#tc-subtlety

1 messages · Page 14 of 1

valid robin
#

Because ”people don’t want to play 1 dance”

#

It was a guy that’s quit the game i think, that made the apl work for it.

rugged solar
#

but i remember testing it in apl, which showed a gain, i don't think there was a big resistence

#

unless uncertenty of results

valid robin
#

Well i know for a fact that it was a regular community member that did the apl work for it

#

And not a ”TC” per se

rugged solar
#

possible

#

again, i don't remember any ressistence to testing things, if anything i tested a lot of things that ended up not worth it over the years

#

if there was any ressistence, i am sorry, it was likely not intentionally but either misscommunication or just not interpreting too much into a more casual discussion

turbid pulsar
rugged solar
#

chuffa and smoker both occacionally helped

#

both are rly nice people

#

we had quite some short term or one time contributors over the years

valid robin
rugged solar
#

e.g. idk i had to surgically go over one of the russian apl's in s3 tww

valid robin
#

I know joey helped me find some weird ass gain at least

rugged solar
#

oh, i completely forgot joey

#

Neverless, i can just iterate, if there is something in need of testing or if there is some uncertenty, just ask

#

many things are quickly checked, and often help get a better general picture of the game

#

(i sadly can't always respond/sim immidiatly, but i usually come back at it as soon as i can make time)

valid robin
#

I think this time it was a case of needing to rewrite a decent chunk of the apl and doing just minor adjustments didn’t help. And chuffa, i think, spent a lot of time testing apl things. Usually ”main tcs” are so busy with other shit that they just test 1-2 things and then move on. If that happened back then, it would have never been mainstream either. So that’s why we need people that dedicate time to weird shit, because sometimes you need to make bigger changes..

#

And saying, to Lament, (without properly reading what he was saying) that it’s just bad and he should move on. Triggers me.

turbid pulsar
valid robin
#

Joey helped me a lot with cleave scenarios. I think he made a custom kurog sim

#

Mostly

#

(If i look at our dms. Sadly cant look at the sims he did anymore)

rugged solar
valid robin
#

Ye it’s not about the specific idea. Many ideas are shit.

#

But in science you don’t settle for a certain truth

#

And iteration is important

rugged solar
#

i never did

#

e.g. in shadowlands, somone came in the sub chat and asked why we used blades pre pull

#

and i was like "wait, do we need to", and it ended up better to not do so because it was a hasted gcd

#

apl and gameplay is like academia/science, there is no ultimate trueth and things might change quickly

valid robin
#

Yea there are bugs, unforeseen interactions etc

rugged solar
#

exactly

#

and they often get discovered in logs accidentally

#

or by players who are not in discord

#

e.g. remember snake eyes

turbid pulsar
#

fwiw eleem does spend quite alot of time testing talents

rugged solar
#

which showed up very well in a log early on, but after implementation and optimization was fairly mediocre

turbid pulsar
#

ingame

rugged solar
#

i usually do too

valid robin
turbid pulsar
#

i used to live at dummies during beta/ptr cycles

valid robin
#

Same

turbid pulsar
#

now not so much

valid robin
#

And dk too

#

Had a LOT of nerds that did testing. I could spend an entire day and not eat, just hitting a dummy and deep diving in logs

rugged solar
#

@odd shore unsure if you follow this channel, but ill ping you for the conversation we had earlier

#

It seems like the ancient arts optimization stays fine even on higher target counts

#

e.g. here is 6,7 and 8 target simulations

#

so there is no need to do it target specific

odd shore
#

I see

rugged solar
#

let me quick check with dungeon talents too

turbid pulsar
valid robin
#

Ye that was interesting

rugged solar
#

i spend way too much time testing interactions in df beta

turbid pulsar
#

and you making me test DF season 2 sepsis shadowdust

#

ill never forgive that

rugged solar
#

we got the entire secret cold blood crit stuff from it tho

#

XD

valid robin
#

Fucking sepsis man

low breach
turbid pulsar
#

come a long way, and now i might not even be playing rogue next season it is what it is

low breach
#

So trigger AA at higher target counts then 1?

