[Balance] Pawn Stats Weight

[Powerlamer]

Hi guys, might be a stupid question but i want to be 100% sure.
I am setting up Pawn with Manual stats weights, are these correct?

Intellect 1
Haste 0.644
Mastery 0.59
Versatility 0.569
Critical Strike 0.568

If you have any advice on Pawn would me much appreciated
Cheers

[Adramelch]

Hi guys, might be a stupid question but i want to be 100% sure.
I am setting up Pawn with Manual stats weights, are these correct?

Intellect 1
Haste 0.644
Mastery 0.59
Versatility 0.569
Critical Strike 0.568

If you have any advice on Pawn would me much appreciated
Cheers

Those don't seem correct no. How did you obtain them? To the best of my knowledge haste is above intellect, crit is slightly above versa and mastery is the worst stat (although the gap was closed, so to speak, in one of the last tuning passes). I've been using the ones from Cyous' guide, although not sure how accurate/up-to-date they are:

Stat Weights (Normalized) - Weighted 70% / 15% / 15%
Haste (1.00)
Intellect (0.86)
Critical Strike (0.72)
Versatility (0.70)
Mastery (0.62)

In his guide, Gebuz suggests instead: ( Pawn: v1: "Balance Druid": Intellect=1, CritRating=0.9, HasteRating=1.1, MasteryRating=0.7, Versatility=0.8 )

but the idea is prevalent: Haste>Intellect>Crit>Versa>Mastery

[Ilnezhara]

Just sim it yourself.

[Powerlamer]

Adramelch - Cheers. To be honest I couldn't find any boom stats weight, I had to use the ones from AskMrRobot and I was sure they were wrong, so cheers for that :-)

Ilnezhara - I would love to, but I have no idea on how to do it, if you have any link that would be awesome :-)

[Zoopercat]

@Collection I did a quick test for you.

FWIW, Starsurge and Lunar Strike both have increase crit chance from the artifact. Which reduces crit's value relative to versatility. Here's a post where Swol explains how that relationship happens, it's for rogues, but the concept is the same.

AMR weights to get 865 BiS, simmed 315,631: http://www.askmrrobot.com/wow/simula...1970d2f7dbbe9b
Suggested weights from Gebuz guide to get 865 BiS, simmed 312,520: http://www.askmrrobot.com/wow/simula...c6990979bb0541

Those are very close, within 3,000 DPS (about a 1% difference). So either one will give decent gear recommendations, with AMR only slightly winning with that setup.

Just so you know, we do a ton of research on this, and check nearly every thread where someone suspects something might be wrong. We then run tests to see if anything needs to change on our end. And we share these reports so everyone can verify. Nothing is hidden, all assumptions are public.

So if anything does end up wrong, we catch it fast, and run updates every day. (In this case, however, the weights are doing the right thing).

- - - Updated - - -

If you want to do the simulations yourself, head over to http://www.askmrrobot.com/wow/simulator/run

You can run those simple tests without even needing to login or download anything. You can do those all on the website for free

And I have a crash course video here: https://www.youtube.com/watch?v=6ZU4...LIWEa6&index=1

[Tauton]

I used sim craft (25,000 iterations, 300s, 20% range, low world lag - not sure i changed anything else) and was given these weights:

( Pawn: v1: "Fenal": Intellect=7.18, CritRating=4.36, HasteRating=4.49, MasteryRating=4.07, Versatility=4.65 )

I changed a few items based on this and re-ran it, each time things changed slightly but pretty much stayed like that.

(Normalized
Int = 1.00
Vers = 0.65
Hate = 0.63
Crit = 0.61
Mas = 0.57)

A lot of the sims are based on 'if i were at 865' but.. i'm not, so as i'm checking for my current gear, are these stats actually correct and what I should be looking at when gear drops, updating each time I get gear where I assume haste will start to take the lead?

I am doing single target only - how do i get the 70 / 30 split on simulation craft?

[I am Fenal on EU - Silvermoon] However i've been using the addon to make sure everything was up to date, including the artifact

[Zoopercat]

@Tauton Good questions!

So SimC does very localized stat weights, only looking at very specific stat changes with the gear you have. This is good for ranking 1 or 2 items at a time.

For the Ask Mr. Robot weights, we look at thousands of data points over a range of item levels, starting with your character and going up. This makes it good to rank the next 20 upgrades, it lets you plan ahead.

We're actually going to be adding more options to the AMR weights soon that let you choose how far to plan out, starting from 'just the next upgrade' all the way up to 'the next 40 item levels.' So stay tuned for that.

Whatever method you use, you can edit the weights on AMR to rank your gear. Click on the 'view / edit' button just above the gearing strat on the character menu bar. Hope some of this was helpful.

[Tauton]

so by the sounds of it, if I took AMR's values, I will perform at a worse rate than if I took simc's until I get to around 20 ilevels higher?

I'm confused by the info given, everyone that seems to do simming does it for super high ilevel but that's 'theory' - surely we need stat weights for our current ilevel as that's where we are? Haste may be amazing at a higher ilevel because with the extra crit, they synergise (for instance), but at the moment that extra haste isn't as good as, for me, versatility according to the sim.

So my question still stands - for me to perform better in an instance/raid 'today', I should sim in simc and use the values I quoted above and eventually it will sim to say haste becomes better than intellect, once i'm nearer the 865 bis gear level, in a few months time. Surely it's better to perform well now for my raid than base my gear on something I will have in December.

I haven't simmed for a few years but i'm not sure when simming and theory became 'the same', there used to be stat weights for 5 ilevel gaps, so you could plan properly (say, the next few weeks of likely drops) rather than 'this is what you will want to gear for when you are at your BIS level'

Oh also - 70/30 split on simc , how is that done please? The program has changed quite a bit... =)

[Swol]

Stats don't scale in a perfectly linear fashion. Imagine it a a bumpy line going generally up.

The weights that AMR shows you by default smooths that line. The current default weights smooth it from about ilvl 820 to ilvl 880.

The weights you get from SimC don't really smooth it at all. They look at a very localized section of the line and tell you what the current bump is doing.

A problem with using localized stat weights is that people forget they have prerequisites. Yes, you might be able to get a very small increase in theoretical DPS by, say, equipping an item with more Versatility instead of Haste. But, that depends on your current gear already having a lot of Haste. If you just look at the localized weights... how would you know that first you need to get a lot of Haste to even reach this particular "bump in the line?" If you just stacked Vers and never got Haste, you would surely do less damage than the other way around.

The AMR weights tell you what gear you should be trying to get. If you want to squeeze every last bit of theoretical damage out of the gear you happen to get, you should just equip the item you are wondering about and simulate it. See if the damage goes up or down. Calculating scale factors in a localized region is really not that useful. You don't want to spend a ton of gold or effort to get an item that is a very temporary upgrade. But, if you just happen to get an item, try it out. If it simulates to higher damage, use it for now while you go about collecting better gear.

[Tauton]

Thank you both for your explanations =)

- - - Updated - - -

The weights that AMR shows you by default smooths that line. The current default weights smooth it from about ilvl 820 to ilvl 880.

An idea - would it be possible to do as you've done, but have a drop down selection for gear ranges? Preferably everything up to and including normal, then heroic, then mythic raid level gear of that tier rather than the full 820 - 880 range?

I feel that for people who do not do mythic raiding, the gear weights you show them will never be correct. They also won't be correct for most people who do mythic level until a few months down the road. I think adding that option will give people a much more accurate weighting for people at the level they will get to (as well as speed up computational time as there's less items to include).

[Powerlamer]

I agree with Tauton. It would be super cool to have the stats weights based on item level.

Like 800-820
820-840
840-850

Sent from my SM-N9005 using Tapatalk

[Zoopercat]

@Tauton @Collection We actually are putting in an option soon for people to choose the range when running the gearing strategy simulation, to generate stat weights. And I'm glad you brought it up, because the more people show they want it, the higher it goes on our priority list

But to expand on what swol was saying, this doesn't mean the weights are only good for mythic gear. Think of it like a graph. Damage is on the X axis and iLevel is on the Y axis. There's a lot of points that go on the graph. Then we draw a trend line from 820 to 880, to get good weights along the whole path. As I showed with our weights part way along that path, and an 865 gear level, they outperformed the ones from the guide.

AMR weights to get 865 BiS, simmed 315,631: http://www.askmrrobot.com/wow/simula...1970d2f7dbbe9b
Suggested weights from Gebuz guide to get 865 BiS, simmed 312,520: http://www.askmrrobot.com/wow/simula...c6990979bb0541

You might have already understood that point, but just in case I wanted to clear it up. That being said, being able to define the range for yourself is clearly valuable. So it's on our priority list of things to do (and fairly near the top!)

[Tauton]

Sounds great - a trend line including only the gear people are able to get (due to difficulty) will produce a more accurate result. As I said, I used to do this years ago and back then it was a case that (ie) crit was important until x value, then haste overtook it, if you had y mastery and.. yeah. So if you can never reach that crit plateau, haste will never be more important - that's what I feared a lot of people are missing by having '@BIS level, gear this way'.

If yours is a trend line that comes out at almost exactly the same as the very final ilevel figure given by Gebuz, it sounds to me like there are no plateaus and we should all gear the same way..... in which case the "ilevel range option" is pointless Here's hoping that's not the case =)

Still - it's a useful option should plateaus become a thing again. I'm future proofing your algorithms! :P

[thedeisel]

@Tauton @Collection We actually are putting in an option soon for people to choose the range when running the gearing strategy simulation, to generate stat weights. And I'm glad you brought it up, because the more people show they want it, the higher it goes on our priority list

But to expand on what swol was saying, this doesn't mean the weights are only good for mythic gear. Think of it like a graph. Damage is on the X axis and iLevel is on the Y axis. There's a lot of points that go on the graph. Then we draw a trend line from 820 to 880, to get good weights along the whole path. As I showed with our weights part way along that path, and an 865 gear level, they outperformed the ones from the guide.



You might have already understood that point, but just in case I wanted to clear it up. That being said, being able to define the range for yourself is clearly valuable. So it's on our priority list of things to do (and fairly near the top!)

This is awesome, Zoopercat! Love all the fantastic work you guys are doing.

Couple notes on the Gebuz v AMR comparison:

1. In the profile for Gebuz, the cloak isn't enchanted. In the AMR profile it is.
2. Gebuz's stat weights are derived from a mix of Single Target and AoE encounter results. So, it's not that surprising if AMR's profile nips his a bit on single target if it's a regression on single target tests. Is there anyway we can get something similar—an aoe/single target mix of data points for deriving stat weights—from AMR? I really like the way that this is done over a range of gear and would like to see if the weights are robust across a mix of encounter types.

(2a. It's not surprising that Gebuz's weights have more value to haste because haste gets a lot of value from Shooting Stars in multi-target situations.)

[Synadrasa]

How can AMR stat weights be off by so much compared to everyone else?

[Ilnezhara]

@Tauton Good questions!

So SimC does very localized stat weights, only looking at very specific stat changes with the gear you have. This is good for ranking 1 or 2 items at a time.

For the Ask Mr. Robot weights, we look at thousands of data points over a range of item levels, starting with your character and going up. This makes it good to rank the next 20 upgrades, it lets you plan ahead.

We're actually going to be adding more options to the AMR weights soon that let you choose how far to plan out, starting from 'just the next upgrade' all the way up to 'the next 40 item levels.' So stay tuned for that.

Whatever method you use, you can edit the weights on AMR to rank your gear. Click on the 'view / edit' button just above the gearing strat on the character menu bar. Hope some of this was helpful.

AMR is still telling me that armory is not updated for 110, I clearly see my character as 110 on the armory. Any news on when this will be updated.

[huth]

How can AMR stat weights be off by so much compared to everyone else?

Because they're pure single target weights, while the others include about 30% of other scenarios.

[Zoopercat]

@Ilnezhara The armory updates are on Blizzard's end, not ours. But they are working on fixes. You can use our addon as a workaround, here are the instructions: http://blog.askmrrobot.com/2016/08/a...alor-upgrades/

@Synadrasa We've answered that a bit, here's the TLDR version.
1) AMR calculates weights over thousands of data points over a range of item levels to get you weights that are very good at ranking gear, like your next 10 or 20 upgrades.
2) Weights generated from SimC use 2 very local data points and tell you what your next 1 or 2 upgrades is, and then you have to get new weights after that.
3) Gebuz's weights look at AoE stuff.
4) When doing ST tests, AMR's weights win out, but not by enough to make a big deal over or worry about.

[Scarjr]

Hi, I would like to know what the Weighted 70% / 15% / 15% means? What does that mean exactly? Ty