handsome wrote:

do slider 100s really play that much of an impact such that they have a higher penalty than circle 100s even on higher-end accuracy? even though its a really small difference i found it a little weird :/

https://osu.ppy.sh/b/442221?m=0

ppdiff = 393.16 - 392.63 = 0.53

combo ratio = 467 / 468 = 0.9979

if combo linearly gave pp from 0 to max then this would give 393.16 * 0.9979 = 392.33

393.16 - 392.33 = 0.83

but ppdiff is only 0.53 meaning that (assuming linear combo scaling) it scales at 0.53 / 0.83 = 64% which is within the range of 45-65% combo scaling displayed in this thread: (http://www.reddit.com/r/osugame/comment ... combo_and/)

pretty sure nothing changes just because it's a slider 100 and that everything is working as intended

Bauxe wrote:

Yauxo wrote:

Dunno if that's been suggested already, but Iam not gonna read through 100+ pages.

Can we give FL some sort of lenght scaling?
I was kinda suprised that the FL score by rrtyui on https://osu.ppy.sh/b/541990?m=0 only gave 235pp, which is pretty much nothing.
I know that FL gives a big bonus on aim and that the map apparently isnt all that aim-y, but it is long and there is a shitload of things to remember, which makes it extremely difficult.

In general, more objects are obviously more difficult to remember. Easy/Normal would probably still give the usual FL pp, Hard a bit more, Insane and Extra would receive a bonus. Doesnt sound too bad to me.

The sliders are what makes that map difficult. I guess changing the way FL is calculated on sliders due to visibility would make sense, but even so, flashlight is mostly memorization, not physical skill. Don't take that the wrong way, all I am saying is that with enough time, any map that can be full-comboed can be done with flashlight. The buff it gives can't be too big imo.

The fact that memorizing a map with FL makes it FCable doesn't mean FL should be rewarded less, especially considering that, with the amount of retries put into FL, it wouldn't be so hard to get a high acc HR FC on a map you previously couldn't pass either. What Yauxo said is legit; the longer a map is, the harder it is to FC with FL, regardless of memorization or not. Even if you have the map memorized, you gotta keep in account that you have to aim at nothing, and the longer you have to do that the higher your chances of missing are.

There's a lot more to FL than just memorization, but sadly most players don't seem to agree.

I had an idea regarding more info porn. Is it viable to split pp rankings into 3 sub categories using the 3 weighted metrics (Aim, Speed, and Acc) for calculating pp? I know it'll put a little more load on the servers if we did this...

Let's say Player X is an HDHR player with 10209 pp and their sub categories consist of 1900 Aim, 1200 Speed, and 2200 Accuracy

And then there's Player Y that is an HDDT player with 10209 pp as well and their sub categories consist of 1900 Aim, 2200 Speed, and 1200 Accuracy

From this we can see that even tho Players X and Y are equal in pp, they have their own noticeable strengths and weaknesses.

Woobowiz wrote:

I had an idea regarding more info porn. Is it viable to split pp rankings into 3 sub categories using the 3 weighted metrics (Aim, Speed, and Acc) for calculating pp? I know it'll put a little more load on the servers if we did this...

Pretty much what osutp.net did before it changed, though things would have to be adjusted for ppv2 of course

http://web.archive.org/web/201311051214 ... et/players

Woobowiz wrote:

I had an idea regarding more info porn. Is it viable to split pp rankings into 3 sub categories using the 3 weighted metrics (Aim, Speed, and Acc) for calculating pp? I know it'll put a little more load on the servers if we did this...

Let's say Player X is an HDHR player with 10209 pp and their sub categories consist of 1900 Aim, 1200 Speed, and 2200 Accuracy

And then there's Player Y that is an HDDT player with 10209 pp as well and their sub categories consist of 1900 Aim, 2200 Speed, and 1200 Accuracy

From this we can see that even tho Players X and Y are equal in pp, they have their own noticeable strengths and weaknesses.

This would triple the load on the servers (ty based tom) and hence it's not viable to implement as of now.

I don't understand why Tom is so up in arms against how much PP World's End gives. It goes for over 6 minutes, has large jumps through the whole map, has high OD and spaced streams as well. It seems pretty logical that it gives the amount of PP that it does; it's not because long songs give too much PP, easy to read songs give too much PP. If it was a hard to read, it probably wouldn't give enough for the difficulty. Song length does not need a nerf, and neither do the common HR maps.
Also Tom mentioned that he want to make PP less combo based. IMO this is a radical change, I think it's fine as it is. Thoughts?

