forum

Performance Points feedback and suggestions (Standard)

posted
Total Posts
2,750
show more
byfar
sorry I'm gonna have to say skill wise overall, remote control DT is > airman HR. 1) being that it has crazy 250 bpm full screen jumps (for people who have experienced playing >240 bpm, they will know how much harder it really is compared to under say 220 bpm. Above that becomes exponentially harder) and second being 10.3 ar
Vuelo Eluko

Purple wrote:

I think it's the hardest thing anyone has ever pulled off in this game (aside from Maffalda DT)
Far from it. It's nothing but aim, there's very little demanded from the player from either speed (200bpm is nothing) accuracy (low BPM + simple 1/2 spam and one 200bpm stream is not hard for acc even at high OD) and reading (low density, low complexity). I'd say remote control DT is harder because it's more demanding in that respect.

Probably one of the most impressive plays in terms of aim alone, but overall? not close imo.
Purple
Well ok I guess there's some differing opinions on the matter, I thought the number of FC's in one map vs the other made it obvious which one was more difficult to get, but apparently not. Honestly I don't really *like* Airman as a map, even with my 600 plays on it, so that may have something to do with it.
DroidBass
I am not sure at all, but 90ish% FC accuracy at OD9 seems to not have a considerable bonus compared to OD7 90s% acc FC. Both are being rewarded somehow too similar but most smashers cant smash at OD9. The first one demands more real skill to smash than the 2nd one and OD pp differences starts being considerable at 95% and above.

Another topic, patterns with complicated streams and tricky patterns like Shounen Radio - neu at gold difficulty are not well rewarded because being just OD7 and FC. The general ranking on that map is somehow bad by allowing me doing a 94,30% acc with just 1 miss at the last circle, allowing me to get the #529th place on the map ranking. I can be sure that really low persons have this map at best performances list as a no mod play because being too harsh and low rewarding one. To be honest, I find that map a lot harder and less rewarding than maps like Redstar or even Nuko World insane at being less effort efficent, mostly because they have more tappable elements and OD7.5 instead of just OD7.
GhostFrog

DroidBass wrote:

I am not sure at all, but 90ish% FC accuracy at OD9 seems to not have a considerable bonus compared to OD7 90s% acc FC. Both are being rewarded somehow too similar but most smashers cant smash at OD9. The first one demands more real skill to smash than the 2nd one and OD pp differences starts being considerable at 95% and above.
You get pretty much no accuracy pp for 90%, even on OD9. 90% is really low acc and remember, 90% doesn't even mean 90%: pp only looks at your circle accuracy. If a map is 2/3 circles, 90% overall accuracy is only 85% circle accuracy.

Another topic, patterns with complicated streams and tricky patterns like Shounen Radio - neu at gold difficulty are not well rewarded because being just OD7 and FC. The general ranking on that map is somehow bad by allowing me doing a 94,30% acc with just 1 miss at the last circle, allowing me to get the #529th place on the map ranking. I can be sure that really low persons have this map at best performances list as a no mod play because being too harsh and low rewarding one. To be honest, I find that map a lot harder and less rewarding than maps like Redstar or even Nuko World insane at being less effort efficent, mostly because they have more tappable elements and OD7.5 instead of just OD7.
You're really not getting much accuracy pp for 94.30% on OD7. This kind of problem could be solved a bit by having aim, speed, and accuracy separate like in osu!tp (shounen radio would be worth it because it would net you a bunch of speed pp), but that can't happen at the moment due to some sort of server space or performance concerns or something of that nature. Tricky patterns would still be a problem though. Pattern difficulty is something that's still being figured out with no ETA.
Vuelo Eluko

DroidBass wrote:

real skill to smash
Please read what you type
DroidBass

GhostFrog wrote:

You get pretty much no accuracy pp for 90%, even on OD9. 90% is really low acc and remember, 90% doesn't even mean 90%: pp only looks at your circle accuracy. If a map is 2/3 circles, 90% overall accuracy is only 85% circle accuracy.
I am already concient of this, but I guess they need ot enlarge a bit the differences between OD9 and OD7 if we are talking about low 90s%. Meanwhile OD+9 is very commonly seen at +98% accuracy at many peoples best performance list, the 90% acc at OD9 is more rare to see at a performance list compared to a 90% FC at OD7. Most low accuracy FC no mods are frequently OD7 plays.

