Theck's Simcraft Thread - WoD/6.x

Hi Theck,

for the DPS simulations, did you stay with SoL except for the explicitely seal-twisting XX_XX_EmpS_DPS variants ?

If so, I am wondering if taking e.g. SS_SW_Sera how much DPS can be squeezed out in comparison when SoT is used e.g. in an off-tanking situation when the SoL heals are not needed. This could be a fairer comparison for that case.

Extermi

PS: As for the dip when adding haste. in the _65 data files I found a situation where I dont understand why it delays CS:
1:31.295 crusader_strike
1:31.295 shield_of_the_righteous
1:33.199 holy_wrath
1:34.202 judgment
1:34.402 divine_protection
1:35.208 crusader_strike

It seems to wait doing nothing (probably no filler available?) for a second to wait for HW to become ready, and then still chains judgement although CS is virtually ready, pushing it back. At 3:44 it even waits nearly 3 seconds(!) for HW to be available, casts it shortly before CS is ready again and thus pushes it back.

To me it seems this behavior is what kills the HP generation - pushing back a short-CD HP generator to instead cast a long-CD one. This seems inherent to FCFS and probably just hits some odd breakpoints when modifying haste. In the end, 3s (CS), 4.5s (J) and 15s (HW) have multiple collisions over time, and each time they are favoring one of the longer CD ones due to a split second clash, this hurts survivability.

Which leads me to think that the "dip" in the haste curve is actually a better rotation we hit by luck, and could even be reached with higher (and maybe even lower) haste. Is there any way to program more than just FCFS, something like a decision based on a dynamic weight of time remaining until skill ready divided by length of CD ? At least for the skills that generate HP ?

I got a weird question. Yeah, mastery is our best stat, but we're also told to pick haste up to 50%... my question is, is there a sort of ratio that would work best? like for 40% mastery, you should have 20% haste or something like that?

Also, what are the chances they will revisit our attunement given that haste is near the bottom if not the bottom stat once you reac h lvl 100

Seems more likely that they'll revisit the fact that haste is underperforming, as per this Theck post from a few days ago:

theckhd wrote:I know they're aware that haste is underperforming, and that they're not really happy about that. I've also suggested a few tweaks that could help bring haste's value up. So I wouldn't be surprised if some changes get made, but it's unlikely they'll do so at L90. It'll probably be changes on beta that will only take effect closer to WoD release.

Extermi wrote:Hi Theck,

for the DPS simulations, did you stay with SoL except for the explicitely seal-twisting XX_XX_EmpS_DPS variants ?

If so, I am wondering if taking e.g. SS_SW_Sera how much DPS can be squeezed out in comparison when SoT is used e.g. in an off-tanking situation when the SoL heals are not needed. This could be a fairer comparison for that case.

Extermi

PS: As for the dip when adding haste. in the _65 data files I found a situation where I dont understand why it delays CS:
1:31.295 crusader_strike
1:31.295 shield_of_the_righteous
1:33.199 holy_wrath
1:34.202 judgment
1:34.402 divine_protection
1:35.208 crusader_strike

It seems to wait doing nothing (probably no filler available?) for a second to wait for HW to become ready, and then still chains judgement although CS is virtually ready, pushing it back. At 3:44 it even waits nearly 3 seconds(!) for HW to be available, casts it shortly before CS is ready again and thus pushes it back.

To me it seems this behavior is what kills the HP generation - pushing back a short-CD HP generator to instead cast a long-CD one. This seems inherent to FCFS and probably just hits some odd breakpoints when modifying haste. In the end, 3s (CS), 4.5s (J) and 15s (HW) have multiple collisions over time, and each time they are favoring one of the longer CD ones due to a split second clash, this hurts survivability.

Which leads me to think that the "dip" in the haste curve is actually a better rotation we hit by luck, and could even be reached with higher (and maybe even lower) haste. Is there any way to program more than just FCFS, something like a decision based on a dynamic weight of time remaining until skill ready divided by length of CD ? At least for the skills that generate HP ?

I'll take a look at this later today. However, you're not looking at the whole story there. You need to include what buffs were active.

For example, I'm going to guess that windsong_haste became active just after that first CS cast. Without that buff, this would have been a standard CS-X-X-CS-J-X (in other words, the last third of the cycle followed by the first third). There was nothing to cast in the first X, but then Holy Wrath became available for the second one, turning it into CS-X-HW-.