World's End is overrated, it gives way too much pp for how hard it actually is.

GoldenWolf wrote:

World's End is overrated, it gives way too much pp for how hard it actually is.

Seems to be the case, everyone who has HDHR'd it has it as their top play or close to it.

yeah that's pretty obvious, but my point is that song length shouldn't be nerfed just because of World's End.

B1rd wrote:

yeah that's pretty obvious, but my point is that song length shouldn't be nerfed just because of World's End.

https://osu.ppy.sh/s/136862
https://osu.ppy.sh/s/133693
https://osu.ppy.sh/b/293193?m=0
https://osu.ppy.sh/b/447179?m=0
https://osu.ppy.sh/b/184308?m=0

Medium length songs are being grossly overrated too.

Summer time music gives up to 400pp and it's easier than a lot of 300pp maps.

I can understand that a higher combo gets rewarded with higher pp values since it is a matter of focus and consistency but why is it that 2 misses at the very end of the beginning right after/before a huge combo don't really matter while 1 miss at around 50% of the map denies you almost all pp?

Imho the system should still reward a high number of hit circles with high accuracy partially disregarding the max combo during a map. Maybe don't give as high rewards but at least like 70-80% of the pp. Currently a high combo with a worse acc (95%) is rewarded much higher than a low combo with high accuracy.

Infevo wrote:

I can understand that a higher combo gets rewarded with higher pp values since it is a matter of focus and consistency but why is it that 2 misses at the very end of the beginning right after/before a huge combo don't really matter while 1 miss at around 50% of the map denies you almost all pp?

Imho the system should still reward a high number of hit circles with high accuracy partially disregarding the max combo during a map. Maybe don't give as high rewards but at least like 70-80% of the pp. Currently a high combo with a worse acc (95%) is rewarded much higher than a low combo with high accuracy.

This is mostly a flaw of scoring system instead of pp system... And keeping a large combo is way harder than getting 2 medium combos, especially saying from psychological way.

manjumochi wrote:

Infevo wrote:

I can understand that a higher combo gets rewarded with higher pp values since it is a matter of focus and consistency but why is it that 2 misses at the very end of the beginning right after/before a huge combo don't really matter while 1 miss at around 50% of the map denies you almost all pp?

Imho the system should still reward a high number of hit circles with high accuracy partially disregarding the max combo during a map. Maybe don't give as high rewards but at least like 70-80% of the pp. Currently a high combo with a worse acc (95%) is rewarded much higher than a low combo with high accuracy.

This is mostly a flaw of scoring system instead of pp system... And keeping a large combo is way harder than getting 2 medium combos, especially saying from psychological way.

It definitely is a flaw since people also lose rank.

Keep also in mind this is supposed to be a rhythm game. The current system counters the rhythm aspect by rewarding a higher score a lot more than a more accurate and overall more consistent play. This might be just my personal opinion but I like to compare this with a musican who play a rather large and complex piece of composition. Now is he less of a musician if he fails _once_ during the middle of the piece instead of failing several times at the very end and the beginning? Cause this is how the current pp system works.

Also your argument is kinda invalid since tv sizes almost give the same amount of pp currently as a lot longer maps. But I understand what you mean.

Infevo wrote:

It definitely is a flaw since people also lose rank.

Keep also in mind this is supposed to be a rhythm game. The current system counters the rhythm aspect by rewarding a higher score a lot more than a more accurate and overall more consistent play. This might be just my personal opinion but I like to compare this with a musican who play a rather large and complex piece of composition. Now is he less of a musician if he fails _once_ during the middle of the piece instead of failing several times at the very end and the beginning? Cause this is how the current pp system works.

Also your argument is kinda invalid since tv sizes almost give the same amount of pp currently as a lot longer maps. But I understand what you mean.

Mostly tv sizes with same star rating with longer maps are harder, but longer maps are in general harder to make full combo. It's very hard to balance this (the same with mods). We have to remember that there are a lots of flaws in both score and pp system (and mostly they are connected), pp system is newer than score, so most current flaws in general comes from scoring.

i think acc is overvalued

pp: [Tom94] Don't count tag beatmaps in pp calculations of standard mode.

what

probably because people apparently abused the hell out of touchscreens on them (I saw someone whos #2 play was like 60pp with his #1 being a tag map with 400+pp)

lol

I wouldn't really call that "abuse"

I would.

I lost pp due to the update and I still agree with the change because those maps were broken in terms of pp

Sucks for GN and Kuvster but this is probably how it should have been to begin with, it was bound to happen eventually.

