Guest
home
beatmaps
rankings
community
store
help
forum
  1. Forums
  2. osu!
  3. Development

New Performance Points System!

posted
Total Posts
53
Loves
446 posts
Joined December 2012
Topic Starter
Loves
We have a very broken PP system now. Admit it you baka!

So going on, I have thought of a simple but effective concept that should (or would) be much more accurate in terms of skill level.

So here is the concept:

All maps hypothetically give the same amount of PP but the amount of PP gained will scale down as the maps median accuracy and score goes up.
This ensures that easy maps don't give out the same amounts of PP as harder maps since the median accuracy AND score is higher than those of harder maps. (Btw score is counted so you don't get the same skill level from HT/EZ players)

Now a map such as http://osu.ppy.sh/b/200437&m=0 (Childrens Domain) on easy difficulty, which most people get 100% on would weigh almost nothing, unless you're a new player.

The median for this difficulty would be something like:

99.5% Median Accuracy
226 000 Median Score.

Now I do something interesting here. I times 99.5 by 10 which gives me 995. I divide 226 000 by 1000 which gives me 226. I add these two together and I get 1221. This here is the median "performance" of the map itself. Of course all maps are different, some being marathons, some being 30 second maps.

Example 2:

Now a map like http://osu.ppy.sh/b/35015&m=0 (Tag4) would give quite a lot of PP in this system if you got 90%+ because it seems like the average accuracy is way below 80% and the average score is probably 1 000 000. The averages of these (75% times 10 AND 1 000 000 divided by 1000) would come to a median "performance" of 1750.

Now lets say thelewa get's 1800 "performance" on this map. He is considered above average and probably on the top 500 (or even top 50) of the score board. He would get some PP.

Now Lily-Kun gets 2100 "performance" on this map and comes first by a staggering amount. Lily-Kun would get a LOT of PP for a) coming first, b) beating everyone by a lot, c) for playing a map with a low median "performance" from the players.

Rule of the thumb: Times accuracy by 10, divide score by 1000 and plus them together to get a "performance" for a map.

Other rules for this concept:

-First place will give 30% more PP than 2nd place.
-The average global ranking on the song will also impact your PP. If you get 1st on easy difficulty and the average global rank of players for that map is around 120 000, then you won't get much PP, even if you come first.
-PP is only given effectively when there are more than 100 ranks on the song.

This concept would fix:

-Hard/normal farmers.
-Reward people for playing hard maps that don't give PP under "peppys" hidden formulae.
-More stable. I got 200 pp for playing some easy insane from 2009. Not fair but meh.
-People's confusion on how to get PP.

Please also note that this concept would also give similar amounts of PP given now. As you rank higher, PP gain on high median "performance" maps are lowered. If you're a rank 300 global playing on a map with a player global rank of 70 000, you won't get much PP.

Also note that median "performance" varies from map to map. Marathon maps would yield "performance" numbers of up to 20 000, while 1 minute maps would yield 400 or so. Depends on the total score.

Please comment and share this thread so opinions may go wilde~
Bobbias
1,447 posts
Joined March 2011
Bobbias
I actually like your suggestion quite a bit (and I would consider myself highly critical of the current PP system).

There are a few things that should be considered though.

1) There are many players who either don't continually add new maps as they are ranked, or for whatever reason end up with a small collection of maps. Both the current PP system as well as your proposed system would put these players at a severe disadvantage as far as ranking is concerned.

2) A player's skill is not static, and changes over time. Because of this, a ranking system should weight recent scores higher than older scores.

Concerning the first issue, I think we need a separate secondary value which indicates how "sure" our system is of a player's rank. Let's call it confidence. If a player has played a lot of songs and has proven that they are playing very consistently at a certain skill level, then the system can be very confident that a player is ranked accurately. If a player has only played a limited selection, then our system cannot be very confident of how good that player really is, because it only has a limited amount of scores to work with. This would result in some players being ranked higher than they would be now, but with a low confidence rating. That would make it much harder for people to farm ranks by simply playing a whole ton of songs, which would still be possible in your system Lily.

The only issue I see with the confidence rating is that players would probably look at a highly ranked player with low confidence and automatically think they don't deserve that rank. But if they are playing well enough to earn that rank even with low confidence, it's still likely that they are good enough to be that rank, it just means that the system doesn't have enough data to be completely sure they deserve it. The one time i can think of that this system would rank someone highly when they don't deserve it is if a player only has a few songs and they play those songs so much that they basically know the song from memory, and would do significantly worse on songs they aren't extremely familiar with. I think it would probably be easy to spot those players looking at other stats such as comparing the number of total plays vs the number of plays covered by their most played songs (or even the number of plays of their top played songs compared to their performance on each song).
kriers
1,963 posts
Joined April 2010
kriers

Bobbias wrote:

Concerning the first issue, I think we need a separate secondary value which indicates how "sure" our system is of a player's rank. Let's call it confidence. If a player has played a lot of songs and has proven that they are playing very consistently at a certain skill level, then the system can be very confident that a player is ranked accurately. If a player has only played a limited selection, then our system cannot be very confident of how good that player really is, because it only has a limited amount of scores to work with. This would result in some players being ranked higher than they would be now, but with a low confidence rating. That would make it much harder for people to farm ranks by simply playing a whole ton of songs, which would still be possible in your system Lily.

The only issue I see with the confidence rating is that players would probably look at a highly ranked player with low confidence and automatically think they don't deserve that rank. But if they are playing well enough to earn that rank even with low confidence, it's still likely that they are good enough to be that rank, it just means that the system doesn't have enough data to be completely sure they deserve it. The one time i can think of that this system would rank someone highly when they don't deserve it is if a player only has a few songs and they play those songs so much that they basically know the song from memory, and would do significantly worse on songs they aren't extremely familiar with. I think it would probably be easy to spot those players looking at other stats such as comparing the number of total plays vs the number of plays covered by their most played songs (or even the number of plays of their top played songs compared to their performance on each song).
A confidence system making players higher ranked than deserved with a "low confidence" sounds pretty absurd. It would definitely be better to increase your rank as your system gets more confident about your current skill level and rather have some players being ranked lower than they should be due to inactivity or lack of scores. I just don't want to see a random cheater top the performance list with low confidence because he hacked every hard map in the game that evening.
Hula
581 posts
Joined January 2013
Hula
No.

Just, no. Because to make this less flawless than the current PP system right now would involve stupid amounts of math that I doubt anyone would care to do.

The current PP system is fine compared to this suggestion. Too much wrong with it to explain. sorry
Bobbias
1,447 posts
Joined March 2011
Bobbias

kriers wrote:

Bobbias wrote:

Concerning the first issue, I think we need a separate secondary value which indicates how "sure" our system is of a player's rank. Let's call it confidence. If a player has played a lot of songs and has proven that they are playing very consistently at a certain skill level, then the system can be very confident that a player is ranked accurately. If a player has only played a limited selection, then our system cannot be very confident of how good that player really is, because it only has a limited amount of scores to work with. This would result in some players being ranked higher than they would be now, but with a low confidence rating. That would make it much harder for people to farm ranks by simply playing a whole ton of songs, which would still be possible in your system Lily.

The only issue I see with the confidence rating is that players would probably look at a highly ranked player with low confidence and automatically think they don't deserve that rank. But if they are playing well enough to earn that rank even with low confidence, it's still likely that they are good enough to be that rank, it just means that the system doesn't have enough data to be completely sure they deserve it. The one time i can think of that this system would rank someone highly when they don't deserve it is if a player only has a few songs and they play those songs so much that they basically know the song from memory, and would do significantly worse on songs they aren't extremely familiar with. I think it would probably be easy to spot those players looking at other stats such as comparing the number of total plays vs the number of plays covered by their most played songs (or even the number of plays of their top played songs compared to their performance on each song).
A confidence system making players higher ranked than deserved with a "low confidence" sounds pretty absurd. It would definitely be better to increase your rank as your system gets more confident about your current skill level and rather have some players being ranked lower than they should be due to inactivity or lack of scores. I just don't want to see a random cheater top the performance list with low confidence because he hacked every hard map in the game that evening.
Catching cheaters should not be a function of the ranking system. The purpose of a ranking system is to rank players, period. I don't know how peppy currently detects and deals with cheaters, but I'd assume he has a system in place capable of detecting cheaters and removing their scores. As far as ranking systems go though, cheaters should not factor into things at all, provided there's a separate system for catching and dealing with them (and their should be).

As an example of how PP favors players with a large number of scores, LunarSakuya is rank 1,225 as of this post, while I'm currently at 19,990, and yet I'm a much better player than LunarSakuya. This difference is primarily because I don't have nearly as many scores as her (and the fact that I've been inactive in standard since mania came out, but that's primarily because my mouse has tracking issues which cause a lot of trouble when I play standard, so I've just kinda given up until I can solve the issue). I'll admit that under a system like I describe, I would be ranked higher and she would be ranked lower, but IMHO considering the fact that I am better than her, that's how things should be (no offense meant to LunarSakuya, I really like her maps).

PP ranking relies too heavily on the number of scores a player has, and that is what I'm trying to decouple from ranking. If you want ranks to reflect a payer's skill, it should be based on a "running average" of a player's skill, not a value that accumulates based on the number of ranked scores a player has. Again, any calculation which is based on the number of scores a player has will not accurately reflect a players skill. If you want to reflect a player's skill in ranks you cannot create a system which takes into consideration a payer's number of ranked scores. It must be based on some sort of running average of a player's recent scores. A player's skill is generally not static, and if you want a proper skill based ranking system, it must be able to reflect a players increasing or decreasing skill if a player is playing better or worse over a certain amount of time.

As it stands, as long as a player is good enough to rank high enough to receive some amount of PP from a map, they can farm PP rather easily by just playing a bunch of maps. If I went and simply S ranked on all the Hards and Insanes in my collection that I could S rank, I'd be ranked far higher than I am currently. Thing is, I don't want to go back and bore myself playing hundreds or thousands of boring maps. I'd rather challenge myself with harder stuff to improve at the game, not waste hours upon hours farming for ranks. The fact that I have to farm ranks to get somewhere close to where I should be ranked (based on skill) in the first place is absurd.
theowest
osu! Alumni
6,891 posts
Joined January 2009
theowest
Suggestion in General development? Almost something for feature requests, at least that's the place we see a lot of the pp related requests..

Personally I believe a perfect pp system would come out of a perfect difficulty rating. That's where the problem lies.

I also can't see how this topic will change anything tbh.
Metro
693 posts
Joined April 2010
Metro

theowest wrote:

Personally I believe a perfect pp system would come out of a perfect difficulty rating. That's where the problem lies.
Yup this is critical. And it already exists coded by Tom.
lolcubes
osu! Alumni
4,437 posts
Joined November 2010
lolcubes
To be honest, there is another thing which sort of makes people's pp skyrocket and that is very lenient timing window. There are maps where in top50 you can see like 30 SS (with mods), this would have to say that the map isn't exactly hard and shouldn't yield such massive pp, however they still award incredibly massive amounts of pp.

Another problem are maps with no spinners, the problem here is that this is a lose-lose scenario. If it gets changed so only ranking matters, whoever gets the first SS (with mods I guess) will stay on top and everyone else won't be able to get same/similar pp. You can go like "but he was the first so thats all that matters", but it's still possible that people beat that score in less plays than he did, simply cause they probably didn't even know about the map. The other problem with this is that certain maps (which have soooo many SS on the board) are just worth way too much, simply cause those maps are easy to get SS on with 2 most popular mods (hidden doubletime).
https://osu.ppy.sh/b/153895 for example, this map is the prime example of no-spinner-easy-to-100%-with-mods maps which are worth incredible amounts of pp. Currently it's worth lower due to people who were doing this with FL (to get even more pp), but when the doubletime hidden SS was like rank #2, this map gave me double the amount of pp than my current top ranks map at the time, which was about 60 (on 5550 pp level).
This is why I say it's lose-lose because no matter how you handle this problem (unless if you include the too many SS means its easy formula), each side will suffer.
Loves
446 posts
Joined December 2012
Topic Starter
Loves
Don't forget that the median "performance" on easy maps is always around the maximum amount of "performance" meaning that if the "performance" score (median) of an easy song was 1000 (SS, all mods) and I got SS (HD/HR/DT), my performance would only be around 950 or so meaning that I wouldn't get pp or only get a tiny bit.
Hula
581 posts
Joined January 2013
Hula

Bobbias wrote:

Catching cheaters should not be a function of the ranking system. The purpose of a ranking system is to rank players, period. I don't know how peppy currently detects and deals with cheaters, but I'd assume he has a system in place capable of detecting cheaters and removing their scores. As far as ranking systems go though, cheaters should not factor into things at all, provided there's a separate system for catching and dealing with them (and their should be).

As an example of how PP favors players with a large number of scores, LunarSakuya is rank 1,225 as of this post, while I'm currently at 19,990, and yet I'm a much better player than LunarSakuya. This difference is primarily because I don't have nearly as many scores as her (and the fact that I've been inactive in standard since mania came out, but that's primarily because my mouse has tracking issues which cause a lot of trouble when I play standard, so I've just kinda given up until I can solve the issue). I'll admit that under a system like I describe, I would be ranked higher and she would be ranked lower, but IMHO considering the fact that I am better than her, that's how things should be (no offense meant to LunarSakuya, I really like her maps).

PP ranking relies too heavily on the number of scores a player has, and that is what I'm trying to decouple from ranking. If you want ranks to reflect a payer's skill, it should be based on a "running average" of a player's skill, not a value that accumulates based on the number of ranked scores a player has. Again, any calculation which is based on the number of scores a player has will not accurately reflect a players skill. If you want to reflect a player's skill in ranks you cannot create a system which takes into consideration a payer's number of ranked scores. It must be based on some sort of running average of a player's recent scores. A player's skill is generally not static, and if you want a proper skill based ranking system, it must be able to reflect a players increasing or decreasing skill if a player is playing better or worse over a certain amount of time.

As it stands, as long as a player is good enough to rank high enough to receive some amount of PP from a map, they can farm PP rather easily by just playing a bunch of maps. If I went and simply S ranked on all the Hards and Insanes in my collection that I could S rank, I'd be ranked far higher than I am currently. Thing is, I don't want to go back and bore myself playing hundreds or thousands of boring maps. I'd rather challenge myself with harder stuff to improve at the game, not waste hours upon hours farming for ranks. The fact that I have to farm ranks to get somewhere close to where I should be ranked (based on skill) in the first place is absurd.

Average player skill would go against what peppy seemingly doesn't want PP to be, a ranking system which punishes players. Since an average player value would have to punish bad performances, and that's not what Osu! is about.

Active players don't get their PP diminished, yet inactive players do. Players who have played longer, have naturally played more beatmaps, get more scores, but as a result of playing so much, they're naturally going to be better than someone who has only played 1/10th of the beatmaps. At least going by your logic, which seems to clash in a quite disastrous manner.

This is why peppy probably ignores this shit, because he's given a friendly ranking system. Not some bullshit competition ladder shit that makes you sweat and get all nervous about losing precious points.
Loves
446 posts
Joined December 2012
Topic Starter
Loves

Hula wrote:

Bobbias wrote:

Catching cheaters should not be a function of the ranking system. The purpose of a ranking system is to rank players, period. I don't know how peppy currently detects and deals with cheaters, but I'd assume he has a system in place capable of detecting cheaters and removing their scores. As far as ranking systems go though, cheaters should not factor into things at all, provided there's a separate system for catching and dealing with them (and their should be).

As an example of how PP favors players with a large number of scores, LunarSakuya is rank 1,225 as of this post, while I'm currently at 19,990, and yet I'm a much better player than LunarSakuya. This difference is primarily because I don't have nearly as many scores as her (and the fact that I've been inactive in standard since mania came out, but that's primarily because my mouse has tracking issues which cause a lot of trouble when I play standard, so I've just kinda given up until I can solve the issue). I'll admit that under a system like I describe, I would be ranked higher and she would be ranked lower, but IMHO considering the fact that I am better than her, that's how things should be (no offense meant to LunarSakuya, I really like her maps).

PP ranking relies too heavily on the number of scores a player has, and that is what I'm trying to decouple from ranking. If you want ranks to reflect a payer's skill, it should be based on a "running average" of a player's skill, not a value that accumulates based on the number of ranked scores a player has. Again, any calculation which is based on the number of scores a player has will not accurately reflect a players skill. If you want to reflect a player's skill in ranks you cannot create a system which takes into consideration a payer's number of ranked scores. It must be based on some sort of running average of a player's recent scores. A player's skill is generally not static, and if you want a proper skill based ranking system, it must be able to reflect a players increasing or decreasing skill if a player is playing better or worse over a certain amount of time.

As it stands, as long as a player is good enough to rank high enough to receive some amount of PP from a map, they can farm PP rather easily by just playing a bunch of maps. If I went and simply S ranked on all the Hards and Insanes in my collection that I could S rank, I'd be ranked far higher than I am currently. Thing is, I don't want to go back and bore myself playing hundreds or thousands of boring maps. I'd rather challenge myself with harder stuff to improve at the game, not waste hours upon hours farming for ranks. The fact that I have to farm ranks to get somewhere close to where I should be ranked (based on skill) in the first place is absurd.

Average player skill would go against what peppy seemingly doesn't want PP to be, a ranking system which punishes players. Since an average player value would have to punish bad performances, and that's not what Osu! is about.

Active players don't get their PP diminished, yet inactive players do. Players who have played longer, have naturally played more beatmaps, get more scores, but as a result of playing so much, they're naturally going to be better than someone who has only played 1/10th of the beatmaps. At least going by your logic, which seems to clash in a quite disastrous manner.

This is why peppy probably ignores this shit, because he's given a friendly ranking system. Not some bullshit competition ladder shit that makes you sweat and get all nervous about losing precious points.

Where in my concept does it mention losing PP for doing badly?

Also, a player who's played maybe only 1\3 of another players map does not mean that the player who's played less is not as skilled. I mean look at Niko... he's probably the closest to Cookiezi and he's only played half of what other top 50 players have done.
Gilgamesh
14 posts
Joined October 2008
Gilgamesh

Lily-Kun wrote:

-Reward people for playing hard maps that don't give PP under "peppys" hidden formulae.
^This is definitely one of the things I want fixed. Why should I put effort into S ranking an Insane map, get absolutely 0pp, when I could put almost the exact same amount of effort to DT/HD a Hard map to get at least 100-250 rank up? What's wrong is that the current system rewards little or no BP at all if your rank on the map is <#500, its ridiculously difficult (well for me lol) to get a high rank on Insane maps considering there are usually at least 4000 ranked plays for most insane maps. I would already be happy if it rewarded just 1PP for being better than the average score/acc on the map.

Bobbias wrote:

As it stands, as long as a player is good enough to rank high enough to receive some amount of PP from a map, they can farm PP rather easily by just playing a bunch of maps. If I went and simply S ranked on all the Hards and Insanes in my collection that I could S rank, I'd be ranked far higher than I am currently. Thing is, I don't want to go back and bore myself playing hundreds or thousands of boring maps. I'd rather challenge myself with harder stuff to improve at the game, not waste hours upon hours farming for ranks. The fact that I have to farm ranks to get somewhere close to where I should be ranked (based on skill) in the first place is absurd.
I think it would be too complex and time consuming (or maybe even impossible) developing a system that relies less on the number of scores a player has. Osu is simply not some competitive game where you are constantly matched with others to determine who's better like in League of Legends. Unless Osu adapted some sort of online matchmaking and you can only gain or lose rank by "ranked queuing" against other players who are ready to gamble their rank ; it would be almost impossible to accurately measure each person's skill and there really is no better way to do it than measuring it by amount of high scores.
Loves
446 posts
Joined December 2012
Topic Starter
Loves
No news?
Hanyuu
2,223 posts
Joined February 2009
Hanyuu
Well in your idea the player with 100% on the song with the highest score will have the highest PP of the game off a single song.
Teara
osu! Alumni
814 posts
Joined June 2009
Teara
In my honest opinion, to fix the PP system, you first need to accurately determine how hard a map really is, from there you can start a calculating system which will actually work.
Full Tablet
2,542 posts
Joined September 2011
Full Tablet

Metro wrote:

theowest wrote:

Personally I believe a perfect pp system would come out of a perfect difficulty rating. That's where the problem lies.
Yup this is critical. And it already exists coded by Tom.
Tom's formula is not perfect. For example, there is a "level 67" beatmap (Avicii - Levels (Nightcore Mix) [Insane] by Kiyro) where I get consistently a SH with 98+% acc. And the map next to it in the rankings is a map I can hardly ever FC (and I get usually about 90% acc on it) (Kanako Itou -A.R. [Insane] by taka1235). Tom's system doesn't consider how hard is to follow the rhythm of a song.
Loves
446 posts
Joined December 2012
Topic Starter
Loves

Hanyuu wrote:

Well in your idea the player with 100% on the song with the highest score will have the highest PP of the game off a single song.
Well that is the definition of skillfulness... If you get 100% on a very hard insane with like HDHR, you deserve that pp.
Metro
693 posts
Joined April 2010
Metro

Full Tablet wrote:

Metro wrote:

Yup this is critical. And it already exists coded by Tom.
Tom's formula is not perfect. For example, there is a "level 67" beatmap (Avicii - Levels (Nightcore Mix) [Insane] by Kiyro) where I get consistently a SH with 98+% acc. And the map next to it in the rankings is a map I can hardly ever FC (and I get usually about 90% acc on it) (Kanako Itou -A.R. [Insane] by taka1235). Tom's system doesn't consider how hard is to follow the rhythm of a song.
Sure it isn't perfect but:

Cookiezi's top ranks in pp system (no FL):

Dark PHOENiX - Taketori Hishou [Extra]
Renard - Blue Night [DoKo]
Hommarju feat. R.Cena - Chousai Kenbo Sengen [Insane]
Jun.A - The Refrain of the Lovely Great War [Lunatic]
The Quick Brown Fox - The Big Black [WHO'S AFRAID OF THE BIG BLACK]
xi - FREEDOM DiVE [FOUR DIMENSIONS]
Demetori - Wind God Girl [Extra]
IOSYS - Power of Dream (Night Fever Refix) [Power of Stream]
SuganoMusic - Imademo... [Lunatic]

Cookiezi's top ranks ordered by Tom's calculator:

221 - xi - FREEDOM DiVE [FOUR DIMENSIONS]
212 - The Quick Brown Fox - The Big Black [WHO'S AFRAID OF THE BIG BLACK]
165 - IOSYS - Power of Dream (Night Fever Refix) [Power of Stream]
149 - Jun.A - The Refrain of the Lovely Great War [Lunatic]
137 - Demetori - Wind God Girl [Extra]
135 - SuganoMusic - Imademo... [Lunatic
128 - Hommarju feat. R.Cena - Chousai Kenbo Sengen [Insane]
120 - Renard - Blue Night [DoKo]
116 - Dark PHOENiX - Taketori Hishou [Extra]

It's almost reversed... Implement it along Tom's calculator and it'll be a really good system.
[Luanny]
8,248 posts
Joined April 2011
[Luanny]

Teara wrote:

In my honest opinion, to fix the PP system, you first need to accurately determine how hard a map really is, from there you can start a calculating system which will actually work.
^this
Just making sure it can be read in this page
Tom pls.
Tsukimi Luna
726 posts
Joined July 2010
Tsukimi Luna
I totally support this~ <3 <3 <3
Big YEAH YEAH
I hate pp farmers.... damn them

I really hope those rankwhores lose their rank one day....
Oinari-sama
osu! Alumni
3,403 posts
Joined July 2010
Oinari-sama
lol I thought this thread would've been moved a week ago...
MillhioreF
Pro Tester
5,004 posts
Joined July 2011
MillhioreF
It is about moving osu! forward, so I've let it stay. Perhaps it would be better in Feature Requests, but...
Dexus
982 posts
Joined July 2009
Dexus
Essentially focus on the difficulty ratings and then focus on the rankings system. Seriously I've not payed attention to stars since I've joined and that was 4 years ago. I just look at the BPM, AR, and jump distances. Note intervals and how clusted things get can be calculated in as well. With Tom's calculations it doesn't really take BPM and not clusters into consideration. Health drain should also be taken into consideration.

again to say, NEW DIFFICULTY RATING FIRST then worry about the rankings.
peppy
19,311 posts
Here since the beginning
peppy
I appreciate your suggestions, but as I have said many times before, I am constantly changing pp based on how it reacts to player actions, and won't currently be looking for external input. Feel free to keep the discussion going though :).
Metro
693 posts
Joined April 2010
Metro
/peppy/thread
Hishida
18 posts
Joined August 2011
Hishida

peppy wrote:

I appreciate your suggestions, but as I have said many times before, I am constantly changing pp based on how it reacts to player actions, and won't currently be looking for external input. Feel free to keep the discussion going though :).
ok, i have suggestion,
why dont u just using rank score 1-500 rank to give pp?
in this 5 day, i got rank about 50 song hard+insane..
im just got 40-50pp..
the last 2 day im just got 4 pp after 80 song play..

i think to get perfect (SS) with mod so hard, i need to play and play again about hundred time just for one song to get SS with mod..

so,back to my suggestion, how about pp follow rank 1-500..
1 : 50pp
2 : 49pp
3-9 : 48-41pp
11-20 : 40pp
21-30 : 30pp
31-40 : 20pp
41-50 : 10pp
400-500 : 5pp

so after this we just need to get 500 rank on beatmap and so easy to understand..
no need long explaination like this pp system now just benefit for pro player only..
nothing good come to new player if u using this pp.. my friends told me this new system can give a chance for new player to chase old player,
but what i see now is,many new player will be leave far behind old player because of big different of skill..
even old player still cannot to chase pro player..

im play osu for fun with score rank, but i dont see any fun with this pp system and i know u got many complain from player about this..
i think im stress to play osu because of this pp things..
not all players play osu for get SS every day..
i have job and just can play osu for 1-2 hour p/day..
to get SS i need to play about 6 hours..

in the time before, im just play around 2 hours and can get 20-50 ranks up..

p/s: i dont see this new system pp is suck or just failed idea from u, but i think people play for fun.
almost every game in the world is determined by score and not by skill..

if u want to do so,just make new pp system that said about skill..
etc : skill rank : 50 of 2,515,089 player

please consider my request~
thanks :)
Sup A Noob
555 posts
Joined November 2009
Sup A Noob
To have a perfect difficulty rating, we need to have a mapper/BAT/MAT-defined difficulty level.

Not by stars. They do jack shit in the game.

I'm talking about actual values. "On the value of 1-100, how difficult do you think the map is?"

When the mapper maps something, he'd rate the map's difficulty based on how difficult an average player would find it. Then the BAT/MAT in charge of that player's map will also give his rating, and the mapper will alter it slightly based on the average of the ratings given.

The mapper could ask for more people who he deems is professional enough to not give a troll rating, to further polish the actual rating of the map, but the final rating lies in the BAT/MAT, because I know some mappers just can't be professional enough.'

That final rating would then be the map's true difficulty value, and will be used in calculations.

If the rating is still deemed wrong after ranking, simply change it.

However, to implement this would mean a slow and painful process of rating every single ranked beatmap there is out there. It would take years, decades even, for a dedicated team to rate them all. Maybe AIMod could be programmed to do such ratings, after considering jumps, streams, overlaps, returns etc. Perhaps now the Mapping Techniques forum page could be of some use, if AIMod could be programmed to recognise mapping techniques.
Twaintoss
141 posts
Joined January 2013
Twaintoss

Sup A Noob wrote:

To have a perfect difficulty rating, we need to have a mapper/BAT/MAT-defined difficulty level.

I'm talking about actual values. "On the value of 1-100, how difficult do you think the map is?"
I would agree to that. There in the end simply is no AI that can tell you about the maps difficulty because there is no way to put
into numbers what I would call "flow" of map. If the map is easy to read, VERY HARD but maybe possible to predict with numbers.
The "flow" is how the map feels while playing. There is no accurate way to explain it since it's about the feeling itself. That's too hard
(or downright impossible) to predict for a system so Modders are certainly needed...

But I'm the last one who wants to increase stress-levels of MAT/BAT/Mappers any further
so a new system would be a nice improvement.

Also: Let Mappers Pick Star Rating for Maps is a pretty nice request about this idea.
I'll ad my personal opinion here and say that I think even with judgement difference from person-to-person, adding [the picking of diffs by mappers] would be beneficial over what we have, but may not be the best way moving forward if the star rating is to be improved in the future.
ppy's words
Naikaze
2,401 posts
Joined July 2011
Naikaze

Sup A Noob wrote:

To have a perfect difficulty rating, we need to have a mapper/BAT/MAT-defined difficulty level.

I'm talking about actual values. "On the value of 1-100, how difficult do you think the map is?"
I think Tom's calculator is pretty accurate for this kind of thing..
maal
maal

Teara wrote:

In my honest opinion, to fix the PP system, you first need to accurately determine how hard a map really is, from there you can start a calculating system which will actually work.
^


Also, why don't every new beatmap that is ranked, get a certain "x" amount of pp on it? This pp is determined by a handful of BAT's/MAT's or something, as a part of the ranking process. Then it would make it a lot better.

Just a suggestion. :D
Metro
693 posts
Joined April 2010
Metro
Human defined pp per map is stupid.

Replying to a thread that peppy has deemed as useless is stupid.

Only good point here is that Tom's calculator should be used for pp criteria.
Twaintoss
141 posts
Joined January 2013
Twaintoss

maal wrote:

Also, why don't every new beatmap that is ranked, get a certain "x" amount of pp on it? This pp is determined by a handful of BAT's/MAT's or something, as a part of the ranking process. Then it would make it a lot better.

Just a suggestion. :D
PP are about competition too, so there can't be a fixed amount of PP for a map. A "nice" map gets much PP, but if nobody played the game/song/music to begin with (or isn't that popular), said map would just score PP for the few players who actually played the map...

Metro wrote:

Human defined pp per map is stupid.

Replying to a thread that peppy has deemed as useless is stupid.

Only good point here is that Tom's calculator should be used for pp criteria.
This is just your opinion.

And I was showing this link to exactly SHOW peppy's opinion on that in the first place. So it isn't useless as it at least shows information.
buschauffeur
13 posts
Joined February 2013
buschauffeur
I don't think you guys even realise how hard it is to even program such a system, if the current pp system will be replaced i call tons of bugs and glitches to fool and or trick it.

It's far better and cost efficient to update the current system like peppy's doing atm rather than design a whole new one.
maal
maal

Twaintoss wrote:

maal wrote:

Also, why don't every new beatmap that is ranked, get a certain "x" amount of pp on it? This pp is determined by a handful of BAT's/MAT's or something, as a part of the ranking process. Then it would make it a lot better.

Just a suggestion. :D
PP are about competition too, so there can't be a fixed amount of PP for a map. A "nice" map gets much PP, but if nobody played the game/song/music to begin with (or isn't that popular), said map would just score PP for the few players who actually played the map...
and when other players discovers that this map is played by a few, and it gives pp, then other players will as well for pp.

see the circle?
boat
2,369 posts
Joined February 2010
boat

Lily-Kun wrote:

We have a very broken PP system now. Admit it you baka!
Never! You're the baka! Aho! Baka!
Oinari-sama
osu! Alumni
3,403 posts
Joined July 2010
Oinari-sama
Instead of having fixed amount of pp per map, it would probably make more sense to have a "price pool" just like any tournament/lotteries.

Let's call the total pp available per map the "pp pool" for simplicity's sake. The pp pool grows as more players play the map, ie for a map played by 1000 players it will have a pp pool 10x larger than a map played only by 100 players.

From there the pp can be rewarded as a % of the pp pool. As to how many % is awarded, that can be based on the number of players, normalised by player rank and difficulty offset such as the current system or a more stringent system where the map difficulty better defined.

Say for each additional player, a map gains an extra 0.2pp in the pp pool. Then:
  1. For a map played by 1000 players, the total pp pool is 200.
  2. For a map played by 1000 players, the total pp pool is 20.
pp will be rewarded based on [Total pp pool] x [Map Rank Multiplier] x [Player Rank vs Difficulty Offset] / [Number of Player per Map]

[Map Rank Multiplier] is similar to current system, where rank above certain threshold are rewarded with pp and good ranks gets higher dividends. eg:
  1. Rank 1 = x50
  2. Rank 2 = x20
  3. Rank 3 = x15
  4. Rank 4 ~ 40 = x10
  5. Rank 41 ~ 600 = x1
  6. Rank 601 and below = x0 (no pp reward)
The [Player Rank vs Difficulty Offset] is a multiplier that adjusts pp rewarded based on the player's current pp rank vs map difficulty. I will leave that discussion for another day (or person) and focus on pp pool in this particular post. Ideally the [Player Rank vs Difficulty Offset] should give something roughly like this:
  1. Low Ranked Player on Harder Difficulty = higher than standard multiplier
  2. High Ranked Player on Easier Difficulty = lower than standard multiplier
So for the sake of simplicity let's consider this example:

Player 1: Current pp rank #1000
Player 2: Current pp rank #100000
Map 1: 1000 Ranked Players [Normal]
Map 2: 10000 Ranked Players [Insane]
[Player Rank vs Difficulty Offset] for Player 1

pp reward for Player1 to achieve #1 on Map 1 = 200 x 50 x 0.1 / 1000 = 1pp
(Assuming [Player Rank vs Difficulty Offset] for Player 1 on this map is x0.1)

pp reward for Player2 to achieve #1 on Map 1 = 200 x 50 x 2.0 / 1000 = 20pp
(Assuming [Player Rank vs Difficulty Offset] for Player 2 on this mapis x2.0)

pp reward for Player1 to achieve #1 on Map 2 = 2000 x 50 x 0.5 / 1000 = 50pp
(Assuming [Player Rank vs Difficulty Offset] for Player 1 on this map is x0.5)

pp reward for Player2 to achieve #1 on Map 2 = 2000 x 50 x 4.0 / 1000 = 400pp
(Assuming [Player Rank vs Difficulty Offset] for Player 2 on this map is x4.0)

Yes those numbers are arbitary, only for the sake of demonstration.

But the idea is to have a Pool paying Dividends according to how well a player perform. The hard part is to have all the "dividends" to sum up to the total pool, but it allows a more dynamic distribution of pp.

Refer to payout structure of some poker tournament if you want to see more example of this kind of system in work.
Sup A Noob
555 posts
Joined November 2009
Sup A Noob

Vrael wrote:

I don't think you guys even realise how hard it is to even program such a system, if the current pp system will be replaced i call tons of bugs and glitches to fool and or trick it.
Actually, we do. But if everyone thought the way you do, no one would come up with any ideas.
Zexous
713 posts
Joined July 2012
Zexous
-First place will give 30% more PP than 2nd place.
This would really screw up the balance in the top 1000 global ranks. A person who gets 1 HD HR #1 would be higher than say someone with 3 #2 HD HR ranks, and I don't see that as balanced.

inb4 Cookiezi 10,000 pp
ChipZ
18 posts
Joined May 2013
ChipZ

Lily-Kun wrote:

We have a very broken PP system now. Admit it you baka!
LOL, no one's an idiot here dude XD
DJKero
231 posts
Here since the beginning
DJKero
I think PP System should be so much complex that it's right now, and what Lily-Kun is throwing to us...

Besides, the base of this idea is currently what I've said already in #Osu and other channels talking about scoring system+performancepoint progress itself...

I think if we do Lily-Kun idea with more variables we should get an excellent filter for all the problems we got right now...

- SEE POST BELOW FOR THE UPDATED FORMULA -

This should be a good system to evade Farming, to weight more the REAL harder songs in PP amount and much other improvements...
Last edited by DJKero , edited 6 times in total.
Full Tablet
2,542 posts
Joined September 2011
Full Tablet

DJKero wrote:

pp system proposal
I think PP System should be so much complex that it's right now, and what Lily-Kun is throwing to us...

Besides, the base of this idea is currently what I've said already in #Osu and other channels talking about scoring system+performancepoint progress itself...

I think if we do Lily-Kun idea with more variables we should get an excellent filter for all the problems we got right now...

Basically I should propose something like:

---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

New Performance Point System [Trademark: Kero]

Accuracy x 10 + 'Score' / 1000

Result of above + 1000 x [ESDS]

ESDS (Exact Song Difficulty System):

(HP Drain Rate Value + Circle Size Value + Approach Rate Value + OverAll Difficulty Value)

Result of above x [ (Circle Notes x 10) + (Slider Notes x 8) + (Spinners x 5) ]

Result of above / [ASL]

ASL (Active Song Lenght):

Active Song Lenght should be the Active time on the beatmap.

This excludes Spinners and Break time, So:

Song Lenght in Seconds - (Spinners+Breaks Lenght in Seconds)

---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

This should be a good system to evade Farming, to weight more the REAL harder songs in PP amount and much other improvements...
That would basically the same as the ranked score system... that would be 100% about farming, and unbalanced by marathon maps (quadratic score).
  1. A = Accuracy x 10 + 'Score' / 1000: Accuracy bonus is overwhelmed by score here, unless it is a very short map (less than ~250 max combo)
  2. B = A + 1000 x (HP Drain Rate Value + Circle Size Value + Approach Rate Value + Overall Difficulty Value). Here ESDS overwhelms "A" unless it is a long map (more than ~1500 combo). Calculating ESDS as the sum of those values is just naive.
  3. B * ( (Circle Notes x 10) + (Slider Notes x 8) + (Spinners x 5) ) / (Active Song Length): This wouldn't make much difference between maps, it would make stream maps slightly more weighted than jump maps though.
Last edited by Full Tablet , edited 3 times in total.
DJKero
231 posts
Here since the beginning
DJKero

Full Tablet wrote:

DJKero wrote:

pp system proposal
I think PP System should be so much complex that it's right now, and what Lily-Kun is throwing to us...

Besides, the base of this idea is currently what I've said already in #Osu and other channels talking about scoring system+performancepoint progress itself...

I think if we do Lily-Kun idea with more variables we should get an excellent filter for all the problems we got right now...

Basically I should propose something like:

---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

New Performance Point System [Trademark: Kero]

Accuracy x 10 + 'Score' / 1000

Result of above + 1000 x [ESDS]

ESDS (Exact Song Difficulty System):

(HP Drain Rate Value + Circle Size Value + Approach Rate Value + OverAll Difficulty Value)

Result of above x [ (Circle Notes x 10) + (Slider Notes x 8) + (Spinners x 5) ]

Result of above / [ASL]

ASL (Active Song Lenght):

Active Song Lenght should be the Active time on the beatmap.

This excludes Spinners and Break time, So:

Song Lenght in Seconds - (Spinners+Breaks Lenght in Seconds)

---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

This should be a good system to evade Farming, to weight more the REAL harder songs in PP amount and much other improvements...
That would basically the same as the ranked score system... that would be 100% about farming, and biased by marathon maps (quadratic score).
No it's not, it won't require to you to get #1 first of all the people or be so farmy like to the current one, since as far as I know, the current system doesn't includes the variables of notes and song lenght that would impact drastically at the time of songs like Marathons that now have the same weight that a non marathon short one with the same amount of notes...

And I don't see the farm-ability, if you want, Wait and re-check the formula in my above post later, i'll add a interesting variable that i've saved in case someone argues about farm-ability of my formula...
boat
2,369 posts
Joined February 2010
boat
Nothing about the current system is broken, it's the users who "break" the system and then complain about it.

Easier difficulties aren't weighted high by default, they get weighted this way because they are far more contested than the outmost insane stuff.

PP does take into account who's score you beat, which is the point of it all. Grinding easier difficulties lasts for a while, but once you get up high you'll be forced to beat the scores of players ranked higher than you. Getting a #1 on a map only contested by those by miles lower ranked then you eventually wont give you anything. The system measures performance, not skill, and as such if you want a high rank you're forced to perform better than others.

Albeit it can be saddening to get good scores and not be rewarded for it, that's how things are and I quite frankly don't even mind it one bit as it's perfectly fair.
DJKero
231 posts
Here since the beginning
DJKero
Updated Formula:

DJ Kero wrote:

New Performance Point System [Trademark: Kero]

(Accuracy x 10) + (Score / 1 000)

Result of above x [ESDS]

Result of above / Play Count until Record / 10 000

ESDS (Exact Song Difficulty System):

(HP Drain Rate Value + Circle Size Value + Approach Rate Value + OverAll Difficulty Value)

Result of above x [ (Circle Notes x 10) + (Slider Notes x 8) + (Spinners x 5) ]

Result of above / [ASL]

ASL (Active Song Lenght):

Active Song Lenght should be the Active time on the beatmap.

This excludes Spinners and Break time, So:

Song Lenght in Seconds - (Spinners+Breaks Lenght in Seconds)
Obviously this is to be added to the current system calculations, not this alone coz it would be not taking in count some things that must be take in count...

I think the concept of ASL is the Most important of all the idea...
Last edited by DJKero , edited 2 times in total.
Full Tablet
2,542 posts
Joined September 2011
Full Tablet

DJKero wrote:

No it's not, it won't require to you to get #1 first of all the people or be so farmy like to the current one, since as far as I know, the current system doesn't includes the variables of notes and song lenght that would impact drastically at the time of songs like Marathons that now have the same weight that a non marathon short one with the same amount of notes...

And I don't see the farm-ability, if you want, Wait and re-check the formula in my above post later, i'll add a interesting variable that i've saved in case someone argues about farm-ability of my formula...
It would be farmable since one would just play a bunch of Hard/Easier-Insanes only once or twice each to get ranking fast (without mods to make it easier, unless you are at a skill level where it is almost certain you would score more with mods on the first/second try than without).
About Marathon maps, with that formula, they would give a LOT of score (since the formula would return a amount of pp near equal to the the quadratic score system the game uses). Marathon maps are fine with the current pp system (they do tend to give a considerable amount of pp, if you beat other's score in the rank list).

Edit: About the new formula:
  1. That new variable would make it even more farmable... maybe we have a different definition of farming.
    Farm: Effortlessly and quickly getting a reward using a certain strategy, this usually implies doing the same strategy over and over again since it gives a lot of reward.
    Grinding: Doing the same thing over and over again to get a reward.
  2. Now that map difficulty is multiplied instead of added together in the formula, score doesn't overwhelm difficulty in marathon maps (but still, marathon maps would give an unbalanced amount of pp). There is still the problem that that way to calculate difficulty is not accurate (for example: A map with "35" difficulty (an Insane difficulty with 10OD, for example) difficulty would be much harder than a map with "33" (the same Insane with 8OD) difficulty).
DJKero
231 posts
Here since the beginning
DJKero

Full Tablet wrote:

DJKero wrote:

No it's not, it won't require to you to get #1 first of all the people or be so farmy like to the current one, since as far as I know, the current system doesn't includes the variables of notes and song lenght that would impact drastically at the time of songs like Marathons that now have the same weight that a non marathon short one with the same amount of notes...

And I don't see the farm-ability, if you want, Wait and re-check the formula in my above post later, i'll add a interesting variable that i've saved in case someone argues about farm-ability of my formula...
It would be farmable since one would just play a bunch of Hard/Easier-Insanes only once or twice each to get ranking fast (without mods to make it easier, unless you are at a skill level where it is almost certain you would score more with mods on the first/second try than without).
About Marathon maps, with that formula, they would give a LOT of score (since the formula would return a amount of pp near equal to the the quadratic score system the game uses). Marathon maps are fine with the current pp system (they do tend to give a considerable amount of pp, if you beat other's score in the rank list).

Edit: About the new formula:
  1. That new variable would make it even more farmable... maybe we have a different definition of farming.
    Farm: Effortlessly and quickly getting a reward using a certain strategy, this usually implies doing the same strategy over and over again since it gives a lot of reward.
    Grinding: Doing the same thing over and over again to get a reward.
  2. Now that map difficulty is multiplied instead of added together in the formula, score doesn't overwhelm difficulty in marathon maps (but still, marathon maps would give an unbalanced amount of pp). There is still the problem that that way to calculate difficulty is not accurate (for example: A map with "35" difficulty (an Insane difficulty with 10OD, for example) difficulty would be much harder than a map with "33" (the same Insane with 8OD) difficulty).
Dude, did you read about ASL? ¬¬"

It's Just the opposite of what you say... It would give minus to Marathons, and Isn't FARMABLE because if you retry a song over and over again it would give you minus and minus and minus PP when you get a decent score in it, so... I don't see your point...
DJKero
231 posts
Here since the beginning
DJKero

Metro wrote:

Human defined pp per map is stupid.

Replying to a thread that peppy has deemed as useless is stupid.

Only good point here is that Tom's calculator should be used for pp criteria.
Also you're a bit closed minded.
boat
2,369 posts
Joined February 2010
boat

DJKero wrote:

because if you retry a song over and over again it would give you minus and minus and minus PP when you get a decent score in it, so... I don't see your point...
If that's so, then your proposed "system" is incredibly stupid. If somebody actually got the score, then they're obviously at the level of getting it and shouldn't be penalized for the time it took them to reach it, those who want to and can get a higher rank should have to prove themselves on harder maps. It's contradictory because how you've put it you're forced to farm maps you can fc to score high rather than strive for a score on a map that is challenging, because apparently improvement isn't worth anything because it's not a first try SS.
Full Tablet
2,542 posts
Joined September 2011
Full Tablet

DJKero wrote:

Full Tablet wrote:

It would be farmable since one would just play a bunch of Hard/Easier-Insanes only once or twice each to get ranking fast (without mods to make it easier, unless you are at a skill level where it is almost certain you would score more with mods on the first/second try than without).
About Marathon maps, with that formula, they would give a LOT of score (since the formula would return a amount of pp near equal to the the quadratic score system the game uses). Marathon maps are fine with the current pp system (they do tend to give a considerable amount of pp, if you beat other's score in the rank list).

Edit: About the new formula:
  1. That new variable would make it even more farmable... maybe we have a different definition of farming.
    Farm: Effortlessly and quickly getting a reward using a certain strategy, this usually implies doing the same strategy over and over again since it gives a lot of reward.
    Grinding: Doing the same thing over and over again to get a reward.
  2. Now that map difficulty is multiplied instead of added together in the formula, score doesn't overwhelm difficulty in marathon maps (but still, marathon maps would give an unbalanced amount of pp). There is still the problem that that way to calculate difficulty is not accurate (for example: A map with "35" difficulty (an Insane difficulty with 10OD, for example) difficulty would be much harder than a map with "33" (the same Insane with 8OD) difficulty).
Dude, did you read about ASL? ¬¬"

It's Just the opposite of what you say... It would give minus to Marathons, and Isn't FARMABLE because if you retry a song over and over again it would give you minus and minus and minus PP when you get a decent score in it, so... I don't see your point...
Let's quickly analyze the formula for a marathon map:
- The first value is basically the score, which is quadratic with time.
- Next, you multiply it with the number of objects, that is lineal with time. That makes the proposed points cubic.
- Next, you divide by ASL, that makes the points quadratic with time again. So marathon maps give an unbalanced amount of points (too much).

The points proposed would make things too similar to the old ranked score system, but even with more problems.
In that system, a player would just play for about a year or so, to get good enough to be able to FC Hards or easier Insanes reliably on the first try (ignoring really hard maps). Then, when the player wants to increase his rank, playing in his account wouldn't be efficient since his account is now "dirty" with the plays he has done.

In that system, the players who get high rank would be the ones that manage to multi-account to farm more easily (one account for training only, and the other to farm). Alternatively, they could make unranked copies of the maps to train on them (for example practicing maps changing the OD to 10 before playing the ranked copy once).
Primula_old
Primula_old

DJKero wrote:

It's Just the opposite of what you say... It would give minus to Marathons, and Isn't FARMABLE because if you retry a song over and over again it would give you minus and minus and minus PP when you get a decent score in it, so... I don't see your point...
Did you think this through? Am I reading this correctly? You want to reward people less the more they retry?

This would be worse than the current PP system when it comes to people who make new accounts.. it would encourage more people to do so when they're unhappy with their retry count because they know they'll get rewarded more for clearing in one try on a new account.

Ranking in Osu is all about retrying until you succeed, mods like HR and FL require you to constantly retry.
show more
Please sign in to reply.

New reply

0
1 / 53