Restore taiko accuracy to ScoreV2 values

[smoogipoo]

This has been brought up as an inconsistency several times:

• Closes Inaccuracies in taiko imported scores #25882
• Fixes osu!taiko "Good"s give less accuracy/score compared to osu!stable #9087

Note that this doesn't fix the accuracy circle/grade issue - that's an issue for all rulesets.

[bdach]

What does this mean for server-side usages of Judgement.ToNumericResult()?

https://github.com/ppy/osu-queue-score-statistics/blob/6d91b7738ad7adba513c86c382fb83b261dae5d6/osu.Server.Queues.ScoreStatisticsProcessor/Commands/Queue/ImportHighScoresCommand.cs#L652
https://github.com/ppy/osu-queue-score-statistics/blob/6d91b7738ad7adba513c86c382fb83b261dae5d6/osu.Server.Queues.ScoreStatisticsProcessor/Commands/Queue/ImportHighScoresCommand.cs#L741-L742

It appears that the import high scores command relies on the method to calculate accuracy, which will presumably be wrong for taiko scores now?

Isn't this the same sort of problem that we had with adding an AffectsCombo flag to Judgement, which implies that it is now necessary to construct all judgements on a beatmap to calculate the accuracy of a score correctly?

Also I'm not even sure why Judgement.ToNumericResult() isn't even so much as obsoleted if it's gonna be wrong for one ruleset out of four. It's a footgun in the making as soon as one of us forgets it's not supposed to be used anymore.

[smoogipoo]

because of course there'd be something that references the raw values, why wouldn't there be...

[bdach]

Well to be more constructive than the above, I think this is possibly salvageable if the method has a signature of GetNumericResultFor(HitResult) rather than accepting the judgement. Then the server-side usages can just construct the requisite score processor and go through that.

The Judgement.ToNumericResult() point still stands tho. Method should not be exposed outwardly as static on Judgement anymore if it's not universally valid.

[smoogipoo]

Perhaps it's time to bite the bullet and not expose it publicly...