Meanwhile Co-Op Mode is still ranked -w-
Next Up: 50% pp for Autoplay
AiAe [Wafles SHD] (100.00%) +Auto - 389pp

Seriously though, I really dont like the way the ranking is going right now.
Really good players have lower ranks than Co-Op Boosters and at the same time,
people (myself included) get HT-pp for maps they cant even dream of passing. (Which would be fine if DT was ranked, too)
And then theres people who could DT those maps and they dont get shit for what theyre able to play.
We just get random updates and changes to what gives pp and what doesnt, but that wont fix the core problem.
Sadest thing about that; except for (a minor part) the community, noone seems to really care at all.
Well played.

Wasn't co-op disabled from ranking to avoid confusion with account sharing?

co-op isn't ranked

Ovnize wrote:

co-op isn't ranked

Oh lord, didn't know that. Thought it's still ranked since everyone who used Co-Op to boost themselves still has the pp for it.
My bad there, sorry

Is easy mod nerfed?
I kinda saw in the G&R main forum that it was nerfed in osu!mania

What do you mean nerfed ?

Alright as far as the individual mods go, this is my opinion. The right bracket represents the max PP achievable, while the left brackets represents an arbitrary PP. As far as comparisons go, don't compare a mod's range to another (like hardrock and doubletime), it is intended to compare to no mod only.

No mod:

No Fail:

Easy:

Halftime:


Hardrock:

Doubletime:

The rest of the mods do little to modify difficulty.

lol'd
mods should disable pp, regardless of what mod is used

I wouldn't mind DT giving pp. Other than that, it's a no-no for me.

Choofers wrote:

lol'd
mods should disable pp, regardless of what mod is used

idk, I think DT would be reasonable at least. o!m's PP system is too acc/7k biased imo

I guess you guys are all for mods that raise difficulty having PP. I understand that there is little to know point on lowering the easy mod's PP, for instance, because odds are the player will wind up with PP that gets weighted to nothing. In my view, however, those mods should be ranked and by doing so penalties/awards should be given accordingly. It just has been unfortunate that however the PP for the HT mod calculated because it's worth quite something under the right circumstances. From what I found, it is worth about 50% the original PP. I have included a comparison of similar accuracy achieved on AiAe:

abraker wrote:

I guess you guys are all for mods that raise difficulty having PP. I understand that there is little to know point on lowering the easy mod's PP, for instance, because odds are the player will wind up with PP that gets weighted to nothing. In my view, however, those mods should be ranked and by doing so penalties/awards should be given accordingly. It just has been unfortunate that however the PP for the HT mod calculated because it's worth quite something under the right circumstances. From what I found, it is worth about 50% the original PP. I have included a comparison of similar accuracy achieved on AiAe:

SPOILER

HT pp is just calculated by the new star rating after HT is set. Which is why aiae ht gives a lot of pp since it's still around 5.5*.

Then I guess that's a weird coincidence. I can't find many HT records to compare to, so I could only assume it's around 50%

HT, NF, and EZ should disable pp

mania just needs to have a better star rating system, and the rest will fix itself automatically

I've just noticed loads of players are doing AiAe+HT, for almost 400pp. LOL

Redon wrote:

mania just needs to have a better star rating system, and the rest will fix itself automatically

i wish LN maps can have a higher star rating..


maybe mappers can set their own star rating about this

maybe

Cozzzy wrote:

I've just noticed loads of players are doing AiAe+HT, for almost 400pp. LOL

That's because AiAe SHD's star rating is overrated, not because HT is inherently overrated.

The SHD, among many other maps, is overrated because easy-to-play but high NPS patterns like this and this are overrated, as all the mania star rating algorithm really seems to take into account is peak note density. If you removed only half the notes on the few spots where these patterns pop up, and left the difficult quadstreams and all in place (which I did to save you the trouble), the diff's rating drops from 6.77 to 5.56 stars. According to Shoegazer, a lot more experienced 4K player than I am, it should realistically be rated around 6 stars. Needless to say, as a 6 star map, AiAe SHD on HT wouldn't be worth nearly as much pp as it is now.

