ohyou wrote:

Will there be an update to the current ingame or website UI in order to integrate the PP metrics?

What i mean is:
- A number before the map's title in the "Best Performance" section that shows how much pp you got from a play;
- An expansion to the "Best Performance" section (10 plays isn't that much);
- A way to see the "difficulty line" regarding to one or the other mod (i.e. to see how "hard" the song is if i play it with DT);
- A way to see how much pp you got from a particular play in game;
- A way to "sort" records by pp (like you can do it now - by country or by particular mod);

Basically most of the things the osu!tp has now.

Sorry for my english in advance.

What osu!tp has been doing is not working that easily here, since the processor has to deal with far more than 100times as many scores. A per-score pp display is being considered, ingame and in the web. I'll probably propose an ingame difficulty display, incorporating mods, at some point as well when I have time.

Kert wrote:

Tom94 wrote:

As for the HR rank vs your DT rank. I believe they are both very good and will keep tweaking the difficulty algorithm. It's hard to calll one of them both better than the other with good reasoning.

I've spent more than a few hundred tries (not all at once of course) to FC that HR map, but managed to do DT in around 15
I really believe it's because of very small circles which makes you miss a lot more often (it's CS6 with HR)

Ah, btw I forgot mentioning TAG maps
It's probably hard to take them into a count, but I'd like to see this tweaked too

TAG4 maps will be fixed as soon as per-hitobject data is available. Before that there is not much I can do due to the extreme difficulty spikes.

Also let me repeat what I said in the other thread for everyone:

Just a quick note for everyone who experiences a big drop in ranks. In the previous pp system players who were inactive were not counted. This is not added to the new system yet, so please wait a little while. Your rank drop might not be that bad - potentially it even rose?

RaneFire wrote:

YayMii wrote:

Just a question: will the new PP difficulty algorithm be replacing the current in-game star difficulty rating in the future?

And is there any way to integrate this with the client for beatmaps existing/already downloaded? So we can sort by difficulty in song select and it would be accurate?

Or is that what you meant?

That's exactly what I meant. The new difficulty algorithm rating is already visible on the beatmap pages EDIT: no longer visible . I wanna see them in game.

Finally, a skill-based ranking that doesn't revolve around [Hard] farmers.

This is beautiful. No complaints.

R e d wrote:

ItHR still have a over value?

Lets have Thelewa for example: He is really good Hr Hd but he don't have so much speed and DT accuracy as SnowWhite, Dungeon or DragonHuman.

I Respect Thelewa. But still think he don't deserve 3#

It depends a lot really. Some maps that are practically impossible to aim propely that only a handful of players can hdhr deserves pp as well. Don't you agree?

Speed isn't everything you know!

I think this is a lot more fair than old pp system. One thing I wanted to say is that some scores are overestimated: my score on this map: http://osu.ppy.sh/b/338594&m=0 is 300 combo (with over 800 full combo) and 84,44% acc. Because it's fresh map I have #84, and probably for that reason it landed in my Top Ranks. I wouldn't consider this a good score, so I wonder why it's so high performance.
Also, because of that single score, my overall acc fell from 95,51% to 95,15%, which isn't cool :/

.

What about calculating accuracy with unstable rate? Or is it that hard to calculate on every map played?.

Salvage wrote:

What about considering accuracy with unstable rate? Or is it that hard to calculate on every map played?.

I don't think it is possible since I don't think there is enough information stored to retrieve that value exactly.

It can be estimated though, as long as you ignore sliders (since sliders can potentially increase UR without reducing accuracy, since they don't need to be hit accurately).

DT is better for pp, as FL. If i haв highscore with DT, but i've broken my old DT-record with FL, my old reord must be deleted from base and i lose pp, isn't it?
For exampe, WoobWoofWolf has his the best record for "Saiya - Remote Control" beatmap now. He has 5 misses, but he used DT+HD mode. But SnowWhite played this map without DT, with HD+HR, have FC, but doesn't have this map for Best Perfomance. If WoobWoofWolf plays now map without mods and break his own records, he loses old records and loses pp.
So, best score for beatmap is not best pp. It is better to use DT and don't play other mods without DT. Is it wrong?
update: I see answer in next post now, thank you.

Wojowu wrote:

I think this is a lot more fair than old pp system. One thing I wanted to say is that some scores are overestimated: my score on this map: http://osu.ppy.sh/b/338594&m=0 is 300 combo (with over 800 full combo) and 84,44% acc. Because it's fresh map I have #84, and probably for that reason it landed in my Top Ranks. I wouldn't consider this a good score, so I wonder why it's so high performance.
Also, because of that single score, my overall acc fell from 95,51% to 95,15%, which isn't cool :/

The score rank you get on a map does not matter at all for the new pp. Only your performance together with the beatmap difficulty counts. Maybe the map was really hard? I've seen a C in jesus1412's best performance list and it was a crazy hard DT score, about as hard as FCing mendes with DoubleTime. I'll take a look what's happening with your scores, though.

R e d wrote:

ItHR still have a over value?

Lets have Thelewa for example: He is really good Hr Hd but he don't have so much speed and DT accuracy as SnowWhite, Dungeon or DragonHuman.

I Respect Thelewa. But still think he don't deserve 3#

thelewa only makes us believe he'd be slow. He 99.5%ed Chipscape! That's not slow. His top scores might not be the fastest, but they are an amazing combination of speed, aim and acc. That being said it doesn't mean, that you are wrong. I am just mentioning it.

UE95 wrote:

Will my pp benefit in any shape or form if I keep improving my previous scores? Also, does it have to be a drastic improvement? I'm really sorry if these are dumb questions, I'm just a bit overwhelmed by this new system :/

Score improvements won't always benefit your pp. On a map which is really hard to DT beating your DT score with HR might actually reduce your pp. This is currently a limitation I wish there wouldn't be, but it exists.

Salvage wrote:

What about calculating accuracy with unstable rate? Or is it that hard to calculate on every map played?.

Something similar to unstable rate is used. OD does play a huge role. We are already considering, to update the accuracy displayed in the player profiles to a more sensible measure, while maybe keeping the old value in parantheses.

MosaicII wrote:

DT is better for pp, as FL. If i haв highscore with DT, but i've broken my old DT-record with FL, my old reord must be deleted from base and i lose pp, isn't it?
For exampe, WoobWoofWolf has his the best record for "Saiya - Remote Control" beatmap now. He has 5 misses, but he used DT+HD mode. But SnowWhite played this map without DT, with HD+HR, have FC, but doesn't have this map for Best Perfomance. If WoobWoofWolf plays now map without mods and break his own records, he loses old records and loses pp.
So, best score for beatmap is not best pp. It is better to use DT and don't play other mods without DT. Is it wrong?

You are correct. I wish this limitation wouldn't exist, but there is no possibility of changing it for the moment.

Full Tablet wrote:

Salvage wrote:

What about considering accuracy with unstable rate? Or is it that hard to calculate on every map played?.

I don't think it is possible since I don't think there is enough information stored to retrieve that value exactly.

It can be estimated though, as long as you ignore sliders (since sliders can potentially increase UR without reducing accuracy, since they don't need to be hit accurately).

Including sliders, that was my point, but well i guess it can't be done.


Oh, ninja'd .. glad to hear that then tom .

Full Tablet wrote:

Salvage wrote:

What about considering accuracy with unstable rate? Or is it that hard to calculate on every map played?.

I don't think it is possible since I don't think there is enough information stored to retrieve that value exactly.

It can be estimated though, as long as you ignore sliders (since sliders can potentially increase UR without reducing accuracy, since they don't need to be hit accurately).

Sliders and spinners already are ignored for the part of the calculation that is responsible for accuracy. But yeah, in the end it only is an approximation of unstable rate due to per-hitobject data not being available.

I went from rank 15k to rank 9k. That's...something. I guess it's because I tried really hard to do well on harder songs and didn't farm normal/hards very much at all?

MosaicII wrote:

DT is better for pp, as FL. If i haв highscore with DT, but i've broken my old DT-record with FL, my old reord must be deleted from base and i lose pp, isn't it?
For exampe, WoobWoofWolf has his the best record for "Saiya - Remote Control" beatmap now. He has 5 misses, but he used DT+HD mode. But SnowWhite played this map without DT, with HD+HR, have FC, but doesn't have this map for Best Perfomance. If WoobWoofWolf plays now map without mods and break his own records, he loses old records and loses pp.
So, best score for beatmap is not best pp. It is better to use DT and don't play other mods without DT. Is it wrong?

The old ppv1 system had the same issue. A solution would be storing "best performances" separately from "best scores", but the problem is that would be expensive in term of server resources; and if the system has time degradation of scores, it wouldn't solve the problem in all cases (depending in how the degradation is implemented)

Not sure if this belongs here but I'll ask anyway.

Are there plans on fixing the star star sytem by replacing it with the beatmap difficulty level?

I'm not sure if I'm understand the "all scores are used" thing, but tell me if I'm right with this example:

Score 1: 10,000,000 score, played on normal, FC 100%.
Score 2: 3,000,000 score, played on DT/HD, had 1 miss. 99%.

My best score would show up as Score 1, but my best "performance" would probably be from Score 2, so the PP I get from Score 2 would count for my PP ranking?

kreph wrote:

Not sure if this belongs here but I'll ask anyway.

Are there plans on fixing the star star sytem by replacing it with the beatmap difficulty level?

That sure sounds like a nice idea, considering all the maps already have a difficulty rating calculated by the system.

Izzypookins wrote:

I'm not sure if I'm understand the "all scores are used" thing, but tell me if I'm right with this example:

Score 1: 10,000,000 score, played on normal, FC 100%.
Score 2: 3,000,000 score, played on DT/HD, had 1 miss. 99%.

My best score would show up as Score 1, but my best "performance" would probably be from Score 2, so the PP I get from Score 2 would count for my PP ranking?

All scores being used means that it doesn't have the limitation that tp had where it could only access top 50 score data. PPv2 has access to ranks beyond top 50 which is why they say all scores are used.

Izzypookins wrote:

I'm not sure if I'm understand the "all scores are used" thing, but tell me if I'm right with this example:

Score 1: 10,000,000 score, played on normal, FC 100%.
Score 2: 3,000,000 score, played on DT/HD, had 1 miss. 99%.

My best score would show up as Score 1, but my best "performance" would probably be from Score 2, so the PP I get from Score 2 would count for my PP ranking?

If both scores are on the same map, then the higher score value gets used, even if the other would give more pp. As I mentioned earlier we can not work around this at the moment.
If those 2 scores are on different maps, then how much pp you will get highly depends on how hard the maps are.

Tom94 wrote:

Izzypookins wrote:

I'm not sure if I'm understand the "all scores are used" thing, but tell me if I'm right with this example:

Score 1: 10,000,000 score, played on normal, FC 100%.
Score 2: 3,000,000 score, played on DT/HD, had 1 miss. 99%.

My best score would show up as Score 1, but my best "performance" would probably be from Score 2, so the PP I get from Score 2 would count for my PP ranking?

If both scores are on the same map, then the higher score value gets used, even if the other would give more pp. As I mentioned earlier we can not work around this at the moment.
If those 2 scores are on different maps, then how much pp you will get highly depends on how hard the maps are.

Got it! Thanks!

Damn where's the beatmaps that i have farm. 9k ---> 27k , guess that's because me too mach farm back then :'(

Is it just me or mods aren't affecting too much in the new pp ranking?

well, i think the current system is the closest to perfect ~ time to farm

.

UE95 wrote:

Hey there thank you so much for replying. One last question, you did mention Dt and HR in your example. How well would Hidden fare against those two and against scores without mods?

I can't give definite answers for any mod. All I can say is that it depends a lot on the map. The new system does not just multiply the value of map when adding mods, it calculates a whole new difficulty with the mods. For example, if you're playing a map with DoubleTime, then the algorithm will calculate a new difficulty value of the map, considering it to be 1.5x as fast. With HardRock it considers the circles smaller and the OD higher. Hidden gives a bonus to the aiming-aspect of the map, so it has a greater effect on maps which are harder to aim.
The same logic applies to all other mods.

I'm not sure if it was written anywhere in here, but I would like to be able to see more than just 10 of my top performances. If a player by themselves could see all, or more than just their top 10 scores (Maybe top 100?) it would be very useful.

