Ranking V3 received many updates since it was released but most of these were bandaid fixes or updates that never applied to the whole ranked pool. This meant recalculating the ratings as-is was near impossible. Additionally, there were significant issues with edge cases and the balance of different modes (One Saber) and modifiers (FS, SFS, GN).
Some maps had their pass ratings buffed massively due to leans or palm-ups
One Saber was overweight relative to the standard maps in the pool
Modifiers were balanced for midspeed maps but not for acc or the top end. Additionally, the current accuracy AI struggles with high eBPM maps which leads to them being overweighted, especially for extremely linear cases
buff = 1 + 0.01 * (njs - 24)
Very low note maps are susceptible to abusive gameplay (e.g. pause abuse leading to abnormally high accuracy plays), breaking the accuracy curve
1 / (1 + Math.Pow(Math.E, -1.4 - (data.Count / 50)))
where data.Count is swing countThe Ghost Note modifier provided a flat buff which is easy to cheat
https://docs.google.com/spreadsheets/d/1cm86cEHpcgmcze_5IKURciFNEe1btFtwwgNPa6Qs25I