Scale Factors and You, an explanation of 'stat weights'

[gahhda 95]

Feel the need to post this in every thread that discusses 'stat weights' from simc.

 

Direct from the simc wiki for stats scaling: "Simulationcraft can evaluate scale factors for stats, as shown in our sample reports (see simulationcraft.org). Scale factors are computed by adding a given number of points (deltas) to a character's stat, and then comparing the resulting dps/heal/deaths count with the reference one : the scale factor is equal to the dps/heal/deaths count difference, divided by the delta."

 

When you run a 'stat weight' sim, this is what happens:

 

The sim run your current gear. Then, it adds x rating to one statistic (defaulted to 100), calculates dps, and then finds the amount of dps you gained and divides by x (100). This tells you how much dps per rating you get for that statistic OVER THE NEXT ONE HUNDRED POINTS OF RATING.

 

In other words, it is entirely based on your current gear choices. Changing any single factor on your gear will result in different scale factors ( 'stat weights' ). They have NO RELEVANCE TO WHICH STATS YOU SHOULD PRIORITIZE in the long term, only for the immediate future.

 

To figure out which stats are the best in the long term, we need to use plots. Reforge plots are more accurate at distinguishing the value of statistics relative to one another, but they take an incredible amount of time, so I'll use dps plots to give a general idea for each spec.

 

Demo:

 

http://gyazo.com/0661eedb540d012855bde427d48d87ff

 

This is a stat plot where I set all 3 to 400 initially. So, this plots the value of each stat from 400 rating to 1000 rating, at 25 rating intervals over 24 points each.

 

What this shows is that mastery is the clear winner overall. Crit and Haste are identical until haste starts to plateau around 800 rating, where as crit continues to scale linearly upwards.

 

Reforge plots take an incredibly long amount of time but are even more accurate to demonstrating relative stat values.

 

Overall, mastery wins, and haste and crit should be kept at SIMILAR levels. If one ever gets drastically (300+ rating) above the other, the values will flip all over the place.

 

Destro:

 

http://gyazo.com/10ed431f9a3e1e9d58c665702fa768ba

 

Same conditions as demo, I set all 3 to 400 rating, the plot shows ratings of 400 to 1000 at 25 rating intervals with 24 total points.

 

Crit and Mastery are insanely close, crit narrowly winning, and haste trailing drastically behind both. The reason some of you are getting values of haste at #1 from your scale factors is because there's very little of it on the gear. The impact of this is reduced by switching to haste/crit orange gems in red sockets, increasing the amount of haste on your gear even if you have a mastery/crit heavy gear set.

 

Overall, destruction hasn't really changed, other than haste being the clear last stat now. Mastery is still much better than either for aoe/cleave, and since it's so close ST, I recommend:

 

Mastery >= Crit > Int > Haste

 

Keep in mind, haste is only below int if you have a decent amount on gear, when you have ~400 rating or below it's clearly going to beat int.

 

Affliction:

 

http://gyazo.com/8b58ec63620ba49ae2c43432e9eee2ad

 

Again, all 3 stats set to 400 rating, plots 400 to 1000 rating at 25 rating intervals over 24 total points.

 

Haste pulls ahead clearly ST, they buffed pets compared to 5.4.8, so haste gained a lot of value for all of the specs compared to mastery. Crit is still not good, but less awful than live, you shouldn't avoid it like the plague, it can be advantageous to go for a haste/crit item for example.

 

Because of the nature of partial haste ticks, mastery is still theoretically better for burst cleave.

 

If either haste or mastery get too far ahead of the other, the values will flip.

 

I'm still recommending:

 

Haste >= Mastery > Int > Crit

 

Try to keep your haste and mastery close, breakpoints no longer matter so just eyeball it.

Edited October 13, 2014 by gahhda

[Wolland 1]

This is an awesome explanation, thanks Gahhda!

[Zagam 1,982]

I think the big misleading thing most people do is they look at stat weights and assume that's the most powerful stat - they don't realize that the stat weights are values of DPS per point of secondary stat AT THAT GEAR LEVEL.  Hope this clears things up for most people. 

 

Even still, while it's powerful to recommend the best stat, most playstyles are very close and can be justified in playing a certain style with preference as long as the playstyle adapts to the change.  Demonology in 5.4 had players going C>M>H, M>H>C, H>C>M, and all of them were able to put up extremely high numbers within a percent or two of the other. 

