09-10-2016, 07:52 PM | #21 |
O Derby, Where Art Thou?
Join Date: Jul 2003
Location: Virginia
Age: 32
Posts: 1,011
|
Re: Stepmania Difficulity Correcter [WIP]
|
09-10-2016, 08:02 PM | #22 |
longing
Join Date: Dec 2007
Location: Ontario, Canada
Posts: 2,680
|
Re: Stepmania Difficulity Correcter [WIP]
I'm just sitting here trying figure out how to build a tree of the directory structure, and I realize I just need to compare a given songs parent folder to the root, and recursively create a tree of that, then do the same for the next level.... Shit
Maybe another day, before my head and the program explodes. |
09-11-2016, 02:59 AM | #23 |
FFR Player
Join Date: Sep 2015
Age: 22
Posts: 107
|
Re: Stepmania Difficulity Correcter [WIP]
thanks so much for this! i mean it obviously has issues considering it's pulling diffs from randomnumbergenerator.com and it doesn't take into account higher difficulties, but it's a lot better than nothing
|
09-11-2016, 08:37 AM | #24 |
longing
Join Date: Dec 2007
Location: Ontario, Canada
Posts: 2,680
|
Re: Stepmania Difficulity Correcter [WIP]
Dang from my logs theres a bunch of downloads already, glad people are making use of it.
If anyone is interested in the source code or how it works feel free to pm me and I'll send you the code/explain it. The only thing interesting about it is my use of custom data structures and the way I make use of objects for storing folders/specific songs, but most of it is pretty basic stuff. I'm glad I used java though, ez |
09-11-2016, 06:49 PM | #25 |
lol happy
Join Date: Oct 2005
Location: DESTINY
Age: 33
Posts: 12,193
|
Re: Stepmania Difficulity Correcter [WIP]
Okay honestly this is a really good idea, and it's what I was hoping Mina's difficulty calculator would eventually become. Because if you pair it with a decent algorithm it actually gives the community the ability to patch their entire song folder and be talking about the same difficulty scale. This is something the SM community has been sorely needing for what like, a decade and a half now?
So I really hope you aren't discouraged by people bashing on the algorithm, because that's unrelated to what you've made and the piece you worked on actually looks quite nice. That said I stand by my statement: SML's algorithm is garbage and until that changes I will prefer relying on pack authors to at least made the difficulties consistent with each other within each specific pack. I hope you can find someone with a better algorithm (read: Mina, but maybe there's some other ones out there too idk) and pair this tool up with their work, because I'd use it in a heartbeat if that was the case.
__________________
Last edited by hi19hi19; 09-11-2016 at 06:51 PM.. |
09-11-2016, 11:36 PM | #26 |
FFR Player
Join Date: May 2011
Posts: 13
|
Re: Stepmania Difficulity Correcter [WIP]
Writing an accurate difficulty algorithm is pretty hard, although I imagine it's not as bad for keyboard/spread since you don't have to deal with fingers changing location in the pattern analysis.
|
09-12-2016, 12:11 AM | #27 | |
Celestial Harbor
|
Re: Stepmania Difficulity Correcter [WIP]
Quote:
|
|
09-12-2016, 02:05 AM | #28 |
FFR Player
Join Date: Dec 2007
Location: nima
Posts: 4,278
|
Re: Stepmania Difficulity Correcter [WIP]
you know it's a testament to the quality of my work that it still holds up reasonable results at a first glance
|
09-12-2016, 02:25 AM | #29 | |
FFR Player
Join Date: Dec 2007
Location: nima
Posts: 4,278
|
Re: Stepmania Difficulity Correcter [WIP]
Quote:
I feel like pad charts have a more "sterile" pattern environment and the variation and more importantly the differential contributions to difficulty thereof makes keyboard analysis much more difficult. The hard part I've found isn't accounting for any singular physical reality of the game, it's accounting for many different ones that may present similarly to each other without being either redundant or too explicit. An extremely fast burst might lead into a fast stream, and whether or not it's "super cbrushable" and very difficult or just "mashy then streamy" might depend on the transition point being 30 ms closer or further. Or it might depend on how the ending sequence of the burst forces you to manipulate it relative to which fingers the beginning patterns of the stream focus on. Or it could depend on both, or neither. Or it could depend on the length of the burst, or any of these things. And that's just consideration for whether or not a single frame of time lasting maybe a half second might drastically increase the difficulty of the file or be a negligible factor. And that's assuming you get far enough juggling all the more macro pattern analysis to even get to something localized within half a second. More importantly it's entirely possible that the half second in question is only relevant factor in the file's difficulty. See linus' dengeki tube edit in odi3 for the perfect example of this. Anyway, some time ago there was a pad pack released here. Apocalypse something, I think. I ran it through one of the very early iterations of my kb oriented difficulty calculator and I from what I remember more than 90% of the results lay within +-1 of the values provided by the pack organizers after scaling. That's better accuracy than I get on most keyboard packs by a longshot. But neither of us is going to try to build a calculator for the other and admittedly I know next to nothing about pad, so we can agree to disagree. |
|
09-12-2016, 02:55 AM | #30 |
FFR Player
Join Date: Apr 2012
Posts: 422
|
Re: Stepmania Difficulity Correcter [WIP]
i wouldn't recommend changing the difficulty number because of SMO side effects. The scorelists wont work right and also those numbers are likely to change in the future. Better would be to take the difficulty numbers from files and display them in-game using lua theming. It also frees everyone from requesting sml to calculate the same files repeated times, as you could just share the files with the values.
As it takes a while to process, i'm afraid it might be too consuming for the server to keep replying those massive requests, so it helps for it too.
__________________
My latest works: BCMP-ish (6key) Solo 137 Pack (6key) Jumpstream of Fighters vol.1 Jumpstream of Fighters vol.2 6k key discord |
09-12-2016, 02:56 AM | #31 |
Senior FFR citizen
Join Date: Jun 2007
Location: Leuven, Belgium
Age: 38
Posts: 1,537
|
Re: Stepmania Difficulity Correcter [WIP]
Good initiative. Mfw Ice Candy Pop is rated 21 and Chipstream 10 though. Can this be a problem of averaging difficulty over a file? Because Chipstream is really easy for most of the file and then has a few speed-ups that make the ending hard. You should try your algorithm on something like Aragorn. Of course, even if it rates something like Aragorn as a low difficulty file, it can be argued whether that is necessarily a bad thing, since it depends to what extent difficulty should reflect how easy a file is to FC or AAA or how hard it feels on average. Anyway, good luck tweaking it!
__________________
Sick nature |
09-12-2016, 03:04 AM | #32 |
FFR Player
Join Date: Jan 2012
Posts: 136
|
Re: Stepmania Difficulity Correcter [WIP]
From the looks of it your program is indeed rather neat on its own, the major issue is that it utilizes data that has been generated by an algorithm that makes use of hilariously flawed logic in regards to the situation it is meant to address.
