Register

Question about Rune of re-Origination

Face-rippin fun.

Moderator: Forum Administrators

Posts: 28
Joined: Wed Oct 27, 2010 12:44 pm

Question about Rune of re-Origination

Postby Beatrix » Sun Aug 25, 2013 5:16 pm

Hi guys,

I have read through the posts about Rune and the statements about it being the best trinket in the TOT tier even at LFR level in some cases but I am still puzzled.

Why does AskMr.Robot show it as being worst than the 2 I am currently wearing i.e. Renataki's Soul Charm (522) and Vicious Talisman of the Shado-pan? Are the gear rankings Ask Mr Robot that off-kilter and if that is the case is it true for all the other pieces of gear or only for the trinkets because they are difficult to sim?

Is it better in the LFR version than both of the trinkets I am wearing?

Exalted
User avatar
Posts: 1655
Joined: Mon May 24, 2010 10:21 pm

Re: Question about Rune of re-Origination

Postby Tinderhoof » Sun Aug 25, 2013 5:33 pm

Yes AskMrRobot is wrong. It never got updated to reforge for a rune proc last I checked.

Exalted
User avatar
Posts: 890
Joined: Tue Oct 23, 2012 7:15 am

Re: Question about Rune of re-Origination

Postby raffy » Sun Aug 25, 2013 5:41 pm

I don't know AMR's methodology, but in general, trinkets are too complex to model effectively without simulation or very careful analysis. The decent chunk of a trinket's power comes from alignment with other effects. With the Rune, proper evaluation comes from both a change in play-style (utilizing the proc) and proper reforging (1:1:1 w/mastery.) Without these, Rune "blows."

I do my trinket sims as pairs, to account for alignment synergy:
522: https://dl.dropboxusercontent.com/u/298 ... _Table.png
535: https://dl.dropboxusercontent.com/u/298 ... _Table.png

Recently, we were talking about AoC, which experiences alignment issues at low ilvl:
Issue: viewtopic.php?f=3&t=4574&start=450#p18512
Details: viewtopic.php?f=3&t=4574&start=500#p18687
Results: https://dl.dropboxusercontent.com/u/298 ... Table3.png
Same gear, only AoC trinket is changing, and you get almost 15K DPS from the improved alignment.

Here is how the Rune matches up now:
https://dl.dropboxusercontent.com/u/298 ... eTable.png
Here is how the Rune looks at the end of T16:
https://dl.dropboxusercontent.com/u/298 ... Table3.png (look at beast 510 Rune)

Additionally, you can see how the Rune matches up in other situations:
T15 vs 16: https://dl.dropboxusercontent.com/u/298 ... 5vsT16.png
T16 HotW vs DoC: https://dl.dropboxusercontent.com/u/298 ... thHotW.png
T16 Periodic AoE: https://dl.dropboxusercontent.com/u/298 ... Table3.png

Exalted
User avatar
Posts: 1344
Joined: Fri Nov 25, 2011 7:49 pm

Re: Question about Rune of re-Origination

Postby aggixx » Sun Aug 25, 2013 11:26 pm

AMR does have reforging for Rune but it's not as good as Catus'. As raffy said, Rune is just too complicated to accurately value just via stat weights and AMRs comparison of it's value relative to other trinkets isn't meaningful.
Image

Posts: 28
Joined: Wed Oct 27, 2010 12:44 pm

Re: Question about Rune of re-Origination

Postby Beatrix » Mon Aug 26, 2013 11:53 am

ok thanks for the replies, it kinda confirms what I thought. As long as the rest of the gear optimization works I am ok with that.


Thanks again :)

Posts: 22
Joined: Wed Dec 05, 2012 4:02 pm

Re: Question about Rune of re-Origination

Postby zoopercat » Mon Aug 26, 2013 1:23 pm

Hey guys! (For those of you who don't know, I'm from AMR).

We just updated the site and have some new Rune handling. If you click the 'edit weights' button, you'll see a rune widget. I'll run this thread past our developer so he can doublecheck what's going on with the rune. I'll get back to you soon.

Also - I'll be stalking, i mean, hitting you to talk 5.4 stat weights soon :)

Revered
User avatar
Posts: 325
Joined: Tue Oct 23, 2012 4:19 am

Re: Question about Rune of re-Origination

Postby Alpheus » Mon Aug 26, 2013 1:44 pm

zoopercat wrote:Hey guys! (For those of you who don't know, I'm from AMR).

We just updated the site and have some new Rune handling. If you click the 'edit weights' button, you'll see a rune widget. I'll run this thread past our developer so he can doublecheck what's going on with the rune. I'll get back to you soon.

Also - I'll be stalking, i mean, hitting you to talk 5.4 stat weights soon :)


10 points for Gryffindor.
Image

Posts: 8
Joined: Mon Aug 26, 2013 1:46 pm

Re: Question about Rune of re-Origination

Postby yellowfive » Mon Aug 26, 2013 2:08 pm

Hey -- I wrote the rune of re-origination widget for AMR, thought I would give a little detail on how it works:

Firstly, a note on its accuracy: our new gem/enchant/reforge optimizer will get you the highest possible score almost every time -- you will see a big "guaranteed optimal" indicator in the stats on the right when we are certain that there is no solution that would score higher.

That said... our rune optimizer is an approximation, albeit a very good approximation. We chose to do this because it would take an inordinate amount of effort to create an algorithm that both guarantees a solution and runs in a reasonable amount of time, and we're really close to the end of 5.3 here. People will start replacing the rune very soon, so we wanted to focus effort on 5.4 stuff.

The rune optimizer has to balancing two competing goals: get the biggest proc possible, and get the highest score possible. We put a higher priority on getting a bigger proc, which is why you will sometimes see your score go down if you have the rune.

The last thing to note: it's very hard to compare the rune of re-origination to other trinkets because it requires such a specialized optimization. We actually modify your stat weights internally to do the optimization, which exacerbates this issue. (We modify your weights for the rune because... that's really what you want! The rune effectively makes the worse 2 of the 3 stats much more valuable, as long as their totals stay lower than the best stat.) For this reason, you really need to just know if you want the trinket or not, and lock it in.

When you are looking at an "absolute" ranked list, we do a quick estimate of the rune's value were you to be perfectly optimized for it. Some people feel that this estimate is a bit low... it might be. Particularly for monks, people feel that the very large mastery proc is a lot stronger than stat weights would suggest, and they're probably right.

Revered
User avatar
Posts: 325
Joined: Tue Oct 23, 2012 4:19 am

Re: Question about Rune of re-Origination

Postby Alpheus » Mon Aug 26, 2013 3:27 pm

yellowfive wrote:... People will start replacing the rune very soon ...


Nope.

yellowfive wrote:The rune optimizer has to balancing two competing goals: get the biggest proc possible, and get the highest score possible. We put a higher priority on getting a bigger proc, which is why you will sometimes see your score go down if you have the rune.

The last thing to note: it's very hard to compare the rune of re-origination to other trinkets because it requires such a specialized optimization. We actually modify your stat weights internally to do the optimization, which exacerbates this issue. (We modify your weights for the rune because... that's really what you want! The rune effectively makes the worse 2 of the 3 stats much more valuable, as long as their totals stay lower than the best stat.) For this reason, you really need to just know if you want the trinket or not, and lock it in.

When you are looking at an "absolute" ranked list, we do a quick estimate of the rune's value were you to be perfectly optimized for it. Some people feel that this estimate is a bit low... it might be. Particularly for monks, people feel that the very large mastery proc is a lot stronger than stat weights would suggest, and they're probably right.


I don't see how your current model could effectively support trinkets like Rune (or cooldown reduction, ICD procs, TED - in the future). You are essentially setting up a BiS setup for every class/spec combination and are giving out stat weights in such a way such that every linear increase in stats moves the players closers on the trajectory towards this BiS setup. But this will no doubt immediately diverge for specs which have several vastly different, viable near-BiS setups (ie. mastery>haste>crit vs. 1:1:1 for ferals), even worse if the two different setups have different playstyles (ie. classic bleed uptime/roar vs. taking advantage of high bleed magnitude procs and predictions).

I think Rune optimization for ferals (and perhaps other classes) is not a technical issue, but a user-oriented one. On one hand you have your average Bob trying to decide what to gem/reforge etc. so that he can rest assured that he's taking the best care of his gear when approached for guild raids, pugs, etc; and on the other, the min-maxing raider who is no doubt a couple steps ahead of the currently deployed AMR build and probably just wants a user-friendly way to conduct a lot of math quickly and set up a shopping list.

I would argue that the former player values 2 things: getting a fairly reasonable estimate of BiS reforges/gems/items for minimal player-supplied input and where to go from there. The latter on the other hand would cash out and go all-in expecting the, literally, BEST in slot reforges/gems/enchants for his current gear AND/OR the ability to set his own expectation of desired trinket/weapon/set combination and have AMR gear up around this fixture (I understand that AMR can do something like this right now by locking item slots but it is the least optimized-for approach as far as usability goes).

