The "15" difficulty scale thread

[devonin]

Re: The "15" difficulty scale thread

The issue of scaling the difficulty based on the scores obtained on it is that a popular song with a very hard file is going to be artificially inflated as a bunch of people who suck at the game log terrible scores because they happen to like the track.

Also even if there were no difficulties, categories, numbers or names at all and just a rankig of songs from easiest to hardest, there will be debate over the order. no system will actually avoid the wrangling and disagreement.

[Reincarnate]

Re: The "15" difficulty scale thread

The issue of scaling the difficulty based on the scores obtained on it is that a popular song with a very hard file is going to be artificially inflated as a bunch of people who suck at the game log terrible scores because they happen to like the track.

Also even if there were no difficulties, categories, numbers or names at all and just a rankig of songs from easiest to hardest, there will be debate over the order. no system will actually avoid the wrangling and disagreement.

That's why you take into account the skill level of the people getting particular scores on your file.

Taking raw performance alone will not work. You'll have easy files that everyone can AAA, but tougher files that also have a high percentage of AAA's, but only because most of the scores come from people who know they're good enough to play the file in the first place that newbies won't ever touch much. Just account for that selection bias and you've got a more accurate system.

[devonin]

Re: The "15" difficulty scale thread

Oh that's all. Just account for the selection bias. We'll have it implemented tomorrow!

[Emithith]

Re: The "15" difficulty scale thread

my no-vote is a vote for the 13 difficulty system. I have my reasons, I just can't get them out of my head and onto paper text.

[ScarletSky]

Re: The "15" difficulty scale thread

The problem is, if you try to get the whole communities' input, you end up with a thread like this where no one agrees on anything.

The reason why qqwerf's idea worked so well was that, it implemented the community into a formula where the layout is designed on it's own.

[Reincarnate]

Re: The "15" difficulty scale thread

Oh that's all. Just account for the selection bias. We'll have it implemented tomorrow!

It's just basic math/stats, dude

[jimerax]

Re: The "15" difficulty scale thread

As for difficulty judgement system, I thought of an idea something like subin's other than "making automatic algorithm" and "hiring difficulty judges" ideas.

- Making difficulty subforum.
- Making threads for songs that have suspicious difficulties.
- If more than [XX%] of people who voted want a difficulty change when we get enough votes (I say a month/100 votes at least), difficulty will be changed.

Abusion with multi accounts will be a problem though.

[The_Toymaker]

Re: The "15" difficulty scale thread

The TS algorithm is broken, and we're having a debate over there what we should change it to. Here's the thread, just to add to the thoughts in this one. http://thirdstyle.com/forum/showthre...l-Difficulties =D

[MrMagic5239]

Re: The "15" difficulty scale thread

I just say lets stop taking FFR from it's original roots and leave it how it is. lol Keep pushing it further and further from what it originally was will make it, as usual, an exact replica of of Stepmania, just in flash form.

[devonin]

Re: The "15" difficulty scale thread

If a song on the 1-13 scale is on the border between 8 and 9, and so we make a 100-point scale, there will be songs on the border between 66 and 67.

There are always going to be borderline songs in any system where more than one song gets the same difficulty value assigned.

[stargroup100]

Re: The "15" difficulty scale thread

I initially wanted to keep this information to myself like a greedy **** (not that I'm implying my method is the best or anything) but in all honesty I think you guys should just concretely analyze the difficulty of a chart through an algorithm, and once that has been implemented, use JX's idea and do manual tweaks.

The problem is with statistical analysis is that as much as you can try to account for bias, it's still far too unstable and unpredictable to use as a method of scaling difficulty, and in the end it's still technically abusable.

Unlike strategic multiplayer games, where your decisions and strategies are dependent on the opponent's moves, there is a skill cap for rhythm gaming skill. Strategic games have no realistic skill cap in theory (assuming the game isn't unbalanced) so dynamic ratings are appropriate. However, there is a hard cap of skill based on physical limit for rhythm games. On top of that, nothing about your gameplay changes each time you play a chart, because the game is static. Static games should have static difficulty ratings, especially if there is already an existing cap.

Technically, a "perfect" difficulty algorithm for this game can't be made, but you CAN make one that is almost perfect. In retrospect, compared to a lot of other types of game analysis (such as strategic multiplayer balance), a difficulty algorithm is very simple to make. For some reason nobody has bothered to try. An automated algorithm would basically take care of everything, would rarely need updating, and is accurate. (Inaccuracies are minor but can be taken care of manually with JX's idea.)