of other reasons too, malleability, readability, etc. But it has since stayed in Python mostly because it was already in Python. Changing the language may have required changing the web portal too. Nextormento converted it to Java back in 4.2 or something to play around, but it looks quite a bit different today, and would still take quite a bit of time to port.

    I think this is straying towards Shadowcraft 2.0 post mortem area though. Which seems a little off topic. I'd more than welcome one here in a few weeks once the 5.4 dust settles, but this feels like the wrong time and place.

  12. #32
    Mechagnome intrinsc's Avatar
    10+ Year Old Account
    Join Date
    May 2013
    Location
    Aberdeen Proving Ground, MD
    Posts
    538
    Quote Originally Posted by vertol View Post
    Is this with T16 2p?


    I think I'll stay with more mastery for aoeing, in SoO there are not enough tanknspank encounters
    That is with T15 4pc, still. One piece of T16(legs) and legendary cloak. I like mastery, still, because of the guaranteed DPS increase and the secondaries are so close to each other that I will take that instead of the other two which are chances. I will always run these sims when I get new loot to see how it pans out, though.

  13. #33
    Deleted
    Killed the last 5 bosses in SoO tonight with the 2-set t16 and replaced Renetakki's and Juju with Sigil of Rampage and Ticking Ebon Detonator.

    Went with a Crit > Mastery > Haste build. Was playing around on the advanced tab on ShC and with the 2 set this seems the most beneficial. Energy is certainly far more manageable with crit than with haste.

    I will be waiting for EJ's theories, but I am very happy playing crit with the 2 set, it just feels more comfortable.

  14. #34
    Quote Originally Posted by Kumpi View Post
    Went with a Crit > Mastery > Haste build. Was playing around on the advanced tab on ShC and with the 2 set this seems the most beneficial. Energy is certainly far more manageable with crit than with haste.

    I will be waiting for EJ's theories, but I am very happy playing crit with the 2 set, it just feels more comfortable.
    I'm crazy about get the t16-2p, this must change all the way of pooling energy and stuff, I'm very excited about this change, its a big change

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •