< /br>< /br>< /br>

Ask Mr. Robot - 5.2 Weights

Ask Mr. Robot - 5.2 Weights

Postby Zoopercat » 26 Mar 2013, 13:12

Hey Drye, Woaden and Veiled :)

When 5.2 came out, we spent a week running a lot of SimC trials for each class and spec to get our initial updated stat weights. Now we're trying to track down information from each class/spec's theorycrafters to refine our 5.2 weights. As I mentioned yesterday, Shadow Priests are on our list to review this week!

We've read up on a lot of the 5.2 posts on H2P, and re-ran some SimC trials with a lot of the information you provided. Here's what we're thinking of doing. I'd love to start a discussion about this and get your feedback.

  • Remove T14 stat weights, they aren't needed anymore
  • Keeping the T14+ stat weights, which are really good for pre-15 gear levels and for people without RPPM items. This set of weights will include the Haste breakpoint from pre 5.2
  • Add a new set of weights for high end T15 gear. These work well for people with the LMG, RPPM items, and T15 set bonuses. They are similar to the T14+ weights, but Mastery will be higher than Crit. And we're removing all Haste breakpoints. So the weights are: Int (3.89), Spell Power (3.14), Spell Hit (2.39), Haste (2.32), Mastery (1.75), Crit (1.5).

So these differ from the ones posted here. I want to explain why they are different and see if we end up on the same page.
  • We have hit higher than you do. The idea is that Hit is more important than other secondary stats, up until the Hit cap. After the cap it's worth 0. The idea is to get hit capped. Would you agree with that?
  • We have a larger difference between Crit and Mastery than your weights. The idea is that Mastery is a fair amount better than Crit. So when the weights are further apart, it produces better reforges. Is there a reason you would want us to keep them so close in weight? If it's that close, it might mathematically work out that reforging Mastery to Haste is better than Crit to Haste. This could happen if there is 400 Mastery on an item, and only 300 Crit. If there was 300 Mastery and 300 Crit on an item, the Crit > Haste reforge would win of course.

The idea is to get a few stat weight presets that are good for a majority of the players. I'd love your feedback :)
User avatar
Zoopercat
Member
 
Posts: 55
Joined: 01 Oct 2012, 11:36

Re: Ask Mr. Robot - 5.2 Weights

Postby Drye » 26 Mar 2013, 13:30

I was going to response to the mmoC post earlier but didn't have enough time on my lunch. When I get home ill read this over and get back to you.
Stream: twitch.tv/Dryelulz
Hit me up on twitter @DryeLulz
User avatar
Drye
Moderator
 
Posts: 2238
Joined: 09 Feb 2011, 12:36

Re: Ask Mr. Robot - 5.2 Weights

Postby Kilee » 26 Mar 2013, 14:15

I think that as the average iLevel of the game increases, your stat weights should ideally move to match that. If the new standard is tier 15, then I would pick a set of stat weights that is balanced around an optimal normal mode tier 15 set. My reasoning for this is that the majority of your players will be in this range, and with a program like AMR, I think you should be aiming for the largest portion of your target audience. So in short I would probably pick "option 3" from what you've listed.

------------------------------

15% hit with tier 15 gear will be so easy to get that it will be difficult to get rid of hit. The reason it is given a low value in the forums is not because you shouldn't cap - you should still cap. However, when evaluating whether the next piece of gear is better than our current gear, we have to pick an in-between value between hit's top value and 0. In order to this, we arbitrarily pick a value that is equal to another of our secondary stats that is NOT cappable, so that this becomes our "pool" of stats to trade back and forth in order to balance hit from.

I suspect that your program is not designed to work using the above logic, so that is why you are using terminology like "it is worth 0 after cap". I think that this hit issue is a "potato, potaato" thing, and you should just keep doing what you are doing. It would probably be too much re-coding to make your program work a different way.

----------------------------------------

For haste breakpoints... I think that if haste "averages" out to be your best stat, that you should always have haste be your best stat. However, I would lean closer towards the non-breakpoint value over the close-to-breakpoint value. For example, if the pre-breakpoint value is .8, and the post-breakpoint value is .6. I would have haste be something like .65. Bump it up a little bit to account for the fact that there's a small damage increase coming "eventually", but make it ONLY be enough that it encourages people to favor haste/otherstat items over otherstat/otherstat items.

-----------------------------------------

As for the crit/mastery thing... unless you are talking about a very large spread, it makes very little difference. You can spread the stats farther apart if you want, but the two resulting sets of gear will probably be so close together in DPS that you couldn't tell a difference in game.

I've said this a million times: What generally matters most is the ORDER of the stats. The actual numbers represented make very little difference.

As long as stats are a relatively close representation of the scaling in the current tier, that's all that matters in the end. In other words, if we have .56 and .55 for mastery and crit, and you choose to use .57 and .54, this doesn't worry or bother me at all. A few tenths of degrees of seperation is to be expected, accounting for a standard deviation of error, and so on. If however, you're using .50 or below for crit and .58 or higher for mastery (etc...), I might start wondering if your final "BIS" set of gear will hit the same pieces and setup. Even then it might not really make a difference...

---------------------

Not to get off on too much of a tangent, but one of the things that has bothered me this expac is how "insensitive" and close the stat weights are. In the past I felt fluctuations in stat weights had a larger impact on our dps. In this expansion (MOP), the end set we end up with feels practically pre-ordaned, regardless of stat variations.

What I'm saying is that... it's harder to be "wrong" these days. I wouldn't worry about minor differences too much.

----------------------

I would probably have a testing plan after you set your program. Take a sample of priests from several different sets of gear and stages of progression, run them through your program, and then sim to see how well it is doing. If you start seeing strange combinations or un-optimized sets, I would try tweaking the stats again.

Because you are trying to hit an "average" of many different types of players, you need to keep in mind that the stat weights aren't going to match or be tailored to one specific gear set. It's a bad idea to just sim a high end tier 15 priest and use those stats. There's a difference between "stats that work well in my program" and "stats that work well for this one priest set". Just be aware of that when you're setting them.

It's a shame that you do not have several different sets of stat weights based on iLevel and tier sets, so that you can adjust weights dynamically based on what kind of priest is using the software. That is definitely a feature that I would add at some point (if it is not already).
"Falling from heaven is not as painful as surviving the impact."

Revenant - my guild on Echo Isles
The Old Guard - my guild on Earthen Ring
User avatar
Kilee
Member
 
Posts: 930
Joined: 02 May 2011, 11:14
Location: New Zealand

Re: Ask Mr. Robot - 5.2 Weights

Postby Zoopercat » 26 Mar 2013, 14:50

Kilee, excellent, excellent post. TY :) I asked yellowfive to stop by and join the discussion. He codes our site so he'll be pretty helpful to bring into the discussion.
User avatar
Zoopercat
Member
 
Posts: 55
Joined: 01 Oct 2012, 11:36

Re: Ask Mr. Robot - 5.2 Weights

Postby yellowsix » 26 Mar 2013, 16:04

Hey -- zoop pointed me to some of the discussions she's been having. As some background, I write most of the code that actually performs the optimizations on askmrrobot. Or in other words, I tend to focus more on making stat weights turn into good sets of gear rather than finding stat weights -- we want an optimizer that can take any strategy and execute it.

That said, I do get into the details of tweaking weights for particular classes pretty often, and I wanted to take a closer look at some of the shadow priest stuff.

Kilee, a lot of your comments are right in line with the way that we think about stat weights and optimization in general, and I think that it boils down to this: the weights "in and of themselves" do not matter. What matters is that the weights you choose in conjunction with whatever tool you are using (AMR, reforge lite, random tool xyz, etc.) produce optimal sets of gear.

