1. #1

    AMR's Simulator Tool

    I just learned that Ask Mr. Robot has developed a simulator tool for Legion. Obviously the tool is in beta now, so I assume it's buggy, inaccurate, etc... But the idea of the tool and conceptual design seems really useful. What do you guys think about the idea? Website... Video...

    In particular are any of our local gurus here (Archi, Sarri, et. al.) involved in helping them lay out Warrior rotations for their tool? Does the tool use SimC, or is it AMR's own simulator? I didn't see a reference to SimC, so I suppose they've made their own code.

  2. #2
    It's their own tool, I'd be willing to bet it's worse than Sim-C, but I guess we'll have to find out.

    Worth noting that we may be slightly biased toward Sim-C, as we are the ones who keep the warrior modules up to date.
    My Stream
    NollTvåTre Looking for Raiders

  3. #3
    So I take it you guys have not been involved in setting up Warrior rotations for the tool? If not, then who are "some of the top theorycrafters" they've used for the Warrior rotations (according to their blog)? Or, are they just winging it for Warriors?

    I understand the attachment you might have for SimC, but I hope you guys weigh in on the AMR tool. It seems like they're taking a crowd-sourcing approach to iterating on the tool through their wiki. If they can actually pull off simulating each spec's performance on each individual boss, then I expect it will be hugely popular throughout Legion. It'd be nice to know that you guys are involved to keep things straight.

  4. #4
    I'm in the Beta, but really haven't had time to look at it much. One of the primary AMR guys plays warrior and has always been on top of everything warrior on their site. I suspect he's always taking part in these boards. But it's likely they have input from others too by this point.

    The best planned feature ( in comparison to simc ) is the ability to recreate actual encounters and sim them. IMO
    Armory
    Twitter
    Gladiator's Resolve Blog

    Quote Originally Posted by Von Bosch View Post
    Warriors are hereos that draw thier super human strength from thier relentless fury and thier unstoppeble willpower to fight on til the end of days.

  5. #5
    You do know you can sim encounters in Sim-C as well, and that we've done that for a few expansions now?
    My Stream
    NollTvåTre Looking for Raiders

  6. #6
    Of course...

    The expectation is it will be as easy to do as the normal AMR stuff.
    They plan to let us do custom rotations, spell changes, new spells, etc, all with that kind of simple straightforward code-free approach.
    Armory
    Twitter
    Gladiator's Resolve Blog

    Quote Originally Posted by Von Bosch View Post
    Warriors are hereos that draw thier super human strength from thier relentless fury and thier unstoppeble willpower to fight on til the end of days.

  7. #7
    Keyboard Turner kilrans's Avatar
    7+ Year Old Account
    Join Date
    Jun 2016
    Location
    California
    Posts
    7
    Perhaps AMR's sim tool is wildly innaccurate. Simming Fury, using 903 artifact weapons and 875 gear valuing haste>mastery>crit with the legendary enrage helm and execute ring, the highest dps values I saw were with Fresh meat, Avatar, frothing beserker, frenzy and reckless abandon.

    This simmed out to an average of 508k on the standard patchwerk fight.

    Whereas the standard, endless rage, wrecking ball, carnage/massacre, frenzy and dragon roar build that i've seen on the forums is at around 461k with the same gear on the same type of fight.

    I'm no theory crafter, but i'm curious to hear from someone who is on disparity between these two builds, or if there is something I missed (probably is).

    I'd also like to mention that I took the artifact relics that buffed damage during Battlecry, perhaps the combination of Frothing beserker, Battle cry and Avatar is throwing off the numbers. It seems as though this would be an extremely bursty build if all the CD's aligned, however you would have to pool rage whenever battlecry was down.

    EDIT: I went back and ran the simulation again using the wild slashes relics instead of the unrivaled strength relics and it simmed much much lower, by about 80k actually.
    Last edited by kilrans; 2016-06-22 at 05:04 PM. Reason: Sentence structure

  8. #8
    I truly welcome more and any of a tool/site to give out or help with information to the player who wish to have a better understanding of the class/spec, while having a good user-friendly interface.

