search

ppy / osu

rhythm is just a *click* away!
https://osu.ppy.sh
MIT License
15.33k stars 2.28k forks source link

Strict Tracking mod seems inconsistent #29371

Open ashandelle opened 2 months ago

ashandelle commented 2 months ago

Type

Game behaviour

Bug description

Strict Tracking mod seems inconsistent, the leniency seems to vary sometimes even letting you completely miss slider ends or sometimes missing even when you stayed inside of the slider ball for the whole slider. (on a side note my game experienced an error when exporting the logs, should I make a separate issue?)

Screenshots or videos

https://github.com/user-attachments/assets/162519a1-325a-4978-8e4a-9b0e365bffce

https://github.com/user-attachments/assets/81b939d8-3a9b-442c-9d33-a52f2f327f29

https://github.com/user-attachments/assets/af055ae6-c857-4cd8-b455-834a63616ef6

Version

Lazer 2024.731.0

Logs

image

ssz7-ch2 commented 2 months ago

can you upload the replays

ashandelle commented 2 months ago

These are the replays used in the clips https://drive.google.com/file/d/1VFUWROrfPQKPhlfLzZLdEmZ9-4wyMH0n/view?usp=drive_link https://drive.google.com/file/d/1_rmnsRyUt7jQzczcXEZlvNek7AnSPS-F/view?usp=drive_link

ashandelle commented 2 months ago

Here's some more replays that I haven't checked. https://drive.google.com/file/d/1XcqdS2VKjizwJa2Yr_AFq1LrhNsYX6aX/view?usp=drive_link https://drive.google.com/file/d/1hRcB0R-L-Bp32dBVH2AEZwNpxG73lYkZ/view?usp=drive_link https://drive.google.com/file/d/1OwDRKtpWeK9ryu5Ytn68JTmizNp68V9x/view?usp=drive_link https://drive.google.com/file/d/1Zxi0w5Ts2jnK0AnxlE934Wyfr0Ynadjk/view?usp=drive_link

ssz7-ch2 commented 2 months ago

For the third video, there's nothing wrong with that. The larger white outer circle is the area which the cursor must stay within, not the sliderball.

Managed to reproduce the miss from the first video.

https://github.com/user-attachments/assets/39f6bd56-3f6c-420d-8b0f-2266138db5be

It appears that strict tracking will cause you to miss when releasing a key with the cursor over the slider, even if the sliderhead hasn't been clicked yet. Not sure if this is intended or not, but imo it would make more sense to only allow strict tracking to miss after the sliderhead is judged.

ashandelle commented 2 months ago

For the third video, there's nothing wrong with that. The larger white outer circle is the area which the cursor must stay within, not the sliderball.

In that case I think there should be a setting to change the size of the outer circle, I've been playing Strict Tracking the whole time assuming that I was supposed to stay in the slider ball.

iijj22gg commented 1 month ago

I think this is happening because sliders are still tracked when the sliderhead is not hit, so whatever input is being held down at the time the slider starts (which if there is a delay for leniency it's probably not long enough) is the one that gets associated with the slider, regardless of whether the player intended to use a different input to hit the sliderhead. I assume this is an oversight.

In my opinion as someone who almost exclusively plays with ST, each slider (and hitobject) should only be associated with one input, so if an input was already associated with one slider or hitcircle, it cannot be used for another slider. This approach would solve this issue, but it's behavior with spinners could become a new issue or point of contention.

Otherwise a special exception needs to be made to solve this issue, or a delay added (or increased if one already exists) so that an input is not associated with a slider until after a certain window of time has passed. Maybe that delay could be tied to OD someway.

ashandelle commented 1 month ago

I don't think that's the cause. As I've played more ST and improved the bug seems to happen less, implying that the bug can be avoided by skill, if it was caused by input mismatch then it wouldn't happen any less especially since I always alternate.

To me it seems like under certain conditions the OD is effectively raised temporarily, I think this because I've noticed that when it happens its always with a 100(or 50) on the hit error bar and I get both a 100 and a miss simultaneously. As I've improved with ST so has my accuracy on sliders causing it to happening less. At least that's what it looks like to me.

iijj22gg commented 1 month ago

I don't understand what you mean here... while it is true that, if my understanding is correct, this problem is avoidable with skill since it has to do with when you release an input, skill doesn't disprove anything.

To clarify, I've come to understand through replay analysis and by creating test maps that if your cursor is within the outer circle's radius of the slider head (or the beginning of the track or something like that) at the time point that slider starts, you haven't hit the slider head yet, and you haven't released your previous input, then that input becomes associated with the slider. So when you release that input it will be counted as a miss. It doesn't matter if you singletap or alternate once you're in this situation, since your next input where you maybe intended to hit the sliderhead is a completely different input.

There are plenty of maps where your path from one hitobject to the next slider is intersected by the track of that slider, which is a perfect opportunity for this to happen.

And what you said about getting a 100 and a miss simultaneously... you might just be getting the 100 on the slider head and the miss on the slider end.

I have a lot of weird edge cases though.

ashandelle commented 1 month ago

Sorry I misunderstood your idea, it seems possible though.

There are plenty of maps where your path from one hitobject to the next slider is intersected by the track of that slider, which is a perfect opportunity for this to happen.

I don't think I've experienced the bug on maps like that, but they might just be to difficult to tell.

Ill have to go and test it.

ashandelle commented 1 month ago

https://github.com/user-attachments/assets/3151d5a1-8cbd-4937-b437-dabb1bd4b4fe

If I've managed to understand your idea this time, it doesn't seem to be the case

I tried to see if a slider can only register one input, this is what happens most of the time.

https://github.com/user-attachments/assets/9e7f747e-a285-468e-9fbd-f5ea79620cf4

But if you get it right it does allow for more than 1 input.

https://github.com/user-attachments/assets/7ecc8727-e89d-4549-a631-4c0bb68438a0

Also if you keep going back and forth on the timeline to look at it, eventually it breaks and shows a nonexistent miss. (this is the same exact replay and session as the last video)

https://github.com/user-attachments/assets/e30c8aec-3557-49c7-abb1-269539007e75

ashandelle commented 1 month ago

I made a separate issue for the nonexistent miss #29945

iijj22gg commented 1 month ago

I attempted to prove my theory by recording a clip with the key display blown up really big, just to realize I had come to that conclusion by not releasing the first input and like I thought I was, and was actually releasing the second input. So... no idea now.

Marking this and previous comments as off-topic.

iijj22gg commented 1 month ago

I have also noticed the replay inconsistencies. I have recordings for daily challenges where I get misses that do not always appear in replays, resulting in differing scores. I assume the replay inconsistencies are related to this issue because the inconsistencies appear on sliders where I get the weird misses. I haven't seen any inconsistent replays with other mods either (except in combination with ST)

The inconsistent misses are weird too, sometimes they don't even break the combo and the sliderend hitsound plays normally.