Various slider-related difficulty changes proposal #27303

Givikap120 · 2024-02-21T23:47:12Z

List of the changes:

Sliderless velocity is now calculated correctly, using JumpDistance instead of LazyJumpDistance
Stream spacing now calculated correctly, using JumpDistance instead of MinimalJumpDistance+TravelDistance
Penalty for missing sliderends now calculated correctly
Aim now have 1.08 multiplier for sliderjumps
Speed now have 1.25 multiplier for sliderstreams
HD now awards slideraim 2 times less (oshama scramble EZHD fix)

What to expect:

Oshama Scramble EZHD is nerfed
Sliderjumps maps are buffed (https://osu.ppy.sh/beatmapsets/656918#osu/1391631 is 6.72*)
Sliderstream maps are buffed (https://osu.ppy.sh/beatmapsets/881996#osu/1844019 is 7.17*)

This rework doesn't fix the sliders completely, and it doesn't touch complex slidershapes problem. It's main point is to fix the some flaws of the current slider system, until better system can be made.

to keep the same scale

peppy · 2024-02-22T00:56:45Z

Please change the title to something more descriptive.
Please link discussions where these changes were decided.

Givikap120 · 2024-02-22T01:07:58Z

Please change the title to something more descriptive.

Please link discussions where these changes were decided.

This is not "decided", this is rework proposal (rework proposals are not in the osu PRs anymore?)

added angle and velocity bonuses to sliderjump fixed alternating bonus

Xexxar · 2024-02-26T17:02:22Z

While I do believe that sliders are hugely problematic in their current state, I personally don't have much feedback to give on this PR as I believe a more fundamental rewrite of sliders is needed to properly assess their difficulty anyway, which is an ongoing effort of mine. This likely does improve things from a values standpoint, but it seems like something that would only be a stop gap solution until something more elaborate can be created. With that in mind, I personally don't see the need to push this within the context of what else is being worked on.

Givikap120 · 2024-02-27T18:18:18Z

While I do believe that sliders are hugely problematic in their current state, I personally don't have much feedback to give on this PR as I believe a more fundamental rewrite of sliders is needed to properly assess their difficulty anyway, which is an ongoing effort of mine. This likely does improve things from a values standpoint, but it seems like something that would only be a stop gap solution until something more elaborate can be created. With that in mind, I personally don't see the need to push this within the context of what else is being worked on.

this rework is a free value improve if we won't have new slider aim algo in next deploy

smoogipoo · 2024-03-07T09:38:51Z

!diffcalc

github-actions · 2024-03-07T09:39:11Z

Target: #27303
Spreadsheet: https://docs.google.com/spreadsheets/d/19JQ4wNDhO-6_TDiPRMJqYxBAfXGe8c1rJqdCp14Nles/edit

it used to depend on velocity, now it depends on spacing

smoogipoo · 2024-03-13T06:07:47Z

!diffcalc

github-actions · 2024-03-13T06:08:14Z

Target: #27303
Spreadsheet: https://docs.google.com/spreadsheets/d/1RwZimudCCfN5CMDrN6pjwbBikxYT_vE6IjG_Gcmh3O0/edit

now it's applied correctly

Most notable is on this map (was 8.2* in previous version) https://osu.ppy.sh/beatmapsets/2005921#osu/4172054

smoogipoo · 2024-04-09T13:35:26Z

!diffcalc

Good time to test out some updates I've made to the sheet generator :)

github-actions · 2024-04-09T13:35:55Z

Target: #27303
Spreadsheet: https://docs.google.com/spreadsheets/d/1ybGYfotEza5qVHVxtTbhOTG-jssMt50Db87MDU5xx9w/edit

smoogipoo · 2024-04-26T18:14:14Z

!diffcalc
SCORE_PROCESSOR_A=ppy/osu-queue-score-statistics#263
SCORE_PROCESSOR_B=ppy/osu-queue-score-statistics#263

github-actions · 2024-04-26T18:14:34Z

Target: #27303
Spreadsheet: https://docs.google.com/spreadsheets/d/1iWxnf_4Zu7exOWaaRFn9nlwLHAJA7s2WsBURVV8qRBU/edit

and changed speed sliderstream bonus

osu.Game.Rulesets.Osu/Difficulty/Evaluators/AimEvaluator.cs

tsunyoku · 2024-05-17T11:40:00Z

osu.Game.Rulesets.Osu/Difficulty/Evaluators/AimEvaluator.cs

+                    // Take just slider heads into account because we're computing sliderjumps, not slideraim
+                    sliderJumpBonus = slider_jump_multiplier * aimStrain * sliderJumpsAdjustingMultiplier;
+
+                    // Reward more if sliders and circles are alternating (actually it's still lower than several sliders in a row)


What is the rationale behind this? If we're alternating between sliders and circles then you're saying it's more difficult to aim than being slider jumps only.. but why?

in general - the meaning of buffing slider jumps is accounting for the fact that slider jumps are more difficult to process/read in comparison to normal jumps
but the cases where there's some notes in the middle of slider jump section or vice versa - it's not becoming any easier to read than just all sliders

you can create jumps test map with three cases:

only circles

alternating

only sliders

in my experience last two have similar difficulty, but maybe i'm just skill-issued

I just tested the jumps here:
With alternating bonus:

normal - 3.38*

alternating - 3.48*

sliders - 3.58*

Without:

normal - 3.38*

alternating - 3.45*

sliders - 3.58*

maybe i actually should revert alternating multiplier to 1.0x, cuz even 3.48 still feels too low

even with 2.0x it's still only 3.55*

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

osu.Game.Rulesets.Osu/Difficulty/Evaluators/AimEvaluator.cs

apollo-dw · 2024-05-19T20:05:15Z

osu.Game.Rulesets.Osu/Difficulty/Evaluators/AimEvaluator.cs

+                // Punish too short sliders to prevent cheesing (cheesing is still possible, but it's very rare)
+                double sliderLength = slider.Velocity * slider.SpanDuration;
+                if (sliderLength < slider.Radius)
+                    sliderJumpBonus *= sliderLength / slider.Radius;
            }


rename this variable to sliderDuration as length can ambiguously be understood to mean distance here.

i would be expecting something to do with follow circle radius here (2.4x). did you just not find that to be necessary? your comment does state that cheesing still occurs with this code, so something to consider

sliderjumps is not connected with sliderbody difficulty in any way, so there's no sense in using follow circle radius
cheesing can occur cuz you can hide long sliderbody under the circle, so the slider will appear like a circle
but i haven't seen it in ranked maps so it's good

also, it's indeed length, not a duration
do you misunderstanding something?
velocity * time = distance
velocity * duration = length

sliderjumps is not connected with sliderbody difficulty in any way, so there's no sense in using follow circle radius

but slider leniency will allow short enough sliders to effectively allow the movement as if it were circles. mappers have used this technique for over a decade at this point

also, it's indeed length, not a duration

yeah I think I just get tripped up by people using the term "length of time" in sentences. duration is just less ambiguous IMO

but slider leniency will allow short enough sliders to effectively allow the movement as if it were circles. mappers have used this technique for over a decade at this point

this is literally the point of sliderjumps
even tho technically they allow the same movement - they're still harder because they're harder to read

yeah I think I just get tripped up by people using the term "length of time" in sentences. duration is just less ambiguous IMO

how duration is less ambiguous if duration literally means time, when this is clearly length in pixels

... i think i confused myself on the duration and length thing again. i just read myself and i am not making any sense, disregard

osu.Game.Rulesets.Osu/Difficulty/Evaluators/AimEvaluator.cs

apollo-dw · 2024-05-19T20:08:38Z

osu.Game.Rulesets.Osu/Difficulty/Evaluators/AimEvaluator.cs

+                // If slider was slower than notes before - punish it
+                if (osuCurrObj.StrainTime > osuLastObj.StrainTime)
+                    sliderJumpBonus *= Math.Pow(Math.Min(osuCurrObj.StrainTime, osuLastObj.StrainTime) / Math.Max(osuCurrObj.StrainTime, osuLastObj.StrainTime), 2);


i'm confused by the comment and formula here. the comment talks about notes before the slider, but the slider is osuLastObj. so wouldn't the notes before be osuLastLastObj?

the comment could be less ambiguous too since when i read "slider was slower", i am interpreting this as the slider's velocity. but the code seems to be regarding a slowing down in rhythm

this line exist to not buff the cases like "jumps jumps jumps, 1/2 slider, jumps jumps jumps"
now I'm think about it - it's still have 0.25 multiplier, maybe i should drop it to 0

osu.Game.Rulesets.Osu/Difficulty/Evaluators/AimEvaluator.cs

apollo-dw · 2024-05-19T20:10:38Z

osu.Game.Rulesets.Osu/Difficulty/Evaluators/AimEvaluator.cs

+            double currVelocity = osuCurrObj.JumpDistance / osuCurrObj.StrainTime;
+
+            // This multiplier exists to prevent slideraim having sliderjumps bonus
+            double sliderJumpsAdjustingMultiplier = 1.0;


rename to sliderJumpNerfFactor or similar.

also why not have slider body bonus and slider jump bonus coexist? are there any particular cases you can point to where this doesn't make sense?

