HomePortalRegisterPublic FFRLog in

Share | 
 

 A new FFR Difficulty scale...

View previous topic View next topic Go down 
AuthorMessage
eiddiM

avatar

Join date : 2009-11-22
Age : 24
Location : Puyallup Washington

PostSubject: A new FFR Difficulty scale...   November 2nd 2010, 4:58 am

Like the one we have here, would this be a good idea or a bad one?
Should it be 1-100 instead?
Or is 1-13 good enough?

Opinions?

I personally like and would prefer the 1-50 scale. It's done exactly what it was intended to. I can gauge more easily how difficult a file is, not just a generalization of challenging or very difficult. I wouldn't mind the 1-100 scale that's in the thread either. I just like the idea that a player could gauge easier the difficulty of the file they're playing.
Back to top Go down
http://www.facebook.com/sevennightmareslong
Dossar
Simfile Author
avatar

Join date : 2010-04-14
Age : 23

PostSubject: Re: A new FFR Difficulty scale...   November 2nd 2010, 1:21 pm

I agree with 1-100. That way rather than making a borderline FGO/high FMO look like an 11, it can be 42 and 43 is the lowest FGO etc.
Back to top Go down
Bob bob(Dan)
Simfile Author
avatar

Join date : 2010-05-21
Age : 21
Location : Portsmouth, UK

PostSubject: Re: A new FFR Difficulty scale...   November 2nd 2010, 2:06 pm

1-100 or 1-50 would be fine

maybe 1-100, saves work and all songs are listed in that way
Back to top Go down
Saphira

avatar

Join date : 2009-11-21
Age : 32
Location : Budd Lake NJ

PostSubject: Re: A new FFR Difficulty scale...   November 2nd 2010, 4:44 pm

I think 1 - 100 is a bit much which is why we went with 1 - 50 here, I think that covers enough personally.

_________________
Back to top Go down
http://dragonsfury.forumotion.net
Bob bob(Dan)
Simfile Author
avatar

Join date : 2010-05-21
Age : 21
Location : Portsmouth, UK

PostSubject: Re: A new FFR Difficulty scale...   November 2nd 2010, 5:04 pm

Saphira wrote:
I think 1 - 100 is a bit much which is why we went with 1 - 50 here, I think that covers enough personally.

The reason i say 1-100, is only because:
1. On the diffuculties thread, its said out from 1-100
2.It would be still easy to distinguish files from VC/FMO for example

Back to top Go down
Saphira

avatar

Join date : 2009-11-21
Age : 32
Location : Budd Lake NJ

PostSubject: Re: A new FFR Difficulty scale...   November 2nd 2010, 5:40 pm

If they went from 1-100 I doubt terms like VC or FMO would even be used anymore. And it still wont change the fact that some files are a little high for one but a little low for another. There will always be that debate.

_________________
Back to top Go down
http://dragonsfury.forumotion.net
Scintill
Simfile Author
avatar

Join date : 2009-11-26
Age : 25
Location : Netherlands

PostSubject: Re: A new FFR Difficulty scale...   November 2nd 2010, 6:52 pm

1-100 to stop the confusion on how to unlock otaku heavy
Back to top Go down
eiddiM

avatar

Join date : 2009-11-22
Age : 24
Location : Puyallup Washington

PostSubject: Re: A new FFR Difficulty scale...   November 2nd 2010, 6:56 pm

That would work to.

and Kayla: Like i said i personally prefer your 1-50 scale over 1-100 Very Happy

Epic wins Very Happy
Back to top Go down
http://www.facebook.com/sevennightmareslong
jimerax



Join date : 2010-11-23

PostSubject: Re: A new FFR Difficulty scale...   November 23rd 2010, 12:26 am

FFR has secret 1-100000 difficulty scale already and songs are ordered by this parameter fyi

1-100 sounds good considering the # of songs imo
Back to top Go down
Jae-

avatar

Join date : 2010-04-16
Age : 31

PostSubject: Re: A new FFR Difficulty scale...   November 23rd 2010, 10:09 am

jimerax wrote:
FFR has secret 1-100000 difficulty scale already

o______________o"
Back to top Go down
MrRubix



Join date : 2010-11-05

PostSubject: Re: A new FFR Difficulty scale...   November 23rd 2010, 1:59 pm

Here is a really, really, really basic system that would be better than a purely subjective ranking.