GhostFrog wrote:

You're really not getting much accuracy pp for 94.30% on OD7. This kind of problem could be solved a bit by having aim, speed, and accuracy separate like in osu!tp (shounen radio would be worth it because it would net you a bunch of speed pp), but that can't happen at the moment due to some sort of server space or performance concerns or something of that nature. Tricky patterns would still be a problem though. Pattern difficulty is something that's still being figured out with no ETA.
Yup, but the problem is that this case of shounen radio - neu at gold is a very weird one in which people gain lower accuracy at OD7 patterns than on OD8 patterns of higher bpm compared to some other maps of equivalent stars that I personally find a lot easier. Will be hard to fix but it is always wonderful seing some justice being done.
Xyhalion
Hello there , what is about calculating , how HARD a map is ( i mean Reading ) , best example is scarlet rose by val0108 , the map has 5,19 * , but is way harder to play than other maps in the same * difficult range
Vuelo Eluko
sliders underrated + there's no algorithm for reading difficulty im sure this has been covered at least a dozen times in this thread orz
B1rd
Tom is really stubborn about not touching sliders.
silmarilen
more like he is really stubborn about not completely breaking star rating.
dont you think that if there was an easy way to fix scarlet rose (and other alternating maps) he would have done that ages ago?
B1rd

silmarilen wrote:

more like he is really stubborn about not completely breaking star rating.
dont you think that if there was an easy way to fix scarlet rose (and other alternating maps) he would have done that ages ago?
How is buffing sliders difficult to do or going to break the star rating? Like other people have mentioned, just increase the calculated cursor route from the absolute minimum possible (I think that's what it is now) to something more than that. The only explanation Tom has given for why he hasn't done this is that it would 'make repeat sliders OP' without giving any further explanation about why this is the case.
Although one of the problem with Scarlet Rose is the low spacing, it highlights how much more difficult it to read and aim sliders than the pp system takes into account. https://osu.ppy.sh/b/327164 Or here is another example. I literally cannot FC without getting a 100 on every slider, yet it's only 4.34 stars.
silmarilen

silmarilen wrote:

dont you think that if there was an easy way to fix scarlet rose (and other alternating maps) he would have done that ages ago?
B1rd
SPOILER

B1rd wrote:

How is buffing sliders difficult to do or going to break the star rating? Like other people have mentioned, just increase the calculated cursor route from the absolute minimum possible (I think that's what it is now) to something more than that. The only explanation Tom has given for why he hasn't done this is that it would 'make repeat sliders OP' without giving any further explanation about why this is the case.
Although one of the problem with Scarlet Rose is the low spacing, it highlights how much more difficult it to read and aim sliders than the pp system takes into account. https://osu.ppy.sh/b/327164 Or here is another example. I literally cannot FC without getting a 100 on every slider, yet it's only 4.34 stars.
Granger

B1rd wrote:

The only explanation Tom has given for why he hasn't done this is that it would 'make repeat sliders OP' without giving any further explanation about why this is the case.
I wouldnt think this actually needs any further explaination, but okay, ill give you a few examples and explain.

Using extetreme slidervelocity as example. x3.5 for both sliders.


Okay we have a looong and weird slider and a repeatslider which is just as long but you can hold it comfortably at the middle.

The black line is how actual gameplay might look.

The blue line is the current algorythm, assume minimal movement.
As you can see, the first weird slider more or less plays like a circle for this algorythm, which underrates it greathly. However the Movement for the repeat slider matches the movement from actual gameplay almost perfectly, as it should.

The other extreme, "perfect" movement is exactly as the slider is mapped, in green.
Look at the long slider, its great! The algorythm sees that this slider is hard as f, overrates it a bit though. Then... oh... my... god. The repeat slider seems just as hard for this algorythm, overrating it to space and beyond. It seems several magnitudes harder than it actually plays.

Then the last line in orange assumes some medium. (i assumed medium as half of the current slider followcircle diameter, and then used shortest movement possible on that, which is how a program could do it)
The long slider looks okay. Gets close to actual gameplay. However the repeat slider still seems several magnitudes harder than what it plays like. Whats worse, it actually might seem harder than the long slider to the algorythm, absoulutely unacceptable.

Now you might think, what if we assume minimal movement if there are repeats present and else use medium movement... well that would solve this situation but got other problems in itself, not all sliders with repeats are like those hold sliders, which would mean we just pushed the problem to other sliders instead of solving it.
B1rd
Well I see now, but I don't think that the problem is that hard to solve. I mean, now Tom's talking about making some sort of algorithm for calculating pattern difficulty, but balancing repeat sliders is too hard?

I haven't done any tests, but I can't imagine them being too OP when I think a slightly closer path would look something more like this..

But really the issue isn't fast repeat sliders or the rare high velocity sliders with weird patterns, it's shorter jumpy sliders at higher BPMs where the sliders add way more difficulty to aim and especially reading compared to circles. So I don't care about how the the algorithm calculates the movement within the slider, which isn't a big factor in gameplay, it can stay the same. What I care about is how early the algorithm releases the slider and moves to the next object. This is an example of what I'm thinking of:

Currently slider patterns like the one above give barely any difficulty increase according to the algorithm than circles would. As an example I did a test- I deleted everything except the starting section of Scarlet Rose and replaced all sliders with circles. The star difficulty went from 4.13 to 4.14... wtf!?
So with the change like the once above it wouldn't affect repeat sliders, it would just keep to the slider's path slightly longer. While a small increase, this would greatly amplify the speed of the jumps between high BPM fast slider patterns like the one above to properly reflect the difficulty.
Vuelo Eluko
if sliders ever get weighting im betting emerald sword DT will be the next map everyone tries to do for 600PP. having tried it myself i believe it might actually be possible if people are given the incentive.

make it happen tom
Gigo
Wow, thanks to Granger, I finally see what the problem is. I couldn't really picture it before. Thanks for the detailed explanation.
silmarilen

B1rd wrote:

SPOILER

B1rd wrote:

How is buffing sliders difficult to do or going to break the star rating? Like other people have mentioned, just increase the calculated cursor route from the absolute minimum possible (I think that's what it is now) to something more than that. The only explanation Tom has given for why he hasn't done this is that it would 'make repeat sliders OP' without giving any further explanation about why this is the case.
Although one of the problem with Scarlet Rose is the low spacing, it highlights how much more difficult it to read and aim sliders than the pp system takes into account. https://osu.ppy.sh/b/327164 Or here is another example. I literally cannot FC without getting a 100 on every slider, yet it's only 4.34 stars.
apparently that doesnt work, because

silmarilen wrote:

silmarilen wrote:

dont you think that if there was an easy way to fix scarlet rose (and other alternating maps) he would have done that ages ago?
Yuudachi-kun
Apparently, but why?
DT-sama
Rather than calculating the absolute minimal path movement to complete the WHOLE slider, would connecting the sliderticks within the sliderball area fix the weirdly shaped sliders? Then connecting to the sliderend would use the sliderfollowcircle.
The path from slidertick to slidertick doesn't have to be within the sliderball, what's important is that when the sliderball passes over a slidertick, the cursor must be inside of it.
This wouldn't overrate fast repeat sliders because they have no ticks, and seems fair since that's pretty much how people play sliders (see Granger's post), you can't skip sliderticks without breaking combo unlike sliderends (thus sliderball instead of sliderfollowcircle), and weird sliders are harder to play when they have more ticks.
B1rd

silmarilen wrote:

apparently that doesnt work, because

silmarilen wrote:

dont you think that if there was an easy way to fix scarlet rose (and other alternating maps) he would have done that ages ago?
Apparently, not necessarily.
Stop cluttering the forums with useless posts that don't add anything to the discussion.
silmarilen
you stop cluttering the forum with thoughtless posts where you could have figured out the answer with 5 minutes of thinking then.
Yuudachi-kun
Please be nice; I love you.
B1rd
It wasn't a thoughtless post. I did in fact think there might be a reason why the easy method I had pointed out hadn't been implemented, the reason of my post was to get a reason why it hadn't been. And Granger helpfully established the reason why, instead of repeating the statement that there was no easy way to fix the problem; how easy it is is besides the point, the aim of the discussion is to actually fix the problem.
woqx
This is not about performance points but rather about the star rating system, I just don't know where else I should put it.

This map is split into three silimar parts (125, 138 and 165bpm) and gets harder every time the song speeds up. I took the Insane and cut the three parts into separate maps to be able to play them individually and I noticed this:

-Full map - 5,22 stars
-Part 1/3 - 3,91 stars
-Part 2/3 - 4,53 stars
-Part 3/3 - 5,22 stars

Apparently the star rating ignores two thirds of the map and is only based on the hardest part. I figured this might be interesting since obviously playing the whole song should be harder than playing only a third of it, no matter how easy the rest is. I know that the full map gives more pp than the individual parts would, because they have less objects, but I thought I'd point out that this is not reflected in the star rating.

I don't know if this is something that has been talked about many times before and if it is, I apologise in advance. I didn't want to read through over 100 pages before posting.
Vuelo Eluko
that's known, and it's fine. star difficulty should show the peak difficulty of the map, and no map is perfectly consistent difficulty wise.
uberpancake
So you could have one map that is insane all the way through with a peak and a map that is really easy all the way through with the same peak, and they'd have the same star difficulty?

If they had the same amount of circles, would they also give the same amount of pp despite being very different in difficulty for most of the song?
Vuelo Eluko
B1rd

uberpancake wrote:

So you could have one map that is insane all the way through with a peak and a map that is really easy all the way through with the same peak, and they'd have the same star difficulty?

If they had the same amount of circles, would they also give the same amount of pp despite being very different in difficulty for most of the song?
nah the easier one would still be a lower star rating. the longer the hard part went on the more the star rating would increase, until it eventually got to a point where it wouldn't increase anymore.
I don't think star rating directly affects pp, I think that it's calculated from all the hit objects, so it should give more pp.

I think.
Vuelo Eluko

B1rd wrote:

I don't think star rating directly affects pp
Nah it does, star rating, OD, and map length are the only things that affect PP, aside from subtle things like how many sliders vs circles and their effect on accuracy pp

Oh and also intangible aim bonuses from superficial mods like FL/HD
silmarilen
https://osu.ppy.sh/s/29157
marathon and in the middle of the night are same star rating. so yeah, pretty much.
B1rd
Wolfskin
Hi!

I would'nt call myself a very good player or anything, but maybe my post might be worth reading none the less.I can see where the performance points are coming from and they seem mostly neat, but this bugs me:

Let's say, I've got different scores on my top Beatmap . Now in order to determine which one of these individual scores gets picked to give me as a player performance points, the map with the highest rank (!) is used. Why? Why does the system choose this map, and not - which would make more sense to me - the map which would result in the best pp?

Example:
I had my pp-wise top map on 98% Acc, 269 Combo,1084.000 points, which gave me 46pp.
After replaying it recently: 94% Acc, 276 Combo, 1111.000 points, 38 pp.

So, the system only compares the point value, and I lost 8 pp by beating my record on this beatmap. It picks the 38 pp score, allthough I've got one who is higher, which is quite a downer. Feels in fact a bit as if I get punished for trying to get better on my best maps. You shouldn't be able to lose ranks just by achieving a higher map ranking, right?

cheers :)
sayonara_sekai
that is getting changed 'soon'. When?

It is a mystery.
Wolfskin
Thanks anyway! That sounds nice :D
koromo

-Ruri no Tori- wrote:

that is getting changed 'soon'. When?

It is a mystery.
Not exactly. You will be able to have scores with all mod combinations on the same map, but will get pp from the highest-pp-giving score only, however this will not change anything regarding same mod combinations (or nomod) at all. If your top score is a 99% HR score with a sliderbreak near the end, and then you FC with 98% and get a higher score, you will still lose pp.

Easiest solution would be to switch to a pp based ranking system on maps or something I think. It's kinda weird how our rank itself is determined by pp, yet scoreboards still rely on the ancient scoring system.
GhostFrog
I can't remember if this is a known bug or not, but....

The star rating of something I'm currently mapping is 6.26. However, if I add a single circle more than half a beat away from my final object (at 201.25 bpm), the star rating jumps to 6.31 (?!). If I instead put it 1.875 or more beats away, the star rating shows as 6.32 (regardless of how long of a break I add). If I put the circle half a beat or less away from the final object, star rating stays at 6.26.

I thought it might be a slider-related bug since the last object I mapped was a slider, but adding a circle half a beat away and then another circle more than half a beat away from that circle triggers the same bug.
E m i
well, I've been getting different star ratings from simply changing the offset of a map, so...
Nyxa

uberpancake wrote:

So you could have one map that is insane all the way through with a peak and a map that is really easy all the way through with the same peak, and they'd have the same star difficulty?

If they had the same amount of circles, would they also give the same amount of pp despite being very different in difficulty for most of the song?
Best FriendS and stuff are high rated because of that, literally every top 300 player I know has a Best FriendS DT FC in their top 10

This is how mappers make "popular" maps that get high ratings, because everyone wants easy pp. Nobody plays the harder and more creative maps much because they generally contain sliderspam and therefore give less pp (Talent Shredder is rated sub-5 stars, it's ridiculous). The big issue here is that sliders are way underrated due to that slider leniency overrating fear. I really think that issue should be addressed, it's starting to get annoying. Maybe it's just me though.
show more
Please sign in to reply.

New reply