All that is true for the no-mod version of it too, by the way. As I said, all we need is a proper star rating system, and the rest would mostly solve itself. If I was to map a 75% speed version of AiAe, made it the same as SHD on HT, and got it ranked, it would be just as much of a pp farm with the current SR algorithm.

Elementaires wrote:

maybe mappers can set their own star rating about this

maybe

nah, that would ruin it imo
tho supervising over sr and adjusting it manually is the only way to introduce SV rating in there. something like +0,5* added to algorithm value when SV is ultra-hard (smooooch example)

From what I see, what determines how much (many?) pp you gain is basically accuracy.
If you score lower than previous but have higher accuracy, no pp. Vice versa = lower pp.

I just wish they use persistent scoring, like ones of IIDX where 300g = 2 points, 300 = 1 point, just to make score-to-accuracy look "linear".

Well... it's hard to explain... but I think you get it.

Redon wrote:

Cozzzy wrote:

I've just noticed loads of players are doing AiAe+HT, for almost 400pp. LOL

That's because AiAe SHD's star rating is overrated, not because HT is inherently overrated.

Eh, I didn't say anything about HT being overrated. AiAe's rating is definitely a bit off, but 400pp HT gives me the impression that some values might have just been scaled down with HT, instead of actually recalculated. That, or something has gone horribly wrong!

stryver12 wrote:

From what I see, what determines how much (many?) pp you gain is basically accuracy.
If you score lower than previous but have higher accuracy, no pp. Vice versa = lower pp.

I just wish they use persistent scoring, like ones of IIDX where 300g = 2 points, 300 = 1 point, just to make score-to-accuracy look "linear".

Well... it's hard to explain... but I think you get it.

You don't even know how wrong you are

sooo can HT mod be unrankable now?

Cozzzy wrote:

I've just noticed loads of players are doing AiAe+HT, for almost 400pp. LOL

What's really cool is that if you passed it without mods, but got less than 400pp, you can't play HT and get more pp since the score will be lower.

xwidghet wrote:

Cozzzy wrote:

I've just noticed loads of players are doing AiAe+HT, for almost 400pp. LOL

What's really cool is that if you passed it without mods, but got less than 400pp, you can't play HT and get more pp since the score will be lower.

This is gonna change when per-mod score get implemented, isn't it?

xwidghet wrote:

Cozzzy wrote:

I've just noticed loads of players are doing AiAe+HT, for almost 400pp. LOL

What's really cool is that if you passed it without mods, but got less than 400pp, you can't play HT and get more pp since the score will be lower.

This also confirmed for EZ mod as well

When are the changes going to happen? I'm sick of looking at my 270 BPM beatmap loaded with hard streams that has 4,2 stars and I cant score an A on it (and I almost S'd verse IV)
Star rating is imcredibly broken
Converts give pp
Devs when is it going to change?

TheZiemniax wrote:

When are the changes going to happen? I'm sick of looking at my 270 BPM beatmap loaded with hard streams that has 4,2 stars and I cant score an A on it (and I almost S'd verse IV)
Star rating is imcredibly broken
Converts give pp
Devs when is it going to change?

Your map is actually not that hard, but it deserves a bit more stars, maybe 4.8?

TheZiemniax wrote:

When are the changes going to happen? I'm sick of looking at my 270 BPM beatmap loaded with hard streams that has 4,2 stars and I cant score an A on it (and I almost S'd verse IV)
Star rating is imcredibly broken
Converts give pp
Devs when is it going to change?

sometimes i wonder if devs checked this thread

akuma123 wrote:

TheZiemniax wrote:

When are the changes going to happen? I'm sick of looking at my 270 BPM beatmap loaded with hard streams that has 4,2 stars and I cant score an A on it (and I almost S'd verse IV)
Star rating is imcredibly broken
Converts give pp
Devs when is it going to change?