To avoid any ambiguity: these two scenarios look similar, but it is important that the latter case does not rely on pre-set stat weights. Mis-predictions and stat weight rounding errors are going to give the min-maxing playing a false positive viable setup and in my opinion AMR should run SimC iterations for every possible gem/enchant/reforge combination and disable setting and viewing stat weights. Stat weights are for the average player who needn't be concerned with how item and ability scaling works.
Image

Revered
User avatar
Posts: 325
Joined: Tue Oct 23, 2012 4:19 am

Re: Question about Rune of re-Origination

Postby Alpheus » Mon Aug 26, 2013 4:28 pm

Also, while you're at it, please check out http://theincbear.com/forums/viewtopic.php?f=9&t=1079
Image

Posts: 8
Joined: Mon Aug 26, 2013 1:46 pm

Re: Question about Rune of re-Origination

Postby yellowfive » Mon Aug 26, 2013 4:50 pm

I haven't looked that closely at the T16 trinkets yet -- I just assumed that some of them would be quite good and start replacing T15 trinkets once 5.4 goes live. We usually examine new trinkets closer to patch day because they tend to get tweaked often during the betas. Regardless, the current rune optimizer is quite good -- unless you want to wait for a significantly slower optimization (order of magnitude slower or more), you can't really get any better. I'm aware that there are some programs floating around that do this -- most people who are that OCD about the rune will know where to find them ;) In the worst cases, our rune optimizer will give a proc that is about 150 rating less than the maximum possible. In some cases, it actually beats the other rune optimizers out there.


Regarding your other comments:

I would argue that the former player values 2 things: getting a fairly reasonable estimate of BiS reforges/gems/items for minimal player-supplied input and where to go from there.


This is exactly what our website does best! We give very good BiS sets that are going to be close to optimal performance. If you think that some of our BiS sets seem a bit off, then please let us know. The only cases that are always a bit tricky are trinkets -- sometimes certain special trinkets like the rune will seem better than an automatic optimizer might suggest. But when you get down to it... that's a pretty minor thing overall for the "average Bob".

The latter on the other hand would cash out and go all-in expecting the, literally, BEST in slot reforges/gems/enchants for his current gear AND/OR the ability to set his own expectation of desired trinket/weapon/set combination and have AMR gear up around this fixture


We literally guarantee the BEST gems, enchants, and reforges for a given set of gear in almost every single case. The rune of re-origination is one of the only exceptions, and as described above, it is extremely close to optimal. Even though we don't put the "guaranteed" indicator on BiS sets of gear, the gems/enchants/reforges for the chosen set of gear are almost always guaranteed optimal. It's just the set of chosen items that might not be perfect, but it will always be very good.

So that's one thing to clear up: Mr. Robot literally is the best gem/enchant/reforge optimizer available now. We put a lot of work into being able to say that with confidence, and please let us know if you ever run into any issues -- we will fix it as quickly as possible.

Regarding setting your own expectation of desired trinket/set items -- this is extremely easy and user-friendly on Mr. Robot. We explicitly designed the UI to support this case for the very reason that we know it is difficult to rank certain trinkets and set bonuses. Just click on Best in Slot to get the default recommended set, then click on the item slot you want to manually change, e.g. click on the name of the item in slot Trinket 1. Then choose your desired trinket from the list. You'll see the red "lock" indicator show up below its icon at the top of the gear editor. Press save, and there you go! Mr. Robot will optimize around your chosen trinket.

Let me know if you have any other questions about how to use any of the features. The website will let you easily and quickly build any custom set of gear and get the best possible gems/enchants/reforges for that gear. You can either hand-modify your current gear, or hand-modify your BiS set to create a custom "wish list".

Revered
User avatar
Posts: 325
Joined: Tue Oct 23, 2012 4:19 am

Re: Question about Rune of re-Origination

Postby Alpheus » Mon Aug 26, 2013 5:14 pm

yellowfive wrote:We literally guarantee the BEST gems, enchants, and reforges for a given set of gear in almost every single case. The rune of re-origination is one of the only exceptions, and as described above, it is extremely close to optimal. Even though we don't put the "guaranteed" indicator on BiS sets of gear, the gems/enchants/reforges for the chosen set of gear are almost always guaranteed optimal. It's just the set of chosen items that might not be perfect, but it will always be very good.


I would like to be an extra mile picky on wording here. An ideal best tool would not require a degree of magnitude of abstractions just to optimize for calculation speed. What I wanted to clarify in my previous post is that the min-maxing raider would prefer a fixed subset of gear to be permutated and analyzed based on true sims and not stat weights.

It is indeed the best tool to reforge, gem and enchant given pre-set stat weights. But it does not give the overall best predictions, because it is assuming that the stat-weights are correct. Any reliance on stat-weights is going to give sub-optimal local maxima, thus a min-maxing optimizer should not rely on them. I believe this demographic of players would be prepared to trade performance for precision (similar to the trade-of you did with hit/exp capping).
Further, this should be done without "exceptions". The player has already excluded certain exceptions by setting/locking certain item combinations. A fitting SimC combat list should be chosen (like you no doubt are already doing for at least professions and talent specs) and optimized from there.

