KQM | Genshin Impact
cause-of-the-extra-absorption-hit-on-amcs-hold-skill
399 messages
TiBotSun, February 19, 2023 at 05:07 AM
@Solasel (ping me about AMC) - As an author, it is your responsibility to complete the ticket
Guidelines
- You can rename your ticket with
- If this ticket was created by accident or it can be deleted, you can use
- When you are ready to submit the ticket, compile everything into one message following the format below and pin it. Then type
- To add contributors to your ticket you can use
- The ticket will be scrapped if: no activity >1 week or open for >1 month.
Guidelines
- You can rename your ticket with
/rename <ticket name> or with the button below- If this ticket was created by accident or it can be deleted, you can use
/delete within the first 5 minutes, otherwise ask a Scholar.- When you are ready to submit the ticket, compile everything into one message following the format below and pin it. Then type
/close or click the button below; the ticket will automatically be moved to be reviewed.- To add contributors to your ticket you can use
/contributor add <user>.- The ticket will be scrapped if: no activity >1 week or open for >1 month.
Write-up Format
Theory/Finding/Bug: Title of your submission
Evidence: Explanations with calculations and/or YouTube/Imgur proofs
Significance: Conclusion
Evidence: Explanations with calculations and/or YouTube/Imgur proofs
Significance: Conclusion
Solasel (ping me about AMC)Sun, February 19, 2023 at 05:23 AM
EDIT: this initial message was the original idea, just data gathering. more comprehensive investigation has continued below.
Finding:
As documented in https://library.keqingmains.com/evidence/characters/anemo/traveler-anemo#amc-hold-e-ticks AMC's hold E absorption damage can tick between 5 and 6 times, seemingly randomly.
However, the probability of getting the extra hit is not 50/50, as one might assume at the outset, but is instead closer to 92%. That is, in the vast majority of cases, you will get 6 additional elemental damage, and only relatively rarely will you only get 5.
To be clear, this means that the first tick of elemental absorption damage will occur after the 2nd initial cutting tick ~92.4% of the time, and after the 1st max cutting tick ~7.6% of the time.
Evidence:
I carried out 250 trials against the ruin guard in 12-1-1 during patch 3.4.0. The enemy was infused with electro, either with Fischl's charged shots (1U -> 0.8U) or the electro elemental node (max 3.6U).
Methodology example: https://www.youtube.com/watch?v=3uNWO5Jx0TE
Example 2 with more varied damage numbers showing how the testing can vary with 4VV: https://www.youtube.com/watch?v=fsu-etYwzA4
I then used AMC's hold E, from a variety of distances from the target, and when the target was stunned or active, and kept count of how many times the elemental infusion damage began after the 2nd instance of initial cutting damage, and how many times it began after the 1st instance of max cutting damage. If it was difficult to see, I used OBS replay buffer to review the footage and double check.
Out of 250 trials, 231 resulted in elemental absorption damage beginning after the 2nd instance of initial cutting damage, and 19 resulted in elemental absorption damage beginning after the 1st instance of max cutting damage. No other odd behaviors were noticed. Some handy interpretations of the data are shown below.
231/250 = 92.4% 19/250 = 7.6% 19/250 ~= 1/13.16
Finding:
As documented in https://library.keqingmains.com/evidence/characters/anemo/traveler-anemo#amc-hold-e-ticks AMC's hold E absorption damage can tick between 5 and 6 times, seemingly randomly.
However, the probability of getting the extra hit is not 50/50, as one might assume at the outset, but is instead closer to 92%. That is, in the vast majority of cases, you will get 6 additional elemental damage, and only relatively rarely will you only get 5.
To be clear, this means that the first tick of elemental absorption damage will occur after the 2nd initial cutting tick ~92.4% of the time, and after the 1st max cutting tick ~7.6% of the time.
Evidence:
I carried out 250 trials against the ruin guard in 12-1-1 during patch 3.4.0. The enemy was infused with electro, either with Fischl's charged shots (1U -> 0.8U) or the electro elemental node (max 3.6U).
Methodology example: https://www.youtube.com/watch?v=3uNWO5Jx0TE
Example 2 with more varied damage numbers showing how the testing can vary with 4VV: https://www.youtube.com/watch?v=fsu-etYwzA4
I then used AMC's hold E, from a variety of distances from the target, and when the target was stunned or active, and kept count of how many times the elemental infusion damage began after the 2nd instance of initial cutting damage, and how many times it began after the 1st instance of max cutting damage. If it was difficult to see, I used OBS replay buffer to review the footage and double check.
Out of 250 trials, 231 resulted in elemental absorption damage beginning after the 2nd instance of initial cutting damage, and 19 resulted in elemental absorption damage beginning after the 1st instance of max cutting damage. No other odd behaviors were noticed. Some handy interpretations of the data are shown below.
231/250 = 92.4% 19/250 = 7.6% 19/250 ~= 1/13.16
Significance:
This has significance in TCing rotations for AMC. A few applications are listed below:
Higher precision when calculating:
* The contribution of initial cutting absorption damage to total damage (very minor)
* The amount of Light of Foliar Incision passive procs across damage types - Phys from NAs, Anemo from A1 and E base, and the absorbed element from E
Although getting 6 hits from a hold E does not increase the amount of secondary reactions triggered by the absorption damage, it does make them occur earlier. This impacts issues of timing including:
* Off-field elemental app from teammates and how it interacts with the AMC blender
* What buffs are active when the reactions happen
* Whether or not these reactions get "clipped" if AMC gets knocked out of their hold E or is forced to cancel it for some reason.
The most significant of which is probably that of LoFI's damage spread, although none of these are major.
This has significance in TCing rotations for AMC. A few applications are listed below:
Higher precision when calculating:
* The contribution of initial cutting absorption damage to total damage (very minor)
* The amount of Light of Foliar Incision passive procs across damage types - Phys from NAs, Anemo from A1 and E base, and the absorbed element from E
Although getting 6 hits from a hold E does not increase the amount of secondary reactions triggered by the absorption damage, it does make them occur earlier. This impacts issues of timing including:
* Off-field elemental app from teammates and how it interacts with the AMC blender
* What buffs are active when the reactions happen
* Whether or not these reactions get "clipped" if AMC gets knocked out of their hold E or is forced to cancel it for some reason.
