Slider input does not match osu!stable

[thewildtree]

There are some major differences that I've noticed when comparing lazer with stable:

• Slider head moves together with the sliderball, while on stable it stays in place with sliderball moving independently. Indirectly mentioned in Distracting edge case of hitcircle/sliderball fade out #6088, almost certainly causes replay inconsistencies vs. stable, especially on sliders with high SV. This is best visible when sliderhead doesn't get hit at all (lazer left, stable right, full example here):

  

  By the way, judgement sprites seem to be displayed 'behind' the sliders, stable has them on top (as seen above)

• Slider circle range seems to be a bit different, same replay on lazer gives different results than stable. Video comparison here. Fixed.

Fixing these, especially the first one, should result in getting rid of major replay inconsistencies on maps with high SVs, like this one for example, since right now if a slider has very high SV and gets hit at least a bit late (even still in the 300 hitwindow), in lazer the head might move before getting hit, resulting in a miss.

[Wieku]

About slider head look at this TODO:

osu/osu.Game.Rulesets.Osu/Objects/Drawables/DrawableSliderHead.cs

Lines 41 to 43 in f3bb47f

	//todo: we probably want to reconsider this before adding scoring, but it looks and feels nice.
	if (!IsHit)
	Position = slider.CurvePositionAt(completionProgress);

About slider follow circle, in stable sprite doesn't reflect actual allowable radius. If I remember stable follow circle radius is 2.4 the hit circle radius, not 2. And it's virtual size is changed instantly.

osu/osu.Game.Rulesets.Osu/Objects/Drawables/Pieces/SliderBall.cs

Lines 98 to 99 in f3bb47f

	FollowCircle.ScaleTo(tracking ? 2f : 1, 300, Easing.OutQuint);
	FollowCircle.FadeTo(tracking ? 1f : 0, 300, Easing.OutQuint);

[Wieku]

And I think one more thing to reconsider with high SV maps is to bring back -36 ms offset on slider tail judgements, at least for replays coming from stable. They're visible in this video as white points, they denote if slider is fully completed or not: https://streamable.com/u0gbg

[thewildtree]

Yup, just noticed that TODO as well, thanks. Fixing this probably needs an accompanying change to slider snaking-out behavior though (EDIT: actually seems like it doesn't), I'll try to work on that.

I'm kinda surprised the slider tail offset hasn't been implemented yet, considering how many maps rely on it to make sliders possible to hit without dropping sliderends / combo.

[peppy]

The 36ms hack isn't added because it was a hack. It may be added for legacy replays but should not apply to new plays (for now).

[thewildtree]

For clarity, from discussion in the PR above:

Sliders are moving hitobjects. Lazer adds accuracy. Part of this addition is that if you hit late, the hitbox changes. This is intended for now.

Due to this, imo we should implement the 36ms hack + static sliderheads only for legacy replays (for now) to at least preserve consistency for old scores.

[bdach]

Note that replay inaccuracy is tracked at #5514.

[smoogipoo]

Fwiw, the 36ms offset is actually implemented. Check how nested hitobjects are generated for Slider.

The biggest gameplay difference is the tracking region. Static slider heads can be implemented for legacy skins.

[thewildtree]

Static slider heads can be implemented for legacy skins.

You meant legacy maps, right? Otherwise we'd have skins making a difference in gameplay mechanics and that's a no-go

[smoogipoo]

As mentioned in the closed PR, we aren't decided on that. The tracking region being 0.4 smaller than it should be is an actual bug, however.

[thewildtree]

Sorry, I meant to ask about legacy replays*. My bad.

[smoogipoo]

Potentially, but there's a lot more that we'll have to consider if we're to go down that path. So I'm not interested in doing that for now.

[peppy]

@smoogipoo propose moving this out of milestone for now as I don't think legacy replay compatibility is a priority we have for February/March at least.

[peppy]

Going to milestone this for now because I think we need to come to a conclusion on how to handle snaking. Likely by fixating the hit position until hit or otherwise, even when snaking out is enabled. I have no issues with doing this as long as it looks visually convincing (which it won't without special attention).