.

Really good job, I don't feel like a farmer

UE95 wrote:

Haha no worries man, I respect the secrecy. Your answer cleared my doubts either way, thank you! Once the full criteria is updated on the wiki I'll know more. GL with the new system ^^

Not like I want to keep it a secret. I just don't want to type a few paragraphs now and in the wiki later on. Call it laziness.

.

I feel a bit skeptical about encompassing ALL passed ( I am assuming passed ) play-through scores tallying into your rank at the moment. I haven't gotten a chance to hop into multiplayer yet and get a feel for how other players compare to me but personally, I've lived in multiplayer for the past 2 odd years. I've made it a personal rule never to auto-fail in an MP match because I missed a note or made a silly mistake which means I have thousands of play-throughs with horribad scores becuase mr/mrs host decided to tack on DT or I was feeling dangerous and decided to live in the HD HR side of life. Or I was just with a bunch of friends and we were dicking around playing '08-'09 maps, HR super small circle size maps, troll an insane diff AR5 map, etc...

I have not toiled a lot of effort compiling this post, it is simply a couple ideas that came to mind while playing and gestating around the post about this new system being implemented. I have no intention of trying to get into the nitty-gritty about "X" mod gives less value then "Y" mod, my ACC dropped because of those 50 D ranks I have from 4 years ago, trying to meta-game the system to know why exactly "Z" map gives more points then "W" map. I really could care less about whatever new system comes out next and how it ultimately effects my golden number. I say give the option to choose what you want in osu! options to display the new pp/tp system or your original total ranked score rank.

The concern I'm trying to get at is for those who really only play multiplayer now and its effects compared to those who really only play singleplayer and have the ability to always ESC->Retry to their heart's content. I do not find the excuse of 'then auto fail the map if you don't want your score tallied' since it undermines the purpose of multiplayer to those who may have thought that while reading this.

Should there be a completely different point system for MP playing? (this subject is irrelevant to everything else in this post, but came to mind while typing it all. Kinda came outta left field.)

From your lovely flying fish who lives in MP, day in and day out, FishHook. I hope something in this post of rambling helps you in your advancement of the new ranking system. Cheers.

FishHook wrote:

The concern I'm trying to get at is for those who really only play multiplayer now and its effects compared to those who really only play singleplayer and have the ability to always ESC->Retry to their heart's content. I do not find the excuse of 'then auto fail the map if you don't want your score tallied' since it undermines the purpose of multiplayer to those who may have thought that while reading this.

The bad scores you got while playing MP shouldn't affect your ranking much.
As for your accuracy, you can increase it by getting good accuracy in hard maps.

Would it be possible for osutp.net to obtain scores out of the top 50 range since it has been implemented into official osu rankings? it would be great if we can get an exact value on each of our best plays!

Still don't think it's perfect. My top performances are a joke, I did them in many less tries than some of my top performances in the old system. I think you've done a good job to reduce the effect of farmed hard difficulty maps in some way but in other ways you have ruined a lot of people's hard performances.

­

Yuzuru wrote:

9k ---> 27k , guess that's because me too mach farm back then :'(

Everyone who used to rank up by tricking the system are getting wrecked quite hard now huehue
 
 
More seriously; I always felt like short fast streams were a bit underrated in tp, feels pretty much the same here.
A short 250bpm stream (~11 circles) will be worth less than a long 210bpm stream (17+ circles), but the first one is way harder imo.
 
 
 
Other than that, because it's also close from tp it's a really nice ranking system. Good job there :p

GoldenWolf wrote:

Everyone who used to rank up by tricking the system are getting wrecked quite hard now huehue

...except me!

Liiraye wrote:

R e d wrote:

ItHR still have a over value?

Lets have Thelewa for example: He is really good Hr Hd but he don't have so much speed and DT accuracy as SnowWhite, Dungeon or DragonHuman.

I Respect Thelewa. But still think he don't deserve 3#

It depends a lot really. Some maps that are practically impossible to aim propely that only a handful of players can hdhr deserves pp as well. Don't you agree?

Speed isn't everything you know!

Yes but i can say the same for Fast maps with Dt, Where is impossible for Thelewa beat SnowWhite or Dungeon.

Accuracy isn't everything you know!

Is there multiplicator for NF?

I was #169 I was so happy

now im #161

i hate all of you responsible for this

Well I went from 5.8k to 11k. I did trick the system quite a bit to get to 5k, but not too much. The proof being I've already climbed 4k ranks since I've logged in.

My main issue though is that sometimes I will pass a song, it will end up in my best performances, but I won't get any pp for it. A little strange. Also, osusig doesn't update properly. I've had to change the signature layout twice already just to update it, but that's fairly miniscule.

I noticed the new ranking system favors mods even though some no mod maps are a lot more difficult and reward less.

Nakage wrote:

The proof being I've already climbed 4k ranks since I've logged in.

Don't forget to count the fact that inactive players were purged from the rankings.


I really like the system as it is right now. Counting all ranks in the system instead of top 1000/800 or 500 makes it a lot more motivating to try and beat your older scores. Went up from 3.5k (ppv1) to 1.8k.

muh speedy fingers

Syclasm wrote:

Don't forget to count the fact that inactive players were purged from the rankings.

Right :3 the purge made me climb 2300 ranks, so I still went from 9.3k to 7.5k, still decent imo.

~

I like this change actually. It feels much more skill based now rather than just knowing what maps to farm up. The way it was before, if you were in the top 300 or so, then you could make it up to the top 10 if you played enough. You would have enough skill to do so because all the maps that would give the most PP are right around that level to easily farm.

My only problem is rank updates. I played 2 songs. 1 of them made it onto the very top of my best performances, I went up 30 ranks. The 2nd one made it to the 2nd highest, I went up 800 ranks (Now sitting at 6761). I'm sure the ranks aren't clustered that hard in pp to where there's a 100 pp gap somewhere at rank 7k.

EDIT: If it's still in the process of cleaning inactive users, I should note I also went up ~100 pp by playing the 2 songs, so I think the 800 ranks was from the pp, not the cleaning.

