| Forum Index > General Games |
|
MegaManEXE
United States845 Posts
But it's still not very good because it is inconsistent and vague, and it doesn't help that peppy won't disclose any of the actual details about how it works and how it calculates beatmap difficulty and whatnot Even if you are good at FCing things you don't last long in the top 500, I get 98-99% or higher FCs on a majority of the maps I play and within a month I'm barely staying in the top 500 just because I get outscored by mods | ||
|
Heh_
Singapore2712 Posts
| ||
|
ThaZenith
Canada3116 Posts
Something like: TriGoon achieved rank #26 on Fujita Maiko - Nee (TV Size) [Hard] (osu!) PP +4 | ||
|
MegaManEXE
United States845 Posts
On July 31 2012 05:41 ThaZenith wrote: I wish you could tell wtf is affecting your PP. It's annoying seeing jumps on the graph but not knowing exactly where they're coming from. Something like: Nope we can't have that in the game That would make sense, you see None of that in peppy's osu | ||
|
Bobbias
Canada1373 Posts
On July 31 2012 05:55 MegaManEXE wrote: Nope we can't have that in the game That would make sense, you see None of that in peppy's osu Which is why a hacked better version needs to be released... I dont think I'll be able to do that by myself though. Here's a mini essay explaining my feelings about osu right now. Notes on skill level, and aptitude: A quick note first: In this post I use a lot of statistical analysis concepts, because to design something liek this it's necessary to have some level of understanding. I'll try to explain things in as plain terms as possible. I also talk about "skill level" and "apparent skill level". Skill level - Skill level is the innate ability of that player, which affects how well they can play the game. What sets skill level apart is that you cannot truly know a player's skill. You can only know a player's "apparent skill level". Apparent skill level - Apparent skill level is a metric which can be calculated based on how well the player has been playing recently. The problems with apparent skill. To calculate this value there are a few things that need to be taken into consideraton: You must be able to compare how well two payers do on a map. You must also compare how difficult maps are compared to each other. The problem here is that to compare two maps, you need to have every other variable involved unchanging, and only compare one variable which contributes to difficulty. This means that either you need to use only one player to compare difficulty, or you need to use many players at the same skill level, but with different aptitudes for the different vriables, and find an average aptitude which can be used to correct for each player's aptitude and normalize the difficulty data. Essentially, the problem is that not everyone is good at the same things, which skews the data depending on which aptitudes are required to do well on a map. To correct for these we need to find a way to get a sort of "profile" of the aptitudes of everyone who plays osu (or a good enough sample). when we do that we can see how much better or worse someone is than the "average" osu player, and we can then correct their score based on that difference. If we do that for each aptitude for a player, we can fom a more accure picture of the player's "apparent aptitude". Why do I think a hacked version is necessary? Because peppy's design for osu is flawed as a competitive game. How is it flawed? It is flawed because there are too many variables which require statistical analysis in order to correct for, and present an "apparent skill" metric. AR OD Circle Size Mapping patterns Backgrounds Storyboards Videos Skins Note colors Combo size Accuracy And more. Every one of these possible aptitudes can bias the apparent skill calculation by creating bias towards one player who's aptitudes are better matched to the map they are playing. In short: There are too many variables which may affect the apparent skill level of a player. What changes need to be made? I'm not claiming to completely understand how to fix osu. The only claim I'mmaking here is that there's something wrong in the first place. Ultimately, these variables need to be removed from the calculation equation to simplify apparent skill calculation to a level which makes it useful. My suggestion is to make approach rate and circle size fully selectable, make it so backgrounds, storyboards, videos, and skins are all selectale or toggleable on and off (to a selectable default state). Overall Difficulty should be fixed at some value, except forwhen you use HardRock. When a player is able to customize a variable, it allows them to select the value where they play the best, this removes bias towards whoever is better at the map's supplied default value chosen by the mapper, and allows for comparison between two players playing their best. Overall Difficulty should be fixed, because it affects player's accuracy, and makes comparing one song to another harder, because you would need to do statistical analysis like I described above in order to correct how much effect a score on a song should have on a player's apparent skill. Hardrock needs to be mentioned here because og how it affects OD. I don't believe that HR should be changed in any way, becaue if OD is locked during normal play, we can use a statistical analysis of proficiency at hardrock to to find a correction factor, which is essentially a value saying "how much harder is hardrock's OD from standard OD", which we can use as a multiplier in the calculation for apparent skill. This is only a rough analysis of what is wrong, and how it should be fixed, but I think it's a step in the right direction of figuring out na truly meaningful way to determine one player's apparent skill from another in order to rank players in a meaningful way. As it stands, PP, ranked score, and score in general, are all completely useless metrics. | ||
|
MegaManEXE
United States845 Posts
| ||
|
McFortran
United States79 Posts
| ||
|
Camail
United States1030 Posts
Congrats, I didn't think you could do it, but you did.  | ||
|
MegaManEXE
United States845 Posts
No awesome performance records yet  .First Place Ranks No first place records currently . | ||
|
Bobbias
Canada1373 Posts
| ||
|
wooozy
3813 Posts
On July 31 2012 06:34 MegaManEXE wrote: Best Performance No awesome performance records yet .First Place Ranks No first place records currently . me too, buddy! high five! + Show Spoiler + | ||
|
Hakker
United States1360 Posts
It must only count scores from when the new pp system went up on the Best Performance/First Place things. atleast thats when all mine are from ![[image loading]](http://i.imgur.com/lW1QC.jpg) EDIT: also gunther maps are free pp | ||
|
Heh_
Singapore2712 Posts
On July 31 2012 06:58 Bobbias wrote: I just edited my post, check it again. You hit the nail on a lot of the points there. To add on to your points, I feel that AR, circle size, decay rate and whatever that can be adjusted on a slider be user-controlled. To make things "fair", the score multiplier should also be tweaked depending on which settings you use, for example AR10 gives you 1.10 multiplier while AR0 gives you 1.00. In a sense, you don't need HR anymore as the settings are fully customizable. This also solves the problem of retarded beatmaps with artifically-inflated difficulty, like beatmaps with widely spaced hit circles with the smallest size or beatmaps with long pauses and maximum decay rate. And yeah the pp system is utterly retarded. Sounds like a disincentive for new players when their rank is stagnant. | ||
|
ThaZenith
Canada3116 Posts
I think you need to get a top 40 rank or something for the list to refresh for the first time. I got one higher rank, and my best performances is now filled with maps from months ago. And it was empty earlier today for me as well. On July 31 2012 07:37 Heh_ wrote: To make things "fair", the score multiplier should also be tweaked depending on which settings you use, for example AR10 gives you 1.10 multiplier while AR0 gives you 1.00. In a sense, you don't need HR anymore as the settings are fully customizable. . No. The mapper would set the baseline, 'normal' numbers. Lowering the numbers would give you a score penalty, and raising the numbers would not. But you'd get no bonus for raising numbers. | ||
|
Heh_
Singapore2712 Posts
On July 31 2012 07:38 ThaZenith wrote: I think you need to get a top 40 rank or something for the list to refresh for the first time. I got one higher rank, and my best performances is now filled with maps from months ago. And it was empty earlier today for me as well. No. The mapper would set the baseline, 'normal' numbers. Lowering the numbers would give you a score penalty, and raising the numbers would not. But you'd get no bonus for raising numbers. Hmm makes sense. Raising the numbers will probably cause the top ranks to be dominated by people who max out the sliders. At least the lowered numbers will allow people to play retarded beatmaps at a slight penalty. | ||
|
Bobbias
Canada1373 Posts
Game theory dictates that if you allow a player to chose some settings which may make the game easier or harder for them, to get the best score, they would adjust everything to make the game easier. That means that when two players have tweaked settings, they will have personalized their settings to optimize the score they get. Whether you turn it up or down doesnt mean shit, because we assume that you adjusted it for your best score, so we dont ahve to care about what you adtually adjusted it to. Mods would need some correction factor though. The whole idea here is to completely separate the mapper's control. Instead of defiing default levels for what the variables should be (OD, AR, etc.) and requiring some way of correcting for it, we say that the moment a player changes something from the default, they are optimizing the settings for their best score. I think having 2 leaderboards, one for "no changes from mapper's settings" and one for "settings optimized for score". Score as we know it would also likely need to be changed, because selecting for high combo creates unnecessary bias. Something based on your accuracy and the number of misses you have and some sort of average combo value would be better, because missing 1 note in the middle of a song should not be worth any more or any less than missing a note in the beginning of the song. Also, I'm actively avoiding actual numbers for any of this because I don't feel like I understand this well enough to actually implement some sort of system like I'm describing. I'm simply talking from the viewpoint of someone who understands some of the fundamental concepts for statistical analysis, rather than uunderstanding the processes themselves in detail. | ||
|
MegaManEXE
United States845 Posts
On July 31 2012 07:37 Heh_ wrote: And yeah the pp system is utterly retarded. Sounds like a disincentive for new players when their rank is stagnant. This is one of the biggest problems, the fact that it only updates every 30 minutes It makes ranking up not nearly as rewarding, it was cool to play a song and see +1,234,567 and +0.01%, now you play a song and get nothing, you don't even see your ranked score move because it's so truncated now on the in-game banner thing | ||
|
Bobbias
Canada1373 Posts
| ||
|
Camail
United States1030 Posts
| ||
|
wooozy
3813 Posts
| ||
| ||
|
|
The PondCast
Kung Fu Cup
OSC
CrankTV Team League
Replay Cast
CrankTV Team League
Korean StarCraft League
Afreeca Starleague
RSL Revival
Serral vs SHIN
herO vs Solar
Online Event
[ Show More ] |
|
|
EPT
