Skill Rating & Leaderboards Discussion [Updated Oct 16]

Re: Skill Rating & Leaderboards Discussion [Updated Feb 16]

hmmm im pretty borderline. looks like i gotta go get better scores to make sure that doesnt happen when the formula changes

Re: Skill Rating & Leaderboards Discussion [Updated Feb 16]

so, another thing that I have thought of:

This system needs another level of scaling for AAA equivalence -- particularly as the difficulty goes up. It goes without saying that a person getting mid-20s good count on something 95+ deserves more than a FEO/FMO equivalent. Devil's advocate because yes, they likely could AAA a song in that difficulty easily, but for the sake of balance...

Re: Skill Rating & Leaderboards Discussion [Updated Feb 16]

Yeah, I was about to bring this up...I don't really see how a 30g score is at the 0% mark when you need serious skill to get a 30g score on very difficult songs.

Re: Skill Rating & Leaderboards Discussion [Updated Feb 16]

This design choice was due to the lack of reliable data. The polynomial represented in the OP was carefully crafted using a wealth of community knowledge of how scores are relative to eachother linearly across all difficulties. That abundance of information becomes unpredictable when entering 30-40+ good-count territory (and is partially influence by educative mashing and pattern cheating). In order to modify the formula to adapt to this model, we would have to first collect data of impressive high good-count scores, and compare that data to the player's immediate actual skill. With the skill rating system, we will have the capacity to do something like this. With a concerted effort, we could foreseeable accurately weigh scores above 27g, but their impact would likely be negligible to the influence of SDG and teens scores.

Simply ignoring all songs with a good count of 28 or greater is not a solution, and this is something I want to actively work towards solving.

Re: Skill Rating & Leaderboards Discussion [Updated Feb 16]

You all did some badass work on this, nice job.

Re: Skill Rating & Leaderboards Discussion [Updated Feb 16]

Yay Max snap, come visit us from the dark alcoves of Langley Township. We miss you.

but srsly wow, amazing job on this

Re: Skill Rating & Leaderboards Discussion [Updated Feb 16]

I swear there are a bunch of people who are considered D7 who are now in D6 according to the update

oh well

Re: Skill Rating & Leaderboards Discussion [Updated Feb 16]

That polynomial approximation =3

Anyways, I just wanted to bring up one point made in the OP. It says that it's impossible to decrease your skill rating. However, isn't this still possible considering that scores are still being recorded using combo scoring and this measure of skill uses raw scoring? If someone "beats" their previous score with a better combo score but with worse PA, wouldn't their skill rating drop as a result?

Otherwise, I think this project is off to a great start and it appears to have relatively accurate results! Good job to all of those involved!

Re: Skill Rating & Leaderboards Discussion [Updated Feb 16]

I really think this should be coupled with a switch to Raw scoring on all stats....

Re: Skill Rating & Leaderboards Discussion [Updated Feb 16]

Yo just wanted to pop in and say how sweet this is. I've been off and on this site for like 8 years now(?) and stuff like this definitely makes me want to keep coming back. It's sweet to see FFR improving and while I haven't taken part of any tournaments or anything, I now have Tier Points (what, I'm not good?) and I'm considered Division 2 (again with the bad thing.) So I think all in all this is a dang sweet update.

Re: Skill Rating & Leaderboards Discussion [Updated Feb 16]

hello if i may sugest something i would like the coloring of the me and my clan to be changed aweay from snail entrail green to something else

Re: Skill Rating & Leaderboards Discussion [Updated Feb 16]

so to move up the rankings you actually have to beat your top 4 scores.


now would 2 goods on a 74 be better than say 12 goods on a 78? i'm not about to do all this math lol



or is there a way to see your 4 best scores?

Re: Skill Rating & Leaderboards Discussion [Updated Feb 16]

This new leaderboard system is well-developed; although, the FFR rank on profiles are not being reflected by the recent system update.
Example: My current FFR rank is 1,586; however, my profile says the FFR rank is 5,274.
If someone can look into the matter, I would greatly appreciate it.

Re: Skill Rating & Leaderboards Discussion [Updated Feb 16]

ok this seems like a brilliant idea but can someone explain heuristically why a score of 20 goods is supposedly that much better than a score of 28 goods or even 24 goods, and also why four scores are supposedly sufficient to determine a player's skill level when they could all be weighted towards the same set of patterns (like aim anthem, club, largiloqu-ican'tspellthisthing are all jack-heavy songs)

also your system assumes that the difficulty scale is multiplicative. i mean like if you have 2 people and one of them can comfortably aaa 30's and the other one can aaa 60's and you get the first person to play a 40 and the second person to play an 80 then they'll get roughly the same score. i'm not 100% convinced on this bit but that's just my impression so feel free to convince me otherwise.

sorry if i come across as overly critical, i really like the idea of sorting out ffr division placements once and for all, i'd just like more...... transparency

you forgot when someone lowers the difficulty rating of one of those four songs (as well as reuben_tate's raw scoring point, but that's probably a marginal case)

Re: Skill Rating & Leaderboards Discussion [Updated Feb 16]

So I guess that's a no