Layne wrote:

I lofe the chnge in thisbmpe ok vry god.
thx u

wat

Can't we just use pp rank instead of score rank for every song like the tp website? or keep both so best pp rank and best score rank are independent.

If ppv2 is using the same difficulty calculations on maps as tp, why is it that my highest and 3rd highest tp maps don't make my best performance even though everything below did?

i guess its good. i went from 8.1k to 8.6k and my acc went up by 1% for some reason. no flaws that i can see

Almost wrote:

If ppv2 is using the same difficulty calculations on maps as tp, why is it that my highest and 3rd highest tp maps don't make my best performance even though everything below did?

shh at least you're higher than me

my second highest ppv2 score gave me gave me 0tp.

GoldenWolf wrote:

Everyone who used to rank up by tricking the system are getting wrecked quite hard now huehue
 

It's actually quite nice that easier ranks don't get counted as heavily. Sometimes when I'm tired of playing hard maps over and over I take a "break" where I try getting a top 50 on easier maps. It's a good way to relax a bit if your hands hurt or if you're tired but if you still want to play, but in the old system that'd make your rank go up increasingly. Now I can do the same thing without having to worry about having a "farmed" rank.


I agree with some of the previous posts that it might be nice to see the amount of pp a certain map has earned you, though I understand if that might be annoying to implement.

I have a question about something I've seen in the rankings - next to some ranks, I've seen something like ^3d. Does this refer to the amount of time passed since that score was set? And why does it only show for some rankings, even though I set some other ranks that didn't have that sign beside them?

[AirCoN] wrote:

Almost wrote:

If ppv2 is using the same difficulty calculations on maps as tp, why is it that my highest and 3rd highest tp maps don't make my best performance even though everything below did?

shh at least you're higher than me

my second highest ppv2 score gave me gave me 0tp.

We both know I'm worse though

Good to see that peppy finally decided to let you manage this stuff.

Syclasm wrote:

. Counting all ranks in the system instead of top 1000/800 or 500 makes it a lot more motivating to try and beat your older scores.

Considering that some of my top performances are scores i got with less than 3x100 and 1 miss, it has totally made me lose all motivation to replay those maps for fear of losing pp lol.

­

Hi Tom, your TP and ppv2 systems are really good, but I have some complaints about ppv2.
My userpage: http://osu.ppy.sh/u/1783793
As you can see, it's my best performance:

I don't think these two achievements:
Demetori - Emotional Skyscraper ~ World's End [Lunatic] (97.76% HD)
Dark PHOENiX - Taketori Hishou [Extra] (98.25% HD)
should be posted in it. I mean, I have some 140+ bpm DT HD scores, and they are better than these two above, I think.
I wonder why these two scores can get much pp for me, you know, the acc is not so high and the beatmap is not so diffucult.

Thanks for your hard working

The new System is great how I said last week

I got a ultraboost ... from 24k to 18k after 1 1/2H Playing to 14k

24k PP(v1) -> 18k PP(v2) -> 14k lawl

Gain PP is now harder with the Difficult Easy-Hard and Insane+ became important.
The new System is more Skill-based how I like it

Huge improvements, nice job by all participants!
--
Now that we have difiiculty levels for each map let us browse in osu! and on the website according to these levels, so that one can find a appropriate beatmap easier. Also include a ranking on how fast the map is and how hard the jumps are. This will give people a general idea on who is the better aimer and who has more speed and lets them choose beatmaps according to that.

Don't just take the score with the highest amount of points as the one relevant for pp. A HD+HR FC is worth less than a DT+HD S with one sliderbreak in the middle of the map, but will always award more points.
Probably one could calculate the pp for every score you get and compare it to your current top score. If it was a better performance the new one is chosen as the relevant pp score, even though it awards less points. (e.g. the DT+HD sliderbreak now appears as top rank for pp instead of the HD+HR FC).

A new score system showing the performance of the scores would also work great instead of the suggestion above.
Now that you guys found a way to calculate the actual peroformance nearly accurate, why don't you use this as the "score". So instead of showing the amount of points, the leaderboards could show the pp for each score.
This would support the great pp system and would work as well as the current system for tournaments.
--
Thanks for reading this!
I hope you could get some ideas of it (and that I thought of some new stuff, don't know if others already mentioned it).

best wishes,
Peter

Soarezi wrote:

In my eyes, right now, everything seems just perfect. The performances are basically based on skill, and the hardness of the scores. And that's how it should be, for now, really good job.

I agree ~

New ranking is great! I'm now placed around those who I think has similar skills in mp, so it's a big improvement.

Great job all!

The new system is definitely more accurate. I really like the new system because it actually encourages me to play challenging maps and do well, which is better for my improvement than the previous system where I didn't really have to push myself for ranks.

Rewben2 wrote:

I really like the new system because it actually encourages me to play challenging maps and do well, which is better for my improvement than the previous system where I didn't really have to push myself for ranks.

This. Improving your rank is now actually about improving... not just playing more maps.

_-KirA-_ wrote:

I feel like there is some sort of justice with this update

Not only you.

[Sakagami Tomoyo] wrote:

Hi Tom, your TP and ppv2 systems are really good, but I have some complaints about ppv2.
My userpage: http://osu.ppy.sh/u/1783793
As you can see, it's my best performance:

I don't think these two achievements:
Demetori - Emotional Skyscraper ~ World's End [Lunatic] (97.76% HD)
Dark PHOENiX - Taketori Hishou [Extra] (98.25% HD)
should be posted in it. I mean, I have some 140+ bpm DT HD scores, and they are better than these two above, I think.
I wonder why these two scores can get much pp for me, you know, the acc is not so high and the beatmap is not so diffucult.

Thanks for your hard working

I have to say these two scores are really damn impressive, I could never do them myself, especially not Taketori Hishou. What you consider a good score is also somehow dependent on the player. Maybe for you these scores aren't that good but maybe for other players the 140+ bpm DT HD scores aren't that good.

This song seems really under-valued to me. - http://osu.ppy.sh/b/137166?m=0
I spent quite a while getting a DT HD FC on it and got basically nothing, some of my other top scores though I would say are no where near as hard as it was for me to FC doll house.

For some reason this seems to be giving a lot of PP - http://osu.ppy.sh/b/95431?m=0
It doesn't seem like that hard of a map to me, not easy, but not worth the points it is giving imo.

