buff default small cs & slow jumps #15481

GoldenMine0502 · 2021-11-05T02:26:53Z

one of patch of https://docs.google.com/document/d/14REVGc-2Ubg5f6MHaITKLOY3HarZuL-ozU5Ow6ng2ig/edit

Changing the aim calculation algorithm (Aim.cs)

As you know, we are using the code calculating jump, Distance / StrainTime.
So, The two jump patterns have same value.

In fact, I think the yellow pattern is harder because we should locate the aim more longer.

for example, 4* dt map https://osu.ppy.sh/beatmapsets/338293#osu/753445 gives 372pp with SS.
6* none map https://osu.ppy.sh/beatmapsets/863227#osu/1805627 gives 380pp with SS. (same pp)
I think 6* none is more harder.

In the current ppv2 system, no matter how far the distance between notes is, if strainTime is long, they are regarded as a short jump.

But from now on, if the distance is far between the notes regardless of strainTime, they are given a bonus.

166ms means almost 90bpm(180bpm) jump.

In code, the original calculation result is slightly reduced,
and the static jump calculation is divided by 166ms.

plus, i changed the scalingfactor to give default small cs bonus. Mainly cs7+, It is underprevilieged i think.

As a result, it buffs HR.

https://www.desmos.com/calculator/vrjxscpa91 (old)
result scalingfactor graph
x-axis means CS
y-axis means ScalingFactor

before = red line
after = blue line

https://www.desmos.com/calculator/hc3ffb1koe
result static jump calculation graph
x-axis means BPM
y-axis means aimStrain when distance in px = 200.

before = green line
after = red line
difference = purple line

overall nerfed

JumpPlayers.txt

overall buffed

HRPlayers.txt

stanriders

Reasoning behind these changes seems incorrect to me. Your screenshot does indeed show movements with the same difficulty.
Overall this is a purely "meta change" instead of difficulty calculation improvement, which I think are considered a bad way forward. You're skewing the algorithm towards kinds of maps you think deserve more pp without actually improving parts of difficulty calculation that make those maps underrated. This also lacks any testing or balancing or even low-mid player level consideration - HR is already meta there.

stanriders · 2021-11-12T11:44:59Z

osu.Game.Rulesets.Osu/Difficulty/OsuPerformanceCalculator.cs

@@ -109,7 +109,7 @@ private double computeAimValue()

            double approachRateFactor = 0.0;
            if (Attributes.ApproachRate > 10.33)
-                approachRateFactor = Attributes.ApproachRate - 10.33;
+                approachRateFactor = (Attributes.ApproachRate - 10.33) / 1.5;


Why? How lowering high AR bonus supposed to buff slow jumps?

@stanriders as always they start from pretty old base with conflicts. Current value is

osu/osu.Game.Rulesets.Osu/Difficulty/OsuPerformanceCalculator.cs

Lines 162 to 164 in 593f11d

double approachRateFactor = 0.0;

if (Attributes.ApproachRate > 10.33)

approachRateFactor = 0.3 * (Attributes.ApproachRate - 10.33);

both for aim and speed, so they indeed buffed it

i should apply newer value. please wait a minute.

the reason why i nerfed high AR is to remain HDDTHR similarly, buff HR to pow scaling factor.
we have a lot of top performance of HDDTHR records. so i think we should remain HDDTHR records result.

stanriders · 2021-11-24T10:55:06Z

osu.Game.Rulesets.Osu/Difficulty/Preprocessing/OsuDifficultyHitObject.cs

            {
-                float smallCircleBonus = Math.Min(30 - (float)BaseObject.Radius, 5) / 50;
+                float smallCircleBonus = Math.Min(35 - (float)BaseObject.Radius, 0) / 40;


You give bonus for small circles staring from cs ~4.2. Current bonus starts from cs ~5.5 which makes more sense to me.

it makes scalingFactor according to CS has continuity.

note)Radius = 64 * ((1.0 - 0.7 * (CS - 5) / 5) / 2)

note) and same 54.4-4.48*cs

https://www.desmos.com/calculator/vrjxscpa91

and in my feel, we should buff from 4.2 to 5.4.

