FFR Stats Bomb

[Trumpet63]

Google Sheet

[FF_rules]

In a nut shell. Songs were either too easy or too hard back then lol

Could I request Highest/Lowest NPS vs Difficulty?

[Trumpet63]

Originally Posted by FF_rules

Could I request Highest/Lowest NPS vs Difficulty?

Do you mean Highest divided by Lowest, or two things: highest v difficulty and lowest v difficulty?

[ilikexd]

Really cool, thanks! A minor note: NPS is traditionally calculated by the community as the time between the first and last note divided by the number of notes, but the song lengths listed on the website are those of the entire duration of the music including before the first note. So as a result the NPSs generated are slightly off, anywhere from about .1-.2 to almost 1.4.

Fun requests if you're willing to do them:
average difficulty for each stepartist with >1 file in the game
average difficulty for each music artist with >1 file in the game

[Trumpet63]

Originally Posted by ilikexd

Really cool, thanks! A minor note: NPS is traditionally calculated by the community as the time between the first and last note divided by the number of notes, but the song lengths listed on the website are those of the entire duration of the music including before the first note. So as a result the NPSs generated are slightly off, anywhere from about .1-.2 to almost 1.4.

Fun requests if you're willing to do them:
average difficulty for each stepartist with >1 file in the game
average difficulty for each music artist with >1 file in the game

1. That's a good point, and I'm glad you made it.

2. Yes yes, any requests I can manage.

[One Winged Angel]

now to fuck around with all the difficulties until the first graph has perfect positive correlation

[ilikexd]

Originally Posted by Trumpet63

2. Yes yes, any requests I can manage.

There are actually 273 music artists with 2 or more of their songs in game, 161 with 3 or more, and 114 with 4 or more, so it might be more manageable to just do those with 4/5 or more. Shouldn't be a problem with stepartists though, I think.

[Arch0wl]

This is so awesome hnngh

[Trumpet63]

I have the average difficulties for all of the authors and stepauthors on the sheet, per your request

[FF_rules]

Originally Posted by Trumpet63

Yeah that's exactly what I meant; thanks

So that outlier in the high 60's for max NPS, which song is that?

[ilikexd]

Originally Posted by FF_rules

Yeah that's exactly what I meant; thanks

So that outlier in the high 60's for max NPS, which song is that?

VS Boss Battle

[FF_rules]

Originally Posted by ilikexd

VS Boss Battle

That's what I was half expecting lol.

[Arch0wl]

That average NPS chart is practically a direct correlation

I think a combination of five variables:

min NPS

min sustained NPS (5 seconds or more)

average NPS

peak NPS

peak sustained NPS (5 seconds or more)

would be enough to remove user-rated difficulty all together. The only reason average NPS is sometimes insufficient is when a chart has a low difficulty throughout but a very hard section of notes that tanks your lifebar. But this would standardize difficulty across all lifebar systems.

For pad ratings this would be far more complicated since some movements use way more calories than others to perform. A jump should really be counted as 3 arrows, not 2, since you have to actually lift your whole body off the ground; a jump step should be 5 arrows (the step before a jump adds to the difficulty)

BUT, since movement is barely a factor in keyboard charts (the only exception I could think of is index charts that use complicated crossovers) this effectively means almost every chart can be rated automatically using the above five metrics.

[ilikexd]

Originally Posted by Arch0wl

That average NPS chart is practically a direct correlation

I think a combination of five variables:

min NPS

min sustained NPS (5 seconds or more)

average NPS

peak NPS

peak sustained NPS (5 seconds or more)

would be enough to remove user-rated difficulty all together. The only reason average NPS is sometimes insufficient is when a chart has a low difficulty throughout but a very hard section of notes that tanks your lifebar. But this would standardize difficulty across all lifebar systems.

For pad ratings this would be far more complicated since some movements use way more calories than others to perform. A jump should really be counted as 3 arrows, not 2, since you have to actually lift your whole body off the ground; a jump step should be 5 arrows (the step before a jump adds to the difficulty)

BUT, since movement is barely a factor in keyboard charts (the only exception I could think of is index charts that use complicated crossovers) this effectively means almost every chart can be rated automatically using the above five metrics.

Not quite. Pure NPS-based considerations come close to community-recognized difficulty, but they fail to recognize when many arrows fall on the same column or columns within a period of time, either forming jacks, which well exceed the difficulty of an average stream at the same NPS, or trills, which do likewise. There are also sections that just bias toward a specific 3, 2, or 1 columns; they can be recognizable and nameable as a specific pattern, or just a more or less difficultly-patterned stream. Approaching true difficulty using only NPS-based considerations would need to actually look at those five factors (and possibly more) on each individual column by itself, while also slightly weighing columns 1 and 4 heavier, giving heavier weight to outward movements vs inward movements (21 and 34 vs 12 and 43). There are a few other things such as transitioning between different rhythms and movements, focusing on a singular skillset vs. a diverse one, and other things which don't seem to be very easy to represent with mathematical formulae.

One could come up with a long mathematical formula consisting of many variables that mimics the community thought process though, given enough brainstorming.

There is clearly a direct NPS correlation though, but it just wouldn't be appropriate to use by itself since it fails to predict the outliers that show up based on some of things I said.

[FF_rules]

To add to the 5 variables, wouldn't patterns of rest be a variable we can consider? (consistent 1/16th stream vs 1/16th jacks/gallop are considerably different and possibly harder but would fall as a lesser rating in your 5 variable list)

[One Winged Angel]

I wish it could be simplified as some combination of those five variables but it just wouldn't work. Difficulty based solely on NPS would ignore patterning completely. Assuming a traditional 'spread' setup, an incredibly dense 1324 wall would force a player to hit two one-handed trills for the duration of the pattern. This would be interpreted (incorrectly) as being equally as difficult as a 1234 or 4321 wall, where the player could synthesize [12] and [34] jumps for the wall while still maintaining perfect accuracy, drastically reducing its difficulty compared to the former. 250bpm 16th one handed trills (or jacks lmao) would be treated the same as 250bpm rolls etc. You'd be forced to include patterning in the difficulty calculation and that makes it a whole lot messier.

edit: didn't see (or read) Nathtan's post but I'm just going to assume I was ninja'd

edit2: ninja'd plus some other bits, I like the outward vs inward bit very true

[Trumpet63]

I'd also like to point out that some elements of pattern difficulty and endurance are dependent upon what finger setup you're using.

Loving this thread right now btw.

[TheSaxRunner05]

I think there are a lot of 2-3 point difficulty changes that could be made to charts in the 45-65 range, but FFRs difficulty system is pretty close overall.

[FF_rules]

Would "Song Difficulty vs Amount of 99%+ scores" be a thing that's already in your system or would you have to add entries for player scores?

[Trumpet63]

Originally Posted by FF_rules

Would "Song Difficulty vs Amount of 99%+ scores" be a thing that's already in your system or would you have to add entries for player scores?

I don't have the count of 99%+ scores, but I do have count of AAA. Would that suffice? Otherwise, I don't think that information is easily accessible.