This review isn't meant as much for 5.3 since right now the only true "exceptions" are RoRo/Renataki/Juju and the default SimC list handles the latter two fairly well (correct me on this). I'm trying to provoke a thought experiment which would perhaps enable the integration of general solutions for new concepts like ICD/cooldown overlapping in 5.4 (AoC+berserk on high item level, sigil of rampage) which might in certain cases be more potent than RoRo is in 5.3 but are going to be less "exclusive" because of the larger amount of options; ergo with the lack of specialized, exclusive scenarios you would have a harder time prioritizing or even deciding to implement a specialized solutions, making an even wider gap between a classic trinket setup and an overall optimized one.


Also, bump:
[...] while you're at it, please check out http://theincbear.com/forums/viewtopic.php?f=9&t=1079
Image

Posts: 8
Joined: Mon Aug 26, 2013 1:46 pm

Re: Question about Rune of re-Origination

Postby yellowfive » Mon Aug 26, 2013 7:12 pm

I replied in that incbear thread, we'll see if anyone responds. There really isn't a whole lot of difference between our rage build and what they seem to be citing, other than the priority of crit vs. hit/exp. Doesn't matter too much to me, I usually defer to the popular theory. Seems that it has shifted since we first created those weights.

With 5.4 coming up, we'll be researching/testing/developing new stat weights, so it's a good opportunity to update everything. The more direct feedback we can get, the better.


It certainly would be more accurate to simulate everything rather than use stat weights to score combinations of gear. But it would be so frustratingly slow... it's just not practical. Also, theorycrafters tend to exaggerate the "inadequacy" of using fixed stat weights to optimize gear or even to choose BiS gear. It is actually much better than you would think. We have shown many times that using good stat weights can produce BiS sets that are very optimal, despite the fact that the relative stat weights theoretically drift as your stat levels change.

Another way to state the problem: people try to use "perfect" stat weights, then create a tool that can take "perfect" stat weights and produce the "perfect" solution. It never works very well because "perfect" stat weights tend to be very localized. A very simple example: if you are above the hit cap, your "perfect" weight on hit is 0. But that's not very useful for an optimizer... there's no way that your stat weights would ever get you to the perfect solution.

What we try to do instead is create stat weights that might look a little different than you would expect, but tend to more reliably move you towards BiS solutions. We have multiple presets because one single set of weights just can't cut it, but you can actually get away with 2 or 3 sets of weights that work extremely well for almost everybody. Also keep in mind: stat weights are not as sensitive as people think. Even if the value of e.g. crit compared to mastery shifts as you stack one or the other, the relative priority doesn't change that often. This gives a lot more wiggle room with weights than you would expect.

Anyway -- we actually used to write a simulator for feral druids, way back in the day. It was a fun project, but we found that it was ultimately not the right tool for optimizing and recommending gear. Stat weights are certainly an "approximate" approach, but so is simulation ;)

Our new optimization algorithm basically gives users a very good, very predictable calculator to test gearing theories. You can use whatever method you want to determine what you think is the optimal way to gear... a simulator, a formula, a napkin and pen... then use our optimizer to maximize it. You can basically get any combination of stats that you want if you provide the right weights.

Posts: 22
Joined: Wed Dec 05, 2012 4:02 pm

Re: Question about Rune of re-Origination

Postby zoopercat » Mon Aug 26, 2013 7:24 pm

One thing I also wanted to point out: thanks to Theck, SimC now exports stat weights directly into the Ask Mr. Robot website. And we also provide an export of your gear/gems/etc to simC. We really love integrating with awesome tools that already exist. So while we aren't running a simulation on our site, we do integrate with the best one out there.

Anyway, just a minor note I wanted to throw in there. I'll leave the rest of this discussion to yellow and all of you fluid druid folks.

Revered
User avatar
Posts: 325
Joined: Tue Oct 23, 2012 4:19 am

Re: Question about Rune of re-Origination

Postby Alpheus » Mon Aug 26, 2013 8:20 pm

yellowfive wrote:I replied in that incbear thread, we'll see if anyone responds. There really isn't a whole lot of difference between our rage build and what they seem to be citing, other than the priority of crit vs. hit/exp. Doesn't matter too much to me, I usually defer to the popular theory. Seems that it has shifted since we first created those weights.

With 5.4 coming up, we'll be researching/testing/developing new stat weights, so it's a good opportunity to update everything. The more direct feedback we can get, the better.


It certainly would be more accurate to simulate everything rather than use stat weights to score combinations of gear. But it would be so frustratingly slow... it's just not practical. Also, theorycrafters tend to exaggerate the "inadequacy" of using fixed stat weights to optimize gear or even to choose BiS gear. It is actually much better than you would think. We have shown many times that using good stat weights can produce BiS sets that are very optimal, despite the fact that the relative stat weights theoretically drift as your stat levels change.

Another way to state the problem: people try to use "perfect" stat weights, then create a tool that can take "perfect" stat weights and produce the "perfect" solution. It never works very well because "perfect" stat weights tend to be very localized. A very simple example: if you are above the hit cap, your "perfect" weight on hit is 0. But that's not very useful for an optimizer... there's no way that your stat weights would ever get you to the perfect solution.

What we try to do instead is create stat weights that might look a little different than you would expect, but tend to more reliably move you towards BiS solutions. We have multiple presets because one single set of weights just can't cut it, but you can actually get away with 2 or 3 sets of weights that work extremely well for almost everybody. Also keep in mind: stat weights are not as sensitive as people think. Even if the value of e.g. crit compared to mastery shifts as you stack one or the other, the relative priority doesn't change that often. This gives a lot more wiggle room with weights than you would expect.


The most relevant example I can come up with is that I import my character into AMR, set hit/exp hardcaps, set the stat weights and hit 'Find BiS'. I get a setup without RoRo. Okay, I check RoRo, lock it, make sure the RoRo widget is on and set to mastery and also enable +8 upgrades for BiS tier gear. I click find best in slot again.

It sets up a setup with the ra-den chest, Rena/Roro and darkwood spiritstaff. All good so far. Now to reforges (mind you this is supposed to be the BEST setup). Now the hit/exp ratios are at 2667/2573. That's fine, I believe that's as low as we can get. As for secondary stats, I'm looking at 8295/8197/7693 (29172) agi.

Now what I did next is put the same gear setup into Catus and run a red-red reforge/regem. what I came up with was 2653/2600 hit/exp and 8443/8442/8441 (this already is a huge difference). Now, I have a high item level (well, we're talking BiS anyway) and I prefer running hybrid secondary gems, so I regem orange into red and come up with 2653/2573 and 8825/8824/8824. To be honest, I would've expected a closer match, at least with the first one.

Now the tricky part isn't only the reforges themselves, but my planned BiS setup which I've planned with Catus manually even has 2 different items, namely boots and rings which brings me a reforge of 2653/2552 (closest fit of all combinations) and 8908/8907/8907 (28473 agility).

Now the weird thing is, when I put my hand-picked setup from Catus and put it back into the optimizer of AMR, I get reforges of: 2667/2553 and 8391/7858/8024 (29018 agi). Now what's bugging me is that they are both reasonably equal fits in AMR (trading agility for mastery, secondaries almost match, a bit less exp waste on second) but they give drastically different "overall BEST" setups in Catus (even when gemming red-red).

---

Now I understand that the comparison isn't really fair because Catus is a tool heavily optimized for RoRo and ferals in general where-as AMR isn't - but isn't that the point? Especially since this will escalate even more in 5.4 with further "special proc" trinkets, not to mention RoRo will still most likely be in heavy use far into Heroic SoO.

Expectations:
  • With everything that's been said in your previous post, I would've expected a much closer match
  • Since I knew the reforge/regem/enchanting wasn't optimal, I would've prefered some way of "tweaking" stats so that I could manually reproduce the setup from Catus
  • When there are several very close BiS-optimizations available (we're talking within ~0.5%) it would be very handy to be able to pick from a list of reasonable candidates to make gearing choices easier (instead of 99% of people going for the same items).
Image

Posts: 8
Joined: Mon Aug 26, 2013 1:46 pm

Re: Question about Rune of re-Origination

Postby yellowfive » Mon Aug 26, 2013 8:35 pm

Alpheus -- what is the region/realm/name of your character? I'll try out catus and see if I can figure out what would be causing differences. Also let me know if there are any fancy settings that I need to use in catus -- I've never used that tool before.

edit: you're going to need to give me step-by-step on how to reproduce your case in catus... it's kind of a daunting program for someone who has never seen it before. I would need to know what settings to reproduce your numbers, and... how do I view the recommended gem, enchant, and reforge changes when it is done?

Revered
User avatar
Posts: 325
Joined: Tue Oct 23, 2012 4:19 am

Re: Question about Rune of re-Origination

Postby Alpheus » Mon Aug 26, 2013 8:59 pm

yellowfive wrote:Alpheus -- what is the region/realm/name of your character? I'll try out catus and see if I can figure out what would be causing differences. Also let me know if there are any fancy settings that I need to use in catus -- I've never used that tool before.

