Thread: Sub rogue help

Page 2 of 2 FirstFirst
1
2
  1. #21
    Quote Originally Posted by shadowboy View Post
    There should be no "I have reason to believe" about it. There is no believing. There is "I know it's not right because I downloaded it and read the source code." And in that eventuality, you can create an ticket so they know it's not right. Steering people away from the tool instead of taking steps to correct the tool is not doing ANYONE any favors. Good players use multiple optimization tools--not just one. I use shadowcraft, simcraft, wowreforge, amr--all of them.
    Sorry thats a load of crap. You sounded like you had some knowledge on statistics and modeling, but maybe your expertise is just in the coding side. If you have enough actual data and try to fit your model to it and it is outside your desired interval, you have reason to believe the model is not an accurate representation. You do not know this for a 100% fact because there is a chance. If I generate 1 million numbers from 0 to 1 and all 1 million are equal to 1, can you state with 100% certainty that the distribution isn't normal? Nope.

    Thats why I said I have reason to believe. The data produced, even if you allow a lenient amount for mechanics over what the other specs are showing, was not close to what simc's initial expected values showed. I don't need to see the source code to say that. Now yeah, if I was arguing that it was flat out wrong, I would have to be able to point out an error in the source code or that would be meaningless.

    You do have a good point about using multiple tools. However, if I had multiple tools and I had reason to believe one of them wasn't accurate reflecting real data in a specific instance, I certainly wouldn't be recommending that tool be used in that specific instance.
    Last edited by Sesshou; 2013-01-26 at 08:50 AM.

  2. #22
    Legendary! Ryme's Avatar
    Join Date
    Jul 2009
    Location
    In a field, somewhere
    Posts
    6,153
    So, I'm currently messing around with the following priority order taken pretty much from discussion of addons, simcraft and shadowcraft. I'm getting some nice results in practise (getting very close to assassination)

    Basic Priority order: Keep up find weakness > keep up SnD > keep up rupture > keep up hemo > build with backstab (or hemo if absolutely required) > dump with eviscerate.

    Advanced priority (anticipation usage explained later):

    • Preparation - as long as vanish has more than 60 seconds til it comes off cooldown
    • Shadow Blades - on CD irrespective of find weakness, the increased up time and cp generation is, in simulation, more important
    • Shadow dance - don't clip find weakness, make sure you pool energy; anything around 75 is good, dont go too high and cap though, that's bad news
    • Vanish - it is better have more than 3 cps and to pool to around 50 energy, this will give you optimal usage of eviscerate with the addtional find weakness vanish provides. Once again, do not clip find weakness or master of subtlety
    • Premeditation - optimal to use when you can get both cps it provides in such a way to take you to a straight 5 cps. So, for pure min/maxing it is optimal to take HaT into account, if this puts you to 4cps and you premed, you are losing some dps.
    • Ambush - when you have no anticipation charges
    • Slice and Dice - once it's duration is <3 seconds and you have 5cps - refreshing it at anything else results in decreased usage of your harder hitting abilities and this spec is less passive and suffers more from that.
    • Rupture - to be refreshed with 5 cps when it has less than 5 seconds remaining, ideally less than 3. Reasoning for 5 is that you want to use 5 cps for this, but you do not want it to fall off, the extra 2 seconds of clipping give you a safety net to work with.
    • Ambush - when your anticipation charges <3 and shadow dance has at least 2 seconds left - this is for better energy management.
    • Eviscerate - when you have 5 cps
    • Hemorrhage - when you have <4 cps or when you have <5 and your about to cap energy, only use this to refresh the dot when it's below 4 seconds remaining or you're facing the front of the boss and have to use it.
    • Backstab - when you have <4 cps and shadow dance won't be ready in 7 seconds or when you have <5 cps and you're about to cap energy

    The idea with ambush usage, is to pump ambushes to give up to 4 anticipation charges (note: not 5 as then we lose the benefit of HaT), this lets us get out 1 or 2 eviscerates at once at very high damage output due to find weakness. This part of the simulation is in it's infancy and can definitely be improved - we could start to take into account when procs line up for example, in which case it would be worth running consistently with 4-5 anticipation charges, dumping them only when procs align.

    Further discussion can be found here: http://elitistjerks.com/f78/t130894-...n_Optimization

    Hemo stills needs some minor optimisation to work better with find weakness, but it's such a low dpe, laze took over.
    Last edited by Ryme; 2013-01-27 at 12:36 PM.
    I am the lucid dream
    Uulwi ifis halahs gag erh'ongg w'ssh


  3. #23
    Quote Originally Posted by Sesshou View Post
    Sorry thats a load of crap. You sounded like you had some knowledge on statistics and modeling, but maybe your expertise is just in the coding side. If you have enough actual data and try to fit your model to it and it is outside your desired interval, you have reason to believe the model is not an accurate representation. You do not know this for a 100% fact because there is a chance. If I generate 1 million numbers from 0 to 1 and all 1 million are equal to 1, can you state with 100% certainty that the distribution isn't normal? Nope.
    Simc doesn't use statistics to model. It models ACTUAL mechanics. The statistics are there to demonstrate the expected outcome based on EXACT mechanics.

    The difference between how simc and shadowcraft works is that shadowcraft takes the expectation values of mechanics (determined through either blue posts or mechanics testing) to generate an expectation value of performance and then determines exact scaling (or extremely close average scaling depending on if differentiation is numerical or analytical). Simulationcraft takes the exact values of mechanics (determined the same way as in shadowcraft) to generate an expectation value of performance and an expectation value of scaling. If simulationcraft's expectation values are wrong it's because either the priority list is not effective (easily changed by the end user--and I've been able to tweak the default combat priority list to gain dps at all gear levels), or that it is modeling mechanics wrong and that is very easily verified by simply reading the source code.

    ---------- Post added 2013-01-26 at 11:05 PM ----------

    Quote Originally Posted by Ryme View Post
    Premeditation if <5 cps
    I wouldn't premed at 4 combo points. Premed doesn't add anticipation stacks or spill over into anticipation points (if used at 4) unless it's changed from when I tested this on beta.

    You should only premed at 3 or less.

  4. #24
    Legendary! Ryme's Avatar
    Join Date
    Jul 2009
    Location
    In a field, somewhere
    Posts
    6,153
    Quote Originally Posted by shadowboy View Post
    I wouldn't premed at 4 combo points. Premed doesn't add anticipation stacks or spill over into anticipation points (if used at 4) unless it's changed from when I tested this on beta.

    You should only premed at 3 or less.
    Sorry, I was actually quite drunk when I posted this - Some things are redundant and others need explanation and formatting, I'm editing the post now. As for premed, the idea is that you're heading into this stealth with less than 4 cps, but if one pops up due to HaT, by fact that premed has no GCD or energy cost, it's still worth using than delaying. The actual dps difference is tiny, probably small enough to be lost in the margin of simulation, but I did find an increase with this.

    EDIT: Because I wasn't accounting for subterfuge... Correcting again.

    ---------- Post added 2013-01-27 at 12:34 PM ----------

    Updated, merged CD usage into main rotation, still need to write opener out.
    Last edited by Ryme; 2013-01-27 at 11:08 AM.
    I am the lucid dream
    Uulwi ifis halahs gag erh'ongg w'ssh


Posting Permissions

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