    As long as set information tool is either realistically accurate or mathematically accurate. I couldn't give a s*** about a tool that delivers innaccurate, detailed information to the players, and I wouldn't recommend it to anyone, before it is polished and finished.

    Still looks better than Sim-C and have a more user-friendly layout than Sim-C in my opinion. Wanna go advanced, I would recommend going Sim-C and see what it can do!

    Looking forward to what they come up with, sounds promising!

  9. #9
    I've been working on this project for awhile, though as of late I've not been able to put in as much time as I'd like. It definitely has potential, but like any tool, user input is important - no matter how good the tool is, if it's set up wrong, the results are going to be inaccurate.

    This is what I believe sets AMR apart from SimC, in that it allows users the ability to customize and share their profiles easier. I could set up a Simulation profile and share it with you all, that way you can simply import your character and go, without having the issue of figuring out what variables each person was using and why their results are different. Likewise, another user could script a specific boss encounter and share it with the public for mass use. This really opens up doors that, while possible with SimC, require a lot more work to reach a larger audience.

    However, maintenance of the tool is important, and the area where AMR is going to struggle the most. This type of tool requires millions of lines of code, which AMR's small staff is going to be hard pressed to maintain; even SimC has this problem and it's maintained by a number of individuals working together. Ultimately this will help contribute to SimC remaining the default for "power users", but AMR offers easier options for mass consumption.

  10. #10
    Do you know if https://www.beotorch.com/ is Associated with SimC since it uses that simulator, and are reliable to give the same simulations?

    Edit: Found out that Twintop maintains the priest module at SimC:P
    Last edited by Slayar; 2016-06-22 at 11:51 PM. Reason: Updated answer

  11. #11
    That website uses far to few iterations to get a reliable result, same could be said for AMR as well. Anything under 10k is going to be very prone to RNG, and if you want to be real sure that your results will come out correct you'll want to use at least 100k+. I may be a perfectionist, but I tend to do 50k while doing trial and error editing, and then switch up to 250k when trying to get the final results.
    My Stream
    NollTvåTre Looking for Raiders

  12. #12
    Hey I'm from AMR and wanted to jump in about a few things.

    1. Current testing: the rotations we setup are a first past and not optimized. So in order to REALLY test things out, you'd want to tinker with rotations to make sure they are optimized for each part of your setup (Talent, gear, etc). But keep in mind that procs from legendaries, trinkets and set bonuses are not ready yet (we hope to have them next week).

    2. Maintenance: we wrote a tool to automate all regular data, things that can be ready from tooltips. We've been tracking all of the changes in the beta, and it takes about 1 day to update everything, for every class (by our 1 theorycrafter). So that actually is turning out to be pretty easy and not a big deal. This is actually a misconception I've run into. Since we built the simulator from scratch knowing we'd have to maintain it, we were very smart about how we built it, automating nearly everything from game data. Then we run a comparison from what we had before and what changed, and review those by hand to make sure everything looks good. All within 1 day, for all specs combined

    3. Iterations: we use the # of iterations required to hit your margin of error target. The margin of error is a statistical formula - when we say we get you to 0.25% margin of error, that means there is a 95% confidence that you are within 0.25% margin of error within the mean (using the standards listed on wikipedia, etc). So we know when to stop when we've satisfied that margin of error. Now, once we release the client program, we'll let you target smaller margins of error (like 0.05% or whatever). Running 10,000 iterations doesn't matter, it's the margin of error that you really want to target

    Edit (Forgot to hit on this one): Warrior TC - as said, our dev plays a warrior and toyed with the initial rotation, but I know he didn't have time to optimize it. If anyone wants to talk to us about some rotations optimizations, of course we'd love it!

    Hit me up if you have any more questions!
    Last edited by Zoopercat; 2016-06-23 at 01:58 AM.
    Ask Mr. Robot Human Minion

Posting Permissions

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