Armory link

I have Catus set up for all raid buffs, consumables and my imported talents/glyphs.
For Reforge: Hit/Exp at least 2550, Range 120, No overflows, Checked: Hands, Back, Change Gems, Break Bonuses, Keep Colors
For gemming I have set (inputs left-to-right): Adept, Adept, Glinting, Yellow, Crystalized Dread, Always use (checked), Yellow, Unless socket (checked)

Whenever you make changes hit Regem, Reforge

My manual BiS snapshot I was working with was (hit edit under equipment to paste):

Spoiler: show
96581 +8 >hm :95346 :76670 # Head: [543] Headpiece of the Haunted Forest
96815 +8 >me :76666 # Neck: [549] Quadra-Head Brooch
96583 +8 >hm :76666 :76666 "Greater Tiger Claw Inscription" # Shoulder: [543] Spaulders of the Haunted Forest
98148 +8 :76670 "Superior Critical Strike" # Back: [608] Tigerfang Wrap
95033 +8 >cm :76666 :76666 :76666 "Glorious Stats" # Chest: [549] Chestguard of Coruscating Blades
94998 +8 >me :76680 :76680 :76670 Socket # Waist: [549] Strap of Murderous Strikes
96764 +8 >ce "Fur Lining - Agility" # Wrist: [549] Bindings of Multiplicative Strikes
96580 +8 >et :76666 "Greater Haste" # Hands: [543] Grips of the Haunted Forest
96582 +8 :76666 :76687 "Primal Leg Reinforcements" # Legs: [543] Legguards of the Haunted Forest
96903 +8 >mt :76666 "Blurred Speed" # Feet: [549] Spurs of the Storm Cavalry
95021 +8 >hm :76687 "Greater Agility" # Finger 1: [549] Ra-den's Swift Seal
96740 +8 >tc "Greater Agility" # Finger 2: [549] Sign of the Bloodied God
96741 +8 >em # Trinket 1: [549] Renataki's Soul Charm
96918 +8 # Trinket 2: [549] Rune of Re-Origination
96978 +8 :76658 :76658 Socket "Dancing Steel" # Main Hand: [549] Darkwood Spiritstaff


Edit: after exporting I noticed the headpiece had an upscaled item level, the fixed setup is 2628/2551 with 8701/8700/8699, will probably need to re-run the example. Highest I managed to find was 8836/8835/8835 with Twist-toe, bloodied god and gore-soaked. It's late, more tomorrow :D


The bottom line really is that I like AMR and I would like to use it but I have a feeling that it comes about 5-10% short as far as heroic tier-level play optimization is concerned and it's the lack of this last mile that sort of prevents me from using it full-time. I believe that it is mainly the design decision to use stat weights that prevents this last mile to be achieved. If the solution to end-game min-maxing is manually editing stat weights.... well then that's a problem: it creates a paradoxical cycle - I need a reforged BiS gear setup to calculate weights, and I need weights to calculate gear. And if I'm going to use 2 tools just to get AMR running then we're already beyond fairly reasonable estimate of BiS reforges/gems/items for minimal player-supplied input.

If a feral asks me right now if they should use AMR to reforge, regem I'll tell them "No, use Catus instead. AMR is easy and pretty good but confusing and difficult to optimize.". I hope we can find a way to change that answer into "Yes, absolutely!" Catus will still be the #1 tool I reckon, but for the vast majority of ferals, there will be a shining beacon of hope which should hopefuly alleviate some of the pain with playing ferals in general (and hopefully raise the population a bit as well, *fingers crossed*).
Image

Posts: 8
Joined: Mon Aug 26, 2013 1:46 pm

Re: Question about Rune of re-Origination

Postby yellowfive » Tue Aug 27, 2013 1:43 am

Thanks -- I'll take a look. I think that I have catus set up now to compare.

One thing right off that you'll need to change on AMR to get similar results: you're forcing catus to use orange gems everywhere. You'll need to similarly force AMR to use orange gems. You can more or less do this by setting your agility weight equal to 2x your mastery weight (so using defaults, mastery = 1.8, set agility = 3.6). If you do this, you'll see that our optimization is extremely similar to your current gear -- it scores a little higher, but is a little less "perfect" around the rune proc.

I'll see if I can play with our rune optimizer a bit now that I see what catus is doing. The tricky thing about the rune is that there are a few competing priorities: do I maximize my rune proc size at the expense of agility or accuracy around the hit/exp caps? Or the other way around: do I optimize my agi/hit/exp levels first, then use the "leftovers" to get the biggest rune proc that I can?

Just for kicks I ran 3 quick simulations for alpheus, one favoring agility gems (~880 more agi than your current setup), one similar to your current setup, and another that drops even more agility for a bigger proc (~1100 less agi)

simc:
high agility: ~242800 dps
medium agility: ~243500 dps
low agility: ~242800 dps

catus:
high agility: ~261900
medium agility: ~260600
low agility: ~260700

So interestingly, simc likes the "balanced" approach here -- not really sure why. Catus likes the high agility approach (which is more like Mr. Robot's default weights than alpheus's custom tweaks to always use orange gems).

Anyway -- that was done very quickly, just messing around with trying to figure out what the right priority would be for a rune optimizer. This really simple test would suggest: don't sacrifice agility on gems to get a bigger rune proc (according to catus).