If you want to get fancier with this, though, you'd want to apply some Bayesian weights to prevent high-variance fluctuations for low-data-point situations. You can also apply some standard deviation metrics if you want to exclude certain files from the ranking (for instance, if you slapped vROFL into the main list of songs, it'd hit Difficulty 100 without a doubt -- but maybe you'd want to reserve Difficulty 100 for a more realistic song, so you may auto-exclude things that would otherwise have a high difficulty rating but absurdly low average performance such that it falls past some certain SD mark (say, 2.5-3 standard deviations).


Average performance = (sum of all scores achieved across all players)/(perfect score for song * number of plays attempted on a song)

True Performance of song = Average Performance on song/Average skill of players

Where the skill of a player = Total points acquired by player/Sum of perfect scores from all songs played by player

Difficulty X = 100*(1-(TP-min(TP))/(max(TP)-min(TP)))

Then you can subdivide:
Difficulty 0-10 = Tier 1
Difficulty 11-20 = Tier 2
...
Difficulty 91-100 = Tier 10

____________________


SAMPLE SCENARIO:

Consider Players 1-10 who have skill ratings of 10-100%, respectively.
Consider songs A, B, C, and D where A is really easy, B is tougher but still simple, C is pretty tough, and D is mind-numbingly hard.

Average performance for A: 70% (played by everyone)
Average performance for B: 55% (played by everyone)
Average performance for C: 65% (played by players 6-10)
Average performance for D: 30% (played by players 6-10)

Taking skill percentage as a whole number (sum of 1-10 is 55 and sum of 6-10 is 40, so we say sum of skills of players 1-10 is 550 and sum of skills of players 6-10 is 400), we calculate True Performances:
True Performance for A: 70/(550/10) = 1.27272727
True Performance for B: 55/(550/10) = 1.00000000
True Performance for C: 65/(400/5) = 0.8125
True Performance for D: 30/(400/5) = 0.375

So in this case:
Difficulty A: 100*(1-(1.27272727-0.375)/(1.27272727-0.375)) = 0 (Tier 1)
Difficulty B: 100*(1-(1.00000000-0.375)/(1.27272727-0.375)) = 30.3797466 = 30 (Tier 3)
Difficulty C: 100*(1-(0.8125-0.375)/(1.27272727-0.375)) = 51.2658226 = 51 (Tier 6)
Difficulty D: 100*(1-(0.375-0.375)/(1.27272727-0.375)) = 100 (Tier 10)
Back to top Go down
DarkChrysalis

avatar

Join date : 2009-11-22
Age : 27
Location : FFR

PostSubject: Re: A new FFR Difficulty scale...   November 23rd 2010, 2:23 pm

MrRubix wrote:
Here is a really, really, really basic system that would be better than a purely subjective ranking.

MAAAAAAAAAAAAAAAAAAATH

super easy gaiz

So what you're saying is, make the new system so complex for normal human beings to figure out they stop whining about the current rating system???
































































nah it looks like it would work well. Of course, even with that kind of accuracy in the ratings people would still complain, and you would need to try and take care of certain skill sets that people have. People excel more at JS than they do with jacks, and vice versa. If a song depended on one of these or other things too much it could be skewed towards people good at that particular skill. Everyone that complains about this stuff though needs to realize that we're human and we can't have a perfect rating system lol
Back to top Go down
MrRubix



Join date : 2010-11-05

PostSubject: Re: A new FFR Difficulty scale...   November 23rd 2010, 2:26 pm

Shouldn't matter -- the point is that it would be a dynamic system that adjusts based on average song performance as well as the average skills of the players that play it. One person may disagree with a particular rating because it's a jack-filled song and that player happens to be good at jacks, but the idea is that the ratings should adjust such that most people would agree with them, as it'd all be performance-based.
Back to top Go down
Halogen-

avatar

Join date : 2009-11-22
Age : 25
Location : asdf

PostSubject: Re: A new FFR Difficulty scale...   November 23rd 2010, 3:08 pm

Even if a song were to depend on a specific aspect, there's always people who do poorly on said aspect, and then there's those who are very strong, which can give you a decent balance as long as there isn't an overwhelming group of people who are excelling/failing at a song, from which you could compensate through the averaging anyway.

