Do you agree with my last post about haste scale factor?

People have already reported that the procs from some of the other 25man trinkets do stack when you have both the 10man and 25man versions.

Given this, it seems like a reasonable assumption that the caster trinket can stack too.

How long this situation would last however remains to be seen.

While the exact length of the fight may have *some* influence on haste scale factor generation, I doubt it's all that significant. If you have specific numbers showing the influence to be significant, please post them, including which parameters you used.

So far all you've said is "then I've changed fight time from 300s to 301s and got different numbers", which could easily be expected variations - there's enough RNG involved that repeated runs, even at 10k iterations, aren't going to result in exactly the same DPS.

Can you please point me out to where did you find that info?

I tested several haste values to examine haste scaling at 300s fight (10k iterations):
576 — 8051.4
577 — 8038.8 <----
...
583 — 8055.8
584 — 8051.0 <----
...
592 — 8064.0
593 — 8060.9 <----
...
596 - 8074.6
597 - 8061.0 <----

It seems like we have local dps losses on haste scaling. I think it might be fixed if we implement random lag (+0...10% of constant value for example) after casts.

Lag is currently modeled as a random distribution between +/- 2 std deviations.

EDIT: And fight length is random per iteration because we use target_health and not max_time as the hard stop.

Keep in mind simcraft uses a scaling delta of 250 by default, so the effects you're observing at a delta of 10 aren't necessarily relevant for most users.

Your argument for varying max_time when generating scale factors does make sense in theory, but it's not that easy to control the fight length for any given iteration. The max_length variable is more of a guideline which indirectly controls fight length by setting target_health - it will produce the desired average fight length in the end, but takes many iterations to reach that balance.

250? So much? The most have 500-600 haste and it is 50% of their haste. Is it correct to use such far point to approximate scaling?

Ran it with the default settings, guess that's a no.

Note, possible changes for the profile: Switch the Chest with the 258 Val'Kyr drop.
Regarding the trinkets - If it doesn't work there is always IotDS or Flare to swap with.

Other than that I think it's pretty much fixed.
If anyone has a way better solution at hand - give it a shout.

Large deltas perform three functions:

(1) They reduce variance so that people get the same numbers without running an excessive number of iterations.

(2) They give a better indication of where you are going as opposed to local minima.

(3) 250 Haste Rating is 7% spell haste (actually less when you add in all the other factors).

If you had asked "Is it intended", I would have simply said "Yes".

Since you asked "Is it correct", I'm squirming a little.

#1 in that list is a real killer. That is why smooth_rng looked so incredibly promising because it could reduce the required number of iterations by an order of magnitude. However, we were running into edge-condition problems associated with haste and with very powerful trinket and set bonus procs.

Using smooth_scale_facotrs=1 the deltas are 40%-50% of the standard values, giving a better picture of local effects. However, when examining the slope of a squiggly line, the locations of the sample points can have a rather drastic effect on the slope.

Note: Just noticed I'm using a wrong gem in the neck, too much notepad editing.

With the 3254 Release
Set V1:
9764 100.0% Warlock_T9_00_13_58_v1
Warlock_T8_00_13_58 Sta=0.00 Int=0.38 Spi=1.05 SP=1.64 Hit=4.40 Crit=0.98 Haste=1.51 Lag=0.00

Set V2:
9809 100.0% Warlock_T9_00_13_58_v2
Warlock_T8_00_13_58 Sta=0.00 Int=0.36 Spi=1.03 SP=1.62 Hit=4.42 Crit=1.00 Haste=1.10 Lag=0.00

Set V3: (Skyweaver chestpiece)
9801 100.0% Warlock_T9_00_13_58_v3
Warlock_T9_00_13_58_v2 Sta=0.00 Int=0.37 Spi=1.04 SP=1.64 Hit=4.29 Crit=1.02 Haste=1.52 Lag=0.00

Ad above:
Im using the Lightbane Focus in the V2 already, Chalice_of_Benedictus is the same.

But, keep looking for that Haste anomaly, its getting even more interesting with V3 now:
V1: Spell Stats: power=3190(3190) hit=14.07%(369) crit=26.55%(643) penetration=0(0) haste=16.41%(538)
V2: Spell Stats: power=3188(3188) hit=14.03%(368) crit=24.25%(540) penetration=0(0) haste=17.96%(589)
V3: Spell Stats: power=3193(3193) hit=14.11%(370) crit=24.24%(540) penetration=0(0) haste=20.83%(683)

V1: Core Stats: strength=68(10) agility=75(10) stamina=1340(1163) intellect=1307(1150) spirit=737(566) health=20780 mana=23760
V2: Core Stats: strength=68(10) agility=75(10) stamina=1322(1146) intellect=1296(1139) spirit=823(652) health=20595 mana=23590
V3: Core Stats: strength=68(10) agility=75(10) stamina=1322(1146) intellect=1296(1139) spirit=729(558) health=20595 mana=23590

V1: Warlock_T8_00_13_58 Sta=0.00 Int=0.37 Spi=1.06 SP=1.63 Hit=4.45 Crit=0.96 Haste=1.50 Lag=0.00
V2: Warlock_T8_00_13_58 Sta=0.00 Int=0.37 Spi=1.05 SP=1.63 Hit=4.46 Crit=1.00 Haste=1.10 Lag=0.00
V3: Warlock_T9_00_13_58 Sta=0.00 Int=0.37 Spi=1.04 SP=1.64 Hit=4.29 Crit=1.02 Haste=1.52 Lag=0.00

The 3 sets used, for anyones usage: