#tc-research

1 messages · Page 44 of 1

knotty oriole
#

(and I doubt that's super major but just worth noting)

plucky mural
#

Are people running mongo crit again? Was always fun above 50%. Cps for days.

noble pendant
#

(Not really a convo for this channel though)

narrow crescent
#

I was told to come ask this here. Is there any research that's been done on whether doomblade's extra 5% per bleed is affected by a zandalari using Kimbul as their loa for an extra bleed?

noble pendant
narrow crescent
#

Ty, no more gimmick shenanigans for me

knotty oriole
#

So, somewhat new findings about the AA ring. It appears after some more controlled log testing I was able to get someone to do, the ring bonus doesn't quite work like normal bonus damage and is reduced by the off-hand damage modifier (applied before the 50% penalty.) This will reduce the value of the ring effect by ~25%. I'm hoping to get it into sims tomorrow now that I have confirmed the behavior.

knotty oriole
#

Court of Stars Ring should have an adjustment in now for the OH issue. Will be in the nightly in a few hours.

lavish zenith
#

just wanted to make sure my thinking was right on this, but w the way vendetta 4pc is working, the pandemic timer would also change within the vendetta window correct? ie normally it's fine to re garrote when there's 5s left on garrote cuz the window is 5.4. but during vendetta it's actually 2.7 cuz 9x0.3. so if u re garroted w 5s left in vendetta it's a misplay. was thinking about updating my wa's to reflect that so it doesn't tell me to re bleed until it gets within the reduced window. also idk if this is in the new apl so why i put it here

knotty oriole
#

I've made some APL changes to account for the re-exsanguination logic (added a new !will_lose_exsanguinate expression to check for cases where Vendetta allows re-upping at the same rate)

#

But need to double check the refreshable stuff is working right

#

I think in SimC it won't really need any modifications the way I've implemented it since the refreshable core expression just checks remains() <= triggered_duration * 0.3

#

and I'm modifying the duration at trigger time

#

Unless it works incorrectly in-game (which is possible)

#

(Initially I tried modifying it after cast rather than on application but I thought this might be a problem so moved it to the actual snapshot state creation so expressions work right)

#

But it wouldn't totally surprised me if this was wonky in-game either... since this case has never come up before

lavish zenith
#

okok good good just wanted to make sure and put it out there

knotty oriole
#

But yeah, tl;dr is that SimC should be OK how I did it in theory. Provided it works intuitively as you mention (e.g. assuming pandemic is half normal duration)

#

But definitely a chance the game is buggy and needs to be confirmed

#

Probably need to test both cases of like:
Non-Exsang Duration -> Vendetta -> Refresh
Vendetta -> Apply -> Refresh
with logs to make sure everything lines up as expected

spiral light
#

After testing it out day one, I can confirm that it does in fact halve the pandemic timer as well.

spiral light
#

Not home currently so don't have access to my logs unfortunately.

knotty oriole
#

All is well with the world then!

lavish zenith
knotty oriole
# lavish zenith https://wago.io/Yya4XBbl- anyone know if this wa is legit? like dslice but more...

It is.. kinda. It's still a WIP system on the SimC side. It's relatively new and was created by some other specs that didn't feel like DS was appropriate. But I don't really know if it's really ready for general/primetime consumption yet. I haven't done a lot of testing of it yet until they tell me it's "finished" but they have still been working on it as of the last month or so, so I think it's still in development.

#

My gut feeling is that this won't actually yield significantly different results than DS, but it's an interesting concept and I'll be interested to prove out what differences it does yield (and why) when it's more correct.

#

I'm not sure it works correctly with re-stealthing for Rogues though, which I'll probably have to muddle in and fix at some point.

#

(Jayeasy is a SimC dev for Hunters, FWIW. So the source of the WA is legit. It just may not be ready for general use yet.)

#

The real issue, for me, honestly, is that DS concerns are often about CD alignment and raid event handling in the APL, but at least in the setup of DS the "boss" segment goes first, which ensures CDs are at least available as you'd expect. This seems far less predictable with a "real" route and while I can understand why the accuracy has some appeal my feeling is that this actually will end up having a harder time approximating realistic cooldown usage.

In general, I think there is an over-emphasis on the wrong type of "accuracy" when using sims, since replicating a dungeon pull-by-pull seems intuitively better. But there are aspects of it that are likely worse for the purposes of approximation/generalization and you run into "local optimization" problems where something just happens to sim better because of CD alignment/timing with a specific large pack in the configuration. (This is why DS randomizes overlaps)

So I'm concerned that these profiles may end up giving less useful results than a generalized model. But it's still something I'm interested in investigating when it's ready.

#

It's really tricky to approach something like this and avoid overfit/local optimization issues with the particular static model.

Of note, our goal with sims is not so much to put out an "accurate" DPS number, but to generate appropriate correlations with stats/talents/gear that map to "better" configurations. There's arguments to be made that DS could be improved, but there's also arguments that this type of "pull-accurate" model could result in better or worse correlations. It's not a slam dunk just because it's pull-accurate. So it just needs testing/investigation really.

lavish zenith
#

ah okokok thanks for the messages. i was just curious cuz someone sent it to me and asked my opinion on it but i'd never even heard of it before. look forward to seeing how the wa gets updated over time

noble pendant
#

Koji made a great explanation of the issue behind some of these models, but I wanted to add an anecdote regarding it as well.

I’ve spent a lot of hours on assassinations dungeonslice APL this expansion, trying to optimize for bleeds and when to refresh them in larger target counts, with or without legendaries, etc. (the question of CT spamming in aoe still comes up every now and then)

In order to ensure changes made aren’t figments of timings and specific fight models; I tested every gain I found in hectic add cleave, as well as 8,5,3, and 1 target patchwerk, for both short times, and long times. Now that might be overkill, but I wanted to be sure what I was doing was actually a certified gain to the specs simulations, and not just a “dungeonslice gain” or a “3 target gain” or whatever it may be. It basically multiplied all my sims by a factor of 5, but local gains of that type in simulations is a real issue that we encounter with every optimization.

Half my gains in dungeonslice never really manifested because they were considerable losses in other fight types that I could not work around.

knotty oriole
#

Yeah, this is definitely a good point. Similarly, sometimes there are "losses" in DS that we still put in (say a 0.3% loss) just because we prove out that it's specific to a local timing issue with DS and is intuitively "correct" and a gain in HAC or sustained cleave.

knotty oriole
#

Rogues are somewhat fortunate that our mechanics really don't care about these specifics too much though. I think we can pretty much whip out any composite fight style and get the same gear/talent recommendations because most of them are pretty cut-and-dry.

As an example (just to do some investigation on this) I plopped in the current route for one of this week's MDT routes posted on r.io and ran it in Top Gear with my profile:
https://www.raidbots.com/simbot/report/quJhAzAs7cd7AnUMH5JJKf

Compared to Top Gear with my profile on DungeonSlice:
https://www.raidbots.com/simbot/report/baqVQktUd7xtSvW6CLEkF6

So notable observations:

  1. DungeonRoute sim took significantly longer to run. 2-3 times as long as DungeonSlice. (As expected, it's a longer and more complex sim.)
  2. I'm still trying to figure out why the DungeonRoute sim prefers Resounding Clarity over Finality, since this is exactly the opposite of what I expected given the increased target count. (DS prefers Finality by 0.7%, DR seems to prefer Resounding by 1.5%..) It appears at a glance this is because DungeonRoute doesn't have timed gaps between the pulls and thus there is less buff ER waste between pulls. DR sim appears to have 0 Expiry of any ER buff, while DS expires 2.6 out of 34.4 buffs--6.6% of the total, which seems more realistic. Need to chat with Jayeasy about gaps between pulls I think.
  3. Talents and Conduits suggested are the same, pretty much as expected. (And Conduits suggested are in the same order, Deeper Dagger/Staccato/Execution > Deeper Dagger/PV/Execution in both sims)
grave wharf
#

If so many specs lack an APL capable of handling DS well, I have a hard time believing they will be solved by simming arbitrary fight styles

knotty oriole
#

I'll take a look with other specs also, but at least for my Sub profile there's really no notable difference here other than the Resounding thing (which is explained by the fight gaps I think) in terms of recommendations but the sim takes 3x a long to run.

noble pendant
#

Which is uh, less than helpful

#

There’s also concerns about the whole “hold it for the next pack” argument

knotty oriole
#

Also the damage breakdowns here are remarkably similar other than the minor shift in BP in DR (as you'd probably expect, I guess) and most of that is explained by the better use of Resounding

grave wharf
#

Yeah with this approach the local maxima/overlap issue is a huge problem. Raising your dps so you 'advance' to the next pack faster and don't have a CD available means your total dps probably goes down unless your APL is very smart about timing predictions

knotty oriole
#

So basically it's a lot of additional simming to yield roughly the same results

noble pendant
#

Which is basically as difficult as shooting darts at a board, since the time you would hold depends on the size of the neck pack as well as the length remaining and if that matches up with any lost casts

#

And spec APL’s in General are nowhere near that complexity

knotty oriole
#

Yeah I mean the best most APLs can realistically do in most cases is like "if ST, hold CD for AoE if it is within X seconds"

#

Which is largely the model in the Rogue APL

noble pendant
#

Yeah

knotty oriole
#

You could potentially do "if next AoE event is significantly larger than current"

#

But that gets into a rabbit hole really fast

#

(e.g. what happens if the next next one is bigger than the next one)

#

(or when the next next next one is bigger than the next next one still)

noble pendant
#

The tldr here is the usual “it doesn’t matter how accurate your model is if your tools aren’t as accurate”

#

And frankly they don’t really need to be that accurate

#

Since dungeonslice seems to suggest the same stuff

grave wharf
#

Right, this probably only matters if you find a situation where there is some uniquely degenerate route that somehow doesn't match DS

knotty oriole
#

And let's be honest, WoW is pretty much generally balanced on the WoW-dev-side as like "AoE is 3T+" which means realistically if the class designers do their job the difference between averaging 8T vs. averaging 5T vs. 6T or whatever

#

Should really not matter

#

Other than cases where certain abilities just suck

noble pendant
#

Yeh

#

Echoing in aoe

knotty oriole
#

I do think the DungeonRoute stuff has some potential because I like the way it's set up from a raid events perspective

#

But I actually think the better version of DungeonRoute is actually just a data-driven generalization similar in focus to DungeonSlice, just based more on route pull sizes

#

Where it's some weighted average between route segments across various dungeons

#

Instead of the full dungeon

grave wharf
#

there's an argument to be made that it's trying to be realistic but not being realistic enough since in actual reality you have a full party and your party's dps is not necessarily just a constant factor of your own

#

other people use CDs when yours are down, etc

knotty oriole
#

But ultimately for melee specs I don't really think there's a massive difference between the two, and the results are likely to be really similar

#

Still worth investigation and testing. I'm always for trying to take advantage of better models/tools, just think the argument that this one is "more accurate" just because it is pull-accurate was something I was trying to explain as being slightly misleading.

grave wharf
#

Certainly. Just if you have a venthyr moonkin in your group good luck incorporating that factor into your pull-accurate sim

knotty oriole
#

Just like in math there are definitely more complex ways to do things that are possible to approximate. And technically they are more accurate, just not more accurate enough to matter. So everyone uses the approximation.