The most significant of which is probably that of LoFI's damage spread, although none of these are major.
TiBotSun, February 19, 2023 at 05:47 AM
Ticket closed by @Solasel (ping me about AMC). If there are any issues with it - it can be reopened by the owner or staff by using the buttons below or
/open.Ticket re-opened by @Solasel (ping me about AMC)
Ticket closed by @Solasel (ping me about AMC). If there are any issues with it - it can be reopened by the owner or staff by using the buttons below or
/open.
( ˶ᵔ ᵕ ᵔ˶ ) <3 | CasdelaSun, February 19, 2023 at 07:10 AM
@Solasel (ping me about AMC) do you have video recording of your trials?
can you link it in your finding
Mr. Pillow | PhD. in IttoSimpingSun, February 19, 2023 at 07:21 AM
i think it's unreasonable to demand an hour of recording lol. when i did a similar ticket a long time ago (it had ~100+ trials), i only provided a short clip of a single trial to show my methodology.
Solasel (ping me about AMC)Sun, February 19, 2023 at 07:21 AM
I do have the clips that I used to double check, some of which are a few minutes long, but not the whole thing ofc
https://www.youtube.com/watch?v=3uNWO5Jx0TE
amc absorption damage ticks testing
just a small selection
jump canceled the hold E so I wouldn't swirl away all the electro
the presence of a 5 (13 crit) means that we got a 6 hit, otherwise a 5 hit
different for fischl's electro because of
jump canceled the hold E so I wouldn't swirl away all the electro
the presence of a 5 (13 crit) means that we got a 6 hit, otherwise a 5 hit
different for fischl's electro because of
here's an example
( ˶ᵔ ᵕ ᵔ˶ ) <3 | CasdelaSun, February 19, 2023 at 08:07 AM
a few clips is fine
Solasel (ping me about AMC)Sun, February 19, 2023 at 08:08 AM
so more than one? i'll get to uploading
ok hopefully that's good
nzpiefaceSun, February 19, 2023 at 09:19 AM
ugh
i hate these kinds of tickets just cause of the sheer amount of data to go through
if it was recorded
other than that theres some other factors, which prolly arent too big of an issue, such as ping variance, and then more important things such as aura checks nad how that could be an explanation
Solasel (ping me about AMC)Sun, February 19, 2023 at 09:32 AM
idk why i keep closing the ticket tbh
i should've waited for feedback
yeah, ultimately the goal is to get to the bottom of why this inconsistency happens, but for now having a probability is a good step
if staff etc want me to actually go record a full length hour long video or w/e of me doing a couple hundred AMC E's I'm happy to oblige, but that'd take some time
Ultimately I want to be as thorough as possible, so aspects such as ping etc are on my list of things to do
nzpiefaceSun, February 19, 2023 at 08:46 PM
not really
it would be more helpful for you to dig into the mechanics more to find why
having numbers is cool and all, but not really that important even if we're off by a few %
Solasel (ping me about AMC)Sun, February 19, 2023 at 09:43 PM
yeah that's fair, I'll continue digging
TiBotMon, February 20, 2023 at 02:15 AM
This ticket has been renamed to
cause of the extra absorption hit on AMC's hold skill by @Solasel (ping me about AMC)Solasel (ping me about AMC)Mon, February 20, 2023 at 05:36 AMframes between 1st initial cutting damage showing up and ningguang's screen reaction:
5 hit: 3 video frames
6 hit: 3 video frames
frames between 1st and 2nd cutting damage hits
6 hit: 9 video frames
5 hit: 6 video framesit seems that in the 6 hit case, the 2nd initial cutting damage is, for some reason, delayed by 3 frames.
this presumably allows the absorption to take effect in time for the 2nd cutting damage.sounds like it's possibly hitlag-related
5 hit: 3 video frames
6 hit: 3 video frames
6 hit: 9 video frames
5 hit: 6 video frames
this presumably allows the absorption to take effect in time for the 2nd cutting damage.
the above were based on the damage number, which is inconsistent
I will do more trials where the relevant hitmarks are more visible and record those
I will do more trials where the relevant hitmarks are more visible and record those
in case anyone cares, I'm doing more thorough trials where I'm recording everything, and noting down the relative frames of every given hitmark
of note is the huge range in when the absorption aura check takes place - I suspect that if the aura check takes place on the same frame as the 2nd cutting hit (frame 7) the absorption won't work, but I haven't gotten any 5 hits yet to confirm this
now the question is why on earth is the absorption aura check inconsistent
of note is the huge range in when the absorption aura check takes place - I suspect that if the aura check takes place on the same frame as the 2nd cutting hit (frame 7) the absorption won't work, but I haven't gotten any 5 hits yet to confirm this
now the question is why on earth is the absorption aura check inconsistent
https://www.youtube.com/watch?v=9rLqn14qVpg
AMC hold E (no absorb) framedata
Notable Frames:
0 - initial cutting hitmark (1)
5 - jade screen aura check hitmark (1)
8 - initial cutting hitmark (2)
11 - jade screen aura check hitmark (2)
17 - jade screen aura check hitmark (3)
18 - max cutting hitmark (1)
23 - jade screen aura check hitmark (4)
26 - max cutting hitmark (2)
29 - jade screen...
0 - initial cutting hitmark (1)
5 - jade screen aura check hitmark (1)
8 - initial cutting hitmark (2)
11 - jade screen aura check hitmark (2)
17 - jade screen aura check hitmark (3)
18 - max cutting hitmark (1)
23 - jade screen aura check hitmark (4)
26 - max cutting hitmark (2)
29 - jade screen...
https://www.youtube.com/watch?v=bUEu8Yi91XI
amc hold E (absorb) trials
Notable trials:
Trials 21 (2:54), 22 (3:03), 37 (4:56), and 46 (6:02) skipped the first absorption hit (that normally follows initial_cutting 2) and thus resulted in only 5 absorption hits overall.
Trial 24 (3:17) had the hitmark for aura check and the hitmark for initial_cutting 2 appear on the same video frame (7 if initial_cutting 1 is on f...