rugged solar
rugged solar
#

which seems to be not the case

#

which might be even a valuable point

#

tho the impact is still extremely minor

#

keeping it as an optional opt-in

#

tho the testing was good

#

means the condition can stay as simple as is

ivory kestrel
#

basically to apex coup right

low breach
rugged solar
#

only during bloodlust

#

wait nvm

#

i am stupid

low breach
#

Ok so no bl no matter what targets no AA optimization

rugged solar
#

i showed the opposite

low breach
#

?

rugged solar
#

the "test" profiles are without the apex

#

so you should stop using it when you don't use strike

#

so 3+ targets

#

@odd shore sorry, correction ^

low breach
#

So tldr: strike during lust only to trigger AA up to 3 targets?

rugged solar
#

well technically 4

low breach
#

Ok 4

rugged solar
#

because its still better on 3

#

diffrence is just very small

#

so intuitively, would just use it on st

#

unless you want to do prio

low breach
#

I just want most optimization because sub at baseline does not have that much anymore tbh

rugged solar
#

so it becomes a prio damage option in aoe

image.png
#

Okay, to not confuse anyone reading this is basically the conclusion:

  • Doing the Apex is a minor improvement on Single target
  • Once you have 4 or more targets, it becomes a small loss (not very noticable)
  • There is a Roughly 2-3% single target (priority target) gain to use it with more targets.
    Note: The use is during bloodlust.
#

to investigate it even further

image.png
#

doing the AA optimization is always a priority damage gain with more targets (8 targets), not only during BL

#

this opens a bit of room to optimize around it

low breach
rugged solar
#

if we need more damage on a priority

odd shore
#

I was also mainly referring to the wording used in #subtlety-faq being misleading

rugged solar
#

its a recent discovery

#

so its fairly fresh

ivory kestrel
#

you can also strike once with premed most of the time i think if there are not a lot of targets with little loss of overall

#

as long as it is not first global on dance that being a bp or sec tech

#

for a bit more prio

crimson vault
#

https://www.raidbots.com/simbot/report/aeL2TGTa85B3dBj24Zah1i
https://www.raidbots.com/simbot/report/voHYy8JYM1Ak2n5tW1PVCh
Some initial test result on VERY high haste WITH TRICKESTER:
APL:

  1. is current APL fitting VERY high haste? No, there will be intolerable delay and waste of cds.
  2. is the posted link APL a better one? Yes, to a degree, at least no visible wrong doing regarding cds and casts.
    2.1 is the posted link APL already compatible to current APL? Not really, with haste decrease, it does not reach the same heighth of max dps, but quite there. Certain added lines needs to be removed.

Haste
3. what bring us by VERY high haste? we have more SecTech that cannot be paired with dance and many more instances of clones
4. Is VERY high haste providing further potential? No, current item level does not support any play of VERY high haste based on the experiment.
5. Do you choose 1314 haste rating for a specific reason? Yes, 1314 haste rating providing a natural 'naked' SecTech while keeping the other ones paired with dance without any delay to blades, at least after bloodlust.
6. Have you tried other talent selection after crit exchanged for haste? Yes, tried rotten, no good. also tried brew, no good either.
7. Have you tried exchange mastery for crit based on initial experiment? Yes, DPS increased but not impressive.
8. Have you tried to use storm for CB when crit rating is being set low? Yes, it shows dps gain but the base is too low

APL change:

  1. what is needed for fitting VERY high haste in APL? Current APL does not allow SecTech to be cast without dance which either leads to quite close to blades cast of dance and SecTech at for example 1:08 delaying blades cast OR immediate cast of dance and SecTech without blade when three cooldowns wedge each other, in VERY high haste setting.

Verdict:
Low priority unless we're talking about huge secondary stat gear item level, the similar level we had in 11.25 season.

Raidbots
Advanced Sim - Subtlety - 93,173 DPS - Raidbots
Raidbots
Advanced Sim - Subtlety - 92,798 DPS - Raidbots
valid robin
#

Can you do a breakdown of the benefits of high haste? What stands out. Is it only sec tech?

#

When i first heard about the new iteration of sub i thought haste would be good because.. everything is in dance and haste improves dance uptime. But turns out that was overrated. Still curious if secondary stuff like passive dmg etc getting a boost due to somewhat higher dance uptime, aka more haste, more dmg% increase and so on

crimson vault
#

To your question:

  1. it is quite a load of work to explain or truly break down the benefits of high haste depending on what you're looking for, i.e. the level of break down.
  2. Right now there could be a potential implementation error or something wrong, which leads to a 10gcd dance needs 1800 haste rather than 1600+ range haste, so The posted sim log you can see have only 9gcd with trinket.
  3. To break a gcd threshold would require a lot of secondary stat, which we do not have. So for example, I believe 9gcd normal dance (without trinket) would need 1000+ and that's 400+ stat = ~10%(?) crit or quite a lot of mastery.
  4. we do not have meaningful talent support us to do so like clone damage specific related crit etc, we do have clone damage % talent but those are already selected.
  5. The original dance %buff has been break down to pieces so we are still much stronger in dance but FW, blades took shares from those so plain +1gcd dance not really improve much, to a degree.
  6. We're now almost forced to use "Finality" that pairs with crit, if we try to reach VERY high haste, we'll sacrifice a certain proportion of it, if not mastery.
  7. For a simple concept for breakdown: Evis drop but Evis clone up (crit down, more apex) SecTech instances up but total damage down (same reason) So basically we're reducing base damage for clone that actually require high base to shine.
  8. Since dance is now a fixed cd, the percentage of relative dance uptime is always shrinking: 6->8 we have uptime 33% up from 30% to 40%; 8->10 we have uptime 25% up from 40% to 50%, while we need to invest a huge amount of secondary stat for that.
#

one sentence conclusion: we drop around 15% general crit for a huge amount of haste that we do not have amplifier talents or mechanisms for it.

valid robin
#

Hm. And the bot is ”spamming” with this amount of haste? Have you checked for energy waste

#

Bit confused as to how total sectech dmg drops when you use it more. Is crit that big for sec tech?

#

I understand each instance of sectech is lower. But surely overall sectech is up?

crimson vault
#

for SecTech, this can be elaborated in explaining to you since I checked it specifically.
Total cast increase is not that many, this is the first conecept you should have , like every 90s you probably have one more or almost one more or every 120s you have one mroe ish.
The increased cast of SecTech is not paired with dance, since all dances are already booked for its own SecTech.
And dance numbers are fixed, so the picture is
we have our base SecTech damage from dance-paired SecTech that have lower damage due to crit, while we have more SecTech not paired with dance but the amount of these casts are limited.

rugged solar
crimson vault
#

And for this limited increase of SecTech cast, I can explain a bit further here:
SecTech CDR right now enjoy %haste buff that dance does not.
This means SecTech will have a CDR boost from bloodlust but it only happens once, other things like berserker from troll.
and we have dance every 20s while we have blades 90s
So in sim, we'll face a dilemma that we might have one more cast of SecTech based on the total haste when we cast SecTech and the window of balde (its specific 90s) would restrict cast of SecTech due to the fact we need all cds in blades. So 1314 might be chosen at a very subjective manner but the truth is we cannot actually have an ideal one more SecTech every 90s and also NOT wasting anything, but 1314 made it looking possible for my own tolerance.
So if you click into it and have a look, you might see SecTech casts only increased by 2 for 5min, that's due to this irreconcilable nature of three cds.

valid robin
#

Interesting. What talents are you using for this?

#

Both rogue and sub version

crimson vault
#

The same one with current profile.

valid robin
#

It’s all weird

#

I’ll have to look more at the sims later

crimson vault
#

Yes, please. I might be very prejudiced with my own work after hours on it and cannot spot wrong doing, if you find anything please let me know.

valid robin
#

I mostly meant the numbers in the report

crimson vault
#

Tbh, current work load for sub is very asymmetric due to we're not that familiar with SecTech cd at a certain haste with which buff. So to spot a delay of cd or unnecesary waiting would require heavy calculation on each ability cycle and cd. So if you spot anything that would be great.

crimson vault
valid robin
#

Gotta remember there could be simc bugs as well. But I’m not home so i can’t look at it in depth

crimson vault
ivory kestrel
#

this is similar to what i kinda found with 2 haste on uses

#

it might be a thing on some fights because we have limited time to do dmg but nothing crazy overall

#

kinda as how dust worked

#

like if boss phases at like 2:30 or 2:20 then you have a whatever phase

#

or if there is an amp lasting a long while

cinder root
#

Cheerleader = recognizing other people are working on the APL(?). It’s ignorant to say a single person is doing all the heavy lifting for optimizing a spec in regards to what isn’t deemed meta. I don’t think the appropriate response is to further disregard other tcs (just like Lament) without knowing what people are cooking. I won’t fill the chat with more of this, but just wanted you to recognize what you wrote/are writing

valid robin
crimson vault
rugged solar
#

The economy of 1 dance was not 100% clear

valid robin
cinder root
valid robin
#

It went on for the entire time one dance was good. People came to the discord, very confused why 1 dance was better.

rugged solar
valid robin
#

Not only

rugged solar
#

the secondary thing was just the damage diffrence between dance and outside dance was not high

valid robin
#

Sec tech was really good and you could only have 1 dance with sec tech.

#

So the second dance was kinda empty

#

And dance and sec tech always lined up

#
  • flag duration etc
#

Tons of reasons why 1 dance was better

crimson vault
# rugged solar The economy of 1 dance was not 100% clear

for the record, not really meaningful anymore here, I said things about Seal Fate and other bottom talents' resource gain compared to one dance charge over there.
and months later, you replied me as "prophecy has been fulfilled".

valid robin
#

Bottom line is. Even if tc isn’t up to date I would “know” what was better to play. Simply due to the sheer amount of testing

#

Lots of times when I just ignored the sim cus it made no sense

crimson vault
# valid robin Bottom line is. Even if tc isn’t up to date I would “know” what was better to pl...

Indeed, and that's some very real experience for sub, you can really obtain very good in-game practical experience and form a theory before it can be made to sim. and this is due to the nature of asymmetric of work load between try something in game and work the sim to do the right thing for sub then and even now since Sub has never been designed to be "naturally aligned" spec which leads to challenges in sim APL implementation.

valid robin
#

We’ve had this discussion before. And sometimes it’s hard to translate what you do ingame into a sim. Especially true for old sub with all its cdr shenanigans

rugged solar
#

it often also depends on the complexity of the problem, not everything is trivial to implement.
Additionally, also depends on how well the apl is maintained in terms of simplicity.

crimson vault
#

Yes, even now when I look at my high haste APL, I still can find something that's not right for the cast sequence, and that would be difficult to tell sim to do SecTech here so everything will then be only delayed for some small seconds

valid robin
#

I know pxb/dk was getting massive logs playing stuff he just found out by himself and playing the game. Like the sylvanas raid, he was one of few to play with unholy gem setup, because he was playing around the proc. Which was a loss in simc but he had r1 by a lot doing that

valid robin
rugged solar
#

Its an interessting problem space

#

the apl needs to be carefully maintained, then more complexity you add then harder it gets to find optimizations

turbid pulsar
valid robin
analog island
#

Can we get back to TC? This is a shit show to sort through.

rugged solar
valid robin
turbid pulsar
valid robin
#

Oh ye

rugged solar
#

some other specs or classes had some rly hard to work with apl's

valid robin
#

Fury warrior

rugged solar
#

dh apl is a usual offender

valid robin
#

🤣

valid robin
rugged solar
#

i won't talk about rogue apl's to not upset anyone from our colleges

analog island
#

Yeah and it’s burried with your all having a dick measuring contest. Take it do sub channel or DMs

rugged solar
#

for subtlety, i re-wrote the apl in tww to make it easier maintainable and extendable

#

the current apl is even simpler, just because of how much you could throw out

#

e.g. one example is thistle tea

#

which was a fine balance between energy and mastery before

#

is now entirely removed from the apl, because the optimal way to use it is to let it passively proc

valid robin
#

So not sure why ur mad about it

#

Like the lament thing with haste. Probably requires more background changes

analog island
#

Trying to find a simple answer in here and couldn’t and had to dm fuu for it because you’re all just bullshitting about nothing at this point of who did what.

crimson vault
#

BTW, if we heading back to TC, I think high haste APL update from my post, could be included to APL after adjustment and refinement so all gear sim wont miss out certain level of haste, if not as high as what I've experimented.

rugged solar
rugged solar
crimson vault
#

Later tonight or tomorrow morning, I think I'll have a look at this diff so give you a list of what has been done so this current low priority thing can be done faster later in its intergration as well as its refinement or adjustment.

rugged solar
#

all fine, i can look at it

past fable
#

And it isnt negative 1-3% depending on talents setups/situations

rugged solar
#

ah nice

crimson vault
#

nice

limber terrace
#

Im trying to absorb tc via osmosis here

crimson vault
# rugged solar all fine, i can look at it

some line was used for further dps gain just for that haste I believe, for example the extra apex strike line or the apex strike line change and potentially other ones.

past fable
limber terrace
#

Ah okay so what lament came to a conclusion of, makes sense

crimson vault
#

What Eleem said to that is 99% true but I found cases we might want to see which SecTech between two blades we'd not pair it with a dance so next blade or all SecTech wont get delayed or wasted due to not delay.
I found one case in a sample that before 4th blade, at certain haste it would show a missing Sectech where if cast blade wont get delayed too much. So the adjustment to apl would be subtle so it can pick it up while not affect other situations.

valid robin
#

I haven’t played a lot with the new iteration. But is it possible to increase the amount of apex procs with haste, does it make sense to try line up the out of dance sec tech with an apex proc? Or is that already in there

crimson vault
#

however, we might want to adjust our sht or backstab pace to fit an apex for the coupe following the naked SecTech

valid robin
#

Ye i was mostly looking at energy waste, wait time and casts

crimson vault
#

Energy comparison shows high haste overall waste is above normal haste but not that much.

limber terrace
brave prawn
crimson vault
# limber terrace I was just thinking this, fuu already had a soft pooling optimization in the nor...

that was for a different reason. Initially when I checked the APL I saw unnecessary delay of dance and SecTech due to our new energy requriement of dance and the waiting before dance after good cp.
and that's also there is a condition for bottom finish evis as well as Coup having a condition to not evis when the time is near dance. and surely fuu made them nice and clearn than my first private version.

brave prawn
#

Certainly a gain if you have sectec sht to make it happen quickly

crimson vault
#

sorry for my language if it sounds weird for you.

rough epoch
rugged solar
#

yes, but there might be more apex related optimizations possible

#

the current bloodlust optimization is very minor in impact

#

so ignorable

crimson vault
#

what's Coup apex's in-game damage delay and in-sim delay?

past fable
#

All apex clones have a 0.5 second delay in both

crimson vault
#

great, with high haste, we might have a new play.
Previously I tried to mute consecutive strike where the 2nd is forcing apex when condition matches.
However, we might want to get that in use in high haste for non-dance coup or even all coup if energy allows, or even not with high haste support (we might need some cd condition to this).
I'll look into it and check some initial data if that's meaningful in any sense.

oak crystal
#

Hey everyone, apologies if this has already been discussed, too many msg)
In the current APL, we’re looking at 623 haste https://www.raidbots.com/simbot/report/6TTgEL21LdQ1twbFnqCiqU. That’s roughly a 0.27s margin. Once you factor in coup, it leaves only a 0.07s margin, which is basically unrealistic for the vast majority of players. Should we consider adding a "human factor" offset, similar to how it was handled back in Shadowlands with Animacharged CP?

For instance, we could implement a global/class variable like haste_margin=x (defaulting to 0.2, or 0.4 for Trickster). This would either subtract haste from the base value or forcibly truncate the Shadow Dance duration to make the sim more realistic

crimson vault
# brave prawn Certainly a gain if you have sectec sht to make it happen quickly

I took a 2nd thought on this. I think it is quite subtle in implementation: from practice in game with normal haste (~650), we need to pick up specific Coup to force apex so 1. it is not too close to next dance so extra waiting would be required for cp and energy; 2. we should not wait too long for sht so that a potential reset of UB is then likely wasted from the backstab triggering the AA. I haven't started the coding since I think we might want to downgrade "in-general Coup-apex" in priority list since if we look into the VERY high Haste profile and the difference between Coup counts and Coup apex counts are 25%, which means that could be a) a potential good gain; and b) we might want to see if we can get in-dance Coup apex running before we get into out-of-dance Coups.

