Performance Points feedback and suggestions (Catch the Beat)

posted
Total Posts
767
show more
Topic Starter
Tom94
Hey everyone, I've got an initial iteration of the algorithm to show. Please remember to tell me which version you are talking about when giving feedback. Also please let me know which tuning seems the best to you. ;)

https://docs.google.com/spreadsheet/ccc ... 3emc#gid=1
Drafura
"1.5 direction with converts" seems to actually work the best imo. But hard to say, there's so many maps.

Also some maps seems to get overrated on all versions, like "Can't Defeat Airman [Holy Shit! It's Airman!!]" and "Kokou no Sousei [Collab]".
TheVileOne


I have major issues with this. Big Black and Kokou no sousei are not even close to being equal in difficulty. The difference between 360 BPM and 240 BPM is huge. Big black is also AR 10 and Kokou no sousei maps (Chaos and Collab) are AR 9. Jump intensity is also a huge factor.

This issue affects both models.

I can't beat Chaos. Someone else could evaluate that difficulty and have a more accurate opinion .
Topic Starter
Tom94

TheVileOne wrote:



I have major issues with this. Big Black and Kokou no sousei are not even close to being equal in difficulty. The difference between 360 BPM and 240 BPM is huge. Big black is also AR 10 and Kokou no sousei maps (Chaos and Collab) are AR 9. Jump intensity is also a huge factor.

This issue affects both models.

I can't beat Chaos. Someone else could evaluate that difficulty and have a more accurate opinion .
Referring to the "Chaos" diff:
First of all, simply taking the BPM does not work well. The big black mostly has 1/1 and 1/2 notes while kokou no sousei mostly consists of 1/2 and 1/4 notes, essentially making it closer to 480 BPM if you want to directly compare it to the big black.

Secondly AR is not factored into the difficulty algorithm. It will play a big role when computing pp, but it is not considered for beatmap difficulty / star rating.

My last point is, that the scoreboards indicate, that kokou no sousei is not very much easier than the big black.


The "Collab" diff of kokou no sousei definitely is a lot easier than the big black and should not be on par. It has a few hard parts but is mostly easy. Will look into that one.
Rorona

Tom94 wrote:

TheVileOne wrote:



I have major issues with this. Big Black and Kokou no sousei are not even close to being equal in difficulty. The difference between 360 BPM and 240 BPM is huge. Big black is also AR 10 and Kokou no sousei maps (Chaos and Collab) are AR 9. Jump intensity is also a huge factor.

This issue affects both models.

I can't beat Chaos. Someone else could evaluate that difficulty and have a more accurate opinion .
Referring to the "Chaos" diff:
First of all, simply taking the BPM does not work well. The big black mostly has 1/1 and 1/2 notes while kokou no sousei mostly consists of 1/2 and 1/4 notes, essentially making it closer to 480 BPM if you want to directly compare it to the big black.

Secondly AR is not factored into the difficulty algorithm. It will play a big role when computing pp, but it is not considered for beatmap difficulty / star rating.

My last point is, that the scoreboards indicate, that kokou no sousei is not very much easier than the big black.


The "Collab" diff of kokou no sousei definitely is a lot easier than the big black and should not be on par. It has a few hard parts but is mostly easy. Will look into that one.
IMO Big Black is still harder because it's AR10 instead of AR 9, and the patterns seem slightly easier. The only way it could be worth more is if you account that Kokou no Sousei is about 2 times longer.
Topic Starter
Tom94

Rorona wrote:

IMO Big Black is still harder because it's AR10 instead of AR 9, and the patterns seem slightly easier. The only way it could be worth more is if you account that Kokou no Sousei is about 2 times longer.
As said before, AR does not count for the difficulty calculated thus far. And map length indeed is factored in.
Luna
I'm by no means a good CtB player, so I can't really give feedback on the current difficulty ranking, but I'd be interested in seeing your alogrithm's opinions on these two maps:
https://osu.ppy.sh/p/beatmap?b=25580&m=2
https://osu.ppy.sh/p/beatmap?b=93893&m=2

Both of these use a special kind of pattern that is extremely hard to play but doesn't rely on big or fast jumps, so I think your difficulty calculation method might possibly underrate these?
VelperK

Tom94 wrote:

Rorona wrote:

IMO Big Black is still harder because it's AR10 instead of AR 9, and the patterns seem slightly easier. The only way it could be worth more is if you account that Kokou no Sousei is about 2 times longer.
As said before, AR does not count for the difficulty calculated thus far. And map length indeed is factored in.
Is this the same in standard?

And if so, how come?
Topic Starter
Tom94

VelperK wrote:

Is this the same in standard?

And if so, how come?
Yeah, it is the same in standard. Mainly because it doesn't really connect well with something like "aim" or "speed". Secondly there are many varying opinions - some people saying AR10 would be easier than 9 and others telling me the opposite.
It is factored in for pp calculations, so don't worry about that part.
eldnl

Luna wrote:

I'm by no means a good CtB player, so I can't really give feedback on the current difficulty ranking, but I'd be interested in seeing your alogrithm's opinions on these two maps:
https://osu.ppy.sh/p/beatmap?b=25580&m=2
https://osu.ppy.sh/p/beatmap?b=93893&m=2

Both of these use a special kind of pattern that is extremely hard to play but doesn't rely on big or fast jumps, so I think your difficulty calculation method might possibly underrate these?
Yes, they're hard.
Kitokofox

Kitokofox wrote:

So, I ended up sharing this to Tom94. I want to see everyone's honest and unbiased opinions.