because lost umbrella +100

apollo-dw · 2024-05-19T20:13:18Z

osu.Game.Rulesets.Osu/Difficulty/Evaluators/AimEvaluator.cs

+                sliderBonus = osuLastObj.TravelDistance / osuLastObj.TravelTime;
+            }
+
+            aimStrain += sliderBonus * slider_aim_multiplier;


if the slider jump bonus and slider aim bonus are not meant to be applied together, then i would suggest throwing them into a max function. this would communicate their purpose much better, and it might render sliderJumpsAdjustingMultiplier redundant.

so i'd be expecting something like:
aimStrain += Math.max(sliderBonus * slider_aim_multiplier, sliderJumpBonus);

they meant to be applied together
sliderjumps account for the fact that aiming sliderhead is harder than normal circle
sliderbody is aiming the body after you hit the circle
sliderJumpsAdjustingMultiplier exist because i don't want sliderbody difficulty to be multiplied by sliderjumps bonus

so they are meant to be applied together, but also there is a nerf factor that "exists to prevent slideraim having sliderjumps bonus" and you "don't want sliderbody difficulty to be multiplied by sliderjumps bonus". this seems to be a pretty confusing contradiction on a conceptual basis, and i wouldn't expect anyone coming in with a fresh pair of eyes to be able to comprehend what's going on based on the code alone...

in my opinion it's pretty simple to understand that sliderjump bonus is for aiming to the circle, while slideraim is for aiming the body
aiming the body doesn't gets any harder if previous object was a slider, so i can't just use raw aim strain that includes body difficulty

i fully understand what slider jump bonus and slider aim bonus is, that doesn't really address my point of there seemingly being contradictions in approach here. if there are no contradictions here and i am mistaken, then the changes here need to be made much clearer

apollo-dw · 2024-05-19T20:17:17Z

osu.Game.Rulesets.Osu/Difficulty/Evaluators/AimEvaluator.cs

+                // Reward more if sliders and circles are alternating (actually it's still lower than several sliders in a row)
+                if (osuLastLastObj.BaseObject is HitCircle)
+                {
+                    double alternatingBonus = slider_jump_multiplier * osuLastObj.JumpDistance / osuLastObj.StrainTime;
+
+                    if (osuLastObj.StrainTime > osuLastLastObj.StrainTime)
+                        alternatingBonus *= Math.Pow(Math.Min(osuCurrObj.StrainTime, osuLastObj.StrainTime) / Math.Max(osuCurrObj.StrainTime, osuLastObj.StrainTime), 2);
+
+                    sliderJumpBonus += alternatingBonus;
+                }


i don't think i'm convinced with this being a factor in difficulty -- could you explain?

backread a bit

i can get behind the notion of "slider jumps are more difficult to process" just fine, but i don't agree with alternating between circles and sliders. in my experience, that offers a respite from the increased processing strain more than anything.

but the cases where there's some notes in the middle of slider jump section or vice versa - it's not becoming any easier to read than just all sliders

the strain algorithm works for this purpose. this shouldn't need any additional bonuses

try to do what i did:
compare the difficulty of alternating and sliders-only in test diff
alternating case will be considerably underweight compared to sliders-only

osu.Game.Rulesets.Osu/Difficulty/Evaluators/SpeedEvaluator.cs

apollo-dw · 2024-05-19T20:21:33Z

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

+        /// <summary>
+        /// Normalised distance from the start position of the previous <see cref="OsuDifficultyHitObject"/> to the start position of this <see cref="OsuDifficultyHitObject"/>.
+        /// </summary>
+        public double JumpDistance { get; private set; }


this naming is very confusing, because in theory this is even more "lazy" than LazyJumpDistance since it's ignoring sliderbodies on sliders

rename to HeadJumpDistance? CircleJumpDistance?

i don't like either of those. maybe swapping JumpDistance and LazyJumpDistance makes more sense. might need more opinions

i don't like it cuz it will be confusing for people who already know pp variables
maybe changing LazyJumpDistance to TailJumpDistance?

that's a fair point. TailJumpDistance doesn't quite get there IMO, because it should be general to every object type. something that covers the notion of it being lazy and from the end position would be good. LazyJumpFromEndDistance would be ideal for me, but i don't mind long variable names that are descriptive

Refactoring Changed slow sliders penalty Made velocity change bonus work correctly

Givikap120 · 2024-09-25T13:37:14Z

Making this Draft because I split-up all changes

stanriders · 2024-10-14T08:58:29Z

This PR has been split into smaller ones

Slider aim rework

245a2ea

pull-request-size bot added the size/M label Feb 21, 2024

adjusted multipliers

c037e02

to keep the same scale

This comment was marked as outdated.