Trials 21 (2:54), 22 (3:03), 37 (4:56), and 46 (6:02) skipped the first absorption hit (that normally follows initial_cutting 2) and thus resulted in only 5 absorption hits overall.
Trial 24 (3:17) had the hitmark for aura check and the hitmark for initial_cutting 2 appear on the same video frame (7 if initial_cutting 1 is on f...
nzpiefaceThu, February 23, 2023 at 10:21 AM
so did you get a solid result?
Solasel (ping me about AMC)Thu, February 23, 2023 at 10:23 AM
Key takeaways:
The first aura check's timing varies - assuming the first initial cutting hit is frame 0, the first aura check happens on frames 4-7 (inclusive). If the first aura check fails, the following checks will take place every 6 frames after the first, as shown in the first video.
As for the second video, out of 51 trials, four of them (trials 21, 22, 37, and 46) skipped the first instance of absorption damage, leading to only 5 absorbed hits. In all of these, the aura check happened as late as possible (frame 7).
This implies that when the aura check happens this late, it might not apply to the second cutting hit, leading to a missed hit.
My current working hypothesis is that, for some reason, the aura check's timing varies, and can be delayed far back enough to occur on the same frame as the second initial cutting hit. then the order in which the two hits are processed by the engine (50/50) determines whether or not this absorption affects the second cutting hit.
This is backed up by the fact that of the 8 trials where the aura check took place on frame 7, exactly half of them resulted in the missed hit.
However, there are some issues with this hypothesis. Notable of course is that in the footage, every trial that missed the hit had the second cutting hitmark appear one frame later than the aura check, on frame 8. If the hitmark is an exact indicator of when things are calculated under the hood, this theory is thrown out.
Furthermore, trial 24 is odd, since in it, both the hitmark for the aura check and for the 2nd initial cutting happen on the same frame (frame 7), and that trial got all 6 hits.
Either way, advice and help is appreciated
The first aura check's timing varies - assuming the first initial cutting hit is frame 0, the first aura check happens on frames 4-7 (inclusive). If the first aura check fails, the following checks will take place every 6 frames after the first, as shown in the first video.
As for the second video, out of 51 trials, four of them (trials 21, 22, 37, and 46) skipped the first instance of absorption damage, leading to only 5 absorbed hits. In all of these, the aura check happened as late as possible (frame 7).
This implies that when the aura check happens this late, it might not apply to the second cutting hit, leading to a missed hit.
My current working hypothesis is that, for some reason, the aura check's timing varies, and can be delayed far back enough to occur on the same frame as the second initial cutting hit. then the order in which the two hits are processed by the engine (50/50) determines whether or not this absorption affects the second cutting hit.
This is backed up by the fact that of the 8 trials where the aura check took place on frame 7, exactly half of them resulted in the missed hit.
However, there are some issues with this hypothesis. Notable of course is that in the footage, every trial that missed the hit had the second cutting hitmark appear one frame later than the aura check, on frame 8. If the hitmark is an exact indicator of when things are calculated under the hood, this theory is thrown out.
Furthermore, trial 24 is odd, since in it, both the hitmark for the aura check and for the 2nd initial cutting happen on the same frame (frame 7), and that trial got all 6 hits.
Either way, advice and help is appreciated
There's a lot that I still don't know so please feel free to tell me I'm stupid and have no idea what I'm talking about
oh also, the distribution of the aura check's timing isn't uniform
Frame 4: 8 occurrences
Frame 5: 22 occurrences
Frame 6: 13 occurrences
Frame 7: 8 occurrences
so it's kind of a mini bell curve ig, although tilted towards the left because of frame 5
Frame 4: 8 occurrences
Frame 5: 22 occurrences
Frame 6: 13 occurrences
Frame 7: 8 occurrences
so it's kind of a mini bell curve ig, although tilted towards the left because of frame 5
last thing before I sleep, just considering the probability of the hit being lost overall, and combining this data with my original 250 trials, we get 301 trials, with 23 of them missing a hit
278/301 = 92.3% 23/301 = 7.7% 23/301 ~= 1/13.1
278/301 = 92.3% 23/301 = 7.7% 23/301 ~= 1/13.1
paichlearThu, February 23, 2023 at 12:22 PM
slightly unrelated but since this is about amc e being inconsistent, i'll leave this here in case you want more data https://docs.google.com/spreadsheets/d/1zar01TBuQXlJfhAdBaww94dUARQEbA7WVnUGLbi80P0/edit?usp=sharing (never finished it but the point was to figure out how inputting an action (i.e. dashing) affects whether or not cutting dmg gets turned into storm dmg) 
Palm Vortex Frames
The only sheet here
Moved to https://docs.google.com/spreadsheets/d/1HQ6qM1CVSoZ3gWGPVnAWQq0dMy05ewyecr_sORES3_U/edit?usp=sharing
This is slightly outdated and inaccurate because m...
Moved to https://docs.google.com/spreadsheets/d/1HQ6qM1CVSoZ3gWGPVnAWQq0dMy05ewyecr_sORES3_U/edit?usp=sharing
This is slightly outdated and inaccurate because m...
Solasel (ping me about AMC)Thu, February 23, 2023 at 07:24 PM
Thank you sir!
