Looking at the just-released SC (2218), I still think armor penetration is a little too effective, especially compared to other theorycraft tools. I'll look into why-- it might be the cycle, it might be something else.

Ah..... Good catch! Update will be incoming......

So, the damage I'm getting from best-in-slot pre-Ulduar with Ulduar mechanics is consistently higher in SC than in FBN-- both the bleeds and non-bleeds. I'll try to figure out sometime this weekend whether SC or FBN is in the wrong and let the loser know, probably by checking the calculators (Toskk's, Rawr). Toskk's shred damage also seems low, though. Worst-case I'll do some training dummy tests.

Edit: I've decided I'm happy enough with ArPen. What throws it off of other relative stat values is that it stacks with itself, so significantly more of it makes each point worth significantly more; the calculators and FBN use much smaller stat differences than SC, so they have much less of the effect of ArPen stacking with itself. 250 ArPen changes its relative value a lot.

I've looked into SC those days to find out why there were a big dps difference from my simulation and SC simulations in DPS output.

The result were that I found 2 mistakes and a un-simulated mechanic in my simulation.

First of all I didn't take into account physical vulnerability changes, I was stuck with 2% while now it's 4% (more dps)
Second I made an error with shred glyph, basically I didn't remember to check for RIP so if I shreded with RIP down it generated 1 RIP tick (now fixed, less dps, does it works fine in SC?).
Third I didn't put into the simulation the boss "time to die" (now fixed)
Forth I didn't take professions into account (now fixed)

With those changes:
my simulator dps with BiS pre-Ulduar gear is about 7173 dps (still 46 dps short SC, but a lot less than before).
ArPen value went up (basically I use FB more, due to "kill time").

One main difference with respect to ArPen is that I use a smaller stats gap to compute stats value, the other is probably that I undervalue Grim Toll (I'm looking into it).

There are still some dps loss, possibly:

I should look into rake-energy micromanage (looking at it now).
Do you use temporary buff like haste potion, hysteria, etc..? I don't use them in my sim.

Actually, that's not quite correct. The profiles don't have profession specific bonuses, however, there is a static amount of AP/SP added to account for 2 professions... In the case of Druid_T7_00_55_16 (Shred-only profile), that's 128 AP.

Oh, never noticed that, my fault.

Just to let you know, I've just finished reviewing my code and now the Druid_T7_00_55_16 kind does 7050 dps. The difference doesn't seems high.

I can also verify that the mechanics in SC are repeatable at Toskk's, except for issues over at his site (significantly underestimating Grim Toll as well as unrealistic bleed uptimes). There's still some tweaking that could be done (energy pooling) but there aren't many significant gains left.

I'll reiterate what Nightcrowler said before, though-- scale factors with respect to ArPen need to be tweaked. Because it's a stat that stacks with itself, it has a much different scale factor at current gear level versus 250 points away. Most other stats don't follow this behavior. From my testing, I don't need nearly as large a scale factor to pin down a good ArPen value-- I was using 50 scale at 1000 iterations, and it was pumping out much better values.

Finally, a comment about the scale factors: is there any way to not generate factors for int/spell power/spirit/etc when you're only looking at melee characters? It probably wouldn't save time for the whole-raid sims, but for those looking at smaller groups (i.e. melee-only, feral druids only, etc) it would cut the wait time in half.

We probably need to revisit all of the scale factors. To be honest, I initially picked values more based upon variance reduction as opposed to accuracy..... and as with all "temporary" constructs, they have lived longer than they should.

Regarding unnecessary machine-cycles spent evaluating meaningless stats: I feel your pain. It makes me twitch sitting there waiting for AP scale factors to generate for a Shadow Priest.... I'll do something about this soon.

EDIT: I opened issues for both of these points for tracking purposes.

Old question that I felt like finding the answer to:

Do bleed ticks suffer from crit depression?

I have a small sample, and I'm not sure if it's significant enough, but it seemed conclusive enough for me. 1500 autoattacks, 500 rip ticks against a training dummy. Character sheet crit percentage of 48.64%, although occasional darkmoon card: greatness procs put the average close to 50%.

It's not the cleanest test (shouldn't have had the greatness proccing, more data), but it passes the 95% statistical significance for Rip crits not being ~50%.

In other words, bleeds suffer crit depression against bosses, and it looks like it's similar/the same as normal melee (or greater).

Don't worry, I'm out of issues! Except glances, of course. (2700 autoattacks, 21.6% glances... So probably not the 24% I see in my parses, but I don't have a solid number yet.)