This system would be totally usable if it didn't take so long to piece together and calculate, but I definitely understand it. To be honest, if it didn't require such a large amount of calculation, I would consider it.

_________________
. /l、
゙(゚、 。 7
 l、゙ ~ヽ
 じしf_, )ノ

2010 People: Most Helpful Poster, Smartest Poster
2010 Threads: Best Signature
2010 "The Game": Most Talented, Best Stepfile (Made in Saitama), Worst Stepfile (Myriagon [Oni]), Best Stepartist

2010 Hall of Fame Winner

Back to top Go down
MrRubix



Join date : 2010-11-05

PostSubject: Re: A new FFR Difficulty scale...   November 23rd 2010, 3:14 pm

How is this a large amount of calculation, if I may ask? It's fairly rudimentary and doesn't use anything absurdly rigorous in terms of statistics/mathematics/server-side resources/etc.
Back to top Go down
Halogen-

avatar

Join date : 2009-11-22
Age : 25
Location : asdf

PostSubject: Re: A new FFR Difficulty scale...   November 23rd 2010, 4:02 pm

Oh, I was talking about calculation by hand, not anything involving the server.

As a mathematical equation done server-side, using the scores would be nice - there's just one problem: we don't have a working database at the moment.

_________________
. /l、
゙(゚、 。 7
 l、゙ ~ヽ
 じしf_, )ノ

2010 People: Most Helpful Poster, Smartest Poster
2010 Threads: Best Signature
2010 "The Game": Most Talented, Best Stepfile (Made in Saitama), Worst Stepfile (Myriagon [Oni]), Best Stepartist

2010 Hall of Fame Winner

Back to top Go down
MrRubix



Join date : 2010-11-05

PostSubject: Re: A new FFR Difficulty scale...   November 23rd 2010, 4:05 pm

Here's a fun fact for you, though: There's a 90% correlation between the steps/second metric and the difficulty metric with respect to the current FFR data.

In other words, if you ranked songs by steps/second alone, you'd get a pretty good match.
Back to top Go down
Halogen-

avatar

Join date : 2009-11-22
Age : 25
Location : asdf

PostSubject: Re: A new FFR Difficulty scale...   November 23rd 2010, 4:32 pm

Well yeah, that's a good assumption because steps per second is usually an indication of overall step density. The only issues you run into are songs spikes in speed for short periods (ala Crowdpleaser).

_________________
. /l、
゙(゚、 。 7
 l、゙ ~ヽ
 じしf_, )ノ

2010 People: Most Helpful Poster, Smartest Poster
2010 Threads: Best Signature
2010 "The Game": Most Talented, Best Stepfile (Made in Saitama), Worst Stepfile (Myriagon [Oni]), Best Stepartist

2010 Hall of Fame Winner

Back to top Go down
Sponsored content




PostSubject: Re: A new FFR Difficulty scale...   

Back to top Go down
 
A new FFR Difficulty scale...
View previous topic View next topic Back to top 
Page 1 of 1
 Similar topics
-
» Hot Toys, 1/6th scale batmobile tumbler.
» New Halo toys - at the right bloody scale?!
» Thoughts on Difficulty
» my collection 1/2 scale HULK has new pants, whats your opinion, better or worse??
» Star Trek figures playsets, scale starships & action figures

Permissions in this forum:You cannot reply to topics in this forum
 :: Rhythm Games :: FFR General talk-
Jump to: