Fix Slideraim penalty for missing slider ends #29991
Conversation
|
Seems like if what is written in the OP is correct, this should go in with the current batch on changes? @ppy/osu-pp-committee |
|
@ppy/osu-pp-committee Any updates on what to do with this one? |
|
I'd like to see this in at some point. However, with #27691 being merged I do not see this as something super critical. I'd be more comfortable pushing it in with the current deploy if not for the fact that it causes changes on any non-FC score, since this means we need to consider balancing. |
This reverts commit 8987b21.
| @@ -181,8 +175,8 @@ private double computeAimValue(ScoreInfo score, OsuDifficultyAttributes attribut | |||
| estimateImproperlyFollowedDifficultSliders = Math.Min(countSliderEndsDropped + countSliderTickMiss, estimateDifficultSliders); | |||
| } | |||
|
|
|||
| double sliderNerfFactor = (1 - attributes.SliderFactor) * Math.Pow(1 - estimateImproperlyFollowedDifficultSliders / estimateDifficultSliders, 3) + attributes.SliderFactor; | |||
| aimValue *= sliderNerfFactor; | |||
| double sliderNerfFactor = (1 - attributes.SliderFactor) * (1 - estimateImproperlyFollowedDifficultSliders / estimateDifficultSliders) + attributes.SliderFactor; | |||
There was a problem hiding this comment.
Considering the known issue with this nerf is the lack of cubing resolved by the line below, I don't think touching the core nerf factor should be in scope of this change.
There was a problem hiding this comment.
Considering the known issue with this nerf is the lack of cubing resolved by the line below, I don't think touching the core nerf factor should be in scope of this change.
it's the opposite of what you think
the way I've done this is not touching core nerf factor, when any other variation of this are changing the current curve
https://www.desmos.com/calculator/z2bdvr5iht
Black curve is live, Red line is this PR, Purple line is what you proposed
| double fullComboThreshold = attributes.MaxCombo - countSliderEndsDropped; | ||
|
|
||
| if (scoreMaxCombo < fullComboThreshold) | ||
| effectiveMissCount = fullComboThreshold / Math.Max(1.0, scoreMaxCombo); | ||
|
|
||
| // Combine regular misses with tick misses since tick misses break combo as well | ||
| effectiveMissCount = Math.Min(effectiveMissCount, countSliderTickMiss + countMiss); | ||
| effectiveMissCount = countMiss; |
There was a problem hiding this comment.
What is your reason for changing this?
There was a problem hiding this comment.
the usage of estimators for lazer non-CL scores was added as a bandaid to nerf oshama scramble without merging this change
Part of this PR - #27303
This PR is fixing a mistake made back in 2021, when xexxar forgot that
SliderFactorneeds to be cubed before applying it to pp.The main effect of this is correct penalty for missing slider ends.
Side-effect of this is nerfing all non-FC scores, as they assumed to miss all sliderends. This is unfixable with stable scores but wouldn't be an issue with lazer scores, because you have info about exact amount of missed sliderends.