Revered
User avatar
Posts: 325
Joined: Tue Oct 23, 2012 4:19 am

Re: Question about Rune of re-Origination

Postby Alpheus » Tue Aug 27, 2013 4:59 am

yellowfive wrote:One thing right off that you'll need to change on AMR to get similar results: you're forcing catus to use orange gems everywhere. You'll need to similarly force AMR to use orange gems. You can more or less do this by setting your agility weight equal to 2x your mastery weight (so using defaults, mastery = 1.8, set agility = 3.6). If you do this, you'll see that our optimization is extremely similar to your current gear -- it scores a little higher, but is a little less "perfect" around the rune proc.

I'll see if I can play with our rune optimizer a bit now that I see what catus is doing. The tricky thing about the rune is that there are a few competing priorities: do I maximize my rune proc size at the expense of agility or accuracy around the hit/exp caps? Or the other way around: do I optimize my agi/hit/exp levels first, then use the "leftovers" to get the biggest rune proc that I can?

Just for kicks I ran 3 quick simulations for alpheus, one favoring agility gems (~880 more agi than your current setup), one similar to your current setup, and another that drops even more agility for a bigger proc (~1100 less agi)

simc:
high agility: ~242800 dps
medium agility: ~243500 dps
low agility: ~242800 dps

catus:
high agility: ~261900
medium agility: ~260600
low agility: ~260700

So interestingly, simc likes the "balanced" approach here -- not really sure why. Catus likes the high agility approach (which is more like Mr. Robot's default weights than alpheus's custom tweaks to always use orange gems).

Anyway -- that was done very quickly, just messing around with trying to figure out what the right priority would be for a rune optimizer. This really simple test would suggest: don't sacrifice agility on gems to get a bigger rune proc (according to catus).


Aggix ran sims some while ago proving that pure-red vs. hybrid gemming in high item level (for RoRo at least) is DPS neutral so I went with a setup which has a higher max dps (opposed to statistical average, which should be the same). It's mostly just shifting the distribution curve, not so much changing the median. In the example above I've done both scenarios (hence red-red and red-orange) for comparison but it got a bit botched since my helm was fubar. The numbers you ran below are all within 0.005% so Aggix's theory seems to hold ground. I changed over to this hybrid approach for higher RNG peaks so I could scumbag some high ranks if I get lucky with procs but it's really not a big gamechanger seeing how close the sims were.

I ran the red-red setup first (agi heavy) to compare with AMR and seeing big differences already I went on to my hybrid one and I think tinder is running pure-yellow as well in 550 gear to share gear with tanking pieces while keeping his dps neutral.

I would just like to address one more issue at this point:
You can more or less do this by setting your agility weight equal to 2x your mastery weight (so using defaults, mastery = 1.8, set agility = 3.6).


That's mainly the thing that's confusing to me. I knew the "goal" that I wanted to approach since I simmed it with an alternate tool (Catus), but if I was solely using AMR, how would I know what the right weight is (in this case mastery = 2x agi)? The way I found out with Catus was simple. I changed gems, I maximize reforges, I run 10k sims. I take note, try a different setup. The community, over time, also finds certain thresholds and boundaries which can either increase/decrease or fine-tune dps so I add those as well into my strategy (in this case hybrid gems). I run 10k sims again. Find a high median, take note of all the maxes and pick one which suits my current progression level (ie. lower standard deviation with a higher median for progression and higher deviation with high peaks for ranking on farm content).
Image

Exalted
User avatar
Posts: 1655
Joined: Mon May 24, 2010 10:21 pm

Re: Question about Rune of re-Origination

Postby Tinderhoof » Tue Aug 27, 2013 12:36 pm

I ran the red-red setup first (agi heavy) to compare with AMR and seeing big differences already I went on to my hybrid one and I think tinder is running pure-yellow as well in 550 gear to share gear with tanking pieces while keeping his dps neutral.

I am still rockin the orange gem setup. Aggixx is the one using pure gems to help split time with tank gear.

Posts: 8
Joined: Mon Aug 26, 2013 1:46 pm

Re: Question about Rune of re-Origination

Postby yellowfive » Tue Aug 27, 2013 1:38 pm

That's mainly the thing that's confusing to me. I knew the "goal" that I wanted to approach since I simmed it with an alternate tool (Catus), but if I was solely using AMR, how would I know what the right weight is (in this case mastery = 2x agi)? The way I found out with Catus was simple. I changed gems, I maximize reforges, I run 10k sims. I take note, try a different setup. The community, over time, also finds certain thresholds and boundaries which can either increase/decrease or fine-tune dps so I add those as well into my strategy (in this case hybrid gems).


