Archived

This topic is now archived and is closed to further replies.

revulva

Team Robot Gear, Gem, Enchant, and Reforge Optimizer

92 posts in this topic

LucidityAxel: I checked it and our tool comes out to 16.85% spell hit for him, just UNDER the cap, not over. Just to be sure, I double checked, and we are indeed calculating spell hit accurately

MatsT: I am not surprised that you are able to find an example where the optimization is very slightly worse. Like I mentioned in the OP, we are forced to use an approximate method in order to make the site usable. It is definitely possible for people to come up with slightly more optimal tweaks manually. This particular example is a little larger margin of error than I would like, though. We will see if there is any way to improve the algorithm. Thanks for the info!

Gir: There is no such option, you are correct. If that particular item is bugged in the game, we might be able to hard-code in an exception until they fix it.

Ayrea: Thanks for the link. We will fix that in our update today. FYI, I'm trying to check this thread regularly, but if you post technical bugs like this on our technical support forum (Technical Support) there are multiple people who read that board so we can respond to stuff faster.

Binkenstein: First off, we are aware that the chaotic meta is giving a slight DPS increase, even in it's current "broken" state. We decided to not take the time to implement these meta gems until 4.0.6, since it takes significant effort to implement each meta requirement in the optimization algorithm. Second, I will explain to you where our Elemental stat weights come from, since a lot of research went into them.

There is an Elemental BiS thread here: http://elitistjerks.com/f79/t114460-elemental_bis_gear_set_up_discussion_4_0_onward/

The folks in this thread have cleverly tweaked Elemental gear. Where possible, the method I use to create stat weights is taking known BiS gear sets and working backwards. One of the first things I noticed was that the stat weights SimC creates do NOT get us to the Bis setup. Here is a near-optimal set based on the SimC stat weights:

Mr. Robot - Elemental Shaman - World of Warcraft

This simulates out to about 3.74% lower than the BiS gear setup.

Using your stat weights from http://totemspot.com/2011/01/cataclysm-elemental-guide-4-0-3/:

Mr. Robot - Elemental Shaman - World of Warcraft

This simulates out to about 9.34% lower than the BiS gear setup. Now, it looks like there is a problem with an interaction between these particular stat weights and the estimates we use for the Elemental set bonuses - we generally estimate set bonuses as a DPS increase, so we use the Spell Power weight to assess if they are worth getting. I modified your stat weights very slightly by bumping up spell power to get a more "fair" assessment of them:

Mr. Robot - Elemental Shaman - World of Warcraft

This simulated out to about 4.34% lower than the BiS gear setup.

So, I looked at the BiS Elemental setup and said to myself... why aren't these stat weights leading our optimization algorithm to the BiS setup? Using SimC's stat weights or totemspot's stat weights, the BiS set does not have the highest EP value. This means that the stat weights are not right. I tweaked the stat weights to create gear sets that are more optimal. Using the default stat weights on Mr. Robot's site, you end up with this:

Mr. Robot - Elemental Shaman - World of Warcraft

This simulated out to 1.66% less than the BiS setup. A portion of this can be attributed to the meta gem, which we are willing to accept until 4.0.6 patch. So, this achieves our goal of being within 1.5% of the optimal gear setup with our approximate optimization algorithm. Actually, I did find some stat weights that put me even closer to the BiS t11 heroic set, but then some weird stuff started happening at the t11 and pre-raid level. These weights still get good sets at all gear levels, although sometimes you can end up slightly under the hit cap. People can manually tweak the weights if being slightly under the hit cap with certain gear combinations bothers them.

So, TLDR: I think Mr. Robot's stat weights are more accurate than SimC's stat weights or totemspot's stat weights. We arrived at these stat weights empirically based on what the BiS elemental gear has been observed to be.

Share this post


Link to post
Share on other sites
LucidityAxel: I checked it and our tool comes out to 16.85% spell hit for him, just UNDER the cap, not over. Just to be sure, I double checked, and we are indeed calculating spell hit accurately

Thanks for the response. I checked again, and I must be missing something.

The pre-optimized setup has 1712 hit rating. Both the tool and the in-game paperdoll frame agree on this.

As far as I can tell, the optimization suggests exactly two changes:

head - 67 mastery -> 67 haste (no change to hit)

neck - 50 mastery -> 50 spirit (net +50 hit rating)

This is a net increase of 50 rating; 1712 + 50 = 1762 total, which is over the cap. The in-game paperdoll agrees with this figure, but the tool says the optimized set has only 1727 rating:

Mr. Robot - Character Profile - World of Warcraft

Is the tool suggesting some other changes to hit or spirit that I don't see? Where is it getting the 1727 total from?

Share this post


Link to post
Share on other sites

I'll admit, I had to stare at this one for a little while...

We both agree that the original setup has 1712 hit rating.

My math shows the optimized setup as having 1726 hit rating on gear.

The optimization does the following with regards to spirit/hit:

Head: +121

Neck: +50

Back: -50

Chest: -87

Waist: +40

Feet: -10

Ring 1: -50

Net change of +14.

Share this post


Link to post
Share on other sites

The scale factors generated by SimC (or stat weights or equivalence points or ...) are relative to the fixed gear point at which they were calculated. If one were to start with a naked toon and pick gear to maximize the sum of stat*stat_weight products, then scale factors generated by SimC will have limited use in this scenario.

Share this post


Link to post
Share on other sites

Mr. Robot - Character Profile - World of Warcraft

The enchant upgrades are pretty clearly accurate as well as the gem, however the reforging results in worse stats by a pretty large margin. The optimization lands me a few rating points over the hit cap, a few under the exp soft cap, and then with much less Mastery in favor of about 2.5% crit. You can see my overall score drops 32.53 points despite the gem and enchant upgrades.

You also say you consider Darkmoon Card: Hurricane BiS for DW frost in this thread but the optimizer continues to suggest Impatience of Youth instead.

Share this post


Link to post
Share on other sites

dedmonwakeen: We understand that what SimC and other people's spreadsheets, simulators, formulators, etc. are trying to do when they create stat weights is to tell you the relative value of each stat at that particular gearing point. As stats change, the weights change. What we have been doing is reverse engineering the stat weights we use based on Best in Slot gear configurations. Between item level 346 and 372, stats just don't change that much. So, the weights still work well at all the gear levels, but the site becomes more and more accurate the better your gear gets. We allow for the "major" discontinuities such as soft and hard caps, but we ignore the small, gradual shifts. When I say that the stat weights that SimC creates don't always look right - that is because sometimes the weights would lead me to make changes that the simulator itself is telling me are less optimal.

Kaejin: We're going to put out an update later tonight. Our code that enforces the DMC: Hurricane as BiS for DW Frost DKs needs to be re-implemented. That will be in the update. As far as your example of an "optimization" that slightly reduces your total EP value - it looks like you just happen to have very optimal reforges on your gear already. The drop in EP value of 32.53 is about a 0.17% change, which would be within in the margin of error on our approximate method. (Keep in mind that about 12 of that is because of the run speed enchant being put on your boots reducing your mastery by 15.) That change in value can basically be seen as negligible. I'd say, upgrade your enchants, and leave your reforges where they are!

Share this post


Link to post
Share on other sites

I'll probably continue on with how I've been doing my reforging manually then, exchanging 2~ Mastery for 2.5% crit just isn't within my ability to accept. Not to say I don't like the tool, but I guess I'm far too anal about being at least at the hit and exp cap and then making sure my next best stat is as high as I can manage. :p

It certainly seems great for anyone who doesn't like to putz around with that sort of thing, though. I liked the original Mr Robot DPS simulator quite a bit as well. The GUI was always pretty nice.

Edit: Also, don't the Fleet meta and Lavawalker boot enchant share the same run speed increase? Once the new metas are introduced this won't be an issue, but until then it seems like an oversight.

Share this post


Link to post
Share on other sites
Between item level 346 and 372, stats just don't change that much.

I was actually less concerned with the total statValue pts available then the relative distribution of various stats. It would be nice to be able to input ( InitialStat, StatWeight ) pairs, where your goodness function becomes a sum of StatWeight*(Stat-InitialStat) for each stat.

Share this post


Link to post
Share on other sites
I'll admit, I had to stare at this one for a little while...

We both agree that the original setup has 1712 hit rating.

My math shows the optimized setup as having 1726 hit rating on gear.

The optimization does the following with regards to spirit/hit:

Head: +121

Neck: +50

Back: -50

Chest: -87

Waist: +40

Feet: -10

Ring 1: -50

Net change of +14.

I think I see the problem. Here is the results pane that I see in my browser:

mrrobot.png

I take it the blank sections in green should have actual text in them?

This behavior is 100% reproducible for me in Internet Explorer, Mozilla Firefox, and Chrome.

Share this post


Link to post
Share on other sites

Blank section in green means the optimizer thinks you should remove the current reforge on that slot.

Share this post


Link to post
Share on other sites

MatsT is correct. Perhaps we should consider putting some text like "removed" to be more clear.

Several of the bugs found yesterday have been fixed, including the issue with using too many jewelcrafter gems. Here's a link to the full list of changes:

Change Log

Share this post


Link to post
Share on other sites

I looked at the thread on how trinkets are calculated and I believe that DMC:V is calculated incorrectly because the INT gained from the proc does not improve your mana (only your max mana) the value of the INT from proc is not equal to the true value of INT.

Rather it is equivalent to the spellpower and crit gained from the proc:

INT*(value of spellpower)+INT/(int to crit%)*(crit rating to crit %)*(value of crit)

The second part of the formula just converts INT->Crit Rating.

