Theck's Simcraft Thread - WoD/6.x

Also note that the conditional is "time_to_hpg", not "time_to_next_hpg". I decided to shorten it. Will be available as of tonight's nightly build.

theckhd wrote:

Thels wrote:On a sidenote, SS is the clear winner in WoD it seems, but do you happen to know the state of 6.0 SoO? Considering 4p T16 and end-expansion gear, the values might be different. I guess I can easily sim them myself if you're not aware of this of the top of your head. Is the current build representative?

The current build doesn't have all of the optimizations I've done this weekend. I'll probably commit those changes within the next hour or so now that I have the time_to_hpg conditional working.

I don't have an answer for you yet. My guess is that the T16 4pc makes EF at least competitive, if not stronger. I plan on doing another talent sim today using my imported L60 tank just to see how that all shakes out (probably ignoring SH though, for time).

Answers. This is using my currently-equipped gear (check my armory), though obviously converted to post-6.0 stats. The boss is the T17M boss, scaled down to level 93, so it might be a little on the hard-hitting side. I'll re-run this with the T16H25 boss next.
https://www.dropbox.com/s/pulxek7tu3536 ... .html?dl=0


Seems like EF is still lagging a bit at L90, even with the set bonus. That said, the interaction with DP (and lack of GCD cost) gives higher DPS. Note that I'm not using GoWoG here either, which would increase the EF DPS values a little more.

Awesome, I'm sorry you had to modify the sim to get that in there, but thanks. I'll probably have a chance to try it tonight or tomorrow, but your results look good, and I agree it's splitting hairs at this point.

I was also going to suggest, as you concluded, that we do separate max dps sims to stop the 45/75/100 from getting ridiculously long. If you get to it before me today, I was getting a minor dps bump in short sims (SS_SW_Sera) with HW>all during Seraphim, even more than just HW>all constantly.

One last thought towards the haste breakpoints posted further up: might we expect haste breakpoints with Seraphim due to getting an extra GCD within the Seraphim buff? I didn't have a chance to see at what haste values it was occurring, but there should be discrete points where we get one more ability of during Seraphim and that might explain that behavior.

L90 against a L93 T16H25 TMI boss:
https://www.dropbox.com/s/66iqegekd38kj ... .html?dl=0


Appears that EF is still behind, though not by a whole lot. And a minor DPS boost.

Updated talent sim, with corrected labels and using the new conditionals:
https://www.dropbox.com/s/ctwwgeji5r7ro ... .html?dl=0

Fundus wrote:I was also going to suggest, as you concluded, that we do separate max dps sims to stop the 45/75/100 from getting ridiculously long. If you get to it before me today, I was getting a minor dps bump in short sims (SS_SW_Sera) with HW>all during Seraphim, even more than just HW>all constantly.

You mean "holy_wrath,if=talent.sanctified_wrath.enabled&buff.seraphim.react" at the top of the queue? Was there any TMI cost to that? If you could throw the html report up somewhere (dropbox, box.com, whatever) so I can look at it, I'd appreciate it.

Fundus wrote:One last thought towards the haste breakpoints posted further up: might we expect haste breakpoints with Seraphim due to getting an extra GCD within the Seraphim buff? I didn't have a chance to see at what haste values it was occurring, but there should be discrete points where we get one more ability of during Seraphim and that might explain that behavior.

Yes, we probably can. In 15 seconds, we get N GCDs at H haste given by:

Code: Select all

N=15/(1.5/(1+H)) = (1+H)*15/1.5 = 10*(1+H)

So the breakpoints should be every 10% haste, i.e. 10%, 20%, 30%, 40%, 50% (at which point we hit GCD cap).

Sorry for the delay, just got home. Yes, I added "holy_wrath,if=talent.sanctified_wrath.enabled&buff.seraphim.react" to the top of the que, and it's doing just a touch better in DPS (~60 DPS) than the HW>All APL or default APL using truth I tried. It's also slightly better on TMI (only 0.4k) vs HW>All and slightly worse on TMI than the default APL using truth (0.2k). Again with the splitting hairs, sorry.

I just re-ran it vs the default APL, and it has a slightly reduced TMI (0.9k) vs the default que using insight for a little less than a 300 dps gain. Probably not worth it during actual tanking or survival phases, but possible a minor increase for tank-swap-type situations.

Full Results
Profile

I was going to try taking out some of the higher priority SS refreshes as well, working towards an off-tanking max dps que. Figured with the quick fall-off of resolve, it might not be worth much to be casting/refreshing SS on the other tank. I didn't see any other questions that needed addressing or that weren't already being worked on, though I still find it odd that SS_DP_Sera is beating SS_HA_Sera by almost 4k TMI, while using HolyS and EmpS_Surv it's at most a 1.5k TMI difference between DP and HA. I would have expected Seraphim to be a very poor combo with DP, so I'll have to look at that report in more detail in a bit.

Someone (Thels?) requested a sim where ES and HPr were self-cast rather than target-cast. Here are the results, after much agonizing and adding a "target=self" option to SimC. This also includes a slight buff to LH (it was only giving 7 ticks rather than the full 8... or occasionally 9).

https://www.dropbox.com/s/hrb7xdr76bk0d ... .html?dl=0

Fundus wrote:Sorry for the delay, just got home. Yes, I added "holy_wrath,if=talent.sanctified_wrath.enabled&buff.seraphim.react" to the top of the que, and it's doing just a touch better in DPS (~60 DPS) than the HW>All APL or default APL using truth I tried. It's also slightly better on TMI (only 0.4k) vs HW>All and slightly worse on TMI than the default APL using truth (0.2k). Again with the splitting hairs, sorry.

I just re-ran it vs the default APL, and it has a slightly reduced TMI (0.9k) vs the default que using insight for a little less than a 300 dps gain. Probably not worth it during actual tanking or survival phases, but possible a minor increase for tank-swap-type situations.

Full Results
Profile

Yeah, I don't think it's worth trading 900 TMI for only 300 DPS - generally you want a DPS to TMI trade to be heavily biased towards DPS to be worth it (i.e. 2x to 3x the TMI loss).

That said, the "max_dps" queue in the default profile is basically what you're looking at optimizing, so feel free to tear it apart and clean it up. In that one, you can more or less ignore TMI and just focus on what order gives the best DPS. I've made some guesses based on earlier sim results, but done nothing conclusive to test it.

Also I just noticed from the "Seraphim as a defensive CD" thread:

Meyrinn wrote:There is also an oddity that Parry does not go up with the 1000 Crit Strike Rating from Seraphim.

I jumped on beta and confirmed that my character sheet parry doesn't change with Seraphim active. I hadn't noticed that before. Do we know if it's a bug or intended and I just missed it? I bug reported it just in case. Which way is it working in the sim atm?

Slightly offtopic Theck, but were there some major optimizations passed in SimC recently? Im noticing appx 20C decrease in load temps while simming and Im trying to isolate any possible reasons why (just had to up voltage to stabilize my OC)

Back OT though, I still get Bonus Armor coming out extremely low on my Stat weights, I thought SimC defaulted to scale over TMI for Tanks, but having it so low seems to be counter the stat intent.

theckhd wrote:L90 against a L93 T16H25 TMI boss:
https://www.dropbox.com/s/66iqegekd38kj ... .html?dl=0
http://gyazo.com/a02d6c90b3eb92b2ee4e15b05725cc53.png

Appears that EF is still behind, though not by a whole lot. And a minor DPS boost.

Hmm, so they're pretty close in both TMI and DPS... I guess I'm switching to SS when 6.0 hits, though there doesn't seem to be a wrong choice of the two for those few weeks in between.

theckhd wrote:Someone (Thels?) requested a sim where ES and HPr were self-cast rather than target-cast. Here are the results, after much agonizing and adding a "target=self" option to SimC. This also includes a slight buff to LH (it was only giving 7 ticks rather than the full 8... or occasionally 9).

https://www.dropbox.com/s/hrb7xdr76bk0d ... .html?dl=0

Interesting... HPr_sc is less HPS than HPr? Does HPS include healing others? I'm a little surprised that ES_sc is doing so much better TMI wise, but also doing so much worse DPS wise. I wasn't aware it was that much of a difference. If you pick ES and EmpSeals, you got quite some leeway in switching between offense and defense mode in the middle of a fight.

Fundus wrote:Also I just noticed from the "Seraphim as a defensive CD" thread:

Meyrinn wrote:There is also an oddity that Parry does not go up with the 1000 Crit Strike Rating from Seraphim.

I jumped on beta and confirmed that my character sheet parry doesn't change with Seraphim active. I hadn't noticed that before. Do we know if it's a bug or intended and I just missed it? I bug reported it just in case. Which way is it working in the sim atm?

Almost certainly a bug, and probably related to the elixir/flask bug (crit elixirs/flasks weren't giving parry either). Will report.

xstratax wrote:Slightly offtopic Theck, but were there some major optimizations passed in SimC recently? Im noticing appx 20C decrease in load temps while simming and Im trying to isolate any possible reasons why (just had to up voltage to stabilize my OC)

Not that I know of, but it's possible. Are you sure you didn't just reduce the number of threads by accident?

xstratax wrote:Back OT though, I still get Bonus Armor coming out extremely low on my Stat weights, I thought SimC defaulted to scale over TMI for Tanks, but having it so low seems to be counter the stat intent.

I'm not sure what's happening in your sims then, Bonus Armor has been consistently dominating in mine. SimC doesn't default to TMI stat weights for tanks, you need to set that yourself via the GUI or the CLI command scale_over=tmi

Thels wrote:Interesting... HPr_sc is less HPS than HPr? Does HPS include healing others? I'm a little surprised that ES_sc is doing so much better TMI wise, but also doing so much worse DPS wise. I wasn't aware it was that much of a difference. If you pick ES and EmpSeals, you got quite some leeway in switching between offense and defense mode in the middle of a fight.

Er.... yes, the HPS+APS plot there would be counting healing done on others. You can't trust that plot for LH or HPr as a result. The TMI plot should be accurate, however, as that metric ignores external healing.

If you want to know the pure HPS of HPr when cast on a target, you'll have to sim that actor individually rather than in a raid sim.

Something is wrong with the ES self-cast. It's getting 50 casts with ~86% uptime. ES normal is getting 7 casts with ~16% uptime.

Would be nice to see how EmpSeals sims without full weaving, and even with no weaving to see the actual gains from weaving. Maybe one for each pair of seals, and one for each seal alone.

Theck ran a bunch of sims with various combos of seals during EmpS, including just SoT, SoR, and SoI. It's back here, which is on page 6 on mine. The terminology might look confusing if you're jumping in now and didn't read the original EmpS posts starting here. I'll try to briefly explain the terminology, though I guess some of this shorthand will probably need to be in the glossary at some point.

So looking at the most recent plot:

theckhd wrote:

The combos of capital letters that don't correspond to any of our regular ability abbreviations represent which seals we're twisting, arranged in order of priority. For example, in middle of the top ranking DPS que we see "TRd" which means that we're twisting SoT and SoR. You may notice that the second highest DPS que has "RTd" in the middle which means we're using the same seals, but reversing the priority. The "d" appended to the seals in each of those two examples means we're using the "down" logic, which is just refreshing each buff after it falls off. There are also some followed by a "u" (for example the 2nd lowest TMI que), which means we're using the "up" logic that tries to keep all of the buffs up by refreshing before they fall off. Finally, some of the seal combos are also followed by an "e" (for example at the end of the top two dps ques) which means we're switching seals during an empty GCD.

The rest of the lines are showing where in the priority list (APL) the seal swaps are being inserted. So using the top DPS que again as an example, "(CS-AS)" should mean everything on the default APL between CS and AS (for example judgment and HW with SW talented, etc) is still higher in priority than swapping SoT and SoR when their buffs are falling off. The second portion "(HoW-TRe)" means everything from HoW and lower (for example Cons and HW without SW talented, etc) is a lower priority and will get pushed back for seal swapping. Recall the "TRe" there means we're also putting a seal swap on the end of the APL (lowest priority) to allow swapping SoT and SoR during empty GDCs. So completely writing out the top DPS que would be something like: CS>J>HW+SW>AS+>FW>SS2>AS>TRd>L90>HoW>SS8>Cons>HW>TRe. Phew, hopefully I didn't leave out any abilities there.

And as you requested, Theck also left in the single seal configurations, labelled simply as "seal_of_" insight/righteousness/truth, so you can see how those stack up as well. If you go back to the original EmpS posts I linked to above, there were even more variations that Theck tested, just in case you were wondering where this list of combos originated. Anyhow, hope that helps a bit, but if not perhaps someone else has a more succinct explanation.

marsbubble wrote:Something is wrong with the ES self-cast. It's getting 50 casts with ~86% uptime. ES normal is getting 7 casts with ~16% uptime.

Hm... you're right. Good catch, I hadn't looked at it that closely. My guess is that it's not properly triggering the cooldown since it's technically stay_of_execution rather than execution_sentence. I'll look into it.

I tried looking around in the cd parts of the code and the stay of execution part as well, but I can't see anything obvious :/

Oh, it's pretty obvious what was missing if you know what you're looking for. For example, just look at hammer_of_the_righteous_t to see how I connected it to the cooldown of crusader_strike.

I've fixed it in my local copy, I'll push it to the repository some time later today.

Updated talent comparison, now with Selfless Healer removed, but with stat weights:

https://www.dropbox.com/s/ctwwgeji5r7ro ... .html?dl=0




To give you some idea what stat weights look like, here are some representatives:

SS_SW_EmpS:


SS_SW_HolyS:


SS_SW_Sera:


EF_SW_HolyS:


SS_HA_HolyS:


SS_DP_HolyS:


EF_DP_Sera:


SS_DP_Sera:


You may note that haste is consistently near the bottom of these. So much for our "attunement" stat, but I think that the entire attunement system was a bit of a fool's errand to begin with. Trying to dictate which stat will be best in a system with this many moving parts is tough even with a relatively simple metric like DPS. Stack enough of your attunement stat - say haste - and crit and mastery will eclipse it just because of the multiplicative nature. And if it was unlikely to be at all accurate with a simple metric like DPS, trying to do it for tanks (who may alternately care about DTPS, DPS, or TMI) was pretty much bound to fail.

/rant off

So... back to the data. BonusArmor is consistently the best thing since sliced bread, no real surprise there. Stamina's value will really depend on the boss and your current HP. In this case, we're in T17H gear against a T17M boss, so stamina is strong because each melee is a significant portion of our health.

Mastery and Versatility tend to vie for the next spot. Mastery tends to be stronger with Holy Shield selected for obvious reasons. Versatility still needs a severe nerf in the survivability department, in my opinion. I would rather see stats like crit or multistrike or haste be preferred over the boring "we couldn't come up with another interesting secondary" stat. But, that's not how the numbers work out. I think it's probably safe to rank Versatility behind Mastery in general. It's only when Seraphim is taken that it bumps up to roughly equal Mastery, and even then it's not consistently ahead.

Crit tends to occupy the next spot, though it's often jockeying with Versatility (when Vers is lagging Mastery by a good bit).

Haste and Multistrike bring up the rear, generally with Multistrike leading. That said, there's some ambiguity about how to value Multistrike. Part of Multistrike's allure is that it increases the healing done to you by other healers through Shining Protector. However, TMI ignores external healing - and even if it didn't, there's no external healer in these sims that's bombing you with heals. So one might assume that the Multistrike valuation in these sims is artificially low.

I actually don't think it is that far off, though - at least in a practical sense. Multistrike's valuation for HPS will be low, for sure. In a real situation, it will produce a bunch of extra healing through Shining Protector. But HPS isn't a spike metric, and it's not entirely clear that we even care about raw HPS. It really depends on encounter mechanics - if encounters are tuned such that your death happens over 10-20 seconds due to lack of throughput, then sure, we may care about DTPS and HPS.

I suspect that isn't going to be the case though, because I think that if tanks aren't in danger of dying in 6-7 second periods, tanking is going to be awfully boring, and I'm pretty sure Blizzard knows that. They definitely wanted to get away from the "bosses are killing a tank in two globals" situation we're running into in the end of Heroic SoO, but the triage scale is a continuum. There's plenty of room between "dead in 2 seconds" and "dead in 20 seconds." And I'm pretty sure the sweet spot they're aiming for is exactly the same sweet spot that TMI uses - around 6 seconds.

Given that assumption, there are two major reasons why Multistrike's TMI valuation in these sims is probably accurate. The first is that, on 6-second time scales a tank with an active, continuously-spamming healer isn't going to die. The only time they'll be threatened is a 6-second window where they take an abnormally high spike compared to healing throughput. That means either the boss hit them with a really big ability, or, more likely, their healer fell asleep, died, or what have you, and stopped healing them. And if you're not getting any (or much) incoming healing, then most of that Shining Protector value is lost. What's left is exactly what we're measuring in this sim - the value due to your own self-heals multistriking and proccing Shining Protector.

The other reason is one of mechanics. If someone tops you off with a giant heal, and Shining Protector procs, that proc is useless. We've been promised this won't happen in the triage model, but... call me skeptical. Healers are going to want to keep their tanks nearly topped off if there's any danger of a spike - even a spike that lasts 6 seconds. Taking 70% of your life in spike damage isn't dangerous when you're at 100%, but it is deadly if you're sitting at 65%. So I think we're still going to see a lot of overhealing on tanks, and Shining Protector is no exception. I think it'll be a great throughput effect that helps deal with average DTPS, and might lessen the load on healers, but I'm not convinced it's an effective anti-spike mechanism.

All of that considered, I do think that it's worth bumping multistrike up a little bit - say to a competitive level with crit. I just think that the argument that it's going to be our OMG BEST STAT EVER is making some naive assumptions about how healing and tanking will work. Maybe it will turn out that way, but I sincerely doubt it.

So, my rough valuation of stats based on this data is:

Bonus Armor >> Mastery > Versatility > Crit / Multistrike > Haste

I've left off things like Stamina, Armor, Strength, and AP since we don't generally get to choose between those and secondary stats. However, I think it's fair to say that Stamina and armor are going to be a better choice than Strength. Slot Armor and Strength in-between Bonus Armor and Mastery, and Stamina wherever you like. I feel it would be above STR/Armor, because I think 20-man Mythic is going to be a lot more like 25H than 10H. But again, it will depend a lot on exactly how tank healing and triage play out (example: having a higher amount of stam is meaningless if you're just hovering between 25% and 75% life all the time, with no danger of spikes).

Also, corrected L90 talent comparison, with a working Stay of Execution cooldown:
https://www.dropbox.com/s/hrb7xdr76bk0d ... .html?dl=0

So LH is a clear winner over ES_sc, it seems. ES_sc still heals more than LH, right? I guess LH's smoother healing makes it better for TMI than ES_sc.

Those stat weights don't make sense to me (which, to be clear, is almost certainly because I am missing something, rather than their being incorrect).

Isn't WoD crit functionally identical to MoP parry (for TMI purposes)? If it's significantly more valuable than haste, shouldn't that mean that right now, on live, prioritising parry over haste would result in a lower TMI value? Reasons this might not be true, off the top of my head:

- The level 100 talents benefit more from crit/parry rather than haste (seems unlikely).
- The move from Vengeance to Resolve devalues haste somehow.
- The strength of live EF is responsible for a large chunk of haste's current value.
- The "tank squish" increases the relative value of avoidance (plausible given that SotR will be substantially weaker, but given the strength of mastery, we should expect haste's value to improve as iLevel increases (due to mastery stacking)?)
- Some combination of the above.

It just intuitively seems weird that haste goes from being one of the best TMI stats to pretty much the worst, especially given it will receive a 5% boost from our spec affinity.

Keep in mind that Crit not only converts into Parry, but also remains as Crit itself, increasing (afaik) the effects of both SoI and SS.

AND shining protector :P