Sign in to view

Givikap120 changed the title ~~Sliders mini-rework~~ Sliders difficulty calculation fix Feb 22, 2024

Givikap120 changed the title ~~Sliders difficulty calculation fix~~ Sliders difficulty calculation fix proposal Feb 22, 2024

bdach requested a review from a team February 22, 2024 06:55

bdach added the area:difficulty label Feb 22, 2024

Givikap120 added 2 commits February 22, 2024 10:29

Update AimEvaluator.cs

511e84d

added angle and velocity bonuses to sliderjump fixed alternating bonus

Update AimEvaluator.cs

2057225

Givikap120 and others added 3 commits March 12, 2024 11:01

Merge branch 'ppy:master' into slider_jumps

e4db975

minor changes to sliderstreams

48abd43

Fixed sliderstreams

7ecde21

it used to depend on velocity, now it depends on spacing

Givikap120 added 2 commits March 13, 2024 19:02

fixed high CS bug

3e18c90

now it's applied correctly

Added bandaid to fix one sliderjump cases

964bd63

Most notable is on this map (was 8.2* in previous version) https://osu.ppy.sh/beatmapsets/2005921#osu/4172054

pull-request-size bot added size/L and removed size/M labels Mar 18, 2024

Givikap120 and others added 3 commits March 18, 2024 22:09

Merge branch 'ppy:master' into slider_jumps

14a9b10

Fixed the bug

2812186

now sliderend penalty is applied correctly

2202d14

Givikap120 added 4 commits May 11, 2024 13:47

fixed sliderstreams having 0 spacing

ca61a82

and changed speed sliderstream bonus

fixed HD bonus for slideraim

cbb59b4

Update AimEvaluator.cs

8b3d414

Update SpeedEvaluator.cs

08b2c75

tsunyoku suggested changes May 17, 2024

View reviewed changes

Givikap120 added 3 commits May 17, 2024 15:35

refactoring

723bfd4

Update OsuPerformanceCalculator.cs

09af5f0

Update AimEvaluator.cs

89662e3

apollo-dw suggested changes May 19, 2024

View reviewed changes

Givikap120 added 5 commits May 20, 2024 00:29

Many changes

13864a6

Refactoring Changed slow sliders penalty Made velocity change bonus work correctly

Merge branch 'master' into slider_jumps

7549791

renamings

ffef4c6

multiple changes

a9191b5

added sanity check

358ed91

Givikap120 marked this pull request as draft September 25, 2024 13:37

minor refactorings

bb30113

peppy added the ruleset/osu! label Sep 27, 2024

Givikap120 added 2 commits October 1, 2024 00:55

improved alternating bonus

288c93c

Update AimEvaluator.cs

a0bcf21

Givikap120 mentioned this pull request Oct 8, 2024

Fix sliderdrop nerf adjusting the wrong value #30144

Closed

stanriders closed this Oct 14, 2024

Various slider-related difficulty changes proposal #27303

Various slider-related difficulty changes proposal #27303

Conversation

Givikap120 commented Feb 21, 2024 • edited Loading

This comment was marked as outdated.

peppy commented Feb 22, 2024

Givikap120 commented Feb 22, 2024

Xexxar commented Feb 26, 2024

Givikap120 commented Feb 27, 2024

smoogipoo commented Mar 7, 2024

github-actions bot commented Mar 7, 2024 • edited Loading

smoogipoo commented Mar 13, 2024

github-actions bot commented Mar 13, 2024 • edited Loading

smoogipoo commented Apr 9, 2024

github-actions bot commented Apr 9, 2024 • edited Loading

smoogipoo commented Apr 26, 2024

github-actions bot commented Apr 26, 2024 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Givikap120 May 17, 2024 • edited Loading

Choose a reason for hiding this comment

Givikap120 May 17, 2024 • edited Loading

Choose a reason for hiding this comment

Givikap120 May 17, 2024 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

apollo-dw May 19, 2024 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Givikap120 May 19, 2024 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Givikap120 commented Sep 25, 2024

stanriders commented Oct 14, 2024

Givikap120 commented Feb 21, 2024 •

edited

Loading

github-actions bot commented Mar 7, 2024 •

edited

Loading

github-actions bot commented Mar 13, 2024 •

edited

Loading

github-actions bot commented Apr 9, 2024 •

edited

Loading

github-actions bot commented Apr 26, 2024 •

edited

Loading

Givikap120 May 17, 2024 •

edited

Loading

Givikap120 May 17, 2024 •

edited

Loading

Givikap120 May 17, 2024 •

edited

Loading

apollo-dw May 19, 2024 •

edited

Loading

Givikap120 May 19, 2024 •

edited

Loading