So as you know... AMR isn't a simulation tool. You're always going to need to use some kind of simulator, or read some kind of theorycraft to figure out how you want to gear. We recommend a default that is generally quite good for most players who aren't into doing their own research, but if you are a heroic raider and want to tweak stuff, you'll have to use some kind of simulation tool to test out your theories. That's not a problem with AMR... it is what it is. We have the simc export (and simc now has an AMR export) to make it very easy to optimize, test in simc, repeat. Similar to what you are doing with catus.

As for knowing what the right weights are to pick hybrid gems -- a stat weight optimizer is fundamentally a very simple and intuitive tool. Mastery (and all secondary stats) have exactly 2x the item budget as agility on gems. So if you make the weight of agility more than 2x the weight of mastery, any good optimizer will favor agility gems. If you make it less than 2x mastery, the optimizer will favor mastery. Depending on how close you are to this break-even point, the optimizer will tend to ignore or not ignore socket bonuses that require yellow or red gems. If you set the weights exactly at the break-even point, you are saying that a red agility gem is equivalent in value to an orange agi/mastery gem, so the optimizer is free to use whichever it feels like.

If you have a basic knowledge of how gear works in WoW (which you all certainly do!), you can see how stat weights allow you to easily customize your gearing strategy. That's why stat weights are so popular... they are simple and flexible, and for us programmers, they are an elegant solution to a complex problem. It's just another tool that you have to learn how to use, like catus or simc or whatever else you may be using.

Sometimes I think that because AMR is so ridiculously simple to use for the common cases... that they don't notice there is a lot of flexibility available for the savvy user. Custom weights, custom caps, manual locking -- all of these give you fine control over the results.

If you ever need some pointers on how to customize stat weights to execute particular gearing strategies, just post on our forum and describe what you are trying to do, and someone will try to give you an example set of weights. Once you've done it once or twice, you'll get the hang of it.

Revered
User avatar
Posts: 325
Joined: Tue Oct 23, 2012 4:19 am

Re: Question about Rune of re-Origination

Postby Alpheus » Tue Aug 27, 2013 4:01 pm

Thanks for the clarification. This thread already de-toured a bit but I'd just like to wrap it up with one more question: if I'm not mistaken the stat weights SimC return are actually DPS predictions (e.g. a stat weight of 1.7 for haste is going to give you 1.7 dps for every point of haste). That means that the "overall score" of an AMR setup is usually the predicted average DPS of this setup in raid conditions. Are there any plans to perhaps take advantage of this information when using default weights?
Image

Posts: 8
Joined: Mon Aug 26, 2013 1:46 pm

Re: Question about Rune of re-Origination

Postby yellowfive » Tue Aug 27, 2013 5:37 pm

While the simc weights are technically a DPS prediction... we find that the score doesn't match the DPS estimate that simc produces closely enough to rely on it. It is usually proportional to your DPS (i.e. higher score will give you more damage), but not exactly spot-on. Part of this is due to the fact that we take a little liberty with the score to factor in special cases. For example, we add a bonus to your score if you have indicated that you want to use run speed enchants. We do this so that any solution without a run speed enchant scores worse than any solution that does have one. (This particular case can be turned off by the user, or is ignored if your class has an innate run speed increase.)

There are a few other cases like this that cause the score to diverge a bit from your average DPS. We really only treat the score as a tool for relative evaluation of sets of gear scored with the exact same scoring function.

Posts: 8
Joined: Mon Aug 26, 2013 1:46 pm

Re: Question about Rune of re-Origination

Postby yellowfive » Thu Aug 29, 2013 10:16 pm

I just posted an update that improves on the rune of re-origination optimizer for AMR. Give it a shot -- it should be as good or better than any other rune optimization tool you may have been using. If you run into any issues, please let me know on our forum ASAP and I'll fix it!

This blog post describes some more of the details about how it works and how to use it, be sure to check it out: http://blog.askmrrobot.com/2013/08/rune ... igination/

Revered
User avatar
Posts: 325
Joined: Tue Oct 23, 2012 4:19 am

Re: Question about Rune of re-Origination

Postby Alpheus » Fri Aug 30, 2013 3:55 am

yellowfive wrote:I just posted an update that improves on the rune of re-origination optimizer for AMR. Give it a shot -- it should be as good or better than any other rune optimization tool you may have been using. If you run into any issues, please let me know on our forum ASAP and I'll fix it!

This blog post describes some more of the details about how it works and how to use it, be sure to check it out: http://blog.askmrrobot.com/2013/08/rune ... igination/


o.O

Image

Seriously though, this is really cool. Put in secondaries = agi /2 weights for hybrid reforges or the default weights for agi-heavy. Go (agi / 2) +1 on any secondary if you want yellow gems as well.
Pro stuff! +50 points for Gryffindor! Raffy, you just got a contender :)
Image

Next

Return to Kitty DPS

Who is online

Users browsing this forum: Google Feedfetcher and 6 guests