legit as fuck stepmania competition (idea phase)

[lurker]

I've been using EX Oni % scores in my testing since it's the default DP system for SM5, but that's pretty obviously flawed since it doesn't differentiate between any type of CB. The old DP system obviously has to go since it doesn't distinguish marvs and perfs. The MIGS system is the one we should most likely go with (for those who don't know, it's +3/+2/+1/0/-4/-8 with +6 for freezes), although the bias towards freezes is a little sketchy. SMO currently uses MIGS, which gives us a ton of data to work with (ignoring the fact that a lot of the score tables are broken as fuck). Just looking at a few popular songs, one can tell that 85% MIGS is typically around where people get the highest As and lowest AAs.
With that in mind, here's the formula I've been using to weight scores on rates:
weighted score = (raw score^(1/(rate^c)))*rate
where the raw score is given as a percentage and c is a constant. The maximum possible score on a given rate is, well, the rate itself (i.e 120% on 1.2). The higher c is, the more using rates boosts your score. I was using c = 0.25 for EX Oni %, but since MIGS typically results in lower percentages, it might be too strict. The idea was that you should be able to beat any score on a given rate if you can get a decent AA one letter grade higher, but this becomes stricter as rates go up for c < 1; that is, you need a higher raw score on 1.3 to beat all 1.2 scores than you need on 1.2 to beat all 1.1 scores. As an example, staiain's OML 1.6 score beats most OML 1.4 scores with c = 0.25 (raw MIGS score = 85.54%, weighted = 139.25%). A score of 92.1% MIGS on 1.5 would beat it, however.

I'm trying to make some progress on a difficulty scale as well, and I'm thinking something like this as the "foot rating" (not necessarily how songs would be weighted in the sets, but definitely a deciding factor) could cover everything easily:
(a*NPS+b*SPD+c*BRS+d*JAX)*LEN
NPS is the average NPS from the first note to the last.
SPD is the highest average NPS over a medium-ish length of time. I was thinking either a flat 20-30 seconds or some percentage of the song (15-25%?). Stargroup pointed out that this was too long. 5-10 seconds seems appropriate for this.
BRS ("burs") is the highest average NPS over a short (but not too short) length of time. I think 2-3 seconds would be enough.
JAX is some measure of the amount of jacks or trills. I think the highest NPS in two columns over 3-5 seconds would be decent, although that notably leaves out hand jacks (which makes you a terrible person if you use them) and possibly misses long mini jack sections. That was a really bad first attempt. I think this should be more like finding some weighting system for jack difficulty, then find the highest value for NPS*weighting over 3-5 seconds.
LEN is based on the length of the song. I'm not sure if just using the length in minutes would be good enough since it would hurt very short songs pretty badly, but if none of the songs are less than a minute and a half long it wouldn't be a problem. Thanks to Stargroup and some inspiration from Mad Matt's BP4 theme, I've thought up a better idea for this: LEN is a linear function of the longest period of time for which the average NPS is greater than 0.8*SPD+0.1*NPS. L will denote the length of said period of time.
a, b, c, and d are constants with a+b+c+2d = 1 and a > b > c > d. I was thinking a = 0.6, b = 0.3, and c = d = 0.1. Now I think a = 0.5, b = 0.3, c = 0.1, and d = 0.05 would be a good starting point.
I think the formulas from an SM5 NPS graph generator could be used to figure this stuff out (except perhaps JAX). A little help would be greatly appreciated.

EDIT: In the long run this is turning out to be non-critical for running this competition, but it is something to look into.

[Leo137]

i did somewhat like this before at our stepmania fb group, basically the contestants had to play the songs at SMO anytime in a term of two weeks and the scores were updated and the points were calculated. it went fairly well for being the first time so i think this could work too (local copy of the homepage)

[lurker]

FORMAT (first draft and missing a lotta numbers, based on the 4k community's size):
Qualifiers
First, there will be a qualifying round consisting of 5 songs. Your goal is to get the highest scores possible on those 5 songs in, say, 3-5 days. Rates will be allowed (and are absolutely critical to get the top score on 3 of the 5 files I've already chosen for this). The top 12 or top 25%, whichever is higher, will be the Competitive tier for the competition proper. The Casual tier will be open to anyone who sends in qualifying scores.

The Competition Proper
The official competition will take place over the course of a week. Each player will have to play two 90 minute sets during the week (the sets may be shortened since a lot of players don't need much break time), each with a different song list and point system. Once the qualifiers are tallied, the players in the Competitive tier only will choose the days (most likely in GMT, possibly server time) on which they will play their two sets, in order of their qualifier rank; the Casual tier may play their sets whenever they would like. There will be a cap on how many people may play in a day depending on the number of players (though this is subject to protest). When the competition is over, the scores will be scaled so that the highest score in a given set and tier receives 100 points. The person who has the highest combined score in each tier will be declared the winner.

The Sets
Set 1: Tiers
The song list will consist of 9 triples of files (27 in all) that gradually increase in difficulty. The first tier should be AAAable by the entire Competitive tier, while the last should be nearly impossible for them to AA. Additionally, there may be one or more additional tiers exclusively for the Casual tier, since being forced to play the highest tiers wouldn't be very fun.

Each tier consists of one song in each of these three categories:
Speed: Requires you to play at a high speed for medium-sized periods of time (15-20 seconds). May not necessarily have the highest sustained NPS of the three files, but the highest speed sections will contain relatively few jumps.
Stamina: Requires you to play at peak NPS for long periods of time (at least a minute). Jumpstream is very common. These will generally be the longest files in a tier and most will be over 3 minutes long, but there may be shorter ones with no significant breaks.
Other: Files that contain challenges other than speed or stamina. These will generally have some mix of jacks, unfriendly patterns, and lots of bursts.

Rates will not be allowed in this set. Your score will be the highest score you get in each tier, plus a bonus for each tier that you complete with more than 95% MIGS on every song (this is to promote decent MA since the next set has next to no room for that). The bonuses will increase exponentially per tier.

Set 2: Fundamentals
This will be a 26 song list ranked from A to Z (because ripping off FFRCP2 is totally ok), ranging from slightly easier than the Tier 1 files, to harder than the Tier 9 files. This time, rates are allowed, so the easier files will generally be longer to compensate. I'm currently debating between using a weighting system similar to BP4's Fundamentals where the points you get are scaled based on how hard the files are, and something similar to Osu!'s ranking system where you get 100% of the score on the hardest file you play, followed by 90% of the second hardest, (90%)^2 = 81% of the next hardest, etc., possibly with a final multiplier based on how many songs you get over 95% on. Regardless, unlike Tiers where you'll have to make some decisions about what to play to maximize score, here you only have one option: play harder files.

Rules and Such:
The files will be specifically packaged up for the competition, along with generic GFX to guarantee that you are playing the competition files. This is to make browsing through the files by difficulty easier, and to make the chart difficulty consistent across all files. Fail on isn't going to be a rule necessarily, but obviously it's highly recommended. The only mod which substantially changes the file which will be allowed is mirror.

To enforce the time limit, you will need to timestamp your pictures. There are themes for SM5 which have a clock visible at all times, but I don't know about 3.95. I may recommend using themes which have a visible clock or post how to add one to your theme if yours doesn't have one. To start the set, you must take pictures showing that you have no scores on any of the files yet (as in, take pictures of the entire folder in the song wheel showing that you don't have letter grades on them; this should take 3-4 pictures depending on your theme). Your time starts when you take the last picture you need to establish this. Any scores posted outside of the time limit will be reduced by half (failing within the time limit is easy, dude).

Videos are encouraged, obviously. Streaming is HIGHLY encouraged if you can, but try not to bump elbows with anyone else, ESPECIALLY if you're in the Competitive tier. SMO is usable for Tiers but not Fundamentals or qualifiers (unless you're playing the qualifying files on 1.0; for Fundamentals it would be a waste of time).
Lastly, for fuck's sake, don't be BS. This may be a competition, but it's mainly for fun. People will find out. Bitch.

Song lists will be posted shortly after dates are figured out.

[Dynam0]

Sickkk

(too lazy to read rules, just gonna smash keys randomly wooo)

[Wafles]

I like the rules

[Askur Yggdrasils]

Oh cool =O