This is a continuation thread based on the feedback from the FC implementation thread. Please go read that thread to know why this metric seems to be needed on FFR.
So, what is a AA ?
Basically a AA on other games serves the purpose of determining if you did "well enough" on a given chart, without necessarily being close to perfection on it. On Stepmania for example, this grade is obtained by getting 93.00% accuracy or better.
What is accuracy on FFR ?
Since FFR is frame based, it makes sense for its judgement values (Amazing, Perfect, Good, ...) to be defined by some frame window around the exact note timing. In any case, frames can always be converted to ms timing. That being said, the simplest way one could define accuracy on FFR is by comparing the obtained raw score on a chart with the AAA raw score on that same chart. This allows for any configuration of non-perfects to be a AA, as long the raw score is sufficiently high.
Other games like Etterna will consider even the slightest ms timing difference to compute the accuracy. In other words, a 100% is extremely rare and only achievable on super easy/short charts by very talented accuracy players. In the context of FFR, that would mean considering Amazings in the computation, where an Amazing would result in more points than a Perfect. Given the reputation of FFR to be relatively unstable at that scale, I think it's safe to say that Amazings should not make a difference.
When it comes to a specific % requirement, I believe it can be discussed mostly subjectively. I personally don't see why we couldn't use a round number like 90%. One interesting piece of data that could be useful is how much % can one get on average when just mashing charts. This would give a lower bound to the discussion. I don't have the time right now to compute that, but if someone could, that'd be very helpful.
What about chart structure ?
Chances are you havent asked yourself that question yet. Chart structure matters when you compare two charts, and one has a single difficulty spike whereas the other one is more balanced overall. This currently affects FFR in a major way since the AAA metrics that we use do not account for such chart structure, resulting in what is commonly known as "farm files". Well, the same logic applies to any AA metric, except that it introduces even more of these "farm files".
Although the AA metric would be even slightly more "broken" than our AAA metrics, it is far less of a big deal. The reason is that AA would never be FFR's main skill measure. Knowing that, there's no real need to fix AAA stuff before implementing AA. There's a non-0 chance that if/when difficulty is redefined in the future, AA will have to follow the changes, but that's probably the least of our concerns given the other implications that would come with a redefinition of difficulty.
What would AA give to users ?
Obviously, the first thing associated with AA's would be a completion metric just like the AAA's and FC's (and maybe SDG's eventually). Then, some Tokens and Skill Tokens could be made to use this new tracked grade.
________________________________
So what do you think of this suggestion, and do you have anything you'd like to see regarding this AA grade ?