|
Post by Amefurashi on May 3, 2014 13:13:51 GMT
We need a porn thread!!!!srsly thoLook at this I mean why? Why make difficulty based on a formula? There are tons of things to consider. Section 1 nps: Overall note density (averaged nps) Peak note density (30 seconds) lowest note density (30 seconds) Section 2 notes: (alongside note density) overall note count Section 2.1 patterns: note patterns jackhammer usage spam usage abnormal note placements (you know, those off putting note placements usually made by new charters) Section 2.2 LNs (if long notes are implemented): LN count LN complexity Section 3 Eye candy: Speed changes (and other gimmicks) J-notes (dunnno if these will be fixed) FOE Section 4 Others: Chart length (dependent on length of peak nps considering dry playtime (parts without notes)) (I could miss other stuff)
But think about it, yes if a comprehensive difficulty formula is made, wow. But it will have its flaws *looking at o2jam difficulty system (where something like cannonloid could be level 100+ and a challenging LN chart will be lvl 40. Speed changes are disregarded) I think we need some difficulty ratings board. A group of seasoned FtB players/charters who are reliable (will not give inconsistent ratings) This is just my opinion.
|
|
|
Post by Decon082 on May 3, 2014 17:20:13 GMT
Every system would have its flaws. You've actually summed up many of the key points that I'm trying to take into consideration for the design of this new system.
Ignoring LNs, the system would be based on pretty much everything else you listed. Some things are quite difficult to quantify, such as the types of patterns, so I feel like there will not be a number on a scale for the difficulty rating. All you might see is a number or percentage signifying where it fits in among all of the other charts on the site in terms of difficulty (this song is in the hardest 10% of songs, for example).
Of course all of this is subject to change, since this difficulty system is extremely complicated. But I would like to get something completed, and then it can be determined whether it is a good system or not. If it's complete, it can still be modified later to improve some of the difficulties. If it never gets complete, we're stuck with one of the systems that have been in place previously, which obviously have their flaws too.
tl;dr, I'm working on a difficulty system. Hopefully it will be fairly accurate, but don't expect it to be perfect.
|
|
|
Post by paulnet on May 3, 2014 20:50:26 GMT
I was actually about to make the same complain. And I'll make the same suggestion I made a long time ago, Comparing.
As an example, pick 11 songs to be 0-10. Let's say Silence is 0 and Freedom Dive imp is 10 (meaning anything impossible is 10) and fill up everything from 1 to 9. When we get to judge the difficulty of a chart, we just compare it to the 11 songs. Let's say 5 is Storm World, 6 is decon's senya chart and the chart we want to compare is Jeux D'eau. The result is clearly going to be 6. Of couse 0 to 10 is a too small number so the difficulty jumps will be too high to accurately compare things. That's what BMS's rating system is lacking: clear landmarks for comparing, if it wasn't for it, they wouldn't have so much trouble with charts being harder than one with a higher difficulty rating. And that's what we lacked in the past too.
Making a difficulty formula is the same as trying to make a robot chart. (if you want to know how well it went, look at osu!mania's, it's even based on osu notes) The best way to rate chart difficulties is to rely on humans. Not completely of course, what would be of charters if they didn't know the basics of charting? That's why we need those "landmarks" to help us rate charts.
<end rant>
|
|
|
Post by nivrad00 on May 3, 2014 20:51:57 GMT
I don't have anything important to add but I agree with paul
|
|
|
Post by Decon082 on May 4, 2014 11:58:19 GMT
As an example, pick 11 songs to be 0-10. Let's say Silence is 0 and Freedom Dive imp is 10 (meaning anything impossible is 10) and fill up everything from 1 to 9. When we get to judge the difficulty of a chart, we just compare it to the 11 songs. Let's say 5 is Storm World, 6 is decon's senya chart and the chart we want to compare is Jeux D'eau. The result is clearly going to be 6. Of couse 0 to 10 is a too small number so the difficulty jumps will be too high to accurately compare things. All you might see is a number or percentage signifying where it fits in among all of the other charts on the site in terms of difficulty (this song is in the hardest 10% of songs, for example). I feel like these are similar concepts... Do you agree? If we had a scale from 0-10, the song in the hardest 10% would end up being a 9. Only difference would be that these charts are ranked by a formula and not by these comparisons. The best way to rate chart difficulties is to rely on humans. Not completely of course, what would be of charters if they didn't know the basics of charting? That's why we need those "landmarks" to help us rate charts. But let's look at some of the things that happened when we had players rate the difficulty. - Many people including me didn't know how to judge difficulties, so ratings were inaccurate.
- When people get better, songs become easier, so they may rate them easier even if they are the same difficulty.
- Some songs just get rated extremely easy for invalid reasons. (Dash Hopes 3)
- Multiple difficulties, but only one rating.