Thus, we focus on making stat weights that play well with our optimizer to produce optimal results. We try to balance two things here -- getting good results, and making the weights understandable. If there are lots of hidden rules... generating custom weights turns into a frustrating trial-and-error exercise. If there aren't enough custom smarts... the resulting sets of gear aren't so good.

Enough general stuff -- shadow priests!


We do provide multiple sets of weights for different levels of gear, and we do have a mechanism to automatically choose the most appropriate set depending on properties of the loaded player. The majority of that mechanism is disabled at the moment... for the most part it just chooses between "pvp or pve". We will be re-enabling the more advanced functionality soon... but there are some usability concerns to resolve first. Basically... we don't want Mr. Robot to keep switching strategies on people when they don't expect it. If a user says, "damnit, I want to use the T14 weights no matter what!" We want to make sure to honor that request.

In the meantime, we set one general pve strategy as the default, and try to make it very obvious how to change that on the UI.

For hit -- yeah as stated above, if someone wants to get near the hit cap, they should just put the weight on hit above other secondary stats with AMR. The optimizer takes care of "cutting it off" and not overvaluing gear with hit. We feel this approach is relatively intuitive for users.

The issue of having weights extremely close together: this is a tricky one. In general, the optimizer will do something that makes mathematical sense... but then you have to step back and ask yourself: what am I really trying to do here?

For example, if I look at Drye's gear: just by inspecting his gear and not digging into the details of shadow priest mechanics, I can more or less tell you what he probably wants to accomplish with his gear, in priority order:

1. Get to or near the hit cap
2. Maximize haste
3. Maximize spell power
4. Maximize mastery

Combining that with the default weights floating around here and at mmo-c, I would also add the following observation, which Kilee also stated:

All of the stats are pretty decent for shadow priests, thus sets of gear that generally have more total stats are better than sets of gear that have more of the "correct" stats but less total stats. Which is just a fancy way of saying: prioritize matching sockets.


Given that information, I can pretty easily produce a set of stat weights that execute this strategy (and thus say that Drye's gear is already 100% optimal). For reference, something like this would work:

int: 1
sp: 0.78
hit: 0.56
haste: 0.55
mastery: 0.49
crit: 0.48

The goal of how our weights work is that one can start from a desired strategy like I have enumerated above, and then translate it into understandable weights. In this case, we can clearly see that in action: hit is just above haste, thus the optimizer will pretty much always get you near that cap. Haste is slightly better than int / 2, thus making gemming for haste optimal. Mastery and crit are just below int / 2, thus making gemming for int better than either of those (and consequently increasing spell power -- int is really a "composite" stat, so I rarely think about optimizing int directly). And lastly, I put crit just below mastery, making mastery the priority after haste.

All of the weights are very close to each other, which results in consistent activation of socket bonuses.

Anyway, I wanted to mention this as an example of how you can make AMR execute a very clear gearing strategy using pretty intuitive weights starting from a ballpark that someone has calculated with simc.

Now the weights zoop mentioned above were some new ones that we generated with simc -- note that when we generate stat weights, we run simc, but we do our own calculation of the resulting stat weights based on the results of the simulation. We simply use simc to get the theoretical DPS of a spectrum of setups, then do a statistical analysis of the results. Those weights were:

int: 3.89
sp: 3.14
hit: 2.39
haste: 2.32
mastery: 1.75
crit: 1.5


Notice how these have the exact same characteristics (and also say that Drye is 100% optimal). The relative values are very slightly different -- slightly more spread on these weights. This has very little impact except when trying to build full BiS sets of gear. Even then, it has pretty low impact, and produces results that have near-identical characteristics.


There is really only one significant difference between the set that this produces and the set that we found to have the highest DPS in simc: trinket choices. simc really favors unerring vision of lei shen... these weights do not. But that's not anything new... trinkets are notoriously difficult to estimate with a pure stat weight approach, and simc is also notorious for heavily favoring certain trinkets. Some personal judgment is always going to be in order for trinket choices.