Kitokofox wrote:

Catch The Beat calculation methods

Catch The Beat is different in multiple ways, by how jumps work and how streams work, mostly. Playing on a 2D plane is a lot diffrent from standard osu!, along with the predefined movement speeds of the catcher. Thus, skill based calculations based on the limitations of Ryuuta among other things must be taken in account for.

There are some main factors which determine how difficult a map is for Catch The Beat:
Map Settings This goes back to the basics.
- Approach rate: This should be documented in difficulty by the actual drop rate of the fruits. This also gives AR10 the much needed skill requirement it actually calls for.
- Fruit Size: This is pretty obvious. The smaller the fruits, the harder they are to catch accurately. Skill requirement is based on the field in which you can actually catch the fruit, rather than the size itself.
- Overall Difficulty: I don't know how this is handled in the map, but as far as I know it only affects the HP loss when missing. I don't think that applies too much to skill, but for players who can just barely pass a song, this would come into play.
- HP Drain Rate: Same as above, only comes into play for those barely making a map.
Beat Placement Factors based on the placement of sliders and circles themselves
- Movement: How much the catcher needs to move to catch the fruits. In stream maps like Marisa Stole the Precious Thing, this would become classified as less difficult since you don't have to click and simply inch the catcher, where in the system used for standard, it would be classified as difficult otherwise.
- Jumpiness: This is the general distance between beats, based on how fast the catcher must move. This also applied to sliders and their slider ticks. This also heightens more based on the general theta of jumps in a row (especially hyperdashes in a row) What I mean by theta of jumps in a row, as in a map with periodic jumps is probably more difficult than one with one part of it having multiple in a row (This sort of compares and contrasts between kinds of jumps and what kind of skill and perseverance is required to acquire a good score) Also, generally, faster hyperdashes are harder to chain off of, e.g. Can't Defeat Airman!.
- Speed: This is how fast the catcher needs to move around in succession. This would rate maps like val0108's Talent Shredder or With A Dance Number as high skill examples for this sector, mainly because of how fast the catcher needs to switch one direction to another and how fast of fingers the player needs.
- Long jumps: Originally pixel jumps, these are jumps that are enough for the catcher to reach with the new hyperdash system, but still require precision placement to catch, as well as timing. An example can be seen in Wild Eyes [JAMES DIFFICULTY!!!] As it has long but viable jumps.
Mode Specifics and Other Stuff
- Mode Specific Maps: Don't need to be buffed. They are probably already difficult as it is with the algorithm that's in place. Plus, it was set this way to entice players to play those maps. This pp system focuses on the player's skill, and tries to measure the skill needed for each beatmap, so making bonuses of that sort is incorrect to ranking..
- Fruits before and after spinners: The time before spinners and after spinners doesn't need to be checked because of how bananas aren't necessary to a full combo.
- Accuracy: In general, accuracy is easier to come by in Catch The Beat, and as such, changes may be made accordingly.
Pretty sure I gave this to Tom94 earlier. I hope this better explains my full intent of CtB and it's performance point calculation.
VelperK

Tom94 wrote:

TheVileOne wrote:



I have major issues with this. Big Black and Kokou no sousei are not even close to being equal in difficulty. The difference between 360 BPM and 240 BPM is huge. Big black is also AR 10 and Kokou no sousei maps (Chaos and Collab) are AR 9. Jump intensity is also a huge factor.

This issue affects both models.

I can't beat Chaos. Someone else could evaluate that difficulty and have a more accurate opinion .
Referring to the "Chaos" diff:
First of all, simply taking the BPM does not work well. The big black mostly has 1/1 and 1/2 notes while kokou no sousei mostly consists of 1/2 and 1/4 notes, essentially making it closer to 480 BPM if you want to directly compare it to the big black.

Secondly AR is not factored into the difficulty algorithm. It will play a big role when computing pp, but it is not considered for beatmap difficulty / star rating.

My last point is, that the scoreboards indicate, that kokou no sousei is not very much easier than the big black.


The "Collab" diff of kokou no sousei definitely is a lot easier than the big black and should not be on par. It has a few hard parts but is mostly easy. Will look into that one.
what kind of proof do i have to provide to refute this argument?
TheVileOne
Both are pretty difficult. The extra AR makes it more difficult for me, but I guess it could be seen as easier or harder depending on how well trained your eyes are at reading fast scroll speeds. AR 9 gives you more time (might be a placebo effect) to read the objects. Sometimes it is the case that if you need to read the objects as they are scrolling, you are already taking too much time to hit them correctly.

There are more red fruit jumps in Chaos and it is the longer map, which makes it easier to make a mistake over the course of the run. I'm not sure how the pros consider red fruits in terms of difficulty compared to awkward patterns. Some awkward patterns are more frame perfect than 99% of red fruit jumps.

Do you value hyperdash jumps as more difficult than large jumps that are not hyperdashed? Hypers can be very easy, even the very far ones. They get more difficult when you have to chain them. 1/4th hypers tend to be more precise than 1/2th.
Yukiteru Amano
Just a bit curious but, is there any date estimated for this to be done?
alienflybot
IMO, just DON'T make ctb diffs give too many pp.
-Ryuujii-
Why can't you guys just be patient lol xD
VelperK
Laharl

-Ryuujii- wrote:

Why can't you guys just be patient lol xD
Zak
It's out when it's out, deal with it.
eldnl

alienflybot wrote:

IMO, just DON'T make ctb diffs give too many pp.
This, please, a system bases on real difficulty wouln't need this.
show more
Please sign in to reply.

New reply