Reply to paichlear: slightly unrelated but since this is about amc e being inconsistent, i'll leave this here in case you want more data https://docs.google.com/spreadsheets/d/1zar01TBuQXlJfhAdBaww94dUARQEbA7WVnUGLbi80P0/edit?usp=sharing (never finished it but the point was to figure out how inputting an action (i.e. dashing) affects whether or not cutting dmg gets turned into storm dmg)
Solasel (ping me about AMC)Thu, February 23, 2023 at 07:33 PM paichlear: slightly unrelated but since this is about amc e being inconsistent, i'll leave this here in case you want more data https://docs.google.com/spreadsheets/d/1zar01TBuQXlJfhAdBaww94dUARQEbA7WVnUGLbi80P0/edit?usp=sharing (never finished it but the point was to figure out how inputting an action (i.e. dashing) affects whether or not cutting dmg gets turned into storm dmg) On this note, inside you said that you set your graphics too high so the numbers weren't reliable - are there standards for what graphics to use? I presume as low as possible where you can still see the relevant visual effects
paichlearThu, February 23, 2023 at 08:27 PMam a girl but yw there aren't any official standards atm (there was this but it's kind of forgotten and deprecated https://docs.google.com/document/d/1rhKDnkgkVYU5jLq6i_7FRz43DRwElyALX_j6hW5DwqY/edit?usp=sharing)
i just find it a lot easier to see what's going on when all graphic settings (except fps) are set to lowest, but when it comes to amc's e and q.. well the hitmarks are really hard to see either way. and hitmarks in general aren't as reliable as auras, but when something's on icd you can't rely on auras, so you have to either eyeball the hitmarks or count back from dmg numbers
oh and in case you didn't know, ningguang's screen messes with amc q swirls, so maybe e absorption has something to do with how far away you stand from the screen when casting too (also never finished it but these are the clips that lead to the theory that if the tornado doesn't move you lose 1 swirl https://www.youtube.com/playlist?list=PLSbStW9syipOgGD_sdUi2XUjUyA0rB7eP)
i just find it a lot easier to see what's going on when all graphic settings (except fps) are set to lowest, but when it comes to amc's e and q.. well the hitmarks are really hard to see either way. and hitmarks in general aren't as reliable as auras, but when something's on icd you can't rely on auras, so you have to either eyeball the hitmarks or count back from dmg numbers
oh and in case you didn't know, ningguang's screen messes with amc q swirls, so maybe e absorption has something to do with how far away you stand from the screen when casting too (also never finished it but these are the clips that lead to the theory that if the tornado doesn't move you lose 1 swirl https://www.youtube.com/playlist?list=PLSbStW9syipOgGD_sdUi2XUjUyA0rB7eP)
Solasel (ping me about AMC)Thu, February 23, 2023 at 08:39 PM
The spaghetti goes deeper and deeper
Thank you again though, you've been a great help
Next up on the chopping block is trying to get a few hold e trials where both initial cutting hits apply anemo, with sac sword shenanigans to mess with the icd
nzpiefaceFri, February 24, 2023 at 03:48 AM
watching this is great
Solasel (ping me about AMC)Fri, February 24, 2023 at 04:07 AM
you mean watching someone in real time spiral deeper and deeper into insanity 
rarepossumFri, February 24, 2023 at 04:15 AM
You haven't gotten to changing fps yet have you
Solasel (ping me about AMC)Fri, February 24, 2023 at 04:25 AM
oh don't worry I value my sanity more than that
because windows is dumb, the above footage is actually 30 fps
I'll re-record a lot of trials in 60 fps and make sure not to use windows to trim down the files so it stays 60 fps
I'll re-record a lot of trials in 60 fps and make sure not to use windows to trim down the files so it stays 60 fps
https://www.youtube.com/watch?v=xypNEax__Ck
amc hold E no absorption
Now in 60 fps! (windows is a plague)
The first aura check happened 13 frames after the first initial cutting hitmark in trial 1, and 14 frames after for trial 2. After that, the aura check was repeated every 11 or 12 frames after the initial check.
The aura checks continue taking place for a small amount of time after the max storm damage, si...
The first aura check happened 13 frames after the first initial cutting hitmark in trial 1, and 14 frames after for trial 2. After that, the aura check was repeated every 11 or 12 frames after the initial check.
The aura checks continue taking place for a small amount of time after the max storm damage, si...
first is just the aura check stuff - in 60 fps now
same idea, except that instead of happening every 6 frames (30 fps), it happens every 11 OR 12 frames (60 fps)
same idea, except that instead of happening every 6 frames (30 fps), it happens every 11 OR 12 frames (60 fps)
Rathalos | ask me abt travelerMon, February 27, 2023 at 05:57 PM
I genuinely basically never get 6 hits
Solasel (ping me about AMC)Mon, February 27, 2023 at 07:13 PM
That's so bizarre
Do you use aether or lumine
Because if it's not a difference between the characters then there's something more complicated going on
Rathalos | ask me abt travelerMon, February 27, 2023 at 07:27 PM
I think it's ping
i play at like 270 ping 
Solasel (ping me about AMC)Mon, February 27, 2023 at 07:30 PM
Looks like I have to get a ping modulator
MercuryTue, February 28, 2023 at 05:40 AM
If it's an "easy" test to do, I could fire up my lagswitch and record a couple of trials. Lag options are (approximately) 120ms, 260ms, 500ms. MC is Lumine.
Solasel (ping me about AMC)Tue, February 28, 2023 at 06:14 AM(it's still processing up to HD) done processing
if you feel like it - I'm going to do it anyway on both so don't feel like you need to
just wanted to share my in-progress data - this is 53 trials and the distribution of where the absorption happened relative to the initial_cutting (2) hitmark
so far, the 6th hit fails IFF the aura check is exactly 1 frame before the hitmark
once I get ~100 trials like this to establish a better distribution, I'll fire up the lag machine and try it on different ping, as well as on aether
1
so far, the 6th hit fails IFF the aura check is exactly 1 frame before the hitmark
once I get ~100 trials like this to establish a better distribution, I'll fire up the lag machine and try it on different ping, as well as on aether
1I now have 101 clips and am processing them 
https://docs.google.com/spreadsheets/d/15EUChwWpDYP9Dk-7wjf3cCONoWlIlGj-7eVrWJNrHA4/edit?usp=sharing
Lumine Auracheck Framedata (~69 ping)
hE absorb (processed)
Trial,Absorption Hits,Animation Start,Initial Cutting Swirl Text,Jade Screen Hitmark,Initial Cutting (2) Hitmark,Aura Check (Successful),Aura Check (Unsuccessful),AC and IC1 Diff (Successful),AC and IC1 Diff (Unsuccessful),AC and IC2 Diff (Successful),AC and IC2 Diff (Unsuc...