Your map is actually not that hard, but it deserves a bit more stars, maybe 4.8?

I don't understand how you can compare absurdly long 270 BPM streams and say that it's just a touch harder than Elemental Creation (rated at 4.72). Verse IV and HAELEQUIN generally have the correct rating (though I'd argue Verse IV is a little higher - 5.5), and I'm pretty certain no one will be able to come close to S'ing The Empress 2D if they're barely S'ing those two maps. I seriously don't see how you can even compare something like 270 BPM stream (and they aren't the easiest patterns, either) for long periods and say that it's comparable to something like Elemental Creation, when it is pretty clear that more people will do substantially worse on TE than on HAELEQUIN or Verse IV.

I am ready to pitch my idea in for the Mania star rating system.


1. SR calculated by adding up points per measure and adjust bonuses/penalties

calculate break time and decrease SR based on length (biggest decrease)
calculate length of low difficulty and decrease SR (lower difficulty = bigger decrease)

if difficulty is >= previous measure, bonus difficulty + .2



2. for every additional note on same time receives .5

ex. 3 note chord highest difficulty = 1 + .5 + .5 = 2, as opposed to 3

this gives chord values less of a difficulty than broken chords, rolls and such

for every chord that is the same as the previous will receive a penalty of difficulty*(.9)



3. for every LN = 1.5x value of a normal note, each additional LN during the duration
will add the difficulty by (1 if only one start/ending point is the same)
or (1.25 if both starting and ending points are different) or (.75 if both start
and end points are the same)

(highest difficulty calculated during the time LN are overlapped)
ex. 3 LN same time = 1.5 + .75 + .75 = 3
ex. 3 LN: 1 pair same start different end, 1 different start and end = 1.5 + 1 + 1.25 = 3.75
ex. 3 LN: 1 same start/end, 1 different start/end = 1.5 + .75 + 1.25 = 3.5



4. tighter jacks deserve exponential difficulty depending on tightness. because repetative notes/chords will decrease difficulty
and jacks that carry over multiple measures receive bonus difficulty due to same level of difficulty, the curve will be more balanced.

difficulty = (1 second/(note distance)*c)^2
c=constant (set to whatever feels right)

lets give c=4
ex. distance between notes = 1 sec: (1/4)^2 = .125 points
ex. distance = .2 sec (5x per sec): 1.5625 points
ex. distance = .1 sec (10x per sec): 6.25 points


5. balancing
part 1:
4k start same as 7k, scales higher
7k start same as 4k, scales lower
reason: as songs become more difficult, stamina will be the key factor in judging this section of difficulty, 4k is easier to maintain stamina at same difficulty
part 2:
4k start higher in this category but will scale lower
7k will start lower but will scale higher
reason: easier to learn low key amounts quickly and recognize positions compared to higher key amounts
part 3:
4k start same as 7k, scale lower
7k start same as 4k, scale higher
reason: beginners will usually be face with 1 or 2 ln's at a time, but due to having more keys = more ln's at a time
part 4:
4k start higher, scales higher
7k start lower, scales lower
reason: less columns = more notes hitting on same lane. that being said, 4k starts to focus on speed earlier and is mostly about speed end game


by combining all calculations into a scaled star rating, I think will represent a much more precise
measurement in every mania beatmap

Before I start, that was a bit hard to read. Consider some formatting.

Squishykorean wrote:

1. SR calculated by adding up points per measure and adjust bonuses/penalties

Suppose a different offset for a given pattern. Sampling it by measure will result in varied star rating for each offset.

Squishykorean wrote:

3. for every LN = 1.5x value of a normal note, each additional LN during the duration
will add the difficulty by (1 if only one start/ending point is the same)
or (1.25 if both starting and ending points are different) or (.75 if both start
and end points are the same)

Not sure what the learning curve of inverted maps is and how they will be affected by your version of SR.

The rest I think I can agree. Again, that was a pain to read, so I might have misread something.

abraker wrote:

Before I start, that was a bit hard to read. Consider some formatting.

Squishykorean wrote:

1. SR calculated by adding up points per measure and adjust bonuses/penalties

Suppose a different offset for a given pattern. Sampling it by measure will result in varied star rating for each offset.

Squishykorean wrote:

3. for every LN = 1.5x value of a normal note, each additional LN during the duration
will add the difficulty by (1 if only one start/ending point is the same)
or (1.25 if both starting and ending points are different) or (.75 if both start
and end points are the same)

Not sure what the learning curve of inverted maps is and how they will be affected by your version of SR.

The rest I think I can agree. Again, that was a pain to read, so I might have misread something.

by measuring the smallest unit, it would yield the most accurate measurement of difficulty, but calculating each beat would be too much. I'm not totally understand what you mean by offset will result in varied star rating, if its starts off late by 1/2 a measure, then it will end 1/2 a measure making no difference to the SR as a whole after it averages out. but then again, with using the bonus/penalty from comparing difficulty of the next measure it could impact the overall SR maybe by an insignificantly small margin.

the LN learning curve will be much more appropriate, usually LNs will start out on same lengths if multiple are present yielding smaller value. as songs become more difficult, there will be various lengths of LN making a higher value of difficulty.

all contain 6 LN in one measure, yet all have a different difficulty which using that formula takes care of that issue

as for the value of a LN to be 1.5 instead of it currently being worth 1 note (the worse of 2 scores: start and end release), it makes sense to increase their worth somehow since the scoring isn't completely fair on this. the scoring should be at least the average of accuracy of hitting and releasing
ex. 300 hold, 100 release would give 200 as opposed to 100
ex. 0 hold, 300 release would give 150 as opposed to 0

Squishykorean wrote:

by an insignificantly small margin.

How does a 6.00 star map differ from a 6.10 star or a 1.00 star from a 1.10 star using your system? If you are not worried about such accuracy, then go ahead dismiss what I said.

Squishykorean wrote:

LN

Fair enough.

EDIT: I feel like irregular notes deserve a bit bonus as well. For example, a two 1/4th, but between them is a 1/6th. I'm pretty sure adjust to hit something between half as fast or twice as fast depending on what note it is would have some difficulty aspect to it.

I want to hear someone else's opinion, especially Tom94's.

You know what would be really nice? If someone put together a way to actually test ideas like these out. I don't mean a tool that simply applies on specific calculation but one that can be tweaked easily so variations could easily be tested. It's one thing to just think about it and come up with an algorithm, it's a whole other thing to take a working implementation and test it against various maps and variations on the algorithm.

The single largest flaw in the current SR system isn't even the failure of the implementation to account for certain issues. The real problem is that while this feedback thread exists, there's been no way for the community to actively participate beyond making suggestions here. The SR algorithm hasn't had any changes in months. If it's stable enough to last this long, the algorithm should be public so we can actually examine it rather than make assumptions on how it works (and subsequently how to improve it). Certainly a community (or even small group of community members) looking over the algorithm would be better at spotting flaws and coming up with alternatives than a single person.

Bobbias wrote:

You know what would be really nice? If someone put together a way to actually test ideas like these out.

I would do that if I wasn't still making the analysis tool I said I was some month ago. Maybe I'll give this some priority a bit later. The best we can do is setup short sample patterns and calculate it ourselves until someone make such a thing.

Bobbias wrote:

The real problem is that while this feedback thread exists, there's been no way for the community to actively participate beyond making suggestions here.

THIS ^

just wanted to ask...
Is the pp still at work or is it finished?
This thread hasn't been updated for 4 months already so just thought to bump it a bit.

dennischan wrote:

just wanted to ask...
Is the pp still at work or is it finished?
This thread hasn't been updated for 4 months already so just thought to bump it a bit.

We dont have any clue where Tom94 is with this. My understanding is that he is busy with other stuff. Unless peppy pushes Tom to get something done, I doubt we are going to see something new for a while

And what happened with the mod "DT" in Mania?

No idea... This thread seems to be dead