#

(Or insert debate about coders who put a const for PI to 18 decimal places when they are doing math where two decimal places is sufficient to yield the same result lol)

#

If anything, I'm actually interested to see what actually changes between the two, such as the Resounding Clarity issue. Since that creates something worth investigating. But unfortunately I think that one is an artifact of the fight style configuration.

#

(As I did confirm there appears to be no delay between the packs by default. I think this can be put into the configuration though.)

#

Probably would need some reasonable output from the addon for proximity (likely chain-pulls) vs. longer-distance packs or something

#

But yeah TL;DR--it's fine but not more useful for us specifically quite yet. Probably gives exactly the same results and takes much longer to sim. So I wouldn't "recommend" it over DS for Rogue for now. Might eventually change as it gets refined, but for now DS is still what I'd recommend for personal simming just because it's faster and a bit more tested at this point.

civic mango
#

The project @lean talon and I were working on specifically avoided a lot of these problems by having an interface option in the sim setup to tell the profile exactly which packs/sections of the combat to hold/use CDs on from the start, and would ideally display the results of varying combinations of holding and sending in a top gear style format to know which strategy was optimal. As far as I'm aware jayeasy's model also pulls mob hp straight from mdt and the key level, which runs into issues like viper described where your team mates can be absolutely nuking the pack one pull, and doing tank damage the next, which it won't be able to account for, and could very easily give incorrect results because of that. Me and Vaughn were planning on the user also inputting the combat duration for each section ahead of time to solve that, and then using the standard dslice variance on combat duration to avoid weirdness

warm zinc
#

Could you, theoretically, give a sim the cast/enemies hit by each cast from a log for example and then it simulates your top gear specificly for those casts?
Obviously this would ignore haste as a stat but you could porpoisefully ignore the haste item slots from top gear (which is pretty much only kt shoulders and sylv waist anyway).
I'm thinking about sub specificly.

noble pendant
#

You’d basically be taking out the advantage of doing a run thousands of times by feeding it the same inputs every time, while also not utilizing the apl at all

#

you'd have to rewrite a lot of it to make it work as well

#

and I'm not sure I see the benefit of something like that

warm zinc
#

The benefit i have in mind is, for example if you are practicing a specific dungeon level with a specific route multiple times you could get the optimal gear setup for that specific route/key level

#

It is very situational and not at all useful in a general sense

regal agate
#

Its probably better to just use different fight styles and piece together information from those for estimates.
Having a estimate/sim for every specific encounter/dungeon/etc is just a hell to optimize and likely won't bring you a lot of advantages over a more generalized approach

noble pendant
#

Yeah but I mean even with the same route and the same dungeon on the same key and the same group you’re still getting a bunch of variance

#

Just in kill times alone

#

So then you’re like okay take a dungeon, and then add variance from pull to pull

#

Which is basically just dungeonslice again

#

So the only real difference is that it takes a ton longer cause you’re simming a whole dungeon versus a chunk

#

But like you can increase the pack sizes or decrease the pack sizes in dungeonslice and it really doesn’t change any evaluation

warm zinc
#

Yeah, you are both correct, this was just a curiosity of mine.
Thanks for taking the time to reply.

noble pendant
#

The difference between say halls, mists and spires for example is actually not really all that impressive

warm zinc
#

The only thing that bothers me a bit about DS is how high it puts mastery on the stat priority while in reality you'd much rather be crit/vers soft capped

regal agate
#

you can put in priority_rotation=1 if you want a more funnel damage oriented stat weight

noble pendant
#

I’ve had this talk with koji a couple times if it makes you feel better and he basically said something along the lines of “if you think dungeonslice isn’t accurate to a dungeon, try changing the pull counts” and then I added 2 targets to every add pull and it basically did nothing except change my ring enchants and 2 gems

#

And I was like aight

#

Cool

#

This was a waste of my time

warm zinc
#

Out of curiosity what stat did it change it to, if you remember?

warm zinc
noble pendant
#

This was a while ago, but it changed from crit to mastery

west prism
#

https://www.raidbots.com/simbot/report/nyxk8Srcbaq6SUkTUPendp/simc

How is it possible that the sim is missing 2-3 anima cp buffs per value in a regular patchwerk sim? It's one thing coding the bot to be stupid with backstabs and missing CPs, another one missing so many buffs. It's highly unrealistic for an avg sim no? I might have misread the table or missed something, but it just seems like its flat out failing, which is kinda weird since the bot is meant to present the best case scenario right?

71b2b230f1aefccab2d4c073ccda094f.png
Raidbots
Raidbots
regal agate
#

its due to this:
#tc-research message
The change does lead to more fails and apl conditions could probably be fine tuned more for it.
I did look at annsies backstab improvements and some variable changes but haven't had time to sit down and go into it a bit more in depth.

knotty oriole
#

Max in this case is the max iteration, e.g. the longest iteration compared to the average

#

Buff window shows no expiry, so actually none are getting missed in this sim

#

7.3 uses vs. 7.5 triggers on the easy buffs and 6.7 vs. 7.5 would likely just be end of fight boundries

#

Either way it's using 27.7 out of 30

#

So definitely not 2-3 per value

#

(In the proc view the value of 9 would be the max observed value in a single iteration. Which would be perfect usage in the max fight length variance of 360s)

#

As Fuu mentioned that amount at the end being clipped can potentially be reduced with some changes to bring down the average time to consume the buff (which is 12-13s for the hardest one) but none of them should be getting wasted mid-fight based on those durations. It’s just fight end clipping based on uptime.

west prism
#

Yea alright, thats why i said i might have misread the table

knotty oriole
#

Think the Backstab changes bring the end of fight clipping from like 2.2 (total) to 1.3

#

Since it shaves a few seconds off the average use time of the harder buff matches

foggy quest
#

Hi rogue friends. I have found a terrible, gamebreaking rogue bug that is accidentally giving you guys more dps so need to make you aware of this awful error.

https://www.warcraftlogs.com/reports/NMrw6YJbTQ3n8Vak#type=auras&pins=0%24Separate%24%23244F4B%24auras-gained%241%240.0.0.Any%24211387642.0.0.Rogue%24true%24175732.0.0.Boss%24false%24358090^0%24Separate%24%23a04D8A%24auras-gained%240%240.0.0.Any%240.0.0.Any%24true%24211387642.0.0.Rogue%24false%24347807^0%24Separate%24%23DF5353%24auras-gained%240%240.0.0.Any%240.0.0.Any%24true%24211387642.0.0.Rogue%24false%246940^0%24Separate%24rgb(78%25, 61%25, 43%25)%24auras-gained%240%240.0.0.Any%240.0.0.Any%24true%24211387642.0.0.Rogue%24false%24327037&boss=2435&difficulty=5&wipes=1&source=13

you'll find here that the rogue dagger, Edge of Night's special effect Banshee's Blight stacks too fast when the rogue is subject to a damage redirect effect like Blessing of Sacrifice or Kindred Protection (provided the rogue is taking damage to redirect).

My running theory is that the dagger is checking against enemy HP but because its damage to a friendly its nulling out, which the dagger says 'no hp? apply max stacks!'

you'll find similar behavior pretty easily. top rogue logs with Kindred Affinity buff have high blight stacks while those without don't.

noble pendant
knotty oriole
knotty oriole
#

You are probably right that it has to do with redirection effects somehow. 🍝 is a Rogue way of life.

maiden cargo
#

definitely has something to do with it, but seems like its a bit random, or theres another factor in play as well

knotty oriole
maiden cargo
#

ye we were thinking the same thing

foggy quest
#

yes. look at pull 10 youll see the 1st kindred protection where the rogue took no damage, he didnt increase blight stacks. on the 2nd kindred protection where he was redirecting, blight stacks jumped

proud wadi
knotty oriole
#

Yeah I mean I think there are a few cases where it can fail to trigger a damage event. When I matched up to look at friendy fire damage events it seemed to always proc when doing a successful damage event against the Paladin

#

e.g. in this particular log, but was also looking at Mystlers logs

proud wadi
#

yeah it seems kind of a pain to replicate

knotty oriole
#

In his case here, it failed to trigger on the Paladin 3 times due to absorbs

# 
00:00:20.164    Sylvanas Windrunner Barbed Arrow Alarea Tick (A: 3923) (U: 4407, M: 484, 10.98%)
00:00:23.161    Sylvanas Windrunner Barbed Arrow Alarea Tick (A: 3923) (U: 4407, M: 484, 10.98%)
00:00:26.177    Sylvanas Windrunner Barbed Arrow Alarea Tick (A: 3924) (U: 4407, M: 483, 10.96%)
#

I'm sure there's other jank around triggering of Sac that isn't immediately obvious lol

proud wadi
#

yeah I only tested it with kindred, I'm assuming sac would make it a lot easier to test

knotty oriole
#

I think Mystler was able to get it more consistent by canceling Frost gem absorbs

noble pendant
#

it's consistent on the fate dot

#

twist of fate

#

just goes through the absorbs

knotty oriole
#

I'm sure there's some damage sources that just don't trigger redirection at all because reasons

noble pendant
#

big damage event = good

knotty oriole
#

Think you could just make an absorb cancelaura macro to spam when trying to set it up too

noble pendant
#

I tried it on pain

#

and it was just not working

knotty oriole
#

Could be the environmental ticks don't redirect at all? I didn't look at logs from that fight yet

#

I think there was some similar jank for N'zoth

proud wadi
#

it's possible, but I think the priest tossing atonement on everything may have contributed to that @noble pendant

knotty oriole
#

Looking at some random logs it seems like it triggers from the ticks though

#

(Sac that is)

#

At least the non-absorbed ones

#

Really easy to absorb them though since they are tiny

proud wadi
#

yeah seems like way more consistent with sac

noble pendant
#

it's worked on pulls 9/10/11 on fate

#

wasn't working at all on painsmith

spiral light
#

Push tried it out on our raid and he only got it to work on Remnant and Fatescribe.

#

He got sac from a paladin and only worked on 2 bosses.

#

I assume you just have to communicate with your paladin and ask for a sac when you know you can tank a bit of damage.

#

Like during Painsmith traps, Soulrender puddles/chains and so on.

#

Such a janky bug, who knows what's going on.

maiden cargo
#

I replicated it like 5 times no issue on a HC sylv

proud wadi
tired pollen
#

My guild is a bit twitchy about this stuff, prolly not gonna get a sac exactly when I want one tbh

golden briar
#

i tried it on a few bosses and it seems, as every1 already said, that you need to actually take damage

frozen crater
#

Does dad trick work on m+ bosses?

#

Sac*

foggy quest
# knotty oriole

the same holds true for other on-hit effects. soulbind stuff like Carver's Eye, Hammer of Genesis and First Strike all proc off the redirect damage. unless the hit is absorbed, then it doesnt. I dont have logs on hand for that tho.

warm zinc
frozen crater
#

Goated

knotty oriole
dapper carbon
#

So if playing frost just macro /cancelaura Frostrime ; /cancelaura Accretion to everything?

noble pendant
#

would help, yes

#

but even then it's very difficult from my experience last night

dapper carbon
#

and you were playing blood set, yeah?

#

I wanted to see how it goes with sin tonight

noble pendant
#

yeh I was playing blood

#

I assume with winds shield on top of everything venthyr token etc it's going to be pretty much impossible

dapper carbon
#

I'll give it a shot anyway
worst case scenario I just give up on it sadgeSitSip

eager prism
#

prayge please patch this degeneracy

thorny fog
#

can anyone check if cooking fire dmg works

dapper carbon
#

maybe with the alcoholic thing that 50x the fire damage?

unkempt crescent
#

Isn't there a toy that deals dmg to you?