Trial,Absorption Hits,Animation Start,Initial Cutting Swirl Text,Jade Screen Hitmark,Initial Cutting (2) Hitmark,Aura Check (Successful),Aura Check (Unsuccessful),AC and IC1 Diff (Successful),AC and IC1 Diff (Unsuccessful),AC and IC2 Diff (Successful),AC and IC2 Diff (Unsuc...
here's the data from all 101 trials
here are the relevant graphs
that last graph "Distribution of Diff Between AC and Start Frames" is a bit iffy, since it seems like the entire E (its hitmarks and elemental applications) seem to randomly sometimes be delayed by 1 frame from the start
I'll probably do a few more charts to see if that particular thing changes how this distribution changes
I'll probably do a few more charts to see if that particular thing changes how this distribution changes
Current hypothesis is the following:
In these trials I had ~69 ping. 69ms = 4.14 frames @60fps
The mean of the distribution seems to be ~12 frames after the first initial cutting hit. subtracting 4 frames for ping delay, we get 8 frames after, which is smack dab in the middle of the 16 frames of delay between the two initial cutting hits (15 if you count from hitmark to hitmark rather than anemo app to hitmark)
tl;dr I think the aura check is supposed to take place 8 frames after the first initial cutting hit, but it is delayed by lag. In my case, only the very tail end of the lag was late enough to cause the missed hit, but higher ping would make it more common..
this will be my next test - doing ~50 trials with artificially induced lag. I'll aim for enough lag to make the distribution 50/50 or so, but we'll see how much I can control the ping
I'll move on to aether after I've reached a satisfying conclusion for lumine.
video evidence of all the 101 trials above coming soon (gl to the people who have to verify it)
In these trials I had ~69 ping. 69ms = 4.14 frames @60fps
The mean of the distribution seems to be ~12 frames after the first initial cutting hit. subtracting 4 frames for ping delay, we get 8 frames after, which is smack dab in the middle of the 16 frames of delay between the two initial cutting hits (15 if you count from hitmark to hitmark rather than anemo app to hitmark)
tl;dr I think the aura check is supposed to take place 8 frames after the first initial cutting hit, but it is delayed by lag. In my case, only the very tail end of the lag was late enough to cause the missed hit, but higher ping would make it more common..
this will be my next test - doing ~50 trials with artificially induced lag. I'll aim for enough lag to make the distribution 50/50 or so, but we'll see how much I can control the ping
I'll move on to aether after I've reached a satisfying conclusion for lumine.
video evidence of all the 101 trials above coming soon (gl to the people who have to verify it)
https://www.youtube.com/watch?v=rKFjEgmqh-Y
Lumine Auracheck Framedata (69 ping)
00:00 - Trial 001
00:44 - Trial 011
01:26 - Trial 021
02:06 - Trial 031
02:44 - Trial 041
03:24 - Trial 051
04:00 - Trial 061
04:38 - Trial 071
05:15 - Trial 081
05:51 - Trial 091
06:28 - Trial 101
00:44 - Trial 011
01:26 - Trial 021
02:06 - Trial 031
02:44 - Trial 041
03:24 - Trial 051
04:00 - Trial 061
04:38 - Trial 071
05:15 - Trial 081
05:51 - Trial 091
06:28 - Trial 101
nzpiefaceTue, March 7, 2023 at 05:51 AM
maybe it's delayed by poise?
i don't see how ping would matter
Solasel (ping me about AMC)Tue, March 7, 2023 at 05:52 AM
how do you mean exactly
ruin guard isn't exactly poise broken
like, that was an idea I had too, assuming that for some reason the hitlag was poise-extendable and that was the issue
but the timing between the two cutting hits is 16 frames in all 101 trials, so i'ts not like there was variable hitlag in there
Solasel (ping me about AMC)Tue, March 7, 2023 at 06:02 AM
also this comment is pretty relevant, unless it's diff on aether or something
(assuming he even plays aether)
@Rathalos | ask me abt traveler when you made taht comment was it aether or lumine
Rathalos | ask me abt travelerTue, March 7, 2023 at 07:54 AM
both
higher ping on aether
Solasel (ping me about AMC)Tue, March 7, 2023 at 07:54 AM
ah
(just to avoid dinkdonk) I am in the process of setting up a means to artificially lag
then we gather data and see what changes
then we gather data and see what changes
https://www.youtube.com/watch?v=_K_n5MbCaMk
Lumine Auracheck hold E Framedata (no absorb 110ms)
done with usb tethering on cellular, resulting in ~110ms
I've managed to get ~110ms and will be using this to gather more laggy data
here's the no-absorb version, which doesn't show any significant differences compared to the 69ms version
here's the no-absorb version, which doesn't show any significant differences compared to the 69ms version
did 30 trials on the worse lag, and got this data
it's a lot more hectic, possibly due to the fact that my method of inducing lag also reduced stability
I'll grab ~100 trials under these conditions to get a more proper distribution, but it seems like ping has some sort of impact on this
Rathalos | ask me abt travelerSun, March 12, 2023 at 04:37 AM
Solasel (ping me about AMC)Sun, March 12, 2023 at 04:41 AM
what the hell could be the underlying way this works 
nothing about this makes sense
feels like a lot of the issue here is just how unstable the ping is
vinsetteFri, March 17, 2023 at 01:58 PM
Oh. Aura checks. Good luck.
Solasel (ping me about AMC)Fri, March 17, 2023 at 10:45 PM
Do you have any insight
for context, I'm still working on getting a better controllable lag machine
the usb tethering method I was using proved to be too unstable, so I'll instead be working on increasing latency without causing instability
vinsetteSun, March 19, 2023 at 11:33 AM
Does it vary with positioning?
Solasel (ping me about AMC)Sun, March 19, 2023 at 07:30 PM
That's another idea, but I kind of doubt it
I plan on reviewing the footage with an eye towards a couple of other different factors though
I plan on reviewing the footage with an eye towards a couple of other different factors though
the biggest thing that makes me discount the positioning is that there are two relations involved: the position of traveler relative to both the ruin guard AND the jade screen
the ruin guard was used to determine whether or not the 6th hit happened, and the jade screen was used to determine what frame the aura check took place
I find it highly unlikely that the data would be coherent if positioning was a factor
the ruin guard was used to determine whether or not the 6th hit happened, and the jade screen was used to determine what frame the aura check took place
I find it highly unlikely that the data would be coherent if positioning was a factor
update: still slowly working on getting a way to lag while maintaining stable connection
irl things make progress slow, but we will get there someday
irl things make progress slow, but we will get there someday
TiBotThu, March 30, 2023 at 09:45 PM
This channel hasn't been active in the past week!
This channel hasn't been active in the past week!
is_this_maddieFri, March 31, 2023 at 02:47 AM
@Solasel (ping me about AMC)
Solasel (ping me about AMC)Fri, March 31, 2023 at 04:10 AM
it is now my weekend so I will work on getting the lag machine up and running
lag machine should be up and running by this evening, with testing beginning soon 
stable 1000 ping testing inc
currently recording 100 trials at ~675ms ping
processing probably done by tomorrow, then i'll post the data and recording
processing probably done by tomorrow, then i'll post the data and recording
recording finished
https://www.youtube.com/watch?v=NPmiADIGoP4
Lumine Auracheck hold E Framedata (absorb 675 ping)
00:00 - Trial 001
00:42 - Trial 011
01:23 - Trial 021
02:04 - Trial 031
02:41 - Trial 041
03:17 - Trial 051
03:55 - Trial 061
04:30 - Trial 071
05:08 - Trial 081
05:45 - Trial 091
00:42 - Trial 011
01:23 - Trial 021
02:04 - Trial 031
02:41 - Trial 041
03:17 - Trial 051
03:55 - Trial 061
04:30 - Trial 071
05:08 - Trial 081
05:45 - Trial 091
there are all 100 trials compiled together, and now I will go through each one and record timing data
preliminary results (20/100) put the failure rate at about the same as the low ping example. Additionally, the aura check never happens later than one frame before the hitmark for the 2nd cutting hit.
if it is indeed dependent on one's network connection, I presume it has to do with packet loss rather than ping, but I'll finish off all 100 trials here just to be sure before testing further.
overall it's very strange - why would the timing of the aura check vary? is it being sent in by the server and thus liable to network delay? why would that ever be the case? also, if so, why is the aura check bounded on the tail end?
I see a few possibilities
if it is indeed dependent on one's network connection, I presume it has to do with packet loss rather than ping, but I'll finish off all 100 trials here just to be sure before testing further.
overall it's very strange - why would the timing of the aura check vary? is it being sent in by the server and thus liable to network delay? why would that ever be the case? also, if so, why is the aura check bounded on the tail end?
I see a few possibilities
#1: the original version of AMC's hold E always did the aura check on the same frame as the 2nd initial cutting hit (one frame before the hitmark), but they didn't realize it didn't ever affect the 2nd hit or something, so they fixed it by having the server tell the client to do an auracheck, leaving the initial hard-coded "check on 2nd initial cutting hit" in the game, hence the upper bound.
there are issues with this though - if the server is telling the client to do the auracheck earlier than normal, how does the client know to repeat the auracheck every 12 frames after the initial aura check? a priori the behavior would be one random aura check before the 2nd cutting hit, then one guaranteed aura check on the same frame as the 2nd initial cutting hit (from the original code), and then 12 frames from there, but instead we see the aura check repeated every 12 frames from the first aura check, regardless of where it lands
also, for this to make sense, the "make an aura check on X frame" message from the server must have been sent at least ~40 frames before the actual check, since otherwise the server wouldn't know the client started an E in time due to the delay I'm using. this allows for another avenue of testing - if we have extremely high latency (over ~1000ms) then the server will not have time to send such a packet, which should result in 100% failure rate
there are issues with this though - if the server is telling the client to do the auracheck earlier than normal, how does the client know to repeat the auracheck every 12 frames after the initial aura check? a priori the behavior would be one random aura check before the 2nd cutting hit, then one guaranteed aura check on the same frame as the 2nd initial cutting hit (from the original code), and then 12 frames from there, but instead we see the aura check repeated every 12 frames from the first aura check, regardless of where it lands
also, for this to make sense, the "make an aura check on X frame" message from the server must have been sent at least ~40 frames before the actual check, since otherwise the server wouldn't know the client started an E in time due to the delay I'm using. this allows for another avenue of testing - if we have extremely high latency (over ~1000ms) then the server will not have time to send such a packet, which should result in 100% failure rate
but like, why would the aura check be managed by a packet from the server sometimes, and not others?
and yet, if it was only through the server, one might expect the aura check to not be upper bounded in the case of extreme network issues
#2: it's not actually related to ping, and the game just randomly picks a frame to do the check according to some distribution to make it seem random. In that case however, rathalos' testimony that he "always loses the hit" needs reexamining
🐟Mon, April 10, 2023 at 12:03 AM
UID dependent1
1Solasel (ping me about AMC)Mon, April 10, 2023 at 12:04 AM
rathalos got the latency-related auracheck patch 1
1@Rathalos | ask me abt traveler when you can, I'd love you to double check that you indeed almost never get the extra hit
if so I'll have to play with more networking parameters
although with my stable connection, I doubt I'd somehow lose enough packets for a 7% failure rate at my normal low ping
also if you could tell me how you normally play where you get those results, that would be great (ie wifi, ethernet, is the connection stable but slow or variable, etc etc)
Rathalos | ask me abt travelerMon, April 10, 2023 at 06:44 AM
I'll do some more tests
I hate it
Solasel (ping me about AMC)Mon, April 10, 2023 at 06:50 AM
don't feel pressured to be as rigorous as me
just do like 10 tests and see
Rathalos | ask me abt travelerWed, April 12, 2023 at 08:20 AM
I'll get back to this on the weekends
1
1!rm 2d
(!) Keqing's Eternal AssistantWed, April 12, 2023 at 08:20 AM
Alright Rathalos, I'll remind you about something in 2 days. ID: 55985284
charliex3000Thu, April 13, 2023 at 06:15 AM
Could the seemingly random timing of the aura check work off some global timer? Maybe how it works is all aura checks are somehow synced to some global timer?