I think that speed is being overvalued. Almost all of my top ranks are songs that just put out a lot of notes with not too much difficulty in-between them. Anyway just stuff I've been noticing and don't quite agree with. This system is already miles ahead of PPv1 so I'm happy with it, but it can always be better.

Bad accuracy hd hr scores dont reward much because hr is very accuracy based except on maps like dispel where the aim is the hardest part rather than acc

Tom94 wrote:

The rank which has a higher score will count, that is the one which will end up marked as your best in the ingame scoreboard.

I actually had a song where I had a previous rank of 360 or so on before the changes to pp, I played it again today and got a rank of 470 something and went from 1546 pp --> 1547 pp, played it once more and got a rank of 318 but got no pp at all from it. This system confuses me a bit =x

I had an interesting idea.

I know peppy wants to try to put a lot of emphasis on "Contextually relevant" scores into a ranking system, something like a ladder.

I've noticed an effect amongst certain scoreboards, something I guess we can call Scoreboard Stability.

It's hard to explain well, so examples are best I guess?

Highly "Unstable" Scoreboards (The top scores vary heavily on the same mod tier, few if any FCs or SS):
FREEDOM DiVE
osu! Stream Compilation
Don't say "lazy"
Usually incredibly difficult, long, or new maps.

Highly "Stable" Scoreboards (The top scores on the same mod tiers are incredibly similar):
Sis puella magica!
Chousai Kenbo Sengen
Usually incredibly easy or highly played maps.

I think maps with a higher unstable rate scoreboards are generally more "Contextual" (OMG, who's going to be the first to FC this? Who will be the first to FC this with HDHR? DT? FL? First to SS?)

For example, plays like some of Mesita's Flashlight plays (World's End, Gold Dust) or some of Cookiezi's old DT/HR/SS plays (Leia, Airman, Kokou no Sousei) are particularly impressive because they're pretty much the only ones who have charted with those mods or stats.

On the other hand, getting a first place in something like Chousai Kenbo Sengen, while impressive, is slightly less so because so many others have proven themselves of accomplishing it. Unless it's SS, which in this comparison I guess you could consider a Mod tier.

Anyways, I think if you had a way of comparing scoreboard instability and either compiling a chart or weighting the PP top ranks towards more unstable maps we would have a nice context relevant ranking formula.

The rough scale would range from the feasible infinite instability (Nobody has ever even passed this) to the quite impossible infinite stability (Literally ever user in existence has SS.) and award most possible points if you're the only existing user who's achieved the rank. (Mesita only user to S rank World's End on HDFL or better.)

Am I making sense?

This would give flashlight players something to hunt for as they can get high PP values of maps that nobody else has gotten a FL FC in before.

It also would weight more recent plays higher since scoreboards naturally stabilize more over time with some significantly slower outliers.

In ppv1 I was 8.3k. In ppv2 I was 10.3k. In this iteration of ppv2 I'm 8.3k. No change here.

However I do like it a lot more, looking at the top players it seems like everyone is being given an accurate place, and I like the fact that if I get a slider break in a 1,000 note song I still get considered for getting pp. It makes ranking a lot less stressful because I dont actually care about the scoreboard ranks at all. They're too volitile, and the people who belong there should always be higher skill level than me.

Soarezi wrote:

Bad accuracy hd hr scores dont reward much because hr is very accuracy based except on maps like dispel where the aim is the hardest part rather than acc

yyyup

even in tp if you got a bad accuracy HR score the only thing that you got more of was aim tp, and even then it was like 30 more than nomod

by getting good accuracy you get like 60 more accuracy tp increasing the total tp gain and thus since this system is based on tp increasing the total pp gained

all in all low accuracy HR is useless now unless the map is really really hard to aim already

DT is pretty stupidly weighed, but I guess that's fine as it's one of the harsher mods. Maybe reduce the amount on DT just a tad bit while bumping a tad bit up on HD HR since HR is a bitch.

[Sakagami Tomoyo] wrote:

Hi Tom, your TP and ppv2 systems are really good, but I have some complaints about ppv2.
My userpage: http://osu.ppy.sh/u/1783793
As you can see, it's my best performance:

I don't think these two achievements:
Demetori - Emotional Skyscraper ~ World's End [Lunatic] (97.76% HD)
Dark PHOENiX - Taketori Hishou [Extra] (98.25% HD)
should be posted in it. I mean, I have some 140+ bpm DT HD scores, and they are better than these two above, I think.
I wonder why these two scores can get much pp for me, you know, the acc is not so high and the beatmap is not so diffucult.

Thanks for your hard working

Could you let me know what exactly those DT HD scores are? I'll look into it then.

EvaRia wrote:

snip

Am I making sense?

This would give flashlight players something to hunt for as they can get high PP values of maps that nobody else has gotten a FL FC in before.

It also would weight more recent plays higher since scoreboards naturally stabilize more over time with some significantly slower outliers.

While a good idea in theory, ppv1 actually had this to some degree, and would make ppv2 open to farming since there are so many maps which people haven't tried to FL, but are certainly possible. It's not that FL is "way too hard" as it is just uncommonly used because it's not that fun.

thelewa wrote:

all in all low accuracy HR is useless now unless the map is really really hard to aim already

What is a low accuracy HR? Anything under 99%?

Comparing here, you'd have to practically SS od10 to get any decent amount of pp on maps which can be commonly SS'd nomod.

Melt3dCheeze wrote:

DT is pretty stupidly weighed, but I guess that's fine as it's one of the harsher mods. Maybe reduce the amount on DT just a tad bit while bumping a tad bit up on HD HR since HR is a bitch.

but this is based on tp

tp doesn't have weighting on mods like DT and HR because those are mods that change the map and the difficulty settings in themselves

I'd assume that all that's done is that the mod changed map is run through the difficulty calculator again, meaning that how much more pp DT and HR will give is based entirely on the map

Tom can correct me on this I have no idea if I've got this right

RaneFire wrote:

thelewa wrote:

all in all low accuracy HR is useless now unless the map is really really hard to aim already

What is a low accuracy HR? Anything under 99%?