coarse laurel
cinder root
oak crystal
#

indirectly this also leads to a bad sim for first dance

rugged solar
#

if you want to emulate a higher input delay, just increase the value from the default

#

it does not lead to bad simulations

oak crystal
#

I know what a world_lag is and how to sim it. I'm just suggesting that we forcibly add a more realistic sim for people. No one can play all the gcds in a dance with a margin of 0.05-0.07

analog island
#

We’re already suggesting a higher haste value to allow for input delays/lag.

crimson vault
rugged solar
#

unsure what else you would want

#

if you increase the world lag, you increase the input latency

#

spell queue does help outside of the first input in dance

#

the problem is also in design to a degree

#

the entire problem space was created by the change of deepening shadows

#

there is an easy solution for blizz to fix it

#

which would be to always round up to the next full second

#

this would give enough extra time to not have such a high impact on player input lag and user error

crimson vault
#

and for reality, if a player needs 7.6s dance to apply 8gcds, he then should use his gear to sim. I dont think that for Sim to add a delay manually and then tell the player what haste he should have.

rugged solar
#

it is an inetresting problem space

#

because you can use dance off gcd too

#

this means there are multiple ways players can fuck up

crimson vault
rugged solar
#

yes

#

guides do mention this problematic

#

there is sadly only so much we can do to bring awareness

oak crystal
rugged solar
#

the sim already includes some amount of lag

crimson vault
rugged solar
#

so what the sim can do, you can do in game

oak crystal
#

Yes, I can tell people: don't trust the sim, you need more haste. But then why sim?

rugged solar
#

i think what you try to argue is just how much input lag the default should be

#

not related to sub or haste

#

because thats what will make or break your assumption

crimson vault
#

that's not "don't trust the sim". that's a guideline for your own play. if a player knows he needs 655 haste, he will get this and then follow the stat distribution rather than saying Sim is not trust worthy.

oak crystal
#

Okay, it's just a suggestion, it's up to you 🙂 I'll just text people as much as they need

rugged solar
#

the best way to have a higher input lag for everyone would be to ask on raidbots to change the defaults

#

or ask in the simc github

#

because the problematic mentioned applies to any class with hasted gcd's or cooldowns

#

and if your assumption is, that the sim gets hugely falsified by not representing the game well

#

its better to actually make things "right" for everyone. Keep in mind, you will need to bring good evidence/data to supprot your claims tho

crimson vault
#

I think it is always to get the most max dps from gears you own rather than specifically target a whole set. there are so many different gears but all in four secondary stats and people already know what to do for their own max dps and that is Sim himself. I dont know why you'd bother to make a comment on if for example fuu's list Sim is trustworthy or not.

#

in addition, you might feel you've made a good suggestion. However, what you actually suggest TC to do? add 200ms somewhere? or manually increase the haste? what if the increased haste still below your people's need?

rugged solar
#

just to re-inetrate what i am saying, simulationcraft has a lot of lag considerations already, as you see there.
We additionally even added lag to the first input in dance for shadowstrike given the stance swap problem.
You can find a full breakdown -> Here

image.png
crimson vault
#

The problem you mentioned cannot be solved universally by one sigle action to the sim. That's why it is a very player-based selection or choice. It cannot be done by giving a standard since everyone has his/her own standard.

rugged solar
#

i understand what you are saying, that some people might be slow to react, use dance during a gcd, etc. But its unclear just how much this should apply to everyone.

#

maybe there is a way to measure this from wcl

#

so there is the possibility to write a 3rd party tool to adjust lag settings in the sim for your personal latency profile

oak crystal
rugged solar
#

the lag is actually very close to your latency

#

it is lower than 0.2

oak crystal
#

u can add -0.1/0.2 its still better than sim with top gear with 613 haste

rugged solar
#

this lag was added to the sim, so we already simulate it on top of the lag variables above

oak crystal
#

I have my own addon and I also checked it on the stream (with ms delay between casts like on screen)

image.png
rugged solar
#

^

#

and adjusted in this commit

rugged solar
coarse laurel
#

fwiw world_lag I think defaults to 100ms in SimC. Previously SimC would effectively allow you to "macro" the instant Shadowstrike with no stance lag, so this matches up with in-game stance delay compared to how sims used to work. That's probably all I'm super comfortable putting in the module unless it's proven out as a bigger delay in-game. But previous testing seemed to be that it looked like typical lag.