Solasel (ping me about AMC)Thu, April 13, 2023 at 07:14 AM
that's also a possibility, but that doesn't explain why I get a bell curve for the timing distribution
one would assume that if it happens on every 8th frame or something, I'd cast my E on some random frame every time, so it'd be uniform
see the new pins for the original, low latency testing - the new testing is almost finished and looks like it'll be mostly the same
i've finally finished counting all 100 trials for the 675 ping video
https://docs.google.com/spreadsheets/d/1dt_IdD2g-Bc9LfUSz-L_xWlXYPPIADrF1I4Rmmj6b7M/edit?usp=sharing
1
(~675 ping) Lumine Auracheck Framedata
hE absorb (processed)
Trial,Absorption Hits,Animation Start,Initial Cutting Swirl Text,Jade Screen Hitmark,Initial Cutting (2) Hitmark,Aura Check (Successful),Aura Check (Unsuccessful),AC and IC1 Diff (Successful),AC and IC1 Diff (Unsuccessful),AC and IC2 Diff (Successful),AC and IC2 Diff (Unsuc...
Trial,Absorption Hits,Animation Start,Initial Cutting Swirl Text,Jade Screen Hitmark,Initial Cutting (2) Hitmark,Aura Check (Successful),Aura Check (Unsuccessful),AC and IC1 Diff (Successful),AC and IC1 Diff (Unsuccessful),AC and IC2 Diff (Successful),AC and IC2 Diff (Unsuc...
1tl;dr it looks like basically the same distribution
rarepossumSun, April 16, 2023 at 09:04 AM
looks like
Thats unacceptable
Perform statistical tests
Solasel (ping me about AMC)Sun, April 16, 2023 at 09:04 AM
i'm googling dad 
N too small probably anyway
rarepossumSun, April 16, 2023 at 09:04 AM
No you can
its fine
even 30 is sually enough
the test you want is a chi-squared goodness of fit
Solasel (ping me about AMC)Sun, April 16, 2023 at 09:05 AM
perfect ty
not too familiar with stats so was feeling quite lost
but what do I even compare in this case?
for "frames between initial-cutting 1 and aura check" and "frames between aura check and initial-cutting 2" just comparing the total per label makes sense
but for "frames between start and aura check" there are a mix of failures and successes on frame 45, so idk what to use for that particular comparison
for "frames between initial-cutting 1 and aura check" and "frames between aura check and initial-cutting 2" just comparing the total per label makes sense
but for "frames between start and aura check" there are a mix of failures and successes on frame 45, so idk what to use for that particular comparison
anyway these are the chi-squared values for all 3 distributions I graphed at least
tl;dr if we accept 0.05 as the cutoff, then in fact all of the distributions are different
this seems to say that I need more data, since the two counts at the tail end are both less than 5, and 2/8 = 25% > 20%
Assuming there is a difference, I presume it's not latency specifically, since then there would be a much larger more noticeable difference
I'll try to think of different networking things that could be causing it, at the moment I can only think of packet loss in particular
I'll try to think of different networking things that could be causing it, at the moment I can only think of packet loss in particular
the other thing that comes to mind is that this was with an overworld ruin guard without the strong electro aura of the abyss - might've been the difference maker
ok while waiting for rath to check that his experience is true, I will do test #3 - normal connection and latency, but high induced packet loss
expect results in idk how long pay me if you want it done faster
Rathalos | ask me abt travelerTue, April 18, 2023 at 05:15 AM
i was wrong
maybe that statement was just confirmation bias back then or something
whatever, i did 22 tests
https://docs.google.com/spreadsheets/d/1pSSZ8jtaCfnZHOqNJaXwd8TzuzsI8PkV1EJqZBE7x5A/edit?usp=sharing
AMC hE test
Data
Trial,Absorption DMG,Time Stamp,Author,Rathalos#2875
Initial Cutting,Max Cutting 1,Max Cutting 2,Max Cutting 3,Max Cutting 3,Max Storm,Video,https://youtu.be/_ee8R8SS9Nc
1,56,47,47,47,47,544,00:05s,Ping,133ms - 142 ms
2,51,84,51,84,588,00:11s
3,84,...
Trial,Absorption DMG,Time Stamp,Author,Rathalos#2875
Initial Cutting,Max Cutting 1,Max Cutting 2,Max Cutting 3,Max Cutting 3,Max Storm,Video,https://youtu.be/_ee8R8SS9Nc
1,56,47,47,47,47,544,00:05s,Ping,133ms - 142 ms
2,51,84,51,84,588,00:11s
3,84,...
TLDR:
on 133ms - 142ms Lumine
idk how I feel about that
do i just calc with 0.5 dmg on the initial cutting?
let me know if u need more
Solasel (ping me about AMC)Tue, April 18, 2023 at 05:37 AM
so just to be clear, you got the extra hit about 50% of the time?
Rathalos | ask me abt travelerTue, April 18, 2023 at 05:37 AM
yeah
Solasel (ping me about AMC)Tue, April 18, 2023 at 05:38 AM
looks like you play on a pc, is it a laptop? are you connected through ethernet or wifi?
Rathalos | ask me abt travelerTue, April 18, 2023 at 05:38 AM
laptop, wifi
Solasel (ping me about AMC)Tue, April 18, 2023 at 05:38 AM
0.5 or w/e works, but apparently it gets really weird with lofi procs
Rathalos | ask me abt travelerTue, April 18, 2023 at 05:38 AM
oh no
why
Solasel (ping me about AMC)Tue, April 18, 2023 at 05:38 AM
Rathalos | ask me abt travelerTue, April 18, 2023 at 05:39 AM
does it consume lofi procs even without dealing dmg?
Solasel (ping me about AMC)Tue, April 18, 2023 at 05:39 AM
probably not enough to matter either way anyway
unsure
GO disc probably has the answers
Rathalos | ask me abt travelerTue, April 18, 2023 at 05:39 AM
what's the issue 
Solasel (ping me about AMC)Tue, April 18, 2023 at 05:39 AM
I was just told "it gets weird with lofi and other hit-based procs if you do fractions of a hit"
maybe it rounds up and assumes a hit?