However, if windsong_haste procs after CS is cast, it won't reduce CS's cooldown (which has already started) but will reduce the GCD caused by Holy Wrath when it's cast. That creates a short (maybe 10-100 ms) gap between the GCD becoming available after HW is cast and CS coming off of cooldown. A gap that Judgment is happy to fill, pushing CS back.

It's not clear yet if that's what's causing the dip or not. Maybe it's happening all throughout the rotation, and that particular window of haste values somehow removes it from happening. I haven't had time to investigate it yet (maybe this weekend). The easy way to test for it is to repeat the haste plot sim with bloodlust turned off and a different weapon enchant, which is what I plan to do - if the dip is still there, then it's not one of those two factors.

Another thing that could fix that behavior is re-introducing the wait lines after CS and J. In other words,

Code: Select all

/wait,sec=cooldown.crusader_strike.remains,if=cooldown.crusader_strike.remains<0.3

Which would force the sim to wait up to 300ms for CS to come off of cooldown, and prevent it from pushing CS back in these sorts of cases. I took those lines out because with the merging of spell and melee haste, we shouldn't have a disparity between GCD times anymore. But it may still be necessary due to dynamic haste effects.

Klaudandus wrote:I got a weird question. Yeah, mastery is our best stat, but we're also told to pick haste up to 50%... my question is, is there a sort of ratio that would work best? like for 40% mastery, you should have 20% haste or something like that?

Not that I'm aware of. Haste is worth ~75% as much as mastery in a gear set with roughly equal amounts of each. There's obviously some synergy between them, so stacking mastery should have some effect on that. But I haven't yet seen haste>mastery in any sim results at L90. Should be easy enough to test for it though - take your own mastery-optimized character, or mine, and sim some stat weights.

Aaaaaand now I have to rescind that last statement:
https://www.dropbox.com/s/b9zhxbfnk451w ... .html?dl=0




So apparently there is some balance to be struck. This profile has the following stats:


I will run a reforge plot to see if I can find some expression for the optimum balance.

Oh, I was logged out with HA and EF taken. Odd. The reforge sims will clarify things though. Entirely possible this is another dip causing the anomalous stat weights, or that it's something about HA, EF or their interaction. Need dataz!

Well, I wasn't expecting Haste to beat Mastery, given Mastery's synergy with holy shield, was just curious if it was MASTERY ALL THE THINGS or a more balanced approached (So many of mastery for one of haste) was recommended.

Reforge plots suggest that this was something to do with HA/EF.

Profile with HA/EF:


Note that we add +70 haste to determine stat weights, so we're jumping from that plateau into the ditch, hence getting anomalously high haste:mastery ratio. Again, it looks like this is caused by these periodic dips, so I guess I need to make time to investigate those further.

That said, there is an overall downward slope (note how the plateaus between the dips keep getting lower), which suggests that at least for EF+HA, there's some gain to converting mastery back into haste if you have significantly more mastery.

If we switch to DP/SS:

Much flatter, and no apparent gain (at least, not a significant one). Again, there may be a small negative slope, but it's less clear.

Running a simple stat scaling plot reveals:


So it appears that once you have enough mastery, haste does catch up - at least to the point that it's essentially even with mastery, if not a little better. I'm running another reforge plot with a wider span to see what that tells us.

Interesting.

Nope. Definitely some balance point:
https://www.dropbox.com/s/hierefqqn7uof ... .html?dl=0


In this gear set I have ~55% mastery and ~29% haste. So we're looking at a 1.89:1 ratio of mastery to haste. Of course, there's no guarantee that this ratio is constant across gear sets; It could be a fixed ratio, or it could be that you always want about 25% more mastery than haste, or (more likely) it's a nonlinear relationship and the ratio just changes depending on how much you have.

I'll need to run more sims at specific gear thresholds to figure out exactly what that relationship is. Unfortunately, that's adding one thing to about 15 others that I need to do, so it may be some time until I get to it. If anyone else has some time and wants to run these sims (it's just a basic sim with a reforge plot of mastery and haste, Reforge Amount=500, Step Amount=25), that would help. Just report the haste% and mastery% that gives you close to the ideal balance. Ex: if you have 30% haste and 30% mastery (raid-buffed) and the minimum of the parabola is at -100 haste + 100 mastery, then that means the ideal ratio is somewhere around 30%-100/20=25% haste and 30%+100/23=34% mastery (roughly).

https://www.dropbox.com/s/o1ksw59dtiw1r ... .html?dl=0

Haste=29.75%
Mastery=42.18%

Will do now a different gearset.

I tried doing a diff gear set but the TMI graph came out like this!?

That seems incorrect, to say the least.

theckhd wrote:That seems incorrect, to say the least.

I have no idea what happened... here's the file if you wanna check it out. https://www.dropbox.com/s/msmn6a0diuih7 ... .html?dl=0

https://www.dropbox.com/s/z1jssb9zw3wxv ... .html?dl=0
Haste = 29.75%
Mastery = 45.36%

Ok, I got the new Simcraft version.


https://www.dropbox.com/s/36kj8ykhvu1w2 ... .html?dl=0
Haste = 29.73%
Mastery = 48.70%

Not that it matters for L100 (and 90 at this point is irrelevant), but I am finding that the Haste buff in raid is not a flat 5% that is increased by our attunement, but is first added as 100 Haste to the character sheet total and then given the 5% attunement, meaning lowered haste values on gearsets above 42.62% Haste unbuffed would make sense. More than 42.62% and you are raid buffed above 50% which would cause haste to plummet in value.

That may not change the weights much, but it could have a role. Its why Ive dropped Windsong, along with the anecdotal "Colossus feels much better at lowering my spike damage."

https://www.dropbox.com/s/49bsm1x1yblyf ... .html?dl=0
Haste = 36.71%
Mastery = 42.65&

https://www.dropbox.com/s/h17ihcb8l2bbt ... .html?dl=0
Haste = 33.25%
Mastery = 45.96%

Update on the haste dip issue: It's definitely due to unintentional rotation clashing, but not solely due to Bloodlust or Windsong:

Same profile as before, with Bloodlust turned off:
https://www.dropbox.com/s/u3s34ioeuqrj0 ... .html?dl=0


Same profile, with BL and Windsong turned off (Dancing Steel used instead):
https://www.dropbox.com/s/03k7y501f8dnk ... .html?dl=0


As you can see, we're still getting anomalous dips in that haste curve. However....

Same profile, with BL & Windsong turned off, and a "/wait,sec=cooldown.X.remains,if=cooldown.X.remains>0&cooldown.X.remains<0.35" added for both Crusader Strike and Judgment:
https://www.dropbox.com/s/03k7y501f8dnk ... .html?dl=0


The waits clearly fix the issue and smooth out the line, telling us that the problem was caused by accidental/unintentional HPG push-back. It's not yet clear whether it was J being cast early and pushing CS back or some other spell being cast early and pushing either of those two back. I'll run some more sims to suss that out. Right now I'm testing to see whether the problem re-appears when WS or BL are turned back on with the waits in place.

Klaudandus wrote:https://www.dropbox.com/s/h17ihcb8l2bbt ... .html?dl=0
Haste = 33.25%
Mastery = 45.96%

Ideal balance there looks to be ~ -125 haste +125 mastery. After attunement & raid buff are accounted for, that's ~26.36% haste, ~51.39% mastery.

Klaudandus wrote:https://www.dropbox.com/s/49bsm1x1yblyf ... .html?dl=0
Haste = 36.71%
Mastery = 42.65&

-150 haste (about), or -8.27% haste. Net is then ~28.44% haste, 49.17% mast

Klaudandus wrote:Ok, I got the new Simcraft version.
https://www.dropbox.com/s/36kj8ykhvu1w2 ... .html?dl=0
Haste = 29.73%
Mastery = 48.70%

-50 haste, or -2.76% haste. Net is then ~26.97% haste, ~48.92% mast

Earlier plots tell more or less the same story. We're always looking at a little less than a 2:1 ratio of mastery:haste. It would be interesting to see if that holds for a significantly different gear set (i.e. much lower ilvl, or just take off 4-5 pieces of gear). That said, I'm not sure it's much of a practical concern. Even in these gear sets, we're basically at the point of "stack as much mastery as you can, and then you might barely reach the point where haste catches up."

In other news, the waits appear to successfully remove the dips in the haste plots:
https://www.dropbox.com/s/jvpf0xdnfd5z5 ... .html?dl=0


https://www.dropbox.com/s/o84x09kxk0in0 ... .html?dl=0


So I'll be adding them to the default profile shortly. I do want to see whether it's CS or J causing it first, though. It would also be nice to actually understand what was going on, in case (as Extermi mentioned a few pages ago) there's a slightly more ideal rotation. Or, as may be the case, if those dips are the ideal rotation where we aren't accidentally pushing back CS/J, and this effect was happening more often outside of the dips.

So basically MASTERY ALL THE THINGS.