#

Deathwings simulation?

tired pollen
#

Can you use that in a boss encounter?

lilac basin
#

It’s just tricky on certain bosses but very easy on others. On some like sylv you can just stand in shit during KA on others it’s the matter of communication with your “sac person” like you would probably want them to delay KA a bit for raidwide damage or like ask for sac during traps on Painsmith for example.

#

Also correct me if I’m wrong you don’t insta get 4 stacks if you just take a single hit. You need to take some damage over a few seconds to get full stacks

knotty oriole
#

It’s one stack per damage event to the Paladin by the looks of it.

west prism
#

cant they just

#

spam light of the martyr

#

during the sac

#

to remove any shields they might have

wanton notch
#

so it doesn't have to be damage specifically you redirect

proud wadi
#

I believe push is more referencing burning any absorbs that could be on the pally

gusty ridge
#

is priest shadow mend good enough

tired pollen
#

Damn thats fucking smart

#

does the shadow mend actually have to heal you in order for it to deal damage to you though?

#

(ie would it work at full hp)

gusty ridge
#

I would assume so, just not sure if the damage is big enough (if you have small shields and stuff)

tired pollen
#

Probably not, but you could just do that to cheese it at the start when nothing is going on yet

#

I dont get a shadow mend debuff when i heal myself at full hp, only when <full

gusty ridge
#

not very useful then I guess

golden briar
#

on a slightly different topic:
does any1 know if swapping out the daggers mid-combat drops their debuff from your target?

tired pollen
west prism
#

You think it will deal full damage despite not being equipped? It’s the daggers doing the actual damage no? The debuff just indicates how many instances? Or

mystic jolt
#

The tooltip on the daggers reads as if they just apply the debuff, and don't actually proc the damage

tired pollen
#

Maybe re-equipping it would apply another debuff

west prism
#

can be tested on a dummy

tired pollen
#

i was on my way already XD

#

that it does not

tired pollen
#

Even if the debuff is still on

mystic jolt
#

Ya then it's just a funky tooltip per usual

lilac basin
# thorny fog can anyone check if cooking fire dmg works

Just tested, Lil Rag fire damage with Nitrofuel does indeed work but only if you do it prepull. Looks like you can't get burning debuff in combat(I guess everyone knew it already from m+ cheese pulls). So pretty much you LilRag/Nitrofuel and KA prepull = easy 4 stacks on pull

slow badge
#

is there a reason you need KA? and is it a certain specs KA?

noble pendant
slow badge
#

okay so you just need one of them, gotcha

slow badge
#

also, is this doable on a test dummy?

slow badge
#

you cna split the stacks

unknown.png
#

just got 7 from one use

west prism
#

did you get in combat

#

with them

slow badge
#

no

west prism
#

wth

#

you can actually do it prepull

slow badge
#

this is more cringe than triple dust macro BUT IM ALL FOR IT

knotty oriole
#

Keep in mind, we probably used all of our reasonable WCL karma to avoid the blacklisting of all the existing Deathly logs. I can't imagine this one survives if it is hotfixed based on their normal policy.

#

(Just as a note for the purposes of parsing. I'll probably experiment with this too just for luls next reset if it's still around, but I'm keeping my expectations for long-term parsing in check haha)

west prism
#

Do you happen to know how long banshee debuffs linger on a boss

#

koji

knotty oriole
#

Forever afaik

west prism
#

wondering if you can bring in extra paladins just for prepull sac

knotty oriole
#

Like I've had to deal with that when dealing with dummy testing in the past

#

Because they get stuck on the dummies

#

(If you get them at low health and they heal up)

west prism
#

Yea ive seen that

knotty oriole
#

Hmm

#

Actually it looks like it just has hidden duration

#

The actual duration on the buff appears to be 10 minutes

# 
Name             : Banshee's Blight (id=358090) 
School           : Shadow
Spell Type       : None
Range            : 40 yards
Duration         : 600 seconds
Stacks           : 4 maximum
Proc Chance      : 101%
Triggered by     : Banshee's Blight (357595), Banshee's Blight (358135)
Attributes       : ........ ........ ........ ..x.....   ........ ........ ........ ........ 
                 : ........ ........ ........ ........   .......x ........ ........ ........ 
                 : ........ ........ ........ ........   ........ ..x..... ........ ........ 
                 : ........ ........ ........ ........   ........ ........ ........ ........ 
                 : ........ ....x... ........ ........   ........ ........ ........ ........ 
                 : ........ ........ ........ ........   ........ ........ ........ ........ 
                 : ........ ........ ........ ........   ........ ........ ........ ........ 
                 : ........ ........ ........ ........   
                 : Aura Is Debuff (26), DoT Stacking Rule (103), Do Not Display Duration (170), Aura Points On Client (268)
gusty ridge
#

does the buff stay after you get in combat though, or does it reset like most things

west prism
#

resets if you wipe

#

nvm read that wrong

slow badge
#

so theres an ICD on how often you can drink nitro fuel and still get the debuff (debuff being 8? seconds and the ICD i think is maybe 10), but sac being 12 seconds i was able to get 8 stacks on two targets barely......im also extremely drunk now so i doubt that extra stack on second target is worth

unknown.png
knotty oriole
#

Actually lol.. this explains the issue with Sylvanas.. it's not the phase transition that causes the stacks to drop, it's the fact that it times out at 10 minutes 🤣

west prism
#

oh no

#

that is so jank

knotty oriole
#

I just assumed it was the transition but this makes more sense now

gusty ridge
#

wait, do you actually need to play drunk if you want to do that, or is there a way to get sober

knotty oriole
#

This doesn't happen when using the sac jank because it forces constant refreshes

slow badge
#

you dont get too drunk from the first drink

#

but the second drink does kinda get your screen fuzzy

knotty oriole
#

I was curious why the sac logs didn't exhibit this behavior but this "makes sense" in this regard now

slow badge
#

i think i spam drank just then to try to get the debuff to come up and drank like 5

knotty oriole
#

Duration doesn't refresh at stack increments, just after it has reached max stack

eager prism
#

is this doable with just sac or do you need sac and ka

slow badge
#

just one

#

im doign it in org with a pally buddy of mine

knotty oriole
#

Honestly curious if it works with some of those odd effects that damage other players in the raid lol

eager prism
#

so it's basically accessible to every rogue Cring

knotty oriole
#

Or if it's just redirection

slow badge
#

guys iom fucking dying

#

dark runes work

#

they aplpy 1 stack

#

15 minute cd

#

but it applies 1 stack while youre in blessing of sac

lilac basin
gusty ridge
#

better stock up I guess

lilac basin
knotty oriole
#

I believe that seems to be the case yeah

#

At least from the logs I’ve seen

wanton notch
#

When would be a good time to refresh with second sac? Second p2 burn?

#

First one is prob ok too

slow badge
#

seems like the prepull does pull boss the way i was doing it, not sure if theres a way to do it without pulling.

#

so a sub rogue is going to do the triple macro on 2-3 and then have a sac on them already and pop that juice basically on pull (i think theres a small window of time where if someone pulls youre still out of combat very briefly?)

wanton notch
#

I just tell my pal to sac at 3 seconds prepull and press dust+nitro macro at like 1-2 seconds

#

And if I see moonkins projectiles fly towards boss I do it earlier xd

lilac basin
#

Does it pull if you’re not targeting the boss? Because I never pulled during testing, maybe that’s the reason

#

Oh actually I was getting sac after pull, yeah thats what it was

#

So you nitro/dust pull boss and get sac first second of the fight

slow badge
#

thats big science, thanks for the brainstorm

prime narwhal
proud frost
#

did it today, got the first sac on p1 and a second one on p3, to refresh it

unknown.png
prime narwhal
proud frost
wanton notch
#

I had it work for a few pulls and then on the kill it just didn't give me stacks lol, think someone trolled and put absorbs on me xd

white elk
#

Wait so how does it exactly work? You need to get a sac when you have 4 stacks, and then sac again later to reapply the 4stacks back to sylvanas?

#

Hm ok I misread. So you need to take damage while you have sac on to apply the debuff faster?

proud wadi
white elk
#

Any damage? Even ticking damage or something direct like standing in a puddle or whatever

proud wadi
#

Any damage at all, my guild decided to try it out last night and it was consistent

tender scarab
#

Sounds like you need to become friends with your local Intel DPS. So you can pre-Pull sac> Use trink>trink swap macro!!!! I don’t know if it will work but maybe worth a shot🤷🏻‍♂️

IMG_5953.png
white elk
#

Don't think it'll work since you need to be in combat to add the stacks, but you can't change items while in combat and vanish won't let you change equips either mid fight

slow badge
#

go do mythic Crucible of storms and loot the trinket

vapid tusk
#

tried it with KA on a dummy and didnt work, with Sac it did copium

regal agate
#

did you get KA from a guardian druid, it doesn't work with restro iirc

vapid tusk
#

yea guardian

vapid tusk
#

already fixed that logs doesnt count 😅

west prism
#

they did, or are you hoping they do?

prime narwhal
#

i dont see anything and i just did it today

vapid tusk
#

they did i guess

#

wait a sec

west prism
#

i dont think that means its removed in logs

#

just means the rank isnt locked in yet no? because there might be more kills before the day is over

#

or something

vapid tusk
#

i dont know, had this never before and did the sac pull today.

#

so idk

west prism
#

Tbh those things tend to happen very early on in a new log partition or a new raid tier.

#

So it's a bit weird that it happens now

prime narwhal
#

it makes sense what you said push

#

that its just not locked in rank yet

vapid tusk
#

already 100k more banshees dmg than other rogue in p1, stonks

proud wadi
#

sometimes it takes a bit longer to lock in

prime narwhal
#

this is some good shit

unknown.png
vapid tusk
knotty oriole
#

not locked in just means the rank is still in the queue or awaiting processing

#

If it was blacklisted it would have a dark red background

foggy quest
#

tisnt dynamic

tough hornet
craggy zephyr
#

14.4% blight dmg on m sylvanas lol

white elk
proud frost
proud wadi
lavish zenith
#

i rlly don't understand why ppl going for parses in the last week of 9.1.5 instead of just saving this for 9.2 so u don't get benched >< they gonna fix it boys and then we gonna be sadge

maiden cargo
#

they would fix it anyway my man

#

probably one of the most high profile bugs rn lol cause it would have drastic implications on rwf

#

thinking it would go live untouched into 9.2 is super naive lol

quartz geyser
maiden cargo
#

literally the moment it was publicized in this channel, the bug was on the track to be fixed

spiral light
#

Yeah, the person who made the dagger says that he's known about it for a while now and are actively finding ways to get it fixed.

#

It might be a bit tough to fix, with the problem being the redirection mechanic in itself. Working on stuff like First Strike etc.

#

But they're definitely aware of it and actively trying to get rid of it.

#

Have fun with it while you still can. :P

lavish zenith
#

nah but ppl here say that all the time, meanwhile there are tons of bugs that have been on blizzes radar that don't get fixed. it just all pertains to their prevalence and impact on the game.

#

there are literally tons of examples of bugs they know about that they let slide cuz it's just not worth the time. if not a single person used this thing, then the odds of them fixing in time for raid would be really low. but ppl use it cuz game theory i guess or i honestly don't know why at this point in the patch lol

spiral light
#

Not the place.

#

But yeah, who knows. We'll just have to wait and see what happens. Scarizard himself has said that he's working on it, so I'll take his word for it, but who knows.

wanton notch
#

Xd but basically it's gonna be hard to make a lot of people not talk about such a major thing

#

Unrealistic to keep it under the radar

maiden cargo
#

A bug that gives sub 900 dps is not gonna slide by

quartz geyser
#

Youre insane to think it would have been let to slide

#

That’s not some minor bug that completely shifts up the dps ranking and overall balance

#

And for early prog would be extremely strong

#

Do you have examples of those bugs that are worth nearly 1000 dps?

cunning cipher
#

Does sac on first painsmith traps works? do you get 4 stacks? rooThink1

maiden cargo
#

not a question for this channel

#

90% of the stuff in here past couple days isn't supposed to be in this channel lol

lavish zenith
#

monks we're bugged for more than 1k dps during 9.0 race to world first

#

limit still abused them and they weren't fixed in time

#

ret paladins were also bugged this tier iirc but that might have only been on dormazain i cant actually remember but that bug was massive

#

hunters also were bugged for 9.0 rwf and also for mdi, i dont recall if that was 1k ST dps, but it was certainly more than 1k aoe

white elk
#

But were those properly discovered and outed in time

lavish zenith
#

they literally weren't

white elk
#

And the dagger one is

lavish zenith
#

they were fixed after the race to world first

wanton notch
#

Were the bugs discovered like a month or 2 before the race?

lavish zenith
#

monk one was. paladin was discovered during testing so maybe slightly later. not sure abt hunters i think hunters was during or after the race but i didnt have the same knowledge then that i do now so not entirely sure

grave wharf
#

monks have never not been bugged in one direction or another. either way I don't think you're going to get anywhere arguing that we should collectively hide all the 'good bugs' until after prog, it's just not the way things work around here

slow badge
#

on a macro scale, yeah its futile to try to hide it, but on a micro scale its not entirely unreasonable to think that blizzard wouldnt have looked into the dagger had they assumed sylv dagger wont be an issue next tier

#

if no one really talked about it/ used it and blizzard assumes it wont be an issue once everyone grabs the new dagger, then they wouldnt have to worry

golden briar
#

does any1 know if spirit link totem would work like sac/kindred for the dagger bug?

#

or does that not count as redirecting dmg

#

now that i asked that question, the answer is obvious to me...

knotty oriole
#

Realistically this is peak parsing time since it's reaching the tail-end of the tier. This is where people lock in their all-star points for the tier/segment.

#

Most people in RH are really not overly concerned with the meta for the first week or two of RWF. Maybe a few folks will be involved in it. Majority really won't care at all. But the majority really has no vested interest in "saving" something for the first week of next tier's prog.

#

But either way, this is kinda the perfect timing for people being interested in parse strats since the last month or so of any given tier is really when people try to lock in their ranks and finish the tier with strong all-star points. So this is honestly not unexpected.

#

Anyway, let's try to get this channel back to just posting actual findings or research if possible. Thanks!

noble pendant
foggy quest
noble pendant
lavish zenith
#

any apl changes changes looked into to rtb outlaw for skull and crossbones? w changes to 4p could change something. also higher sleight condy=slightly higher probabability to fish for SaC, although the latter probably doesn't matter

wintry reef
#

Are you talking about possibly finishing at -1 max combo points?

knotty oriole
# lavish zenith any apl changes changes looked into to rtb outlaw for skull and crossbones? w ch...

Haven't heavily looked into RtB priority yet because there are still some uncertainties on how the 2pc is intended to interact with buffs. Currently there's still interactions with Opportunity (and GSW, etc.) and it's been stated there shouldn't be, but maybe those plans have changed. Was kinda waiting for clarification because it changes the valuation of things enough that it is worth waiting for better information.

lean talon
rocky fern
#

(9.2 PTR) Wasn't sure where to leave this info,but it looks like they might have pushed some Deathspike fixes to charge on death resets. Testing combo point generation at the moment.

rocky fern
#

Combo point generation appears to be fixed as well.

knotty oriole
#

Good to know, thanks for the updates!

#

Will adjust SimC to not have those for PTR sims tonight then.

rocky fern
#

Still works off GUID and the cleave range is tiny, but yeah those fixes should make a noteworthy change to sims of it.

knotty oriole
#

Well the charge change is fairly large for its usability in AoE situations so still a nice improvement

rocky fern
#

Absolutely, night and day difference. All you really need for full efficacy is a mouseover for bone spike and now it'll actually give tons of CPs quickly, and charges back reliably on death of targets.

lavish zenith
#

anyone looked into any conditions for the # of shadowstrikes itd be worth to delay a 6 stack pv backstab if u couldnt be behind ur target? ie if u had 6stack pv but couldnt get behind the target for 1 global, itd be worth to overcap pv stacks for 1 global with a shadowstrike and then backstab behind the target? any sims u could possibly run for that?

#

also was wondering if anyone did the min number of targets where it becomes a prio loss to press 6pv backstab

civic mango
#

what scenario are you in melee range but can't get behind your target in time that you need to choose between backstab or strike?

lavish zenith
lavish zenith
civic mango
#

also isn't it always worth pressing a 6 stack backstab no matter what scenario if your goal is purely priority target damage?

lavish zenith
#

so u do a backstab, then have to do a ss to get back to evisc

#

so over the course of a funnel fight u lose a lot of cdr and eviscerates in favor of the backstab

#

also the dagger damage

#

from the cp gen as well

civic mango
#

ok i guess my point is that you asked about a minimum target

lavish zenith
#

and the higher 4pc proc chance form increased eviscerates

civic mango
#

oh nevermind i see your point

civic mango
lavish zenith
#

well this isnt obviously a main thing i just thought of the first example off the top of my head. its a niche thing but id say it occurs enough that i had the thought to think about it lol

#

on first boss theres this extreme dam thing and a safe zone, but u face the boss in the safe zone

#

have to wait for the safe zone to phase to be able to get behind the boss again

#

its obviously not a meaningful thing maybe its like 5 dps increase, i was just mostly curious. the 2nd one the funnel one was more of an impactful question

civic mango
#

does it though? Like sure, if there's a boss that you have to stand in front of the entire time it's worth thinking about, but like, on almost every target i would say the correct answer is just not press a button at all, run behind the target, and then press backstab

#

and if the model is gigantic and running would take so much time that you'll cap on energy, just shadowstep to teleport behind them

lavish zenith
#

yeh i mean idk it was more of a curiousity question the first one. obviously it has occurred enough, not being able to be behind that i thought of the question lol. but its probably not massive either way was just wondering if anyone thought about it

#

sometimes on sylvanas for instance u cant get behind boss

#

but u could 1 second late

#

and i backstab the boss from the front

civic mango
#

yea so just don't press a button lol

lavish zenith
#

theres a more practical expmple i suppose

civic mango
#

and run behind her

lavish zenith
#

well then u overcap energy

#

in lust in the opener

civic mango
#

in 1 second?

lavish zenith
#

i mean pretty much

civic mango
#

it's worth it

lavish zenith
#

well intuitively it is

#

of course

#

i knew that

#

that isnt the point of the question then lol

#

like at what point of overcapping would it not be wroth?

#

2 seconds?

civic mango
#

well you specifically laid out the scenario of delaying a single gcd

lavish zenith
#

3 seconds?

#

idk

#

ok well u can read between the lines the information i was aiming for i guess

civic mango
#

but if you start asking about 2, 3 seconds, what possible example can you be in melee range for 2, 3+ seconds but not able to get behind the boss in that time, but is also a boss that can be attacked from behind at all

#

and also don't have shadowstep available to teleport to the back

lavish zenith
#

vigilant guardian

#

like i said

civic mango
#

i thought you couldn't be behind that boss at all

#

meaning you can't delay/use shadowstrike until you can hit from behind

lavish zenith
#

u can be behind the boss almost all of the fight

#

just not during the expose phases

#

cuz u have to be in the safe zone facing outward, and he is on the edge of the safe zone facing inward

civic mango
#

did the boss change? I don't remember a single thing about that fight except fighting adds then stacking in front of the boss to burn it at the end

lavish zenith
#

i mean the results dont really matter it was more of a question "if anyone has thought about x and ran a simulation, lmk. if not, ignore it." but the second one was far more practical

lavish zenith
civic mango
#

oh, i guess i missed that boss on mythic? I don't have any memories of it

#

what boss was taht on the same day as

lavish zenith
#

mythic u just funnel the boss the whole time its basically m+

civic mango
#

oh i see

lavish zenith
#

i think it was anduin + guardian for mythic i believe

civic mango
#

well, i actually really hate this answer because it defeats the purpose of TC, but it is the 1st boss and nit picky optimization is completely pointless if it literally only applies to that one boss

#

feel like that should be stated at some point, but it completely ignores that an answer to your question can be found

lavish zenith
#

that one just popped into my head instantly

civic mango
#

i think you would have a very hard time finding examples to be honest. It would need to be an enormous boss model that somehow forces you to stand in front of it, and also requires excessive use of shadowstep to execute mechanics. Eye of the Jailer comes to mind, but it is definitely rare

lavish zenith
#

yeh its for sure rare

#

the second question has many more implications. first question was curiosity for the most part.

civic mango
#

feel like it would be answerable by running a 6 target sim with perforated veins and without perforated veins and see if it's higher priority target damage to not use it at all and just spam storm evisc compared to consuming backstab stacks when you get 4pc procs

lavish zenith
#

yeh that would do it pretty sure

civic mango
#

because if it doesn't win at 6 targets, it definitely won't win at fewer than 6

#

the non perforated veins setup, that is

regal agate
#

i haven't read the entire discussion but if the question is about priority target damage in aoe and pv, it is worth it to use.
I still have a apl change pending that i need to push for it.

lavish zenith
cyan rivet
#

if there's enough AoE/funnel would it not be worth it to run PE>PV?

regal agate
#

pv and pe is similar in prio

cyan rivet
#

ig not for funnel since PV is prio

#

but if you're using SS a lot less

regal agate
#

i did that comparison with the change, so you essentially can play both and decide wether you want to have more damage on single target/low target (PV) or spread the damage a bit diffrently with PE

knotty oriole
#

When it boils down to it

#

Which should be easy to test. Can just set position=front and compare the line being disabled or not

lavish zenith
#

well they were 2 separate questions. i actually think the 2nd question is actually more applicable to most ppl. like how much dps do i gain/lose by pressing backstab w/ 5pv stacks doing prio dam on 6 mobs

#

like if it's 1 main boss and 5 adds, if i can play pe/dd/lashing and just storm evisc 24/7, or pv/dd/lashing and backstab whenever 4p procs. what's the dif in prio dam. was wondering about if ppl had done that one yet

lavish zenith
#

anyone looked into if its worth symbolsing a 6stack pv? or 5? or symbolsing if blades is active? not sure how often the sim runs into that since its timings are perfect but it can happen sometimes. like symbols off cd when dance ending when u got a 6stack

cyan rivet
#

You mean Symbols->backstab before going into a second Dance? Or just raw Symbols without another Dance

lavish zenith
#

symbols backstab before going into dance

#

as opposed to plain backstab -> symbols + dance + shadow strike

#

symbols + backstab -> dance + shadowstrike

#

on 5/6 pv w/ or w/o blades

dapper carbon
#

Pretty sure fuu said if you have 5/6 pv stacks you can symbols -> pv

#

not sure on the blades thing though, you might wanna tag or ask fuu/push/annsie

cyan rivet
#

Yeah I feel like I remember reading that Symbols to juice a pv backstab is good

regal agate
#

i did some tests a while back, but i would need to spend some time fiddeling with the apl to see if i find a consistent damage win in doing so

golden briar
#

6pv stab is WAY stronger than a strike

#

i think even at 3-4 stacks, a pv stab would be worth consuming the crit buff with (but don't quote me on that)

lavish zenith
#

yeh thats what i figured

knotty oriole
#

This is only really relevant on a B2B Dance right? (Otherwise you're not gonna have PV stacks going into a Dance) Otherwise you'd be delaying Symbols too much most likely or having bad Dance coverage in general

#

But mostly a question of what timing it would actually come up for, since usually you'd be using SoD at the start of a double Dance sequence, if you wanted to capture the crit on the PV BS in the middle, you'd have to delay SoD by a full Dance duration which really might not be worth it

#

Not really as much about the value of the Crit buff as much as the value of the full SoD buff duration and what it overlaps with

#

Like you may gain 2-3k damage from the Crit on the PV Backstab but lose out on SS/Evisc coverage during Dance so hard to say. Fuu might be able to find something but usually things like this don't end up pulling ahead if they require significant resequencing. This will only naturally happen pretty rarely I think.

mystic jolt
#

you can have dance->sod backstab -> dance pretty consistently and reliably as kyrian though

knotty oriole
mystic jolt
#

no you just use dance 7-10 seconds before sod cd is up

#

probably not 10 sec but around that range

knotty oriole
#

I'm not really convinced it's totally worth re-aligning all that stuff to get an extra 2k damage or something off a PV crit vs. a Shadowstrike but I'm sure Fuu will find out 😛

lavish zenith
golden briar
#

There is no realigning, you swim in dances as kyrian

#

I even do this as venthyr, double dance on pull, reapply rup and snd then dance again 8 sec before sod is back up

#

Then sod stab and dance again

finite sparrow
#

in simc, is it possible for the outlaw 2pc to trigger during the casting animation of a dispatch/bte and therefore miss your targeted anima cp for example?

#

im trying to find a moment where it happens but it's uhh gonna take a while

#

oh nvm, i guess cp is consumed on cast, and not on the damage instance per se

gusty ridge
#

I think its time for the kyrian sub treatment of reacting to cp procs for outlaw as well kekw , for sub you can at least control the procs

knotty oriole
#

Sub only works since it’s pretty much deterministic

gusty ridge
#

maybe add a tracker to the proc if there isn't one already and consider yourself on the previous cp if you procced it within X seconds, which would lead to alot of thinking you hit 2 cp animacharge but instead hit a 3 cp finisher, which might happen alot ingame as well

finite sparrow
mystic jolt
finite sparrow
#

unlikely that is useful, as bte'ing another target does nothing unless you plan to swap to that target, or something like phial happens to be ticking on it. and then perhaps bte is on cooldown for when it actually satisfies the above conditions to use it

knotty oriole
#

BtE is still higher DPET and the difference in buff uptime by doing this is extremely minimal so I don't think it's related to debuff uptime

#

(There's only a 0.1% change in debuff uptime in some of these profiles)

#

Or I suppose Dispatch and BtE end up pretty similar DPET if RP is down so yeah, it probably is just slight CtO skew

#

I'll test a few more conditions just to make sure it's not a loss for other setups but looks good in general

finite sparrow
#

hmm yea if the dpet is the same without rp then must be cto

hearty olive
#

Dumb question what’s dpet ?

spiral light
hearty olive
#

thanks cev wolfLol

lavish zenith
#

is there a command to force the sims to use axe?

#

on the trinket?

#

or is that baked into the new trinket apl or something

twilit granite
knotty oriole
#

Axe on 2T+ and Wand on ST is best for Sub and Outlaw

lavish zenith
#

oh

knotty oriole
#

Assassination is a little strange and taking a bit to figure out because of 4pc oddities

#

The haste snapshot makes utilizing the Haste buff with all the timing offsets extremely difficult even though it should be good in theory.

#

use_item,name=cache_of_acquired_treasures,if=buff.acquired_axe.up&spell_targets.shuriken_storm>1|buff.acquired_wand.up&(spell_targets.shuriken_storm=1&raid_event.adds.in>60|fight_remains<25)

#

For Sub

twilit granite
knotty oriole
#

That was not what I was finding when I tested

twilit granite
#

and with my profile it also sims higher

regal agate
#

i did test a bit just some minutes ago, and it seems for subtlety at least that sending it on cd is better and optimizations like for add spawn timers/etc seemed neutral or a loss

twilit granite
#

What was the timer for the adds spawn

regal agate
#

i did use the conditions from above with 30/60

twilit granite
#

in 4t it hits for more than x3 times st damage

regal agate
#

yes, might need more testing to see if there is a condition that holds up

knotty oriole
#

When I looked at the sim linked in the Sub channel it seemed the copies were wrong but idk

#

I’ll have to look at this again but I ran this specifically against the default profile and found the opposite like 2 days ago.

regal agate
lavish zenith
knotty oriole
# regal agate wdym wrong?

copy="trinket_1,default"
trinket1=",id=188265,bonus_id=4800/4786/1524"

copy="trinket_2,default"
trinket1=",id=188265,bonus_id=4800/4786/1524"

#

The two copies are just the same thing

twilit granite
#

he forces it in the apl

regal agate
#

oh ye

#

that one

knotty oriole
#

Yeah but it’s not comparing anything

regal agate
#

it is comparing to the default

twilit granite
#

ye

regal agate
#

but yes, i see what you mean

knotty oriole
#

Well the default just uses on CD

regal agate
#

yes

#

the sim lievet linked however does use the same trinket

#

so even tho i made a mistake by not correcting the slot, it is not point of the discussion

#

but i can compare the above condition to simply using it on cd, as this might lead to a more healthy discussion about it

knotty oriole
#

Well the comment after that sim was that it was better than the previous best but the previous best line wasn’t in there so wasn’t quite sure what it was comparing.

regal agate
#

previouse best trinket

#

it was comparing power level of trinkets

#

in this case it did just replace iqd two times

#

what is like you mentioned wrong

knotty oriole
#

Right

regal agate
#

but it does not rly matter in that the apl change is considered better than the above line. but giveme a minute

knotty oriole
#

I don’t totally see why you wouldn’t have seen a gain from delaying for adds though. I definitely did the other day.

#

For both HAC and DS

regal agate
#

i was surprised too, thats why i mentioned it

twilit granite
knotty oriole
#

Honestly for Venthyr I wouldn’t be surprised if this was just accidental timing. So be careful about that.

#

Whatever syncs with the Vers buff will be best. Almost certainly.

#

Might need a custom condition for Obedience

#

Since the order is deterministic now instead of random like it was the other day (fixed in a recent build, always starts with Sword) it could be leading to specific timing overlaps.

#

So just have to be wary about overfitting

regal agate
#

alright to showcase it:
patchwerk:https://www.raidbots.com/simbot/report/moVhKaJXZwV1P11T9kfmUr
dungeon slice: https://www.raidbots.com/simbot/report/jAEtbzCDragKA4cKxNXZgY
hectic add cleave: https://www.raidbots.com/simbot/report/pH8NfpH8anJ6o79MXSn9d7
on cd condition: actions.cds+=/use_item,name=cache_of_acquired_treasures,if=buff.acquired_axe.up
default koji: actions.cds+=/use_item,name=cache_of_acquired_treasures,if=buff.acquired_axe.up&spell_targets.shuriken_storm>1|buff.acquired_wand.up&(spell_targets.shuriken_storm=1&raid_event.adds.in>60|fight_remains<25)

unknown.png unknown.png unknown.png
knotty oriole
#

This really feels like overfit to me after Jayeasy’s timing fixes. But I guess I’ll look at it again tonight.

regal agate
#

ye, it felt a bit weird that neither of the optimizations seemed to have any effect

knotty oriole
#

Like there’s no way it should be better to just use on CD in a 5 minute fight since it’s the same number of uses no matter what

#

For HAC

#

The DPS distribution is wonky if you look at it

#

The graph is very skewed

#

So I think there is definitely a secondary factor here

regal agate
#

agree, somehting seems not correct in the sim. wonder what it is

knotty oriole
#

Yeah I will debug a bit more locally tonight

knotty oriole
#

I think the deterministic timing with the fact that 3m cooldown doesn't get any additional uses on a 5m sim with 20% vary length and the fact that HAC adds stop a set percentage into the sim just result in a 5 minute sim often missing the Axe buff on the last add wave

lavish zenith
#

wonder if axe is strong enough to merit an apl change on 2t with the axe active for more damage incidents. like storm on 2t in dance, or bp if both have fw, etc

#

to apply the axe more frequently

knotty oriole
#

Checked in the APL change tonight to always prefer Axe for GV sims tomorrow

lavish zenith
#

i meant like

#

normally on 2t u shadow strike in dance

#

but since the axe works off of instances of damage

#

storm + instant poison would give u way more

#

and if it works like dp on spriest

#

could be ok idk i didn't check it out

knotty oriole
#

It doesn't proc from Instant Poison apparently (sims thinks it does but it has some funny flags and doesn't appear to)

#

I was testing this tonight

#

It only appears to trigger from foreground class abilities

#

Does trigger from Storm, but not from IP for whatever reason

lavish zenith
#

well maybe it's worth to alternate strikes then

#

to refresh it on both targets

knotty oriole
#

Which means Outlaw doesn't get it on secondary targets other than from Blade Rush or the instant damage of BF

#

@regal agate could possibly look into some APL changes for 2T when Axe buff is up yeah. Alternating could make sense.

#

Although the duration is quite short and it pandemics

#

You'd probably need a rather high crit chance for it to sim as a notable difference

#

But could be worth looking at

#

Probably in this way it's slightly oversimming right now until we figure out why it can't proc from certain things...

#

Although still pretty good anyway

knotty oriole
#

I'm gonna have to start banning anyone that links Anduin parses

civic mango
#

O_O

#

Not exactly a parse

knotty oriole
#

😛

civic mango
#

😋

knotty oriole
#

Gonna be dealing with 6 months of people linking Anduin pad until WCL removes it from all-stars 😛

civic mango
#

i mean, could just filter bonus damage and maybe small adds too

#

but i didn't press black powder a single time on the kill

knotty oriole
#

But yes it's very good for mass AoE for sure

#

But probably not as good as it sims since it's refreshing from poisons and stuff

#

Also probably doesn't apply from Tornado storm ticks

#

Or Black Powder Shadow damage

knotty oriole
#

Ok, as a note. I think I figured out a way to sort out what can and cannot proc the Cache trinket. Implemented some basic rules but didn't make the nightly.

regal agate
#

I did ask for a nightly rebuild, so should be in the nightly now on raidbots 🙂

cyan rivet
civic mango
#

the thing is, those adds are actually extremely dangerous, and killing them is very important

#

it's not like you're wasting boss damage by blasting the adds, the boss damage is not a priority at all and is basically a given, the adds are what wipe you

#

anyway, trinket is awesome, but i am not sure i agree with wand being better than axe on ST for kyrian

#

can you link me the specific sim for that fuu?

knotty oriole
#

It’s the case for my gear personally with Kyrian

civic mango
#

I see, i'm getting basically identical damage values from wand and axe in my venthyr and kyrian sets

regal agate
#

so wand is on pure st seemingly better for non venthyr

civic mango
#

Thanks Fuu!

regal agate
#

also fun observation

knotty oriole
#

I think Sword is potentially better with some magical perfect alignment RNG but getting things to align properly without waste is extremely hard consistently

#

I tried messing with the APL for it for over an hour and found some improvements but nothing that made it strictly consistent or better yet

#

With 1 minute Vendetta cycles and Haste snapshots prior to casting Vendetta and also the DoT refreshes during Vendetta the potential for getting high vs. low value is very notable

#

The 4pc haste snapshotting behavior is really problematic considering the weapon buff is on a timed cycle that can't just be popped like GCB at the right time

knotty oriole
#

Does seem neutral without CtO and without set bonus (probably because it's exceptionally rare to even happen)

#

But does seem to be a gain with the 4pc and normal setups

hearty olive
#

Thanks koji and solo loviko

knotty oriole
#

Also included Fuu's similar change for checking the Ghostly Strike debuff duration

#

Which is smaller, like 0.2-0.3% but the two add together fairly well

hearty olive
#

Must've missed it what was that change ?

knotty oriole
#

Just checking the debuff was below 3s

hearty olive
willow knoll
#

What specifically do both of these optimizations entail to actual gameplay?

finite sparrow
lean talon
#

ah just checked the sim u linked you did test different times, sorry. any chance that BtE falsely pandemics in simcraft? seems a bit weird to me that the difference of 5s and 4s is that significant

finite sparrow
#

i dont think so, the difference was within margin

#

well just outside of margin for the first profile i posted and marginal for the second

knotty oriole
knotty oriole
#

(Honestly, I think it should pandemic. Always felt like an oversight on Blizzard's part.)

willow knoll
#

an oversight? for outlaw? from blizzard? 😲

finite sparrow
#

What exactly is desired_targets

#

in actions.cds+=/use_item,name=windscar_whetstone,if=spell_targets.blade_flurry>desired_targets|raid_event.adds.in>60|fight_remains<7

knotty oriole
#

Just means the targets specified in the sim

#

e.g. the number of patchwerk targets

#

All that basically means is "if we have higher than the static number of targets"

finite sparrow
#

So in dslice desired_targets is 1?

#

Throughout the entire sim duration

knotty oriole
#

Yes

lavish zenith
#

if ur never behind ur target a single time, does pv still beat pe/ss as venth sub on ST? is there a custom apl line for that or somethin?

west prism
#

there is a command to hit the boss directly in its face

#

always

torn dome
#

position=front

#

back is default

random forge
#

Any tc/sims of not using PV? I know in pve most rogues always pick it, but was wondering howmuch the dps loss would be in %

willow knoll
#

You can just sim something like that on top gear

knotty oriole
#

Think PV goes from like 2.2% ahead vs. 1.5% ahead my next option with my current 4pc setup (without Akaari, which will push it further)

lavish zenith
#

does anyone know if the halondrus trinket Earthbreaker's Impact collects all three little fragment things always in the sim? and is there a line to force it to not collect any, collect only 2, etc?

hearty olive
#

there's the line here

regal agate
#

shadowlands.earthbreakers_impact_weak_points=2 for reference for mobile users.
also yes, its 3 by default

regal agate
#

The outlaw buff should be:override.spell_data=effect.136652.base_value=50If you want to sim it.
(It is roughtly 5-10% depending on the target count)

dense mountain
#

in subtelty, when nightfae, when adrenaline rush triger, with celerity ring equiped, the buff indicate a 8% damage bonus.

lavish zenith
#

does cache underperform in dslice cuz it's not implemented correctly or cuz the target count doesn't accurately reflect the avg # of mobs in a real dungeon? or both?

finite sparrow
#

im guessing pack duration has the most impact

#

majority of dslice pulls dont last for the 25s of axe, especially the larger pulls, but also iirc the subtlety APL does not make an effort to hold cache for large pulls, it uses axe as soon as it sees 2+ targets

noble pendant
#

you also don't see big 25 mob packs like you would in ardenweald

knotty oriole
# lavish zenith does cache underperform in dslice cuz it's not implemented correctly or cuz the ...

I mean realistically it's a little of everything. The rotating timing issue makes it difficult to line up in any smart way. Usually it's not worth delaying unless the pulls are huge, which isn't usually an issue for DS overall value because there are larger pull overlaps but not ones you can "plan for" in the way that cache alignment might require. And the duration will skew on higher keys vs. lower. Something like Whetstone was relatively straightforward because there was no time-based "rotation" to plan around, just use it at the start of the pack.

I suspect most of how players use Cache is actually not optimal, but can be offset by using on really really large/aggressive pulls since it's effectively uncapped/unscaled.

Considering Cache is like the 2nd best simming trinket in DS anyway though I honestly don't think that's really far off the average value for "average", mid-range keys which is what DS is more geared to emulate.

#

Don't think it's really a huge deal though, since anyone who is doing giga pulls probably knows enough to use Cache or Chains since they're essentially uncapped.

#

That being said, I think Fuu did have some ~1% improvements for the APLon 2T/3T for Sub by actually changing the rotation (Shadowstrike conditions) that he sent me while I was out of town.

#

Which I'll probably get in over the next day or two

civic mango
knotty oriole
#

This is kinda what I mean by a lot of people use it in a somewhat inefficient way. I've seen a lot of logs where people generally hold it too long, hoping for alignment at the start of a big pull. Only to have it not sync up the Axe buff until it's too late. Then they hold some more. Then they hold some more.

civic mango
#

I mean, in a dslice sim you're only going to get 2.2 uses on average? Getting 2 guaranteed on high target count packs will be better every time

knotty oriole
#

And then they luck out one time on some 6 mob pull and it makes up for it

civic mango
#

i've never had an issue lining up the axe in a real key with at least a decent pack

#

sometimes you have to blow it on 4 targets when the very next pack might have 7 or 8 just to make sure you get the axe buff right when you want it/get it back for a pack later on

knotty oriole
#

See a lot of damage profiles like this where almost all their value is only coming from 3 pulls

civic mango
#

ok but i personally don't think bad usage cases are good data

#

just don't be bad at using it

knotty oriole
#

I mean this is honestly just the reality of it unless you have all your pulls really well timed out in advance

#

It's not a straightforward trinket to use

civic mango
#

i sort of disagree. I think it is actually quite straightforward, it just takes practice and planning

knotty oriole
#

If it takes practice and planning it's not straightforward 😛

civic mango
#

hm, i get your argument, but i've had the trinket for several weeks now and it's very straightforward for me to know when to use it and get great value

knotty oriole
#

Straightforward is you macro Earthbreakers to SoD and hardly don't care about it at all.

#

And it'll do a consistent 4.5% of your damage at least with minimal effort

civic mango
#

ok, but i could just as easily say that using it on pull 1, 6, 9, 15, and 19 is equally straightforward

knotty oriole
#

Not saying Cache is 200 IQ, but it's not quite as easy to sync up on high value pulls unless you already know your route and timings

civic mango
#

100% agree with that

#

no one will use it perfectly their first time

knotty oriole
#

It's pretty hard in pugs because who knows what they are going to do lol

civic mango
#

but i also think it doesn't take a huge amount of effort or experience to see what to do better with it

#

but as long as you're using it on at least 4 targets, it's basically guaranteed to do 4x more damage than using it on single target

#

if you can spend the entire 25 seconds alternating storm powder, it's going to do amazing

knotty oriole
#

DS is meant to replicate a somewhat chaotic timing pattern because that is what allows for more fair modeling of CD usage in general, however I would say the moderate duration coupled with the cycle mechanic of cache and long CD of the trinket is probably not ideal for having enough intersections of the great use case.

civic mango
#

Yea, that's kind of my point, with this trinket specifically it should just wait for the 1st pull after the boss, but it's not because of any flaw with dungeon slice

#

it's just the trinket is kind of odd

pastel sundial
#

@knotty oriole does the trinket start with "random" rotation in ds? bc if you bring it on cd before you start the key it will always start sword into axe so you have a guaranteed first pull axe (not sure if this is common knowledge - i didnt catch up with all the stuff, sorry)

knotty oriole
#

Maybe can add an option to force

pastel sundial
#

should push up the value by quite a bit , considering many dungeons have a big first pull, right.

knotty oriole
#

Not sure how much it would help with DS since the start is the boss however if it is deterministic at the start instead of random it could help with the later alignment issues being more consistent. I’ll look into it.

golden briar
#

i feel like DS should have boss at the end 😄

boreal patrol
#

It was a very practical decision as most people can be expected to use cooldowns on bosses. Having it at the start makes that easy without putting much fightstyle specific logic into the APL. Plus reduces the potential awkwardness of the sim yoloing all CDs into the very first trashpack - whatever size it's defined to.
So, yeah, in the dungeon you don't see a boss first but for the sim it's most practical to take a slice from boss through trash to next rather than from trash to boss.

lavish zenith
#

yeh i just wondered if the same thing was gonna happen with the jailer chain spike. like if it performs much better in real keys than sims i mean. i got the 285 scars of fraternal strife and tried that and that thing was actual trash, but admittedly that thing was way harder to time than the axe. probably will just use that as an anduin parse trinket lmao. anyways appreciate the discussion

knotty oriole
#

Yeah Jailer one is honestly even more obnoxious to use in some ways

#

At least since it doesn't currently AoE when it expires or the target dies

#

But in a huge pull I'm sure it still has value

chilly shard
#

Does anyone know why ghostly strike doesn't show up on WCL

#

At least on debuffs?

knotty oriole
#

Debuff is flagged in the spell data not to be logged unfortunately

chilly shard
#

so you have to extrapolate uptime from just the cast times? sadge

solar swift
#

is there a condition in the sim to use the last (5) scars of fraternal strife only at the end (ttd <=30-35sec)?

knotty oriole
tired root
#

would this work to sim the kyrian hunter buff from resonating arrow? # external_buffs.Pact_of_the_Soulstalkers=0/50/100/150/200/250

regal agate
#

This is the timing i use for my sims of pact:
external_buffs.pact_of_the_soulstalkers="0/56/112/168/224/280/336/392/448"

tired root
#

thanks

rigid thicket
#

ok so this is a bit of a hot take here: what if (with master assassin talented) you dont immediately envenom after vanish (during vendetta with every dot running) but sit in stealth for 2-3 seconds and let the +50% crit rate hammer and only come back into the fight when your bleeds are about to drop off?

unknown.png
remote granite
#

you'd be missing out on melees + poison procs as a result of those melees

spiral light
#

This has come up multiple times in the past and it has never shown gains, even on multiple targets. The reason everyone is chasing the Signet of Melandrus is because of how good it is on assassination and how much damage our melee does, and you standing in the vanish would stop your character from auto-attacking completely and causing you to lose a lot of damage from the melee and poison, just like Odowla said. It also causes you to cap energy, which is a massive waste as well.

