Well I guess that it could approach any value without ever reaching it. What makes you say that value is 2000 and not some other number?Full Tablet wrote:
The thing is, not even with an infinite amount of 100 raw pp scores your total pp will past over 2000.TheVileOne wrote:
There's no bottom limit to the formula. Theoretically you could get an unlimited amount of 100pp scores and each score would add something to your pp. Only after awhile the pp given is so small that it could take 100s of performances to get a single pp from it.
Theoretically:
10 100pp scores: 862.4 total pp.
100 100pp scores: 1988.75 total pp.
1,000 100pp scores: 2000 - 10^(-19) total pp.
100,000 100pp scores: 2000 - 4.36*10^(-2225) total pp.
The values in practice could be different because of rounding errors with floating-point calculations, but even then the errors would be incredibly small.
forum
Performance Points feedback and suggestions (Standard)
posted
Total Posts
2,750
Geometric series have easily calculatable limits, and it's 20 for 0.95.
100pp*20 = 2000pp limit
/E: To be more specific, the limit is 1/(1-x), with x being 0.95 here
100pp*20 = 2000pp limit
/E: To be more specific, the limit is 1/(1-x), with x being 0.95 here
Maths...
I used to be good at it. >.>
I used to be good at it. >.>
Would it be possible to sort the 'First Place Ranks' by PP as well? Browsing the user page of the likes of WWW is quite disorganized.
Why would this even be necessary?NotThat wrote:
Would it be possible to sort the 'First Place Ranks' by PP as well? Browsing the user page of the likes of WWW is quite disorganized.
I also prefer it as "recent" as it is currently.-Scylla- wrote:
Why would this even be necessary?NotThat wrote:
Would it be possible to sort the 'First Place Ranks' by PP as well? Browsing the user page of the likes of WWW is quite disorganized.
What could be done is to add a sort option just for the first place ranks, but top performances should already cover that... so I don't see a point either.
The point is when viewing his top PP ranks, you have to click 'show me more' a bunch of times. After each time fastest way to get to the 'show me more' button again is to scroll to the bottom of the page. What ends up happening is you accidentally click the 'show me more' of the first place list instead of the overall plays list. That's fine and dandy because if you're viewing his profile for top plays there's a good chance you came there to search his first place play of some map to see how much PP it gave him, which means you can find it on his 'first ranks' list just as well. If the list is unsorted and he has many first place plays, you gonna have a mess real fast.
Perhaps there's another way to go about this. What I'm after is I'm curious how much PP a map awards with 100% on certain mods. This will help me figure out where to focus my efforts. If my play on map A had 98% accuracy and it gave me 150PP, and my play on map B had 96% and it gave me 143PP, which map has more potential for me for improvement?
I assume there's a direct correlation between that and the PPv2 'level' of the map with said mods selected. This gets me thinking. What is the purpose of using a PPv2 'level' to designate map difficulty? Wouldn't it be just as possible and more relatable to use 'max PP' as a measuring stick?
Perhaps there's another way to go about this. What I'm after is I'm curious how much PP a map awards with 100% on certain mods. This will help me figure out where to focus my efforts. If my play on map A had 98% accuracy and it gave me 150PP, and my play on map B had 96% and it gave me 143PP, which map has more potential for me for improvement?
I assume there's a direct correlation between that and the PPv2 'level' of the map with said mods selected. This gets me thinking. What is the purpose of using a PPv2 'level' to designate map difficulty? Wouldn't it be just as possible and more relatable to use 'max PP' as a measuring stick?
Topic Starter
Mind elaborating what you mean by 'max PP'? Where would you derive them from if not from beatmap difficulty?NotThat wrote:
The point is when viewing his top PP ranks, you have to click 'show me more' a bunch of times. After each time fastest way to get to the 'show me more' button again is to scroll to the bottom of the page. What ends up happening is you accidentally click the 'show me more' of the first place list instead of the overall plays list. That's fine and dandy because if you're viewing his profile for top plays there's a good chance you came there to search his first place play of some map to see how much PP it gave him, which means you can find it on his 'first ranks' list just as well. If the list is unsorted and he has many first place plays, you gonna have a mess real fast.
Perhaps there's another way to go about this. What I'm after is I'm curious how much PP a map awards with 100% on certain mods. This will help me figure out where to focus my efforts. If my play on map A had 98% accuracy and it gave me 150PP, and my play on map B had 96% and it gave me 143PP, which map has more potential for me for improvement?
I assume there's a direct correlation between that and the PPv2 'level' of the map with said mods selected. This gets me thinking. What is the purpose of using a PPv2 'level' to designate map difficulty? Wouldn't it be just as possible and more relatable to use 'max PP' as a measuring stick?
The PP score awarded for SS'ing the map. I assume that's basically what the 'Level' number from osu!tp website is with some conversion, except the maximum PP value obtainable from a playthrough of a map is a more relatable number to players. It would be nice to know that I got 145PP out of 163PP obtainable on a map with no mods, etc.
At the moment, it seems like N is more common for Easy diffs than the intended E, I'd suggest to put the minimal value for an N a bit higher.
Examples:
http://osu.ppy.sh/s/6257 http://osu.ppy.sh/s/41379 http://osu.ppy.sh/s/87630 http://osu.ppy.sh/s/155457 http://osu.ppy.sh/s/81557 http://osu.ppy.sh/s/102307 http://osu.ppy.sh/s/152786 http://osu.ppy.sh/s/119359 http://osu.ppy.sh/s/134220 http://osu.ppy.sh/s/150242 http://osu.ppy.sh/s/150784 And many more.
Examples:
http://osu.ppy.sh/s/6257 http://osu.ppy.sh/s/41379 http://osu.ppy.sh/s/87630 http://osu.ppy.sh/s/155457 http://osu.ppy.sh/s/81557 http://osu.ppy.sh/s/102307 http://osu.ppy.sh/s/152786 http://osu.ppy.sh/s/119359 http://osu.ppy.sh/s/134220 http://osu.ppy.sh/s/150242 http://osu.ppy.sh/s/150784 And many more.
Something to keep in mind is that a lot of "Easy" difficulties are actually Normals, though.
Topic Starter
That couldn't possibly take mods into consideration which again would make it not quite right, because mods have different effects depending on the beatmap.NotThat wrote:
The PP score awarded for SS'ing the map. I assume that's basically what the 'Level' number from osu!tp website is with some conversion, except the maximum PP value obtainable from a playthrough of a map is a more relatable number to players. It would be nice to know that I got 145PP out of 163PP obtainable on a map with no mods, etc.
But it's a nice idea. It'd make for a very good alternative measure, but I doubt the big change would be worth the benefits.
Why not just make the top ranks go something like : "193pp/230pp", where the latter is the max possible pp gained with the mod being used?Tom94 wrote:
That couldn't possibly take mods into consideration which again would make it not quite right, because mods have different effects depending on the beatmap.NotThat wrote:
The PP score awarded for SS'ing the map. I assume that's basically what the 'Level' number from osu!tp website is with some conversion, except the maximum PP value obtainable from a playthrough of a map is a more relatable number to players. It would be nice to know that I got 145PP out of 163PP obtainable on a map with no mods, etc.
But it's a nice idea. It'd make for a very good alternative measure, but I doubt the big change would be worth the benefits.
Or maybe you could just make it so that information available elsewhere, as to not clutter the top ranks, but I agree that it would be interesting to know what the max pp gained from a song is. Maybe some kind of pp calculator to test different scenarios without having to play the maps ourselves? For example you select a beatmap and input all variables into the result yourself (300s/100s/50s etc....), then it shows you how much pp you would gain?
Topic Starter
I plan on making all the underlying formulas open in the wiki, so anyone will be free to implement his own pp calculator. I'm not sure if or how that would fit into the game or the website, though. I'd say it'd be better as a 3rd party program.mcdoomfrag wrote:
Why not just make the top ranks go something like : "193pp/230pp", where the latter is the max possible pp gained with the mod being used?
Or maybe you could just make it so that information available elsewhere, as to not clutter the top ranks, but I agree that it would be interesting to know what the max pp gained from a song is. Maybe some kind of pp calculator to test different scenarios without having to play the maps ourselves? For example you select a beatmap and input all variables into the result yourself (300s/100s/50s etc....), then it shows you how much pp you would gain?
Showing the maximum possible pp for a given mod combination sounds juicy. It likely won't happen unless the pp formulas get implemented into the client, though. Online adding a mod selection or even worse: a display for each possible mod combination would be far too much clutter in the current state.
Keep in mind that those limitations can always change, so I wouldn't rule it out completely. I like the idea.
So new scores always replace old ones even if the pp value is the same? I had a 51 pp play as my #1 play for the longest time, but it was just passed [and consequently weighted less] by an equal pp play. Is it because the more recent one has higher accuracy? Is it because it gives a higher amount of pp but in decimals? Or do newer but equal pp plays always replace older ones?
edit: could it also be because of DT?
edit: could it also be because of DT?
Now that we can see the PP gain for each song, I can't help it but think that jumps or some other thing is being overrated, CONTROL REMOTE is overrated, I was ok with that score being rrtyui's top performance, even being www's or dragonhuman's best, but the difference is way too much, what I mean is... 542pp ?? (rrtyui's), seriously? then, his second best is 454pp. I KNOW that Control Remote on DT is just freaking hard, and just a few can do something like that, but I still feel like something is wrong there...
Mind linking such a map? Never heard of Control Remote!pold10 wrote:
Control Remote
it definitely overrates jumps hence why remote control is harder than just about every 0108 map according to it
https://osu.ppy.sh/s/53857- [ U z z I ] - wrote:
Mind linking such a map? Never heard of Control Remote!pold10 wrote:
Control Remote
not overrated at all. those jumps with DT are bloody difficult.
Oh but that's Remote Control!!! Not Control Remote!!!TMoI wrote:
https://osu.ppy.sh/s/53857
Haha.
Yeah those jumps on DT are a whole new thing, especially with that OD.
Yeah, as it is now, it's OD9 AR9 with jumps that stretch across the screen at a decent BPM and normal CS.
So I'd assume it'd be worth some.
So I'd assume it'd be worth some.
Lol, my bad... xD- [ U z z I ] - wrote:
Mind linking such a map? Never heard of Control Remote!pold10 wrote:
Control Remote
Indeed it's not easy at all, can't deny that, but I still feel like it's a little too much.TMoI wrote:
Yeah, as it is now, it's OD9 AR9 with jumps that stretch across the screen at a decent BPM and normal CS.
So I'd assume it'd be worth some.
Not sure if this counts as a suggestion or a feature request, but it would be nice to have a separate sub-ranking and/or best performance list for selected mods
For example, one would be able to choose a combination of mods (e.g. HDFL only) so that only scores using those mods will contribute to pp. That way we can compete with people who use specific mod combinations, as well as see where we ourselves are positioned.
An alternative for the second feature (best performance) would be to display the mods used to achieve that score next to the pp value. That way we can distinguish between HDDT and HDHR scores (because as of now, we'll never know what mods a player used to get that C in his best performance).
For example, one would be able to choose a combination of mods (e.g. HDFL only) so that only scores using those mods will contribute to pp. That way we can compete with people who use specific mod combinations, as well as see where we ourselves are positioned.
An alternative for the second feature (best performance) would be to display the mods used to achieve that score next to the pp value. That way we can distinguish between HDDT and HDHR scores (because as of now, we'll never know what mods a player used to get that C in his best performance).
Topic Starter
It's a known issue, that jumps scale too much in the very high end. Since this only concerns a veeeery small amount of scores (e.g. the ones that come close to FCing remote control with DT) it doesn't have a high priority, though.
Thats a nice idea, though actually most normals are easy anywayPriti wrote:
At the moment, it seems like N is more common for Easy diffs than the intended E, I'd suggest to put the minimal value for an N a bit higher.