Share this post


Link to post
Share on other sites
I looked at the thread on how trinkets are calculated and I believe that DMC:V is calculated incorrectly because the INT gained from the proc does not improve your mana (only your max mana) the value of the INT from proc is not equal to the true value of INT.

Rather it is equivalent to the spellpower and crit gained from the proc:

INT*(value of spellpower)+INT/(int to crit%)*(crit rating to crit %)*(value of crit)

The second part of the formula just converts INT->Crit Rating.

I would agree that the portion of the intellect stat weight which accounts for increased mana pool should not be considered.

We used a sort of rough estimate for DMC: Volcano right now because everything I've read seems to unanimously agree it is the BiS trinket for damage casters. It's just that good. Over-estimating it slightly at this point won't actually make a difference in gear selection. I actually prefer it being over-estimated for now because I want it to always be selected.

Share this post


Link to post
Share on other sites
I would agree that the portion of the intellect stat weight which accounts for increased mana pool should not be considered.

We used a sort of rough estimate for DMC: Volcano right now because everything I've read seems to unanimously agree it is the BiS trinket for damage casters. It's just that good. Over-estimating it slightly at this point won't actually make a difference in gear selection. I actually prefer it being over-estimated for now because I want it to always be selected.

If we are to believe SimC:

From my calculations using stat weights from BiS 359 from SimC (which may not actually be BiS), it is true that Theralion Mirror and DMC:V come out at BiS. (with TM as 1 and DMC:V as 2) Though DMC:V is very close to Bell and depending on stat weight changes it may be possible for Bell to surpass DMC:V.

Additionally the reason I bring up the calculation is because depending on where people's gear is currently at stat weights change significantly and when entering custom stat weights it may be beneficial to have a more accurate modeling of DMC:V.

TL;DR...I'm just being nitpicky and GREAT JOB on the SITE =)

Share this post


Link to post
Share on other sites
For SV hunters some WotLK enchants can be better than Cataclysm enchants, specifically

Retribution has the same issue; secondary stat enchants are rated below the AP enchants from WotLK. The default stat weightings Mr. Robot uses for Ret agree that

The WotLK enchants are not recommended when appropriate: example.

Share this post


Link to post
Share on other sites
It treats spirit/hit as identical stats for moonkin, elemental, spriest. We can probably update the display to make that more obvious. It will always favor spirit over hit, though - just because spirit at least does a little something for your out of combat regen.

You may want to change how it treats spirit/hit for human priests - we get an additional 3% spirit from racial so spirit is always preferred over hit. Using my character it is telling me to remove hit->spirit reforging which is not optimal.

Share this post


Link to post
Share on other sites

In what I believe is an issue regarding only enhancement shamans it seems that one issue I have found is that it is not detecting elemental precision being talented, so it is not calculating spirit value at all. This is an issue because the stats enchants to chest is also free hit rating if that is talented which needs to be accounted for. I imagine this is because it is an elemental talent, and the robot site may not even grab talents. My suggestion would be to have a box to put in 0/1/2/3 points in elemental precision to calculate the actual value of the spirit and get more accurate hit rating reforging.

Share this post


Link to post
Share on other sites

Smeags: I see what you are saying. I didn't realize it was so close a decision for mages. We will have to think about how we can improve the estimate slightly.

CaseyTheRetard: We'll have to add in those enchants in a small update.

Alinna: It removes that reforge since spirit and hit are weighted the same in the stat weights. I see what you are saying that as a human you can actually get 1.5 extra spirit by reforging hit to spirit... Technically that is not a zero-value reforge, although it is so minor you could argue it's not worth doing. We'll see if we can add something in to stop the site from recommending that you remove it.

Passislisk: We are actually taking elemental precision into account. When you load a character we check how many points are in the talent. Can you give me an example where you think it is not being recognized?

Share this post


Link to post
Share on other sites

Another hit question for you.

My 'more stats' pages gives me the following (as a boomchicken)


Armory Optimized Best in Slot
Spirit 1734.00 1477.00 1797.00
Spell Hit 19.80 % 17.06 % 17.08 %
[/HTML]

Armory/paperdoll is reporting 17.01% hit. Is there some buff you are taking into account I'm not thinking off (as you stated this is full raid buffed)? Regardless I'm struggling to see how these conversions work - even attempting to take into account the 468 pure hit I have on gear.

Mr. Robot - Character Profile - World of Warcraft

edit: Ahh, OK, thanks guys makes perfect sense. Now can you make the DPS version drop to avoid this kind of confusion for me in the future? ;)

Share this post


Link to post
Share on other sites

Lab, Mr. Robot is averaging in the on-use Spirit from your Dream Owl. That is probably a mistake for DPS.

Share this post


Link to post
Share on other sites

Yup, it's averaging in the spirit proc from your trinket. We'll update it to ignore spirit procs for the hit calculation. That's one of those situations where you don't expect DPS'ers to even use a trinket like that ;)

Share this post


Link to post
Share on other sites

  • Recently Browsing   0 members

    No registered users viewing this page.