[Kazistrasz 216]

Obligatory post:

As always, secondary Stat changes will never be responsible for more than 10% of all the DPS improvement you stand to make. 90% of the room you have to grow comes from analyzing logs to determine how to fix mistakes in playing your spec.

Similarly, your efforts and focus should be split no more than 10% on gearing and at least 90% on playing better. If you read Zag's Demo guide and the section you spend the most time thinking about is the stats section, you need to reconsider your priorities.

[Zagam 1,982]

Stats are always the popular discussion because it's the easiest for people to understand.  Without reforging, someone will claim that a Mastery Crit piece is better than Haste Crit piece and think it's a huge deal while it might only be worth 30-40 DPS better - or even worse if they don't play the spec ideally.  Everyone still asks 'what's the best stats' instead of 'when is the best time to use Soul Fire?' which makes me sad.  Eventually we will get everyone to see.

[gahhda 95]

Obligatory post:

As always, secondary Stat changes will never be responsible for more than 10% of all the DPS improvement you stand to make. 90% of the room you have to grow comes from analyzing logs to determine how to fix mistakes in playing your spec.

Similarly, your efforts and focus should be split no more than 10% on gearing and at least 90% on playing better. If you read Zag's Demo guide and the section you spend the most time thinking about is the stats section, you need to reconsider your priorities.

 

I would argue that an affliction lock stacking crit would probably do 10% less than one minimizing crit. The point is still relevant though :p.

Edited October 13, 2014 by gahhda

[gahhda 95]

I'm going to be updating this with reforge plots hopefully today.

[Zagam 1,982]

I think I'll make Kaz play Destruction and make Locky play Affliction in SoO just so we get a good mix of what specs look like on various fights.  I'm not sure what Locky wants to do, but I know Kaz has a Demo boner too.

[Liquidsteel 279]

I'll flip flop between Demo and Destro. I get bored with only one spec. I'll likely play Affliction too if a fight warrants it.

[gahhda 95]

I think I'll make Kaz play Destruction and make Locky play Affliction in SoO just so we get a good mix of what specs look like on various fights.  I'm not sure what Locky wants to do, but I know Kaz has a Demo boner too.

 

Fight has lots of movement: demo

Fight doesn't has lots of movement: destro

 

Kappa.

[Zagam 1,982]

ORRRR

 

Fight exists: Demo. 

[BaleFire 0]

Can you plot multi strike yet? Did not know.

 

Also, are these plots for level 90 or 100?

[gahhda 95]

Can you plot multi strike yet? Did not know.

 

Also, are these plots for level 90 or 100?

 

Multistrike is only available on a very tiny amount of items at level 90, not really worth the time to plot. It's obvious you want to get multistrike if you can, if you have 0 of a statistic it instantly becomes the most valued statistic.

[Kazistrasz 216]

Gahhda, I wold in return argue that someone playing affliction with buttloads of crit probably has serious play issues like 20% uptime on DoT.

[Liquidsteel 279]

Gahhda, I wold in return argue that someone playing affliction with buttloads of crit probably has serious play issues like 20% uptime on DoT.

A couple weeks ago I played Affliction with Mastery > Crit > Haste on a few fights. Felt wrong, but damage wasn't terrible!

[gahhda 95]

Anyone on the forums familiar with converting csv files into surface graphs, I have no bloody idea how to work gnuplots :D 

 

http://pastebin.com/mVH5272p

[JvChequer 43]

TL;DR:

 

Demo: Mastery > Crit = Haste focusing on Mastery gems/enchants

Destro: Mastery = Crit > Haste focusing on Mastery gems/enchants due to cleave.

Affli: Master = Haste > Crit focusing on Mastery gems/enchants due to cleave.

[Furyio 3]

Nice thread. As someone who really likes the tool AskMrRobot...but never trusts its stat prior...this goes a long way to help me create custom weights for 6.0

[Kazistrasz 216]

There's really no point in AskMrRobot right now. There's no longer a hit cap to worry about hitting. There are no longer haste breakpoints to worry about getting to. There's no longer reforging to try calculating. All of its functions are now obsolete.

[Liquidsteel 279]

It's great for making gear sets in tandem with Simcraft though.

[Furyio 3]

There's really no point in AskMrRobot right now. There's no longer a hit cap to worry about hitting. There are no longer haste breakpoints to worry about getting to. There's no longer reforging to try calculating. All of its functions are now obsolete.

 

That is a bit severe. Personally I used it heavily for finding BIs lists, and also managing gear lists. 

[gahhda 95]

Turns out 3d reforge plots out of simc have exporting issues, meh. Generally concept still stands.