This implies that the people that will be most open to validating your efforts will, as you have, generally tend to express a certain degree of leniency under the simple pretense that it's better than nothing. Which is perfectly fine and understandable all things considered assuming that's indeed all they want, but only as long as you don't aspire nor attempt to be taken seriously by people that actually do value reasonable results being produced while conforming to a logically-sound margin of error. As it currently is, SML's algorithm inflates difficulty by orders of magnitude based on pattern density - thus jumptrills and chordjacks/handjacks tend to be quite prevalent in the highest-rated files difficulty-wise. You may nonetheless believe the results to be relatively reasonable when looking at Nuclear Blast packs due to their comparatively lower differences in pattern density across files, but what happens if you take more content into consideration? If you happen to be curious enough, then I would like you to attempt to AA the songs I have listed below in red, or at the very least obtain scores within the same range. Hint: If you are a low to mid D5 player then I guarantee you this won't be happening. As you can see, I have paired up files together based on their relative proximity in SML-generated difficulty values, surely this must mean you should be perfectly capable of producing roughly similar scores... right? The answer may surprise you. Diavolo (Light) 1.0x rate - Yolomania 3 - 10.27 Jackhammer Madness 1.0x rate - Hard Songs Megapack 1 - 10.25 HAELEQUIN 1.0x rate - Nuclear Blast 5 - 19.34 Never Had A Friend Like Me 1.0x rate - midare megapack 3 - 19.43 Tower Of Saints 1.0x rate - Nuclear Blast 4 - 21.59 Loli fishing next 20000 leagues (Oni) 1.4x rate - FFR Community Pack 4 (FFRCP4) - 21.42 Paraclete (Oni) 1.0x rate - Nuclear Blast 5 - 22.71 Sandstorm 1.0x rate - Hard Songs Megapack 1 - 21.70 There's most certainly people out there that would be capable of explaining this far better than I have attempted to just now, not to mention that there are countless other factors pertaining to why SML ratings are pure garbage (yes this is merely only ONE facet of it), but I don't think that would be worth the time and effort to elaborate further given what I have already pointed out so far. At any rate, my intention with my statements is not to disparage your efforts in their absolute entirety, but rather to elaborate upon where exactly the backlash is coming from, and moreso why it is justified. By taking even a brief moment to put critical thought about what the data you are working with truly amounts to from a logical standpoint, you should be able to realize why the number of people willing to take it seriously isn't anywhere near as high as you had initially hoped it to be. Last edited by DeadSignal; 09-12-2016 at 03:14 AM.. |
09-12-2016, 09:08 AM | #33 |
longing
Join Date: Dec 2007
Location: Ontario, Canada
Posts: 2,680
|
Re: Stepmania Difficulity Correcter [WIP]
Lol man its so garbage to be honest, it got me motivated to play some disgusting songs though like destroyer of the universe however
There's this song in nuclear blast with captain Falcon as the stepper logo or w/e that legit destroys me and it gets rated as a 21. I dunno man I don't do this for people to get hyped about it, it is what it is, honestly I'd rather have a discussion about it like what's going on above. Ill read more later gotta go to class. I'll just say the algorithm isn't the focus unfortunately (I wish I could make my own or open source one, but it's above my level right now :/) LOL and YoshL saying he doesn't put me on blast, get outta here bro ahaha Last edited by Dinglesberry; 09-12-2016 at 09:09 AM.. |
09-12-2016, 09:38 AM | #34 |
FFR Player
Join Date: Apr 2012
Posts: 422
|
Re: Stepmania Difficulity Correcter [WIP]
That the algoritm sucks or not ain't your problem, is out the focus of your program and honestly you are just doing the best that you can with the availiable resources. Everyone who is critizing the algorithm is derailing massively from the thread topic, so that's sad ;(
Of what i noticed from the screenshots, it doesn't work quite right for files with multiple difficulties, the only solution that comes to mind is to process the same file for each diff, stripping all of them from the file excepting the current one to process
__________________
My latest works: BCMP-ish (6key) Solo 137 Pack (6key) Jumpstream of Fighters vol.1 Jumpstream of Fighters vol.2 6k key discord |
09-12-2016, 03:37 PM | #35 |
longing
Join Date: Dec 2007
Location: Ontario, Canada
Posts: 2,680
|
Re: Stepmania Difficulity Correcter [WIP]
Either way would be possible, I think I'll end up "stripping" the lower difficulties and then doing the hardest, and realistically adding the easier ones to it unchanged.
Once I implement an actual file selection menu, could even perhaps have what difficulties you want selectable. I don't know if the algorithm scales down to lower difficulties though, so I dunno EDIT: ooh I just thought, I should make songs themselves their own object with more data like a key, the File itself, and stuff like # of difficulties, that'd be the easiest way to implement the other stuff I feel... Last edited by Dinglesberry; 09-12-2016 at 03:40 PM.. |
Currently Active Users Viewing This Thread: 1 (0 members and 1 guests) | |
|
|