Use proportion of difficult sliders to better estimate sliderbreaks on classic accuracy scores #31234
Conversation
6ac1067 to
b259e8d
Compare
|
I've updated this branch to the latest |
|
!diffcalc |
|
Difficulty calculation failed: https://github.com/ppy/osu/actions/runs/14682429178 |
c376097 to
ceda902
Compare
61da19c to
6ac1067
Compare
6ac1067 to
4c08c14
Compare
4c08c14 to
9dacc57
Compare
9dacc57 to
1bbb3f9
Compare
|
I've fixed up the compilation errors caused by dodgy rebases/merging. Please be careful with pushing your changes up in future. |
|
!diffcalc |
|
Difficulty calculation failed: https://github.com/ppy/osu/actions/runs/14774814568 |
|
!diffcalc |
|
!diffcalc |
|
!diffcalc |
|
!diffcalc |
Works out individual top weighted strain count for hitcircles and sliders, and then multiplies countmiss by 1 + the proportion of top weighted strain on sliders to account for misses on sliders sliderbreaks. This assumes that the player breaks equally on difficult hitcircles and sliders, which is the most fair assumption in my opinion.
The number of sliderbreaks is then scaled with combo, to account for the fact that higher combo scores have less places for sliderbreaks to hide in.
Spinners are handled incorrectly (they're included in the hitcircle strain count) but because spinners are 0 difficulty anyways there is negligible values impact, so making a better solution would probably be more effort than it's worth.
TopWeightedSliderFactormay be infinity on maps with no circles. This doesn't have any adverse values effects, becauseestimatedSliderbreaksis capped atcountMeh + countOk, but I'm unsure if it's going to cause any other weirdness.This probably wants some form of miss penalty or
effectiveMissCountadjustment, because currently every affected score is losing.Sliderbreaks may be double dipped, as it uses
effectiveMisscount + estimatedSliderbreaks, whileeffectiveMissCountmay already contain some estimated sliderbreaks. I'm not really sure of a good way to resolve this, and a little doubtful that resolving it would have any significant effect.