I am somewhat surprised that Renai Circulation also got pp taken away from it but I would suppose that's more for completeness (so that people can't claim it's arbitrary) rather than because it's broken

I was wondering - to solve that issue where maps like Talent Shredder give like 160pp for an SS (among others) you could maybe change that "shortest possible distance between 2 notes" thing for sliders that are longer than 1/3 (so, 1/2 or longer), so that those sliders would have distance counted as if the sliderend were a regular hit circle with no slider leniency.

I'm not sure how great of an idea this is but I remember your reasoning for this somewhere was that you didn't want sliderstreams, or other patterns with short sliders at a distance from each other that are pretty damn easy to play, to be overrated. Even though this is a legit argument, slider streams really only happen at 1/4 sliders, so I figured that for any slider that's 1/2 or longer that could never possibly be an issue, and it would ramp up some of the really hard slider maps in terms of pp without overrating the easier sliderstream maps.

It's just a suggestion, but I think it'd be cool if we could focus on solving this issue. It's pretty ridiculous that Talent Shredder gives 160-170pp for a nomod SS while Coloring HDDT gives like 215 pp. It has HDDT SSes all over the leaderboards while Talent Shredder doesn't have a single nomods SS - the highest being 99.92% (1x100) at 159.86pp. If it gave at least 200-210pp for an SS that'd be a lot more fair than the way things are now, so, even if my idea isn't viable, maybe if we focused on finding a reasonable solution with multiple people we might get to something.

Aren't the sliderstreams in big black 1/2s? I suspect it'd get way overrated if this change was impkemented.

Tess wrote:

those sliders would have distance counted as if the sliderend were a regular hit circle with no slider leniency.

point is theres still slider leniency so it would be a bad change and overrate sliders. slider leniency is still there, if someone leaves that slider before the last tick it becomes a 100 not a combo break, and ppv2 still gives him the pp for the play except with a single 100 which it doesnt know where that 100 came from. this cant be implemented until theres more information available.

basically you cant give people credit for hitting jumps that didnt get to.

I'm not even sure how exactly sliders are treated now. Like distance and time between the sliderhead and next note and minimum distance to complete slider is added to the jump? Or min distance to complete slider+distance from that point to next note?

Either way if it's merged into a single jump that is timed between sliderstart and the next note, it'd mean that sometimes the actual jump you have to do is faster, since when you don't have to move much due to leniency, you still have to stay there for half the duration (between sliderstart and next note) before you can move on to the next note, unless the direction of the slider and jump is not too different + distance is kinda even, or they're close enough so that it doesn't matter.

Also being too strict on leniency is like calculating jumps with the shortest possible distance between the very edges of circles, it's not realistic to play like that in practice. You don't wanna risk slider 100s or misses.

Such treatment of leniency is perfect when the path you'd take to play the sliders as singles falls within the sliderfollowcircle anyway, like plenty of big black sliders, but stuff like talent shredder is a different story.

So I heard certain maps don't grant pp anymore, and that's because someone or some people don't approve of a certain playstyle. Why isn't this strange?

Drezi wrote:

I'm not even sure how exactly sliders are treated now. Like distance and time between the sliderhead and next note and minimum distance to complete slider is added to the jump? Or min distance to complete slider+distance from that point to next note?

Either way if it's merged into a single jump that is timed between sliderstart and the next note, it'd mean that sometimes the actual jump you have to do is faster, since when you don't have to move much due to leniency, you still have to stay there for half the duration (between sliderstart and next note) before you can move on to the next note, unless the direction of the slider and jump is not too different + distance is kinda even, or they're close enough so that it doesn't matter.

Also being too strict on leniency is like calculating jumps with the shortest possible distance between the very edges of circles, it's not realistic to play like that in practice. You don't wanna risk slider 100s or misses.

Such treatment of leniency is perfect when the path you'd take to play the sliders as singles falls within the sliderfollowcircle anyway, like plenty of big black sliders, but stuff like talent shredder is a different story.

Tom said that the game calculates the shortest possible route through sliders that the player can take. Obviously that isn't realistic and needs to be changed so sliders are properly weighed.

uberpancake wrote:

Aren't the sliderstreams in big black 1/2s? I suspect it'd get way overrated if this change was impkemented.

http://ask.fm/Tom94/answer/122374575566


Also, I've been reading some things about per-mod highscores. Could anyone give me a summary or link on that? I'm curious.