Problem #4 can easily be fixed in FtB4. Problems #1 and #3 can be fixed by having a trusted group of charters rate the difficulty based on the landmarks above. Personally I think the members of this "rating team" would need to agree on a difficulty while a chart is still in Pending, before the chart can get out of judgement. But then we have more things to think about. Who would be on this team? How much effort and cooperation would it take? What if people don't agree on the rating? What if the landmarks change like in Problem #2 above? Could an accurate formula rate songs in the same way so that this team isn't needed? I'm just rambling at this point, but I do have an actual idea to add in. Since we're all used to changes here, and having things be tested constantly... What if we try both methods at the same time? For a while at least. This would require every chart on the site to be manually rated again by the "rating team", which would take a lot of effort, but once that happens, we can compare how the formula does in comparison to how the team has rated each chart, and then take an average. If the formula and the team end up consistently giving the same rating, then we can try just using the formula or keeping an average. If there is a lot of variation, then we can get rid of the rating team formula because it is bad, or continue to take an average. The point is that we won't know until we try them both.
|
|
|
Post by Amefurashi on May 4, 2014 13:41:29 GMT
Lemme join the rating team hehehehe (Im the only who played all charts)
Slightly off topic: I remembered the DJMAX diff system lol Fermion MX 8b (FC) level 15 Nightmare NM (barely passed with a C) level 14 8b Space of Soul HD (B) level 12 8b Grave consequence MX (A) level 14
I lol everytim
|
|
|
Post by paulnet on May 4, 2014 20:10:30 GMT
- Many people including me didn't know how to judge difficulties, so ratings were inaccurate.
- When people get better, songs become easier, so they may rate them easier even if they are the same difficulty.
- Some songs just get rated extremely easy for invalid reasons. (Dash Hopes 3)
- Multiple difficulties, but only one rating.
Problem #4 can easily be fixed in FtB4. Problems #1 and #3 can be fixed by having a trusted group of charters rate the difficulty based on the landmarks above. Personally I think the members of this "rating team" would need to agree on a difficulty while a chart is still in Pending, before the chart can get out of judgement. But then we have more things to think about. Who would be on this team? How much effort and cooperation would it take? What if people don't agree on the rating? What if the landmarks change like in Problem #2 above? Could an accurate formula rate songs in the same way so that this team isn't needed? Problem #2 is automatically solved by the system itself. The songs we're going to compare the charts to will CONTINUE the same, remember we're not comparing charts to numbers, but to charts, which means 19/30 wouldn't change with time. I agree on having both of the methods being tested though. Even if I'm sceptic about a formula that actually works, I'm not sure how it'll end up being. Also we can easily just make a chart go out of pending ONLY if it was already rated on difficulty. As of now, before changes are made, we can make an unofficial difficulty rating thread and go with it.
|
|
|
Post by nivrad00 on May 4, 2014 21:35:56 GMT
Lemme join the rating team hehehehe (Im the only who played all charts) okay hold on now
|
|
|
Post by Decon082 on May 4, 2014 23:15:35 GMT
Lemme join the rating team hehehehe (Im the only who played all charts) okay hold on now 102% completion... well he has a point!
|
|
|
Post by Amefurashi on May 6, 2014 12:44:59 GMT
The problem with #2 are the old charts. I mean what was the hardest chart that time? Boss Rush? 3y3s? I mean the diff rating given by the board will be consistent and unchanging (when official) because we have reached the point of inpossiblity, that is Freedom Dive (Circus Gallop also lol). So the diff ratings shouldn't change.
|
|
|
Post by Decon082 on May 6, 2014 14:19:41 GMT
The problem with #2 are the old charts. I mean what was the hardest chart that time? Boss Rush? 3y3s? I mean the diff rating given by the board will be consistent and unchanging (when official) because we have reached the point of inpossiblity, that is Freedom Dive (Circus Gallop also lol). So the diff ratings shouldn't change. So we're in full agreement that Freedom Dive will always be 30/30. (Even though there was that one person who gave it a 29 on the old ratings...) And obviously joke charts like Tullsespam would also be 30/30 because they are impossible as well. In theory, yes, there can be charts more difficult than Freedom Dive. But if they are humanly impossible to pass, then they would all deserve the same maximum rating. There was never that impossible chart before, which is why a lot of older charts that were 18 difficulty, I could pass with 98 or 99% (most of their ratings were given several years ago and not changed). While charts that were given 18s more recently were not even passable for me.
|
|
|
Post by Zach72 on May 8, 2014 5:02:47 GMT
I still hold the opinion that a specific few team of qualified raters would be the best way to determine difficulty . . . going off the reasons said above about the many variables involved making formulaic analysis close to impossible, and the fact that the opinion of many is super subjective (not to say that the raters wouldn't be, but specifically chosen to rate, I'm assuming they would try to be as least biased as possible and maintain consistent standards). gooddiscussionthanksforpostingtho
|
|