Rathalos | ask me abt travelerTue, April 18, 2023 at 05:40 AM
Solasel (ping me about AMC)Tue, April 18, 2023 at 05:40 AM
sheet op for this anyway
optimizer stinky
Rathalos | ask me abt travelerTue, April 18, 2023 at 05:40 AM probably just assume everyone gets the extra proc
probably just assume everyone gets the extra proc I'll go fix the configs then make a submission
i have so many things to do 
Solasel (ping me about AMC)Tue, April 18, 2023 at 05:41 AM
I mean you're probably gonna get the proc later anyway, it's basically just 15 electro damage 1
1Man after watching your clip I have even less idea what could possibly be causing this
Rathalos | ask me abt travelerTue, April 18, 2023 at 06:52 AM
Solasel (ping me about AMC)Tue, April 18, 2023 at 12:47 PM
@Rathalos | ask me abt traveler can i see your amc build that you were using for those tests? Mostly interested in talent level and arti
Rathalos | ask me abt travelerTue, April 18, 2023 at 12:48 PM
4pc deepwood ER/Dendro/CritR, 10/13/13
why?
Solasel (ping me about AMC)Tue, April 18, 2023 at 12:50 PM
I've thrown out all preconceptions
I will embrace the spaghetti
Oh what weapon
Rathalos | ask me abt travelerTue, April 18, 2023 at 12:51 PM
sac
Solasel (ping me about AMC)Tue, April 18, 2023 at 12:52 PM
I think I'll do a few tests without the Jade screen there too, just to see if that's affecting it
oh god
what if it's talent level
what if it's talent level
surely not
Rathalos | ask me abt travelerWed, April 19, 2023 at 12:47 AM
Solasel (ping me about AMC)Wed, April 19, 2023 at 04:27 AM
did a couple dozen trials without the jade screen and with a 1U aura rather than the 2U that I did at high latency, and I'm still seeing ~90% success
cpu throttling, packet loss, and talent level are the remaining possibilities I can see
otherwise, UID, server, character name, the way the wind is blowing, etc
cpu throttling, packet loss, and talent level are the remaining possibilities I can see
otherwise, UID, server, character name, the way the wind is blowing, etc
nzpiefaceThu, April 20, 2023 at 08:23 AM
im still waiting on an explanation on this
Solasel (ping me about AMC)Thu, April 20, 2023 at 11:25 AM
then go do like 25 trials and see your success rate 
(and tell me how potato your pc/internet is)
I have found a way to specifically induce packet loss and will come back with my findings
idk probably by the weekend? maybe earlier if I have determination
idk probably by the weekend? maybe earlier if I have determination
I have also crowned my amc E, so we'll see if that makes a diff 
1
1I have finally gotten around to programming my computer to do packet loss, and I'll start doing trials this weekend... Probably
after trying 20 times without changing any network settings (only real diff being E talent level) the distribution is the same, so we're at least not in the worst timeline
ok so there's an issue:
on highly lossy connections, sometimes damage numbers simply get dropped altogether
so i'll need to find some other way to test whether or not the electro is happening
alternatively, if lossy connection causes other issues (like hits getting dropped regardless) then perhaps this isn't exactly the best point of comparison
on highly lossy connections, sometimes damage numbers simply get dropped altogether
so i'll need to find some other way to test whether or not the electro is happening
alternatively, if lossy connection causes other issues (like hits getting dropped regardless) then perhaps this isn't exactly the best point of comparison
given that the health bar didn't move, it seems that the attack was simply lost, so looks like sometimes hits are dropped regardless
i can find out a prior for hits being lost, but overall this seems like a poor environment for testing whether or not the extra hit actually lands or not
i can find out a prior for hits being lost, but overall this seems like a poor environment for testing whether or not the extra hit actually lands or not
I'll take a look at rath's footage to see if anything similar was happening to him (ie hits other than the first possible absorption hit getting lost)
but otherwise, throttling my cpu will be the next step
doing a couple dozen tests with a very throttled cpu feels the same, so it's probably not that
I'm speeding up my testing - once I find what I think is probably the cause, I'll do more trials to check
guess I'll make an asia acc to see if it's server-related
welp after doing a couple dozen attempts on a fresh asia acc, it's about the same as usual, so it's not the server
guess it's back to the drawing board 
I will now look at rathalos's video and die a little inside while I figure out wtf is going on
I guess I'll try some combination of ping and packet loss
@Rathalos | ask me abt traveler if you can find out your packet loss rate so I can exactly emulate your setup, that would be epic
also your network jitter
if you don't know how to do it I'll help you some other time, but for now, it is bedtime 
tl;dr next i modify jitter and then some combo of stuff, maybe replicating rathalos's network environment, and then if that doesn't work it's not network related as far as I can tell
Could also be packet corruption or something esoteric like that
Rathalos | ask me abt travelerSun, May 7, 2023 at 12:31 PM
how do i do that
Solasel (ping me about AMC)Sun, May 7, 2023 at 04:33 PM
Are you on Windows?
@Rathalos | ask me abt traveler if you're using windows:
open cmd prompt as administrator and run:
that should ping the asia server, double check that it's the right address
then do:
that'll ping it 100 time, then send me the results - should look something like this: (note that i'm pinging na so the ip is diff)
open cmd prompt as administrator and run:
ping 47.91.24.239that should ping the asia server, double check that it's the right address
then do:
ping -n 100 47.91.24.239that'll ping it 100 time, then send me the results - should look something like this: (note that i'm pinging na so the ip is diff)
if it doesn't work for some reason lmk and i'll help you debug
ideally export the results from all 100 pings into a txt file or w/e
just to show you what it looks like
preferably i'd have you do more than 100, maybe up to 10k to get a good sample, but that'd take quite awhile, although you can leave it in the background
anyway up to you, gib results when you can
anyway up to you, gib results when you can
(also if you could tell me your computer's processor that would be nice)
@Rathalos | ask me abt traveler what fps are you running when doing these tests 
all of my tests are done on 60 fps, but to the best of my sight, your video is in 30fps
when I lower my fps to 30 I see a much closer distribution to what you're seeing
i s2g if you've been saying it's failing and been playing on 30 fps
in case you weren't aware, it automatically lowers fps to 30 if you select "lowest" in graphics settings
sharaths21312Fri, May 12, 2023 at 08:02 PM
fps shenanigans
Solasel (ping me about AMC)Fri, May 12, 2023 at 08:26 PM
I honestly hope he messed up and it's just fps
otherwise it's more unknown than ever