Examples:
http://osu.ppy.sh/s/6257 http://osu.ppy.sh/s/41379 http://osu.ppy.sh/s/87630 http://osu.ppy.sh/s/155457 http://osu.ppy.sh/s/81557 http://osu.ppy.sh/s/102307 http://osu.ppy.sh/s/152786 http://osu.ppy.sh/s/119359 http://osu.ppy.sh/s/134220 http://osu.ppy.sh/s/150242 http://osu.ppy.sh/s/150784 And many more.
I guess remote control with dt will be the next big thing to grind, seems many people already are trying to get good scores at itTom94 wrote:
It's a known issue, that jumps scale too much in the very high end. Since this only concerns a veeeery small amount of scores (e.g. the ones that come close to FCing remote control with DT) it doesn't have a high priority, though.
Though that's just because it's an X difficulty.. being done with DT... I think if any other X were able to be fc with DT similar thing would happen.
..im just assuming the new star system is directly related to ppv2
I just wanted to throw out this idea I've been thinking of. I've been worrying about losing pp when I play a song with mods because I get more 100s. Perhaps we should be allowed to choose whether we want to submit a ranked score if it rewards less pp than our current personal best. It would certainly alleviate my concerns about replacing a SS with something not a SS but with mods.
Also, how much does not FCing due to 100s on sliders hurt you? (i.e. x1000 v x999 due to getting a 100 on a slider)
EDIT: *hurt pp gain
EDIT: *hurt pp gain
it usually kills meTMoI wrote:
Also, how much does not FCing due to 100s on sliders hurt you? (i.e. x1000 v x999 due to getting a 100 on a slider)
sameGoldenWolf wrote:
it usually kills meTMoI wrote:
Also, how much does not FCing due to 100s on sliders hurt you? (i.e. x1000 v x999 due to getting a 100 on a slider)
although I should have phrased that better, I should have said hurt pp. Will edit.
I find that fast single tap maps (bpm 210-250) that aren't jumpy are underrated.
For example, I think http://osu.ppy.sh/s/45422 is at least about as hard as http://osu.ppy.sh/s/43466 , but the stacks and small jumps make dear you more highly rated.
The difficulty comes from tapping and snapping at 220 bpm, and there are a few chains of up to 10.
Low spacing with higher bpm--> Aim-Algorithm sees as being easier, because the average cursor speed is lower.
220bpm single tap --> Speed-Algorithm sees this as being easier, because 220 bpm 1/2 is nowhere near 172 1/4.
For example, I think http://osu.ppy.sh/s/45422 is at least about as hard as http://osu.ppy.sh/s/43466 , but the stacks and small jumps make dear you more highly rated.
The difficulty comes from tapping and snapping at 220 bpm, and there are a few chains of up to 10.
Low spacing with higher bpm--> Aim-Algorithm sees as being easier, because the average cursor speed is lower.
220bpm single tap --> Speed-Algorithm sees this as being easier, because 220 bpm 1/2 is nowhere near 172 1/4.
-Chronopolis- wrote:
Low spacing with higher bpm--> Aim-Algorithm sees as being easier, because the average cursor speed is lower.
220bpm single tap --> Speed-Algorithm sees this as being easier, because 220 bpm 1/2 is nowhere near 172 1/4.
Tom94 wrote:
Another reason at least for some of these maps to be underrated is, that they feature quick single passages with low spacing which gets underrated in the current algorithm. Couldn't really find a way to fix that without completely breaking spaced streams yet.
To add to it, Tom already said that to get a proper algorithm that considers every possible aspect he'd need to have per-hitobject data, which is currently not available.
Getting combo and decent accuracy on OD 10+ fullscreen jumps doesn't sound like something many people would bother grinding. Remote Control already requires a bit of luck (or a ton of skill if you can do it consistently) to FC without DT.Bassist Vinyl wrote:
I guess remote control with dt will be the next big thing to grind, seems many people already are trying to get good scores at it
Though that's just because it's an X difficulty.. being done with DT... I think if any other X were able to be fc with DT similar thing would happen.
..im just assuming the new star system is directly related to ppv2
The idea has been suggested before. Not sure if there has been feedback, but it sure would be nice to compare how much difference combo/accuracy would make xD.TheVileOne wrote:
I just wanted to throw out this idea I've been thinking of. I've been worrying about losing pp when I play a song with mods because I get more 100s. Perhaps we should be allowed to choose whether we want to submit a ranked score if it rewards less pp than our current personal best. It would certainly alleviate my concerns about replacing a SS with something not a SS but with mods.
Here's my proposal. It involves trying to treat streams and singles separately. I'm aware of the general points of the current pp system, but obviously I may be ignorant of lots of things.mcdoomfrag wrote:
-Chronopolis- wrote:
Low spacing with higher bpm--> Aim-Algorithm sees as being easier, because the average cursor speed is lower.
220bpm single tap --> Speed-Algorithm sees this as being easier, because 220 bpm 1/2 is nowhere near 172 1/4.Tom94 wrote:
Another reason at least for some of these maps to be underrated is, that they feature quick single passages with low spacing which gets underrated in the current algorithm. Couldn't really find a way to fix that without completely breaking spaced streams yet.
Stress rating of pattern as calculated as a stream:
Speed stress calculation: Stream bpm and length contribute. If average mouse velocity is high (spaced streams), that also contributes to speed.
Aim stress: Average cursor speed. The exact ratio of average_cursor_speed_stream: speed_stress_stream can be different than average_cursor_speed_singles: speed_stress_singles.
Turns add to this based on how fast they change in degrees and the average cursor speed. Change in spacing between stream-distanced spaced notes can also add difficulty.
Singles:
Aim: Primarily based off of average cursor velocity.
Seperate idea: make change in spacing add somewhat to aim difficulty:
Effective distance from note to next note = distance from note to next note + 0.5*difference in spacing from this note to the last note.
Speed stress calculation: Chain length and bpm contribute. Single bpm contributes much more from 200>280-300 1/2 notes than it currently does in pp (which doesn't discriminate between singles and streams for speed calculation).
Very high speed singles calculation (280-300+):
At this point virtually everybody is alternating, so one can take that into consideration when constructing an algorithm. I don't know what the relative difficulties:
Aim: The same as current.
Speed stress calculation: For a base line, the speed difficulty at the boundary from 279 to 280 should be continuous. The bpm contribution to stress probably should go up slower (because aiming and mouse speed become the limiting factors to my knowledge https://www.youtube.com/watch?v=rPlCPRMM2g8).
Judging whether something is singles or stream:
If notes are farther than X distance apart AND have have an average change in angle greater than Y from note to note:
-If <280 bpm: treat as singles
-If >280 bpm: treat as very high speed singles (which tries to estimate difficulty based on assumed alternation.)
Otherwise:
-treat as streams
are sliders still undervalued? http://osu.ppy.sh/ss/1440488 got 0 pp for this even thought it was by far the hardest FC i have. yeah the accuracy is low... but still
Sliders are ignored in accuracy calculations so your acc was like maybe 93% which does hurt a lot. Also as mcdoomfrag pointed out, fast singles that aren't far spaced are underrated for pp. The score was worth 162pp AFAIK on your profile.Soulg wrote:
are sliders still undervalued? http://osu.ppy.sh/ss/1440488 got 0 pp for this even thought it was by far the hardest FC i have. yeah the accuracy is low... but still
Soulg wrote:
are sliders still undervalued? http://osu.ppy.sh/ss/1440488 got 0 pp for this even thought it was by far the hardest FC i have. yeah the accuracy is low... but still
Tom94 wrote:
The reason is, that the current way sliders are dealt with is very generous in terms of the slider's favor. Currently there is no measure as how hard it is to actually follow a slider. The minimum distance you have to move to complete the slider is added up to the jump to the next hitobject, but that's it atm. This "minimum distance" is required to not give fast repeatsliders, or even worse: slider-streams, ridiculous pp amounts.