#

Fitting things into buff windows will always be an issue for optimal play, so world_lag is definitely more applicable to just Dance, although I get that was what was being looked at specifically here. GCDs within Metamorphosis for DHs was a similar issue when it was found that Eye Beam had channel lag. But then people figured out you could use a nochannel macro to get it down to normal world_lag, so there were some discussions about which should be modeled. And for the purpose of sims and trying to figure out optimal rotations, went with the world_lag version.

#

This type of stance lag is not really anything we can do anything about though afaik right now so it should be fairly "static" under typical conditions

#

We also do model the higher GCD on CdG

rugged solar
low breach
#

We are now simming this high how?

rugged solar
#

only if the buffs include everything

#

so, unlikely

rugged solar
#

so basically removing the highlighted lines

actions.cds+=/shadow_blades,if=variable.shd_cp&cooldown.shadow_dance.ready&(cooldown.secret_technique.ready|talent.deathstalkers_mark)|fight_remains<=10
low breach
rugged solar
#

the talent check basically means the option before that will get ignored

cinder root
#

(deficit 2 instead of 3, and energy 50 instead of 40)

obsidian ibex
#

110 soon prayge

rugged solar
#

which is good

#

the change makes the apl simpler

#

and increases the result

#

@obsidian ibex what was your question in general

#

you wanted to check if evis for snd was relevant in aoe?

#

or did i missunderstand

obsidian ibex
#

its a bit complicated so let me apologize beforehand

#

i put a check in the blades condition that the previous cast was shuriken storm. since the st profile would cast eviscerate before using dance (dont know why), it would begin dance at 0 cp. when i ran the sim at 8 targets, no finisher is used after the first shadow strike cast, and thus the combo points are at 7, so the shuriken storm is never cast, and then secret technique goes on cd before the blades condition is met, which caused blades to get pushed back by one dance. i was wondering if it would be good to force it to use evis as the first finisher like it seems to do in st, which would also solve this issue (but i am also not sure if using a previous cast check for blades is the most elegant solution to begin with)

rugged solar
rugged solar
rugged solar
crimson vault
# rugged solar

I think we might want to check a bit of "backstab" condition:
In trickster, we set a condition for Evis so it does not delay blades due to Trickster cp requirement.
This changes to DS as DS would love a low cp to enter blades.
The sim posted shows at least 8s of unnecessary of blade delay, or safely say at least 6s blade delay for my late time checking and potentially miscalculation.
However, backstab is different from Evis due to its filler nature, so we might want to loosen blade condition on cp, maybe, given a hard time threshold "stopping" backstab + cp condition would mess up with each other.

#

I have no solid confidence on this, but might worth trying considering the workload I pressume. This might not have a particular dps increase in practice however might be some gain in Sim due to random fight time and casts of blades.

#

if you guys busy, I might have a try after sleep +6.5 hours.

rugged solar
crimson vault
#

0.3% is solid!

rugged solar
#

Thats just a quick hack tho

#

i need to look more into it

#

maybe it would make sense to wait for the patch

#

to have the correct spell data too

rugged solar
#

Just a super quick test i did, will probably need to look more into it.

#

Quick Test for BP/Strike on 3/2 targets:

coupdegrace Trickster:

⭐Conclusion:
With Single Target focused setup (and only switching between potent powder and veiltouch/finality):

  • It is better to use strike on 3 targets
  • It is better to NOT use BP on 2 targets on Trickster without Potent Powder
  • It is better to use BP on 2+ Targets on Deathstalker

🫡 Other findings:
Trickster:

  • Potent Powder seems to be only worth it on 3+ targets over VT.
    Deathstalker:
  • Potent Powder seems to be worth with 2+ Targets over Finality.
    Deathstalker seem slightly better in sustained cleave on two targets.
    Trickster seems slightly better on cleave with three targets.