Anyway -- does the general gearing strategy enumerated above seem to be the right one for T15 shadow priests at the moment? What we really look for from theorycrafters is the strategies, and then we handle the details of making weights (and optimizer codez) that execute those strategies effectively.
User avatar
yellowsix
Member
 
Posts: 7
Joined: 26 Mar 2013, 15:07

Re: Ask Mr. Robot - 5.2 Weights

Postby aica » 26 Mar 2013, 19:15

If mastery and crit are extremely close, with mastery coming out on top, as the stat weights on this site are saying, then yes, you will find a situation where it will be better to reforge mastery to haste instead of crit, and you should do so.
Last edited by aica on 26 Mar 2013, 22:07, edited 1 time in total.
{[( )]}
User avatar
aica
Member
 
Posts: 609
Joined: 11 Feb 2011, 18:20

Re: Ask Mr. Robot - 5.2 Weights

Postby Kilee » 26 Mar 2013, 20:24

Re: Yellowsix

Are you just simply asking for a "blessing" on the second set of stat weights you've posted? If so, I would defer to Twintop or Woaden, who have done more research this tier than me.

----------------

In addition to the spread, I would say another difference is the precision. I think if you can increase the level of precision, you can work with smaller spreads between each number. Given that it cuts at 2, perhaps a larger spread is best. (i.e. I think I agree with putting a bit more space between crit and mastery.)

I thought that simcraft gave us 3-4 points of precision for stat weights. Usually when I'm making my own lists, I don't round them back down... I always find it odd when I see people doing that. If using "normalized" stat weights (like we are doing on this website), I'd rather see at least 3 past decimal rather than 2.

--------------------------------

Trinkets are generally a bit subjective. In the past, this is how I handle them. I divide trinkets up into three categories.

1. Trinkets with raw stats. Trinkets with random-procs which proc raw stats. And trinkets with on-use procs that proc raw stats, and the on-use does not synch up with PI.

For these trinkets, I apply stat weight conversions to them. Proc trinkets are evaluated based on their average up-time.

2. Trinkets that have damage procs, that cannot have stat weights applied to them.

For these trinkets, I find World of Log samples, as close as possible to a Patchwerk-style fight (i.e. The priest has near 100% activity.) I then sim the priest to get their raw, non-scalled intellect weight. Then I take the total damage done by the proc, divide by the length of the fight, and then divide by intellect's scale factor in order to get a PP value for the trinket's damage component. I generally do this around 5 times for 5 different priests and then take an average of the samples. I like taking samples from priests who I know are playing well.

3. Trinkets that are on-use, and match up with PI.

A trinket whose cooldown synchs with PI is far, far more valuable than the calculation of its stat weights. I generally will increase these trinkets' value by about 10% or so. And always keep them for Patchwerk style fights, or fights that require specific "burst" at key points.

I don't know if this helps, but I see a lot of sites around the net that tend to evaluate trinkets poorly, and generally speaking it's because they aren't following the above "strategy".

---------------------------

(You can also get an above-average performance out of a trinket that has a proc that times itself with the length of the fight. For example if I have a trinket that process every 120 seconds (2 minutes), and I'm in a fight that lasts 6 minutes and 30 seconds, that trinket will have an above-average uptime for that specific fight, and will be far more powerful than listed. Knowing things like this can be the difference between good dps and great dps, although I don't know how you would incorporate that into a program, it's just simply important to know that proc trinkets, in general, are always a tad stronger than the projected average. So simply assigning a raw stat weight to them isn't a fair evaluation.
"Falling from heaven is not as painful as surviving the impact."

Revenant - my guild on Echo Isles
The Old Guard - my guild on Earthen Ring
User avatar
Kilee
Member
 
Posts: 930
Joined: 02 May 2011, 11:14
Location: New Zealand

Re: Ask Mr. Robot - 5.2 Weights

Postby Veiled » 26 Mar 2013, 21:13

I've been out of state for the weekend but I just wanted to stop by and say that these guys are the ones to listen to for theorycrafting stuff. Numbers freak me out, so I make them do all the work. >:)
Image
Image
[personal twitter*] -- [site twitter] -- [facebook] -- [youtube]
*my tweets are my own and do not reflect H2P.
User avatar
Veiled
Site Owner & Admin
 
Posts: 2213
Joined: 06 Feb 2011, 16:13
Location: Virginia

Re: Ask Mr. Robot - 5.2 Weights

Postby yellowsix » 26 Mar 2013, 23:03

That's pretty much what we do for trinkets -- we take fights over a spread of common fight lengths (I think 5-8 minutes or something like that), at every 5 second interval therein, and simulate each trinket. Then we average the result to give you an average score for the trinket -- it can be hard to predict how long any particular fight will last, so we find this approach to give the best average value for people.

The only thing you mentioned that we don't do, is a modifier based on how trinkets line up with particular class cooldowns. That's something that we want to look into doing. For shadow priests, is the main thing to factor Power Infusion? We could probably take a few "simple" cases like this and tweak estimates appropriately.


I wasn't really looking for a "blessing" as much as verifying that I was reading all the information correctly, and that the strategy stated is actually the strategy that you guys are going for -- making sure that I didn't miss something or misread something. Sounds like we're on the same page though, and that our simulations show the same thing that yours are showing.


As for precision of stat weights -- we don't go out to 3 points of precision for several reasons:

1. The tool used to generate the weights isn't accurate enough for all 3 points of precision to be meaningful. i.e. we just don't think simc, by its nature, is accurate enough to be that precise with stat weights. All theorycraft in general is fundamentally an approximation.

2. That level of precision almost never has an impact on gear choices made by the optimizer -- optimizer tools just aren't that sensitive.

3. Gear is discrete in nature (you get stats in big chunks, not continuous amounts). For this reason, 3 points of precision is far beyond the level required to make accurate gear choices.

4. Performance. For technical reasons, keeping the precision limited results in much faster optimizations. Given 1-3, there's no cost in accuracy for doing this, so it's a win-win.
User avatar
yellowsix
Member
 
Posts: 7
Joined: 26 Mar 2013, 15:07

Re: Ask Mr. Robot - 5.2 Weights

Postby Woaden » 27 Mar 2013, 00:56

Well ilvl / socket bonus / Meta Gem conditionals might be odd to implement. I'm fairly ignorant of coding so I have no idea how true that is.

Some quick stuff I just wanted to make sure were doubly covered in a tl;dnr fashion:


Gemming
Below ~505 ilvl: Brilliant, Purified and Reckless
Between 505 - 515 ilvl: Reckless, Energized and Quick
Above ~515 ilvl: All Quick except to get Haste and Intellect socket bonuses


Reforging:
No LMG:
Hit to Cap > Haste > Crit > Mastery

With LMG / Unerring Trinket:
Hit to Cap > Haste > Mastery > Crit


Enchanting:
Shoulders:

Back:

Chest:

Wrist:

Hands:

Waist:

Legs:

Feet:

Weapon:

Off Hand:

Finger:


I'm not sure if it's something you're doing already but if you could add a conditional for 'cheap' enchanting for gear that is lower than 476 ilvl. Like Windsong for a Blue 463 ilvl weapon and Jade Spirit for 476 or better weapons, no Living Steel belt buckles for < 476 gear... you get the idea.

I think that would make it more accessible and give suggestions that people are more willing to follow.
User avatar
Woaden
Moderator
 
Posts: 1762
Joined: 22 May 2012, 07:46
Location: Probably Sleeping


Return to Theorycraft



Who is online

Users browsing this forum: No registered users and 0 guests