It is not a gain and not recommended whatsoever.

knotty oriole
#

@willow knoll @round canyon @finite sparrow don't suppose any of you managed to test/summarize the various results of the state of the set bonus/Pistol Shot interaction changes? I can test myself some more but if you guys have done some testing already it would be very helpful. Looking to get this in sims today if possible.

willow knoll
#

Haven't checked really no

#

Solo might have but I'm not great at changing actions in sims

knotty oriole
#

Mostly just curious what the interactions with the buffs and procs and what's expiring/benefitting from what.

finite sparrow
#

I haven't rechecked everything that was previously established

#

but i know the free pistol shot from the 2pc:

  • consumes opportunity
  • doesnt benefit from the damage bonus of opportunity (or quick draw)
  • still gives +1 cp with opportunity+quick draw
  • still doesnt consume greenskins or blunderbuss (believe loktark checked this too)
  • still doesnt fire the 4pc between the eyes
  • still benefits from the cp and damage with broadside
  • now properly blade flurries and procs combat potency
#

i haven't checked the timing when opportunity is gained and consumed

#

i.e if you SS double hit, and the first SS hit procs 2pc, does the resulting opportunity still get consumed

#

or is the resulting opportunity gained after the 2pc happens in that instance

#

i do know that it is possible to SS double hit, and the resulting opportunity is immediately consumed, but possibly only if the second SS hit proc'd 2pc

#

or a melee during that animation

round canyon
#

the first hit causing a 2pc proc should be able to consume opportunity since opportunity is gained before the MG proc would go off

solar swift
knotty oriole
#

E.g.

#

actions+=/use_item,name=scars_of_fraternal_strife,if=!buff.scars_of_fraternal_strife_4.up|fight_remains<=30

fervent vector
#

hi, how may I add PI buff for raidbots sim?

regal agate
#

external_buffs.power_infusion="1/121/241"
you can adjust timing for it depending on the spec

knotty oriole
finite sparrow
#

2pc no longer interacts with opportunity (still gives +1cp with quick draw)

knotty oriole
finite sparrow
#

Yeah

#

the only difference between now and patch day is 2pc now correctly flurries + procs potency

fair crane
#

quick question, does anyone know of an APL anywhere for the spec focused more on bleed snapshots with vend?

noble pendant
fair crane
#

thank you, i'm thinking my simcraft must be out of date. my apologies!!

noble pendant
#

you're doing local sims?

fair crane
#

busted... >.> old habits die hard i guess

regal agate
knotty oriole
#

But anyway @strong gorge definitely something I'll keep looking at. I'm curious how much this issue really matters (e.g. if there is a TTD bias towards the Flag target and they die early and you get your 10 stack demise bonus anyway and still get max flag stacks, is it really a notable DPS loss from flag damage itself? Maybe 1-2%?) and it would definitely be the case that your approach of doubling the HP and "diluting" the damage spike would potentially mitigate that specific aspect. But I think it could also cause some other issues so I'm not sure it's the best approach to fixing it or not.

I suppose you could populate a sim with SV Hunters and max out the HP and that could also be potentially interesting but get into some really funny (and even further locally optimized) territory most likely. Unfortunately many other specs have awful support for dynamic fight styles so adding a Warlock is not even on the table. They do basically no damage at all in DR sims which is hilarious.

#

I do agree that likely not much can be derived from this fight style quite yet anyway, which is why we didn't really publish any statements about it. Still very much WIP and in exploratory phase.

#

I will mention @west prism if you happened to have what you felt were more realistic route setups sitting around, that would be really interesting/helpful to have available for testing. I'm mostly just using publicly available routes (since I assume that's what 'most' people will run) and modifying slightly based on anecdotal experience of which pulls end up being dragged together. But something a little more accurate would be interesting for sure.

proud wadi
#

I can send you some routes if you'd like Koji, I'm pretty sure I have a bajillion of them saved in mdt

knotty oriole
strong gorge
#

Back from gym now

#

Is there any way to show how long mobs survive in those pulls in Simc? Because checking things like crippling poison uptime shows pulls being around 20s

knotty oriole
#

Not without hacking stuff in. I'm working on some local changes for this that I'm going to suggest to Jayeasy

strong gorge
#

Could you send me one of your later reports?

knotty oriole
#

Identifying specific enemies and also pull lengths right now is not possible easily

#

Although with some debugging hacked in for my DoS route

# 
    Line 55947: 27.157 Finished Pull 1 in 17.157 seconds
    Line 58553: 61.553 Finished Pull 2 in 24.396 seconds
    Line 61782: 107.836 Finished Pull 3 in 37.283 seconds
    Line 63879: 138.140 Finished Pull 4 in 25.304000000000002 seconds
    Line 72068: 315.631 Finished Pull 5 in 157.491 seconds
    Line 74150: 357.900 Finished Pull 6 in 35.269 seconds
    Line 76549: 410.799 Finished Pull 7 in 43.899 seconds
    Line 77323: 438.366 Finished Pull 8 in 16.567 seconds
    Line 80028: 504.830 Finished Pull 9 in 53.464 seconds
    Line 82090: 561.417 Finished Pull 10 in 44.587 seconds
    Line 85177: 653.880 Finished Pull 11 in 79.46300000000001 seconds
    Line 87957: 732.035 Finished Pull 12 in 78.155 seconds
    Line 90130: 779.600 Finished Pull 13 in 39.565 seconds
    Line 91232: 821.464 Finished Pull 14 in 34.864 seconds
    Line 93808: 871.521 Finished Pull 15 in 42.057 seconds
    Line 96379: 935.636 Finished Pull 16 in 52.115 seconds
    Line 98500: 982.939 Finished Pull 17 in 38.303 seconds
    Line 100685: 1038.000 Finished Pull 18 in 42.061 seconds
    Line 104679: 1156.582 Finished Pull 19 in 104.58200000000001 seconds
    Line 109782: 1304.791 Finished Pull 20 in 137.209 seconds
strong gorge
#

I guess there would be a world where a simc log could be using the same output as ingame logging, allowing upload to WCL

#

Oh, thanks

#

So like, right off the bat, a lot of those feels really short

knotty oriole
#

I dunno if it's really short or not for BiS profile with 20 Tyrannical

strong gorge
#

DoS you said?

knotty oriole
#

But presumably if you were actually in this good of gear doing a 20 Tyrannical you would merge more pulls

#

Although in this route I did actually merge the mobs prior to the DoS boss together

strong gorge
#

Well, might be decent

knotty oriole
#

raid_events+="/pull,pull=05,bloodlust=0,delay=020,relic=urh,enemies="Bladebeak_Hatchling_1":35111|"Territorial_Bladebeak_1":175559|"Territorial_Bladebeak_2":175559|"Territorial_Bladebeak_3":175559|"Bladebeak_Matriarch_1":351118|"RELIC_Urh_Relic":35663|"Weald_Shimmermoth_1":158003|"Weald_Shimmermoth_2":158003|"Weald_Shimmermoth_3":158003|"RELIC_Wo_Relic":35663|"Juvenile_Runestag_1":140447|"Juvenile_Runestag_2":140447|"Runestag_Elderhorn_1":210671|"RELIC_Vy_Relic":35663|RELIC_Urh_Relic":35663|"RELIC_Vy_Relic":35663|"Dealer_Xy'exa":1548959|"RELIC_Wo_Relic":35663"

#

And some others

#

I'd have to check if these HP values are accurate as well, but I assume they should be if they are coming from MDT

strong gorge
#

I think we can assume that yea

knotty oriole
#

I am also talking to Jayeasy about if we can change the hp scaler to be dynamic

#

And have the addon output the max HP

#

and just use the scaler dyanmically

#

atm I think the HP percent scaler only gets used dynamically for Urh

#

Otherwise you have to manually re-scale the HP values to add more actors

strong gorge
#

But checking the quickest here, which has similar pulls. Pull 1 - 30s. 2 - 35s

#

Encounter time on tyra

#

Which is double above

#

But not saying it's wrong

#

It is however quite a bit difference

#

In your pull 1, how much hp does the Risen warlord have?
Around 1,19m?

knotty oriole
#

That's about a 22s first pull

strong gorge
#

ah I just looked here

knotty oriole
#

The pull counts I think count tanks gathering stuff up

strong gorge
#

Yeah

knotty oriole
#

So might be a bit longer

#

Also seems to count lingering debuffs

strong gorge
#

And checking when relic dies is a few sec into fight

knotty oriole
strong gorge
#

From personal experience, on bosses I most often get off a full tornado of funnel before relics die, + 2 gcd for storm + flag

knotty oriole
#

Oh wrong graph

strong gorge
#

The longest line is the risen warlord

knotty oriole
#

Yeah I think they just drag it along or whatever

strong gorge
#

Mm so not a good example here

knotty oriole
#

He's got like double the HP of the other mobs

#

But I think the main thing is the start of the pull here

#

Not as much as the end

strong gorge
#

Their setup loses 0 dps by dragging it along

knotty oriole
#

e.g. the "start" of the pull according to WCL

#

Is like 6s before anyone really touches anything

#

So I dunno how relevant that is

strong gorge
#

The first non-tank damage ability from rogue is at 15.312. Monk is just before that. Warrior charges in later that second

knotty oriole
#

Obviously brings down the combat time DPS but not sure if it would be something we needed to reflect in sims

strong gorge
#

Nah dont think that's important

knotty oriole
#

In that pull in DR sim I'm sure the Risen Warlord would be alone at the end the same as their pull

#

Since it's highest HP

strong gorge
#

But what is really important is the time the rogue (or well actor) is actively fighting mobs

knotty oriole
#

I guess some of it would maybe be more interesting to get a live gear profile

#

Instead of using BiS

#

Realistically BiS gear profile is probably like 6-8% better DPS than most of the gear in these keys in 20s

#

Maybe higher

strong gorge
#

In the case of say Venth vs Kyrian. A fight lasting 27s vs 17s is likely 10s worth of haste/versa from 30 stacks of flag

knotty oriole
#

On this one iteration to get the fight lengths that I ran

#

Over longer sims obviously more variability

strong gorge
#

yea

knotty oriole
#

Still hitting 27-30 "most" of the time

#

Which is a lot better than what it was before I made changes

strong gorge
#

Yeah but the buff persists for 12s after

knotty oriole
#

But certainly the possibility there is something odd happening with the TTD predictions

strong gorge
#

Which makes up quite a bit of the flag's worth

knotty oriole
#

I mean I will say looking at logs it seems pretty consistent to only get 4 finishers in any ST/boss Flags so realistically it will probably just depend on how many Black Powder packs there actually are during Flag

#

Which is kinda why I'd like to see more "competitive" routes

#

I honestly think this is a very real problem for Venthyr when playing in lower keys with casual/pug routes.

#

If the tanks pull small it's pretty hard to get good value

strong gorge
#

I feel that for bosses it's rather easy to get 6 finishers due to relics + tornado

#

On the first*

knotty oriole
#

(This is why I have an aversion to using Tornado with pugs as well because half the time I just sit on my hands like wow that's a great... 3 pull you have there)

#

I did check some logs we were discussing like that Mists log and pretty sure he only averaged 4 finishers within Flag buff on every boss. It was usually 12 finishers per 3 flags.

#

Not sure I remember if he was running NS or SF though

strong gorge
#

Yeah I think bosswise it's a large difference

knotty oriole
#

NS is pretty sluggish on bosses yeah

strong gorge
#

But also.. losing 10% damage for the rogue on bosses is like... 5 seconds time loss

knotty oriole
#

But anyway my main focus moving forward on this is getting more informational debugging

#

Maybe some HTML report section that shows the pull lengths and stuff

#

Also would like to improve the logging to show the actual mob names instead of generic recycled names

#

Since it's really difficult to know if what it is doing is correct

strong gorge
#

mm

knotty oriole
#

I know there used to be some TTD prediction issues (since it uses raw HP) and it's possible in some cases it's just targeting the wrong thing

#

But very hard to confirm that

strong gorge
#

yeah..

#

Tbh it's so much easier to analyze logs :P

strong gorge
#

@knotty oriole Can you send one of your sims with your DR adjusted apls?

#

like just any report is fine and I'll extract it

knotty oriole
strong gorge
#

No worries, getting late as well so I probably wouldn't tinker until tomorrow

knotty oriole
#

Ok, so had a chat with Jayeasy and debugged a few issues. Definitely some oddities floating around. Time to die expressions are not reliable with the current code (not his fault, more core behavior. Just nothing in place to override code behavior like DS or HAC does)

#

Have tested a few things and sent them over to him that seem to improve this significantly

#

He and I also talked about sorting the enemies so that highest HP targets are the default target, which is pretty much better across the board for most APLs

#

Which I tested and had a pretty positive improvement for all the specs (in particular Assassination)

#

This is with the local changes (not ready for prime time, still have to run them by him)

#

Kyrian actually gained a lot from this change also so wasn't just a Venthyr issue

#

(Cleaving off the higher HP target probably gives more CDR from Effusive on average)

#

Assassination still running Dusk here just as a test case so not really an optimal M+ profile

#

This increases the CP Spend during Flag:

#

Rather significantly

#

This is what it looks like with Raidbots right now with the same settings:

#

(Ignore the percentages.. the base profile isn't a Rogue lol)

#

(RIP Shadow Priests)

strong gorge
#

Neat job!

#

Yeah TTD has been quite weird, recall before DS trying to emulate a dungeon run and ran into quite some oddities of it not really being reliable esp on lower health mob packs

knotty oriole
#

Yeah just checked in the changes but not sure if they'll make nightly or not. Probably not. Didn't have enough time after raid.

#

TTD calculations could potentially still be improved. It's a bit of a tricky thing between too much volatility and too much generalization

#

DS and HAC waves are all timed rather than specific HP thresholds so that makes them more reliable. But due to some core code, the TTD values were just "random" since this was using recycled actors via the pet spawner.

#

Which was very unhappy for Assassination in particular lol

#

Sorting enemies by HP also just makes the default APL assumptions a bit better.

#

Most default APL logic basically assumes cleaving off a "boss" (thus the highest HP target) and has logic to specifically target lower-lived things, rather than the other way around.

#

So doing it this way just improves support

strong gorge
#

Maybe I'll have some time during travel today, there's some thing I want to try out based on your ideas. Especially related to target prio and hp thresholds

#

But I would still want one of your reports @knotty oriole :P

#

or just the apl you've been using. So it's a similar starting point

spiral light
#

Amazing work!

robust marten
#

anybody do ML and ever considered building an RNN model for simming.

regal agate
knotty oriole
robust marten
#

yea im not even sure how would be best to implement it. like my first idea is some sort of descision classifier that is tested every gcd with what the APL does, and initializing a model that way via supervised learning and then let it run unsupervised using a loss function that would maximize dps to see if it finds any changes in the APL with better loss. i need project for this ML graduate course im in xD

#

what types of models do you guys use for optimizing the APL then?

knotty oriole
# robust marten what types of models do you guys use for optimizing the APL then?

Realistically it's just trial and error mixed with human intuition/understanding of the interactions. There may be some potential applications for ML training of APLs but the concern I think is that what would get spit out might lead to very oddly specialized or overly complex logic that yields marginal gains but reinforces poor training towards "illogical" approaches.

#

But it's still a potentially interesting topic

#

Just a lot of pitfalls

robust marten
#

very true, ill have some more thinks about it, maybe have a foolish go at it. id have to brush up on simc and how the apl works coz outside of using raidbots i dont really touch it much.

regal agate
#

Simulations are also computational expensive, so evaluation is not trivial.

mystic jolt
#

You could possibly take a CFR approach as a starting point, which should at least get you similar results to the current APL performance with enough training - that being said, the state space would be really large, but you can minimize it with a bunch of one-hot encodings for a lot of states, like having a binary representation for n-1 combo points instead of having 1, 2, 3,...5/6 combo points in a given state.

echo matrix
#

Could be interesting perhaps to throw it at a reinforcement learning algorithm as well have current DPS, energy, cp, cds etc as the state and change in DPS as the reward. Then just let it run. Have each 5 minute sim be an episode.

To risk being accused of using a buzz word but its the kinda thing where Deep RL methods could work really well. Train it on current action lists generated from the current APL.

knotty oriole
# echo matrix Could be interesting perhaps to throw it at a reinforcement learning algorithm a...

The challenge with this is that there's a fine line between positive DPS changes and reasonable APL changes right?
Because, to use a specific example, you could essentially end up with things that are technically DPS positive but significantly overcomplicated. Finding small DPS improvements would need to be weighted against increased complexity. So you'd need to find some way of reinforcing that as well. You don't want to add 10 APL lines to gain 0.2%.

echo matrix
#

Could add a negative reward that scales with adding extra lines?

regal agate
#

complexity like this also can limit potential of further optimizations

viral mist
#

Is there a way to sim draenei/LF Draenei/Tauren/HM Tauren now?

noble pendant
#

race=orc already exists in your simc

#

if you want to compare to another race, make a copy in an advanced sim and use any of the following blood_elf, draenai, dwarf, gnome, human, night_elf, orc, tauren, troll, undead, goblin, worgen, pandaren, pandaren_alliance, padaren_horde, void_elf, highmountain_tauren, lightforged_draenei, nightborne, dark_iron_dwarf, maghar_orc, zandalari_troll, kul_tiran, vulpera, mechagnome

#

but if those charts are anything to go off of, I doubt the damage difference it going to matter at all

regal agate
#

You can paste this below your /simc in advanced sim to sim all races.

📎 race.txt
lavish zenith
#

has anyone done a sim w/ scars where u force it to sit at 4stack (+90 each secondary, and not pop final rune)? just curious what the results would be

#

or u could make it pop final rune for the final 30s of the sim i suppose

knotty oriole
#

This appears to be the best approach, but the trinket still doesn't seem that amazing for Rogues in general

regal agate
#

Outlaw buffs spell data overwrites:

override.spell_data=effect.269972.ap_coefficient=0.42162
override.spell_data=effect.283619.ap_coefficient=0.54335
override.spell_data=effect.290715.ap_coefficient=0.54335
override.spell_data=effect.806205.base_value=60```
regal agate
#

changes are in the nightly too, so no need to use the ovwerwrites

finite sparrow
#

couple minmaxes to consider for outlaw

#
  1. using blade rush only for energy or to spread acquired axe. no longer press it on cd in aoe, except at very high target count

3 targets https://www.raidbots.com/simbot/report/exKKYbDWmBqTmSKVNKR9rR
8 targets (probably the limit) https://www.raidbots.com/simbot/report/g273nKKNKGdBCQGpHKmUvo
dslice https://www.raidbots.com/simbot/report/qeoLjuqsQAZ9sSJkNBN5Fd

seems similar across covenant builds, kyrian dust/venthyr blunderbuss/necro greenskins etc

note that the expression buff.acquired_axe_driver.up returns an error if cache is not equipped

Capture2.PNG
#

need to add a target count to the apl action when a proper limit is determined. perhaps
actions.cds+="/blade_rush,if=energy.time_to_max>2&buff.dreadblades.down|energy<=30|spell_targets>=2&buff.acquired_axe_driver.up|spell_targets>7"

knotty oriole
# finite sparrow couple minmaxes to consider for outlaw

Looks good. I tested some variants of the Blunderbuss finishing and it seems solid for both GS and WM builds. Largely it didn't seem to matter if it was just combo_points>=4 or anything more complex but everything seemed neutral so just merging it in seems fine

#

Blade Rush thing makes sense, but I'll see if I can add the axe driver to the fallback buff creation in the special effect so it works properly

lavish zenith
#

does anyone know if the #s are posted anywhere or have the old sims for the exact % dps loss on ST to rupture in dance? rupture in dance + symbols? rupture in symbols only?

lavish zenith
regal agate
knotty oriole
#

@finite sparrow wrt minor optimizations seems worth 0.2% to add SnC to the reroll exclusion list. Possible TB can also be removed but I need to check if that's just a BB vs. Dust thing

finite sparrow
#

SnC for all or just blunderbuss?

knotty oriole
#

Not sure yet, it's small so I'll test it later.

finite sparrow
#

no rush

knotty oriole
#

Was planning on just running a full RtB matrix again since it's been a while

#

Once I do the other stuff

knotty oriole
#

But it's really quite poor on HAC

finite sparrow
#

what's the average target count for HAC? there is some target count where it is no longer a gain (im not sure what it should be, i guessed 8 targets) which needs to be part of the logic

knotty oriole
#

Obviously on sustained there isn't a "rush" to use it so you'll probably get the AoE value either way, so more about optimizing the energy gain when there isn't much to lose. But it still seems worth pressing in AoE when it's a limited window

#

HAC is 5 adds

#

Realistically it's still using BR quite a bit in sustained obviously due to the energy condition

finite sparrow
knotty oriole
#

19.8 times vs. 30.9 times. So basically it's "worth it" to drop 10 uses for 2 energy/sec between the profile in the sustained ones, but the DPET is still more than high enough to be worth using in AoE.

#

It's just more "if you can drop 'only' 1/3rd of the uses but get a lot more energy that's worth"

#

But in HAC if you use those rules you just miss out on a lot of AoE damage that gets replaced by a ST cast later

#

So most likely it needs some form of raid events condition or targets>desired_targets type of deal

finite sparrow
#

i see what you mean

knotty oriole
#

Tricky tuning on this one lol

finite sparrow
#

i still dont get what desired_targets means

knotty oriole
#

It's the static target count

finite sparrow
#

is it not just 1 for both HAC and dslice?

knotty oriole
#

And yes, it's 1 in most default cases

finite sparrow
#

yeah i copied the raw for 4t and it overrode the dslice setting susge

knotty oriole
#

Strange

finite sparrow
#

i guess im confused how spell_targets.blade_flurry>desired_targets is not identical to the default logic already (use on cd on 2+ targets)

#

for dslice (or hac)

knotty oriole
#

For HAC it will be true when the adds are spawned and not when they are not

#

Regardless of base target count

#

e.g. if you did HAC with 3 static targets it would still only use it when the adds were up for that condition

#

Ok I might have something

#

Just have to run some slow high precision checks lol

knotty oriole
#

Eh, some very odd timing stuff going on here. I'll need to dig into it more tomorrow.

finite sparrow
#

i knew it would be a weird one

knotty oriole
#

Have something that works pretty well for existing fight styles but I think it's mostly just kinda "hard-coded" in a way that the time check seems to just be conveniently aligned with the HAC stuff and doesn't interfere with DS. Which I guess is technically fine but it bothers me since it's not really proper. 😛

#

actions.cds+="/blade_rush,if=variable.blade_flurry_sync&(energy.time_to_max>2&buff.dreadblades.down|energy<=30|raid_event.adds.up&spell_targets>1&raid_event.adds.in>20|spell_targets>=2&buff.acquired_axe_driver.up)"

Is the line that works pretty well but I'm not really happy with it since it's just coincidental imo. Need some way of restricting the energy dumps when raid events are upcoming or something checking duration vs. next or something.. idk. But it would work fine I guess in a pinch.

regal agate
regal agate