Note: Only quick check, so not super in detailed comparison (maybe some other talent combinations work better)

ivory kestrel
#

If it is not much of a bother with how much mastery that changes if possible or if it stays always like that

rugged solar
#

def. needs more testing, the dungeon profile uses a lot more mastery, so could be a good reference

brave prawn
#

it's a 2 target sim and the line is being changed to only bp at 3+

rugged solar
#

yes

#

bp on 3+ means not using it on 2 targets

brave prawn
#

your conclusion line says "worse to use BP on 2+ targets" though

#

it's a 13% gain on ds to BP on 2 targets over eviscerating (with PP, and still a gain without), so it's better to use BP on 2+ on DS

rugged solar
#

ohh

#

my mistake, typo

#

fixed

brave prawn
mellow quartz
#

I'm probably super overthinking this but would doing something like adding Mana Divining Stone to the mix for dance haste snapshotting work? Arcane mage did something like this previously https://wago.io/ABarrFakeQueue - lil background info towards the bottom of the description

Arcane Barrage Fake Queue

Overview This WA try to mimic normal queuing by freezing your client until the spell should have finished casting. T

rugged solar
#

Just wanted to share this too, also needs more research

#

The Deathstalkers mark optimization seems to hold up a bit longer; using it on 4 targets seems like a gain still.
Simulation

image.png
crimson vault
#

is this AA on 4 means Apex Strike on 4?

rugged solar
#

yes, it means using the apex optimization with strike on 4 targets

crimson vault
#

Some things worth mention (or not):

  1. Simc would skip Strike line in builder sector in dance but execute Backstab line, not sure it would be a simc problem or an APL problem (fuu considered it some check before strike, which already fixed after patch update)
  2. Improved SecTech talent increases SecTech's shadow damage by 30% in Simc rather than an in-general 15% (which increases its value compared to other candidate if we're looking for some special build for special task like killing the egg)
  3. An experimental sim has been conducted to search for special task of killing the egg: It seems forcing Apex Strike can provide some gains (not restricted to supercharged combopoint)
    link A: https://www.raidbots.com/simbot/report/p2FRjqJBtUZcYSPJDZRAUb
    link B: https://www.raidbots.com/simbot/report/9XAoMT7sP8DyXgKRJXH3f4
    Comment on lines: just some minor changes to the Apex Strike line and nothing else.
    These does not provide overall gain for ST 5mins, although there could be further optimisation if dig deeper and not sure if worth the time and energy.
  4. For special task like killing the egg, there is an interesting and extremely experimental sim: https://www.raidbots.com/simbot/report/tGbePUGnuH3L6poz5QzdjT
    This build does not have Apex3 and its gear has been manipulated for secondary stat, so it cannot be considered a practical play but only theo work.
    This build's stat has been manipulated for crit and vers to facilitate the use of Strike as well as potion of recklessness (Crit)
    This build always try to cast Strike as long as it has energy for a strike following by an evis.
    This build shows lower but close number in a 35s full buff fight compared to current BIS.
    This build provides a close alternative (and very difficult for gear since its stat has been manipulated) with a much higher Crit in a fixed potion window so it would be more stable if compared to a normal build with ~45% crit.
    And might at the cost of a lower 2nd egg phase damage.
Raidbots
Advanced Sim - Subtlety - 252,069 DPS - Raidbots
Raidbots
Advanced Sim - Subtlety - 252,103 DPS - Raidbots
Raidbots
Advanced Sim - Subtlety - 247,140 DPS - Raidbots
past fable
#

Yeah as you said the first point is already adressed and will be default on raidbots once the ptr changes get applied (which should be when NA servers come up afaik).

#

Improved Secret Technique actually buffs it by 30% ingame, so thats why it works that way in simc.

crimson vault
coarse laurel
#

You can’t entirely rely on the sample sequence for the expiration of the buff because it’s delayed by one tick due to how consume resource works

#

Have to look at the log output

past fable
coarse laurel
ivory kestrel
#

you need to have at least5 sht after the builder fills cp to get the refund on the next finisher

coarse laurel
#

Right but what if you already get the buff, and then the ShT stacks drop from using a non-damaging finisher and then use another builder and drop below 5? (I know this is an edge case, I'm just curious)