something strange need to change, min not max

stanriders · 2021-11-24T11:02:33Z

osu.Game.Rulesets.Osu/Difficulty/Skills/Aim.cs

@@ -29,6 +29,16 @@ public Aim(Mod[] mods)
        private double skillMultiplier => 23.25;
        private double strainDecayBase => 0.15;

+        private double calculateDefaultVelocity(double distance, double strainTime)


Why are you buffing velocity below 90 bpm and nerfing everything above? If you think velocity isn't being calculated correctly it shouldn't be changed here, it should be changed in the difficulty object processing

in OsuDifficultyHitObject, we should calculate it stupidly like
MovementDistance = calculateDefaultVelocity(MovementDistance, MovementTime) * MovementTime;

in conversation, it is.
In fact, I think the yellow pattern is harder because we should locate the aim more longer.
imagine insane HDDT farming maps and 180bpm 6* None farming maps. (the two type of maps give same pp)

apollo-dw · 2021-11-25T15:32:57Z

I completely disagree that the yellow pattern is harder. Furthermore, there is statistics works and experimental data that prove that the deviation of a hit spread on an object (which can be used to model difficulty) scales linearly with the velocity of a jump, which implies that the 2 patterns you posted in fact should have the same difficulty assuming both have the same velocity, see https://docs.google.com/document/d/1pXNl9sduYkQ_vXbO4v8SMM6uBk-zX8tHfDvB9kgX9dg/edit#

GoldenMine0502 · 2021-11-25T17:04:06Z

I completely disagree that the yellow pattern is harder.

give me logical reason. i said why the yellow pattern is harder.
6* jump map versus 4* dt jump map

on high bpm, short jump is also hard, so i applied the value limited to long, slow jump.

experimental data that prove that the deviation of a hit spread on an object

honestly, the biggest problem is there's no code or formula. so i can't say anything about it.

when a player get more greater aim skill, the standard deviation is decreased although the distance is longer.
for example, when playing easy map(example 1* or 2*), expert player has low standard deviation but newbie player has high standard deviation.

to prove your work, we need to compare same map with 1 digits jump player and 3 digits jump player, and we should get there's no deviation difference between them.

and i can't understand the difference between the standard deviation and simple linear function.

last, applying statistical value on not statistical factor is too hard to code.

stanriders · 2021-11-26T18:33:42Z

give me logical reason

It has been statistically proven by collecting replay data on test map by multiple people.

6* jump map versus 4* dt jump map

This is incredibly biased and also makes little sense. In my opinion both types of maps are relatively the same (assuming we're talking about the same mapping style on both ofc)

the biggest problem is there's no code or formula

There is, it's all in the linked docs

The rest I'm not too sure what supposed to mean, except the "graph" though - you pretty much proven yourself that relation is linear.

GoldenMine0502 · 2021-11-26T18:52:59Z

It has been statistically proven by collecting replay data on test map by multiple people.

it's not for the statistical thing. you got wrong point.

and i said, "The better aim skill, The smaller the deviation.".

we should compare expert player and newbie player at same map to prove.

"when a player get more greater aim skill, the standard deviation is decreased although the distance is longer.
for example, when playing easy map(example 1* or 2*), expert player has low standard deviation but newbie player has high standard deviation.

to prove your work, we need to compare same map with 1 digits jump player and 3 digits jump player, and we should get there's no deviation difference between them."

This is incredibly biased and also makes little sense. In my opinion both types of maps are relatively the same (assuming we're talking about the same mapping style on both ofc)

for example, 4* dt map https://osu.ppy.sh/beatmapsets/338293#osu/753445 gives 372pp with SS.
6* none map https://osu.ppy.sh/beatmapsets/863227#osu/1805627 gives 380pp with SS. (same pp)

do you really think 4* dt map is harder or same? i definitely no.

There is, it's all in the linked docs

really i can't find anything except osu-aim-tool. idk. where is pp calculation?

The rest I'm not too sure what supposed to mean, except the "graph" though - you pretty much proven yourself that relation is linear.

i said the statistical thing is wrong before. why should i prove already wrong thing?

bdach · 2021-11-26T19:06:05Z

@GoldenMine0502 Please keep caps lock and bold font off. There is no reason to get heated.

I, for one, cannot logically follow the argument you are trying to make. It's possibly because things are getting lost in translation, so let me ask some questions.

when a player get more greater aim skill, the standard deviation is decreased although the distance is longer.

The standard deviation of what, precisely?

to prove your work, we need to compare same map with 1 digits jump player and 3 digits jump player, and we should get there's no deviation difference between them.

Even if this is true, how does this relate to buffing slow jumps?

and i can't understand the difference between the standard deviation and simple linear function.

I don't understand the point the picture there is trying to make. What is the squiggly line? What is the straight line?

GoldenMine0502 · 2021-11-26T19:09:54Z

The standard deviation of what, precisely?

to prove your work, we need to compare same map with 1 digits jump player and 3 digits jump player, and we should get there's no deviation difference between them.

and i can't understand the difference between the standard deviation and simple linear function.

this is not direct related to my pr. its from the docs, https://docs.google.com/document/d/1pXNl9sduYkQ_vXbO4v8SMM6uBk-zX8tHfDvB9kgX9dg/edit#

bdach · 2021-11-26T19:11:27Z

I don't think it's unrelated if the document in question is being quoted as directly contradicting the changes in this pull request, which as I understand, it is being quoted as such.

GoldenMine0502 · 2021-11-26T19:29:00Z

overall nerfed

JumpPlayers.txt

overall buffed

HRPlayers.txt

I think it works.
We could nerf lengthbonus to prevent buff stream

stanriders · 2021-11-26T21:15:19Z

we should compare expert player and newbie player at same map to prove.

That's what the research does. Data came from a range of players from 3 digits to 6 digits, all running the same test maps.

Maps you linked are very different in structure and difficulty, so comparing them isn't valid.

really i can't find anything except osu-aim-tool. idk. where is pp calculation?

Pp calculation is the end result of the research that will eventually come, but isnt required for the research to be valid. There are formulas in the docs though, not sure how you missed them.

If you really want to see it in the code form you can frost's repository in the pp dev discord, it's pinned in the related channel.

abraker95 · 2021-11-26T22:16:15Z

we should compare expert player and newbie player at same map to prove.

That's what the research does. Data came from a range of players from 3 digits to 6 digits, all running the same test maps.

Just to clarify. While the doc examines data from a single player, data collected from other players demonstrate that the linearity still holds. Data from all players is made available in this spreadsheet. Those files can be loaded either via the aim tool found here or examined via the numpy python library.

smoogipoo · 2022-04-13T07:14:29Z

Closing due to inactivity to lessen the PR burden. Please discuss future changes in the PP discord. @ppy/performance-points-committee poke me if this is an incorrect closure and the change is good-to-go.

GoldenMine0502 added 4 commits November 5, 2021 11:10

default HR buff

0b5c4ef

nerfs high ar

059dbf9

revert gitignore

33aeea9

revert gitignore

5e9b738

pull-request-size bot added the size/M label Nov 5, 2021

peppy added area:difficulty ruleset/osu! labels Nov 5, 2021

change static value from 0.05 to 1.0

eb6b430

GoldenMine0502 changed the title ~~buff default small cs~~ buff default small cs & slow jumps Nov 5, 2021

GoldenMine0502 mentioned this pull request Nov 24, 2021

Make speed skill consider only the shortest movement distance #15758

Merged

stanriders requested changes Nov 24, 2021

View reviewed changes

GoldenMine0502 added 2 commits November 25, 2021 22:42

merge

c3bef43

update osuperformancecalculator

333a337

GoldenMine0502 added 6 commits November 27, 2021 02:06

min to max

0da2d07

50 to 80

5b7b1a4

same

3c9d2cc

adjusted

729686f

adjusted 2

6b0a62b

adjust 3

a203c2d

smoogipoo closed this Apr 13, 2022

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

buff default small cs & slow jumps #15481

buff default small cs & slow jumps #15481

GoldenMine0502 commented Nov 5, 2021 •

edited

Loading

stanriders left a comment

stanriders Nov 12, 2021

Wieku Nov 24, 2021 •

edited

Loading

GoldenMine0502 Nov 25, 2021 •

edited

Loading

stanriders Nov 24, 2021

GoldenMine0502 Nov 25, 2021 •

edited

Loading

GoldenMine0502 Nov 25, 2021

GoldenMine0502 Nov 25, 2021 •

edited

Loading

GoldenMine0502 Nov 25, 2021

GoldenMine0502 Nov 26, 2021 •

edited

Loading

stanriders Nov 24, 2021

GoldenMine0502 Nov 25, 2021

apollo-dw commented Nov 25, 2021

GoldenMine0502 commented Nov 25, 2021 •

edited

Loading

stanriders commented Nov 26, 2021 •

edited

Loading

GoldenMine0502 commented Nov 26, 2021 •

edited

Loading

bdach commented Nov 26, 2021

GoldenMine0502 commented Nov 26, 2021 •

edited

Loading

bdach commented Nov 26, 2021

GoldenMine0502 commented Nov 26, 2021 •

edited

Loading

stanriders commented Nov 26, 2021

abraker95 commented Nov 26, 2021 •

edited

Loading

smoogipoo commented Apr 13, 2022

	double approachRateFactor = 0.0;
	if (Attributes.ApproachRate > 10.33)
	approachRateFactor = 0.3 * (Attributes.ApproachRate - 10.33);

buff default small cs & slow jumps #15481

buff default small cs & slow jumps #15481

Conversation

GoldenMine0502 commented Nov 5, 2021 • edited Loading

stanriders left a comment

Choose a reason for hiding this comment

stanriders Nov 12, 2021

Choose a reason for hiding this comment

Wieku Nov 24, 2021 • edited Loading

Choose a reason for hiding this comment

GoldenMine0502 Nov 25, 2021 • edited Loading

Choose a reason for hiding this comment

stanriders Nov 24, 2021

Choose a reason for hiding this comment

GoldenMine0502 Nov 25, 2021 • edited Loading

Choose a reason for hiding this comment

GoldenMine0502 Nov 25, 2021

Choose a reason for hiding this comment

GoldenMine0502 Nov 25, 2021 • edited Loading

Choose a reason for hiding this comment

GoldenMine0502 Nov 25, 2021

Choose a reason for hiding this comment

GoldenMine0502 Nov 26, 2021 • edited Loading

Choose a reason for hiding this comment

stanriders Nov 24, 2021

Choose a reason for hiding this comment

GoldenMine0502 Nov 25, 2021

Choose a reason for hiding this comment

apollo-dw commented Nov 25, 2021

GoldenMine0502 commented Nov 25, 2021 • edited Loading

stanriders commented Nov 26, 2021 • edited Loading

GoldenMine0502 commented Nov 26, 2021 • edited Loading

bdach commented Nov 26, 2021

GoldenMine0502 commented Nov 26, 2021 • edited Loading

bdach commented Nov 26, 2021

GoldenMine0502 commented Nov 26, 2021 • edited Loading

stanriders commented Nov 26, 2021

abraker95 commented Nov 26, 2021 • edited Loading

smoogipoo commented Apr 13, 2022

GoldenMine0502 commented Nov 5, 2021 •

edited

Loading

Wieku Nov 24, 2021 •

edited

Loading

GoldenMine0502 Nov 25, 2021 •

edited

Loading

GoldenMine0502 Nov 25, 2021 •

edited

Loading

GoldenMine0502 Nov 25, 2021 •

edited

Loading

GoldenMine0502 Nov 26, 2021 •

edited

Loading

GoldenMine0502 commented Nov 25, 2021 •

edited

Loading

stanriders commented Nov 26, 2021 •

edited

Loading

GoldenMine0502 commented Nov 26, 2021 •

edited

Loading

GoldenMine0502 commented Nov 26, 2021 •

edited

Loading

GoldenMine0502 commented Nov 26, 2021 •

edited

Loading

abraker95 commented Nov 26, 2021 •

edited

Loading