Comparing here, you'd have to practically SS od10 to get any decent amount of pp on maps which can be commonly SS'd nomod.

anything under 99% on HR gives very little accuracy points on osu!tp so yyyyep

but instead of getting under 99% accuracy on HR you could just, you know, git gud?

thelewa wrote:

Melt3dCheeze wrote:

DT is pretty stupidly weighed, but I guess that's fine as it's one of the harsher mods. Maybe reduce the amount on DT just a tad bit while bumping a tad bit up on HD HR since HR is a bitch.

but this is based on tp

tp doesn't have weighting on mods like DT and HR because those are mods that change the map and the difficulty settings in themselves

I'd assume that all that's done is that the mod changed map is run through the difficulty calculator again, meaning that how much more pp DT and HR will give is based entirely on the map

Tom can correct me on this I have no idea if I've got this right

Correct.

thelewa wrote:

RaneFire wrote:

What is a low accuracy HR? Anything under 99%?

Comparing here, you'd have to practically SS od10 to get any decent amount of pp on maps which can be commonly SS'd nomod.

anything under 99% on HR gives very little accuracy points on osu!tp so yyyyep

but instead of getting under 99% accuracy on HR you could just, you know, git gud?

Pfft, what you call "very little" might be quite a lot for others. It's heavily based off OD, so having the same accuracy with HR as with nomod will also give a lot more pp. But yeah, the closer you get to 100%, the more points you get. Just look at it this way:
A 98% score has twice as many 100s/50s as a 99% score. The 99% score should therefore give a lot more points.

oh yeah maybe I should've mentioned that part about the 100's

As for the whole topic, tl;dr

Just asking since my rank now seems true, but still I experienced a significant drop compared to tp (maybe it's all about using all scores, not top 50)

Is the weighing the same as in tp? I mean does the number of your good scores matter more now?
Is there a division between aim/speed/acc or PP is acquired based on total points for the map? (in TP when you had low speed points and then got a good speed score you sky-rocketed in rank)
How is it with accuracy? - does it matter even more now?



@EDIT - Maybe this is not the place for such suggestion, but I'd like to know whether it's doable


For those who aren't happy with their HHR (because they did a far harder DT but couldn't beat the HHR) scores or w/e, could there be an option letting you choose among your scores as which should appear in the scoreboard/which one should be counted? That would solve the problem, but dunno if it's possible to implement.

peterhasaparker wrote:

Huge improvements, nice job by all participants!
--
Now that we have difiiculty levels for each map let us browse in osu! and on the website according to these levels, so that one can find a appropriate beatmap easier. Also include a ranking on how fast the map is and how hard the jumps are. This will give people a general idea on who is the better aimer and who has more speed and lets them choose beatmaps according to that.

Don't just take the score with the highest amount of points as the one relevant for pp. A HD+HR FC is worth less than a DT+HD S with one sliderbreak in the middle of the map, but will always award more points.
Probably one could calculate the pp for every score you get and compare it to your current top score. If it was a better performance the new one is chosen as the relevant pp score, even though it awards less points. (e.g. the DT+HD sliderbreak now appears as top rank for pp instead of the HD+HR FC).

A new score system showing the performance of the scores would also work great instead of the suggestion above.
Now that you guys found a way to calculate the actual peroformance nearly accurate, why don't you use this as the "score". So instead of showing the amount of points, the leaderboards could show the pp for each score.
This would support the great pp system and would work as well as the current system for tournaments.
--
Thanks for reading this!
I hope you could get some ideas of it (and that I thought of some new stuff, don't know if others already mentioned it).

best wishes,
Peter

If the difficulty algorithm ever was to change (which it will!) then all the top scores would become moot, since previous scores, which have not been registered might be considered better afterwards. This is why a stable scoring system is required - even though I agree that the current scoring system is not the best option. It'd be impossible to re-calculate the score for all existing scores, which is why this likely won't change.

AmaiHachimitsu wrote:

As for the whole topic, tl;dr

Just asking since my rank now seems true, but still I experienced a significant drop compared to tp (maybe it's all about using all scores, not top 50)

Is the weighing the same as in tp? I mean does the number of your good scores matter more now?
Is there a division between aim/speed/acc or PP is acquired based on total points for the map? (in TP when you had low speed points and then got a good speed score you sky-rocketed in rank)
How is it with accuracy? - does it matter even more now?

@EDIT - Maybe this is not the place for such suggestion, but I'd like to know whether it's doable


For those who aren't happy with their HHR (because they did a far harder DT but couldn't beat the HHR) scores or w/e, could there be an option letting you choose among your scores as which should appear in the scoreboard/which one should be counted? That would solve the problem, but dunno if it's possible to implement.

Due to processing limitations the total score value is used at score stage already. This might change in the future, but until then scores won't be weighted in all 3 categories independently. Regarding your rank drop - let's wait a bit until all inactive players have been removed. Everyone should increase in rank slightly with that.

Regarding your edit - I'm not sure how good of an idea that is. Scoreboards are supposed to represent your highest score, so you'd trade an inconsistency here for an inconsistency there + additional toggles. But I don't know what the future might bring. :p

EvaRia wrote:

SPOILER

I had an interesting idea.

I know peppy wants to try to put a lot of emphasis on "Contextually relevant" scores into a ranking system, something like a ladder.

I've noticed an effect amongst certain scoreboards, something I guess we can call Scoreboard Stability.

It's hard to explain well, so examples are best I guess?

Highly "Unstable" Scoreboards (The top scores vary heavily on the same mod tier, few if any FCs or SS):
FREEDOM DiVE
osu! Stream Compilation
Don't say "lazy"
Usually incredibly difficult, long, or new maps.

Highly "Stable" Scoreboards (The top scores on the same mod tiers are incredibly similar):
Sis puella magica!
Chousai Kenbo Sengen
Usually incredibly easy or highly played maps.

I think maps with a higher unstable rate scoreboards are generally more "Contextual" (OMG, who's going to be the first to FC this? Who will be the first to FC this with HDHR? DT? FL? First to SS?)

For example, plays like some of Mesita's Flashlight plays (World's End, Gold Dust) or some of Cookiezi's old DT/HR/SS plays (Leia, Airman, Kokou no Sousei) are particularly impressive because they're pretty much the only ones who have charted with those mods or stats.

On the other hand, getting a first place in something like Chousai Kenbo Sengen, while impressive, is slightly less so because so many others have proven themselves of accomplishing it. Unless it's SS, which in this comparison I guess you could consider a Mod tier.

Anyways, I think if you had a way of comparing scoreboard instability and either compiling a chart or weighting the PP top ranks towards more unstable maps we would have a nice context relevant ranking formula.

The rough scale would range from the feasible infinite instability (Nobody has ever even passed this) to the quite impossible infinite stability (Literally ever user in existence has SS.) and award most possible points if you're the only existing user who's achieved the rank. (Mesita only user to S rank World's End on HDFL or better.)

Am I making sense?

This would give flashlight players something to hunt for as they can get high PP values of maps that nobody else has gotten a FL FC in before.

It also would weight more recent plays higher since scoreboards naturally stabilize more over time with some significantly slower outliers.

The problem with this concept, is that the people who play [hard] diffs, old maps or top-tier maps are very different and therefore the data we have to analyze will also be different. This was one of the main issues of previous pp iterations, making [hard] and old maps being weighted much more, because statistically they seemed harder than they were since most of the good players didn't play them.

Not really a suggestion or feedback, but pretty dumb question; How hard is it to get points by playing nomod compared to playing with mods?

Good job with the new ranking system.

Do you ever plan on using play statistics to enhance the accuracy of the map difficulty algorithm or do you intend to use map data only?

Tom you should really try to concisely explain to people that this iteration of pp is fundamentally different since a lot of people think of this as just an improved version of ppv1

Spyrunite wrote:

This song seems really under-valued to me. - http://osu.ppy.sh/b/137166?m=0
I spent quite a while getting a DT HD FC on it and got basically nothing, some of my other top scores though I would say are no where near as hard as it was for me to FC doll house.

For some reason this seems to be giving a lot of PP - http://osu.ppy.sh/b/95431?m=0
It doesn't seem like that hard of a map to me, not easy, but not worth the points it is giving imo.

I think that speed is being overvalued. Almost all of my top ranks are songs that just put out a lot of notes with not too much difficulty in-between them. Anyway just stuff I've been noticing and don't quite agree with. This system is already miles ahead of PPv1 so I'm happy with it, but it can always be better.

I personally consider the second map you linked harder, than the first one. While having less demanding jumps, there still are quite a few near the end and it also challenged the streaming aspect of the game while those are pretty much inexistent in the first map you linked. Also the different ApproachRates are considered equally difficult when they lie in the interval 8 to 10, since many players have different subjective opinions about those.

Not saying, that your opinion is invalid, this is just my personal view on it. If the future feedback will go further in the direction you suggested, then I'll tweak the algorithm a bit.

Ziggo wrote:

Do you ever plan on using play statistics to enhance the accuracy of the map difficulty algorithm or do you intend to use map data only?

If I ever get an idea regarding map statistics, that would work, then I'd be more than happy to include it. The problem with it is, that constantly changing difficulty numbers mean a lot more stress to the processor. Using all available scores for the computation would be a lot harder this way.
The most important downside is what I mentioned earlier, though. Since the people who play [hard] maps are not the same people who play top-tier maps, a statistical comparison would just be wrong. Those only work well when you have the same sample data to work with.

thelewa wrote:

Tom you should really try to concisely explain to people that this iteration of pp is fundamentally different since a lot of people think of this as just an improved version of ppv1

I'm really tempted to answer with your current signature, but I'm sure some people would get that wrong.
As stated in the news post a wiki article is planned.

Tom94 wrote:

Due to processing limitations the total score value is used at score stage already. This might change in the future, but until then scores won't be weighted in all 3 categories independently.

I'm not sure if that's even necessary. Comparing the current top50 rankings between pp and tp I feel like the pp one might actually be better. Or maybe that's just me...

Almost wrote:

If ppv2 is using the same difficulty calculations on maps as tp, why is it that my highest and 3rd highest tp maps don't make my best performance even though everything below did?

Please explain Tom. :3

Ziggo wrote:

Tom94 wrote:

Due to processing limitations the total score value is used at score stage already. This might change in the future, but until then scores won't be weighted in all 3 categories independently.

I'm not sure if that's even necessary. Comparing the current top50 rankings between pp and tp I feel like the pp one might actually be better. Or maybe that's just me...

That's not the only difference between tp and pp. It'll all make more sense when I found the time to write the wiki article - but don't expect it too soon. Unfortunately I'm very busy for the next 1.5 weeks.

Almost wrote:

Almost wrote:

If ppv2 is using the same difficulty calculations on maps as tp, why is it that my highest and 3rd highest tp maps don't make my best performance even though everything below did?

Please explain Tom. :3

The algorithms are not the same - they are just similar. I'll look into it, but can't give any definitive answer at the moment.

Tom94 wrote:

Ziggo wrote:

I'm not sure if that's even necessary. Comparing the current top50 rankings between pp and tp I feel like the pp one might actually be better. Or maybe that's just me...

That's not the only difference between tp and pp. It'll all make more sense when I found the time to write the wiki article - but don't expect it too soon. Unfortunately I'm very busy for the next 1.5 weeks.

Alright, I'll wait for the wiki then. But still, I believe getting good aim, speed and accuracy in a single score is more impressive than doing the same thing in 3 different scores. But I guess you are already aware of that and have a solution in mind.

Ninja edit to make thelewa look stupid.

that's the best quoting I've ever seen

My questions:

1. Since it says it both factors in difficulty and checks all scores, will you get pp for, let's say, a low acc pass of Freedom Dive 4D? (Or insert any other appropriately hard map) Or is FCing still heavily important?
   
2. Is there still a huge bonus for SS?
   
3. How much is accuracy weighted compared to ppv1?

Ippikiryu wrote:

My questions:

1. Since it says it both factors in difficulty and checks all scores, will you get pp for, let's say, a low acc pass of Freedom Dive 4D? (Or insert any other appropriately hard map) Or is FCing still heavily important?
   
2. Is there still a huge bonus for SS?
   
3. How much is accuracy weighted compared to ppv1?

In order:

Misses are penalized very strongly if you have more than just a few. Your combo relative to the maximum possible combo also plays a huge role. But you can indeed get a huge amount of pp for lower scores. A prime example would be jesus1412's C score on https://osu.ppy.sh/b/124501?m=0 . It is a 86% 500 combo with DoubleTime in case you are curious. That's an insane score on such a map, you are free to look at Auto-mode play it with DoubleTime.

SS does not have a huge bonus by itself, but due to the high accuracy required to get one (100% is quite high, eh? :p) it indirectly is worth a lot.

Accuracy is still very relevant, however not as much as in ppv1 anymore.

6,000 to 18,000. Woo!

Tom94 wrote:

EvaRia wrote:

SPOILER

I had an interesting idea.

I know peppy wants to try to put a lot of emphasis on "Contextually relevant" scores into a ranking system, something like a ladder.

I've noticed an effect amongst certain scoreboards, something I guess we can call Scoreboard Stability.

It's hard to explain well, so examples are best I guess?

Highly "Unstable" Scoreboards (The top scores vary heavily on the same mod tier, few if any FCs or SS):
FREEDOM DiVE
osu! Stream Compilation
Don't say "lazy"
Usually incredibly difficult, long, or new maps.

Highly "Stable" Scoreboards (The top scores on the same mod tiers are incredibly similar):
Sis puella magica!
Chousai Kenbo Sengen
Usually incredibly easy or highly played maps.

I think maps with a higher unstable rate scoreboards are generally more "Contextual" (OMG, who's going to be the first to FC this? Who will be the first to FC this with HDHR? DT? FL? First to SS?)

For example, plays like some of Mesita's Flashlight plays (World's End, Gold Dust) or some of Cookiezi's old DT/HR/SS plays (Leia, Airman, Kokou no Sousei) are particularly impressive because they're pretty much the only ones who have charted with those mods or stats.

On the other hand, getting a first place in something like Chousai Kenbo Sengen, while impressive, is slightly less so because so many others have proven themselves of accomplishing it. Unless it's SS, which in this comparison I guess you could consider a Mod tier.

Anyways, I think if you had a way of comparing scoreboard instability and either compiling a chart or weighting the PP top ranks towards more unstable maps we would have a nice context relevant ranking formula.

The rough scale would range from the feasible infinite instability (Nobody has ever even passed this) to the quite impossible infinite stability (Literally ever user in existence has SS.) and award most possible points if you're the only existing user who's achieved the rank. (Mesita only user to S rank World's End on HDFL or better.)

Am I making sense?

This would give flashlight players something to hunt for as they can get high PP values of maps that nobody else has gotten a FL FC in before.

It also would weight more recent plays higher since scoreboards naturally stabilize more over time with some significantly slower outliers.

The problem with this concept, is that the people who play [hard] diffs, old maps or top-tier maps are very different and therefore the data we have to analyze will also be different. This was one of the main issues of previous pp iterations, making [hard] and old maps being weighted much more, because statistically they seemed harder than they were since most of the good players didn't play them.

Well, keep in mind that what I'm proposing is a purely contextual metric.

I agree that as far as pure skill is involved, the difficulty of the map itself is the most relevant for determining skill level.

But since you're calculating primarily based on difficulty and not on context, where I'm proposing a metric based on context with no regard for difficulty, they could theoretically exist as different metrics right?

Even shoutouts to unique awesome plays (mostly for top tier players) would make things interesting.

"<NAME> is the first to achieve rank <RANK> or better using <MODS> on <BEATMAP> (<RANKING>)!"

I don't know how exactly it would be implemented right now, but as far as a ladder system or proper context frame goes, I feel like it would be better than whatever's currently popular. It gives a certain focus on undermining scores and record-hunting that I think would be neat to see. It would also be pretty dynamic, I think.

Unstable maps are played to get the score bonus, but as they get played they stabilize again. This leaves the more difficult maps that take much longer to stabilize lingering at the top of the chart and looking through it gives you a handy way of finding the most "relevant" leaderboards.

Maybe this doesn't work the way I imagine it would though?

EvaRia wrote:

SPOILER

The problem with this concept, is that the people who play [hard] diffs, old maps or top-tier maps are very different and therefore the data we have to analyze will also be different. This was one of the main issues of previous pp iterations, making [hard] and old maps being weighted much more, because statistically they seemed harder than they were since most of the good players didn't play them.

Well, keep in mind that what I'm proposing is a purely contextual metric.

I agree that as far as pure skill is involved, the difficulty of the map itself is the most relevant for determining skill level.

But since you're calculating primarily based on difficulty and not on context, where I'm proposing a metric based on context with no regard for difficulty, they could theoretically exist as different metrics right?

Even shoutouts to unique awesome plays (mostly for top tier players) would make things interesting.

"<NAME> is the first to achieve rank <RANK> or better using <MODS> on <BEATMAP> (<RANKING>)!"

I don't know how exactly it would be implemented right now, but as far as a ladder system or proper context frame goes, I feel like it would be better than whatever's currently popular. It gives a certain focus on undermining scores and record-hunting that I think would be neat to see. It would also be pretty dynamic, I think.

Unstable maps are played to get the score bonus, but as they get played they stabilize again. This leaves the more difficult maps that take much longer to stabilize lingering at the top of the chart and looking through it gives you a handy way of finding the most "relevant" leaderboards.

Maybe this doesn't work the way I imagine it would though?

What you're describing is somewhat dimilar to what the previous pp iterations tried to do and you've seen where it led to. Nobody has the time to play _all_ the overrated old / easy diffs to farm and make them "stable".
With a lot of work a system like that might be tuned to be somewhat correct, but from my past experiences (yes, I also tried those kind of things with tp to a small degree) it doubt it'd be feasible.

The shoutouts you mention are indeed very interesting, but that would be even harder to implement with the way scoring works at the moment.

Tom94 wrote:

. Pattern difficulty is not yet implemented but definitely planned.

Planned or actually going to happen?

i will have my old Rank ;_; please give it back
30k-->18k-->5k-->8k-->31k ;(

NaThaNeL wrote:

i will have my old Rank ;_; please give it back
30k-->18k-->5k-->8k-->31k ;(

It's time to learn to play Insanes++ (: