Haste Breakpoints and Their Effect on Stat Weights

[Twintop]

Just posted a new blog post where I play around with Haste Breakpoints of our DoTs to get to the bottom of why Haste was under performing in our stat weights. In short: below the SWP-2 breakpoint, Haste is king. Above, it really doesn't matter if you go Haste/Crit/Mastery.

Below SWP-2 Haste Breakpoint (8085 Haste)
Int > Spell Power > Haste > Hit=Spirit > Crit=Mastery

Above SWP-2 Haste Breakpoint (8085 Haste)
Int > Spell Power > Hit=Spirit > Crit=Mastery=Haste

If you want to look at the guts of the post, it's over here ---> http://warcraft.twintop-tahoe.com/?p=205

[Celentes]

Thanks for figuring that out.
A few questions:
How trivial is 8085 amount?
Do you have any graphs of stat scalings? Does the haste reign supreme significantly below breakpoint? The way simcraft figures these numbers, you need to either change the delta to negative for below-breakpoint char (it'll look backwards to determine scaling) or make a big graph for all of that.

P.S. Make it obvious in the title that you're talking about beta.

[Twintop]

Thanks for figuring that out.
A few questions:
How trivial is 8085 amount?
Do you have any graphs of stat scalings? Does the haste reign supreme significantly below breakpoint?

8085 is only obtainable with Heroic gear. Obviously Goblins will have an easier time reaching their breakpoint.

I'm working on stat scaling graphs but probably won't have anything to post until next weekend.

I would imagine that right below the SWP+1 and DP+1 breakpoints that something similar would occur but haven't done any testing. Same goes for the middle ground. I'm interested in seeing what happens with the T14 normal BiS profile I'm working on.

[Celentes]

Oh, I edited the post a bit too late >.>.

It's just, I really doubt that if you make simcraft not look at the breakpoint to determine stat scaling (by avoiding it with negative deltas, for example, although a graph would be more descriptive) that hit/spirit would fall behind haste.

[yurano]

Instead, they could just do

( (DPS @ X% Haste) - (DPS @ 0% Haste) ) / (X% Haste)

[Celentes]

Instead, they could just do

( (DPS @ X% Haste) - (DPS @ 0% Haste) ) / (X% Haste)

This way it wouldn't be correct. Deltas are a proper way to determine snapshotted values (ok, derivatives would be even better, but not applicable to simulations), but they have their flaws which you should be aware of. Similarly, with this formula you've posted, it is still skewed by the breakpoint.

Graphs are the best thing ever, also the hardest ones on CPU.

[yurano]

This way it wouldn't be correct. Deltas are a proper way to determine snapshotted values (ok, derivatives would be even better, but not applicable to simulations), but they have their flaws which you should be aware of. Similarly, with this formula you've posted, it is still skewed by the breakpoint.

Graphs are the best thing ever, also the hardest ones on CPU.

Deltas are derivatives?

My solution is a total derivative instead of an instantaneous derivative?

[Celentes]

Derivatives are limits with delta approaching zero. From that point of view, a fixed delta is closer to zero than your whole haste amount. It also works for simulations with zero ratings.