Also I'm with b1rd here. I highly doubt that my suggestion would overrate longer sliders, but hey, I'm not a programmer or whatever.

those wrote:

So I heard certain maps don't grant pp anymore, and that's because someone or some people don't approve of a certain playstyle. Why isn't this strange?

It's not like we (or I) disagree with that playstyle. Heck, I have scores on tag4 maps with touch myself. The reason it doesn't give pp is that it's inherently impossible to make a difficulty algorithm that works correctly with touch and cursor-based input at the same time.

Jumps are trivial on touch while insanely hard cursor-based. That's a fact. Same with spaced streams being almost impossible using touch due to not being able to move the cursor while hovering. (That being said, theoretically when using some method that allows hovering during touch it should be superior to cursor-based input unconditionally.)

There are only 3 options. One of them being to rate touch plays differently than cursor plays. Which is impossible to do reliably, because we can't reliably detect input devices used. People can even switch them mid-play during pause. The other 2 options are

1. Make large jumps give little pp for both cursor and touch plays, basically making it impossible to gain good pp while playing with a cursor and reducing pp to be speed/acc based.
2. Remove the only maps (in the case of osu!, that's exactly 2!) where there are patterns which become problematic. Those patterns don't exist in the rest of osu! simply due to not being allowed by the ranking rules - fortunate coincidence.

I think it's obvious to choose the 2nd option here since there is no alternative. I don't like it either. I love gimmicky maps, I love playing with touch.
Now lastly, why did all tag maps lose their pp rather than only the 2 problematic ones? The answer is being consistent. tag maps were never meant to be played by a single player and would have never passed the ranking rules - even back then - when being considered as single-player maps. This simply was a good opportunity to remove them from ranked play altogether. Scoreboards have been kept to let people keep the amazing scores they put tons of effort into and to keep the niche competition that obviously many enjoy.

Something like a tag-ladder would be an amazing thing to happen. There are in fact so many things that would be amazing to happen and which would be allowed to happen if there was the manpower to implement them properly.

Question: If you took a map and removed all the slider ticks from every slider so the combo is lower, would it award less PP?

No, because the potential PP a map can award is not based off of max combo, but on object count.

Well, what if there were separate boards for touch and non-touch players? That way, you could respect the scores that mouse/tablet players earn while keeping track of the touch scores (and the touch scores would not give pp, or give pp on a different scale.)

Also, wouldn't touch players jump directly to the spot at the same time they hit, whereas mouse and tablet players would have a gradual cursor movement? There should be a way to tell the two apart. (Of course, there is dragging, but in that case, the touchscreen has the same limitations as a mouse or tablet.)

Even if it's too hard to program a detection method inside the game, there is always the option of osu scanning for touchscreen inputs. Or, people could just voluntarily say that they are using touchscreens instead of mice or tablets.

gameon123321 wrote:

Well, what if there were separate boards for touch and non-touch players? That way, you could respect the scores that mouse/tablet players earn while keeping track of the touch scores (and the touch scores would not give pp, or give pp on a different scale.)

Also, wouldn't touch players jump directly to the spot at the same time they hit, whereas mouse and tablet players would have a gradual cursor movement? There should be a way to tell the two apart. (Of course, there is dragging, but in that case, the touchscreen has the same limitations as a mouse or tablet.)

Even if it's too hard to program a detection method inside the game, there is always the option of osu scanning for touchscreen inputs. Or, people could just voluntarily say that they are using touchscreens instead of mice or tablets.

I still don't think it can be done reliably, Tom has said so anyways. Also, what can you do about people swapping devices part-way through a play?

Rewben2 wrote:

gameon123321 wrote:

Well, what if there were separate boards for touch and non-touch players? That way, you could respect the scores that mouse/tablet players earn while keeping track of the touch scores (and the touch scores would not give pp, or give pp on a different scale.)

Also, wouldn't touch players jump directly to the spot at the same time they hit, whereas mouse and tablet players would have a gradual cursor movement? There should be a way to tell the two apart. (Of course, there is dragging, but in that case, the touchscreen has the same limitations as a mouse or tablet.)

Even if it's too hard to program a detection method inside the game, there is always the option of osu scanning for touchscreen inputs. Or, people could just voluntarily say that they are using touchscreens instead of mice or tablets.

I still don't think it can be done reliably, Tom has said so anyways. Also, what can you do about people swapping devices part-way through a play?

Why not seperate mouse and tablet rankings too?

ZenithPhantasm wrote:

Why not seperate mouse and tablet rankings too?

Yep

marcostudios wrote:

ZenithPhantasm wrote:

Why not seperate mouse and tablet rankings too?

Yep

Why not separate mouse only/tap only/tap x/click x/ touchpad/ trackpoint/ps3 controller/this too

You can't (not yet, anyway). This applies to mouse and tablet as well.

I was thinking about the problem mentioned on the previous page of sliders being underweighted and I had an idea: What if the difficulty algorithm were to minimize strain across 3 notes at a time instead of 2? Let's call the objects A, B, and C. The idea would be to find the lowest strain way of going from A to B to C such that B and C are each hit within their respective hit windows and assume that the player plays hit object B in that way. This would give you a new starting time and location and you would then consider objects B, C, and D in order to find when and where C should be hit. The main benefit to this approach is that it allows slider starts, ends, and ticks to be treated as individual hit objects without causing problems, which should give more accurate difficulty judgments in maps that use sliders in tricky ways. I think it would also help to make short fast streams more valuable compared to longer slower streams, since higher bpm streams would gain some value from being harder to transition out of.

The biggest downside I would see to it is that the relative value of patterns like this would go down when they're already undervalued as is:

Making that worth an appropriate amount would probably at a minimum require increasing the importance of time over distance for angular cursor movements and I don't feel like thinking about that right now.

gameon123321 wrote:

Well, what if there were separate boards for touch and non-touch players? That way, you could respect the scores that mouse/tablet players earn while keeping track of the touch scores (and the touch scores would not give pp, or give pp on a different scale.)

Also, wouldn't touch players jump directly to the spot at the same time they hit, whereas mouse and tablet players would have a gradual cursor movement? There should be a way to tell the two apart. (Of course, there is dragging, but in that case, the touchscreen has the same limitations as a mouse or tablet.)

Even if it's too hard to program a detection method inside the game, there is always the option of osu scanning for touchscreen inputs. Or, people could just voluntarily say that they are using touchscreens instead of mice or tablets.

How would you do that?...

Gumpyyy wrote:

gameon123321 wrote:

Well, what if there were separate boards for touch and non-touch players? That way, you could respect the scores that mouse/tablet players earn while keeping track of the touch scores (and the touch scores would not give pp, or give pp on a different scale.)

Also, wouldn't touch players jump directly to the spot at the same time they hit, whereas mouse and tablet players would have a gradual cursor movement? There should be a way to tell the two apart. (Of course, there is dragging, but in that case, the touchscreen has the same limitations as a mouse or tablet.)

Even if it's too hard to program a detection method inside the game, there is always the option of osu scanning for touchscreen inputs. Or, people could just voluntarily say that they are using touchscreens instead of mice or tablets.

How would you do that?...

Rewben2 wrote:

gameon123321 wrote:

Well, what if there were separate boards for touch and non-touch players? That way, you could respect the scores that mouse/tablet players earn while keeping track of the touch scores (and the touch scores would not give pp, or give pp on a different scale.)

Also, wouldn't touch players jump directly to the spot at the same time they hit, whereas mouse and tablet players would have a gradual cursor movement? There should be a way to tell the two apart. (Of course, there is dragging, but in that case, the touchscreen has the same limitations as a mouse or tablet.)

Even if it's too hard to program a detection method inside the game, there is always the option of osu scanning for touchscreen inputs. Or, people could just voluntarily say that they are using touchscreens instead of mice or tablets.

I still don't think it can be done reliably, Tom has said so anyways. Also, what can you do about people swapping devices part-way through a play?

Well, a touchscreen can't register a finger unless it's touching the screen. Therefore, when you tap with different fingers to aim, the cursor would skip directly from one position to another, instead of gradually moving there. That's what I meant.

Plus, when an input device is plugged in, it usually has a name. What if osu checked for keywords, or complied a list of touch devices? Then, when a device is connected, osu can see if it's a touchscreen.

gameon123321 wrote:

Well, a touchscreen can't register a finger unless it's touching the screen. Therefore, when you tap with different fingers to aim, the cursor would skip directly from one position to another, instead of gradually moving there. That's what I meant.

Plus, when an input device is plugged in, it usually has a name. What if osu checked for keywords, or complied a list of touch devices? Then, when a device is connected, osu can see if it's a touchscreen.

Guess how rapid motions at low framerate look like? Right, like teleportation, just like touch screens. Not reliable.

I think I don't even need to begin explaining why the keyword method you suggest is also not reliable.

Anything not reliable can not be used for something like scores that require absolute reliability. A leaderboard only makes sense if it's integrity is out of the question. That being said, even if we could detect input device types with 100% certainty - how should we tackle scores that switch between input devices mid-play?