Xcrypt wrote:

Yeah I agree xxbidiao, to some extent.
With the example of stairs vs random hits. It really depends on the player, I am mainly a RD user so patterns like stairs are not my best thing.
But if we calculate it according to the average of all players, I think this is the best we can do. Ofcourse, we'll have to guess a bit here and there, but let's say stairs vs random hits, they weigh exactly the same? For example.
Maybe only weigh things that are clearly harder for the majority of the playerbase, like density, LN, jacks etc.

In all honesty this is painful guesswork right now. We'll just have to see at the first release of the new scoring system and then we will be able to pinpoint some flaws I think.

Yes, exactly. We would have a hard time upon initial release... Maybe on many other issues

And all players average? That doesn't work either, because players of osu!mania is not all the humankind.

Just use 4key vs 7key as example.
We all know that 4key and 7key are difficult to play in different ways. 4K players play 4K insane maps and say 4key is hard; 7K players play 7k insane maps and say 7key is hard. (Please reference to earlier posts on 4key and 7key discussions.)
If you try to average all players, you will most probably get the answer that 7key is harder, because there are more 7key players than 4key ones (Edit: on osu!mania, who can play at least 4k/7k expert maps). But is that the truth?

My point is that don't hope that the ranking system would be 100% accurate (or to say, reliable) - even 10% accurate is nearly impossible to guarantee.
(Edit: Instead of making a complex system that is hard to understand), It would be a good idea to make the algorithm simple and quick enough, even possible to completely substitute star difficulty, regardless of its reliability. Everyone just want to know that 10 star map is harder than 7 star map, doesn't it?
(Actually, woc had considered improvements on star difficulty, but he gave up when finding osu! just lost its response for tens of seconds just calculating every map's SD)

What can I say, I suck at stairs, but they're fun when you can accidentally pass them. I personally get secretly excited and jump with happiness.

Oh, right, on-topic and all: I can't (or shouldn't) complain much since 99,9999% of the maps I played were auto-converts (and hell did they give me pp!) but are there that many mania-specific maps to "cover" all the auto-converted ones? 'Was just wondering that.

how about rank of osu!mania separate become 4 (4k player only,5k,6k and 7k) ?
and every player can only choose 1 or 2 of them

Karuta-Roromiya wrote:

how about rank of osu!mania separate become 4 (4k player only,5k,6k and 7k) ?
and every player can only choose 1 or 2 of them

Already been suggested and heavily denied. But I cant help but ask, why one or two? If a player is good enough to rank high in all keymodes I don't see why he shouldn't be allowed to :/

And like xxbidiao is saying, since difficulty is so subjectional, its impossible to make a completely accurate scale. But that doesn't stop us for making a basic scale to give players an idea about what theyre about to play. So the difference between stair and random should probably nonexistent. The algorithm might be better off focusing more on the bigger factors.

Also @tom, checking for single finger strain simply covers for jacks and hand strain is pretty much irrelevant. I mean there's so many different playstyles out there. I personally play 4k with one hand. In 7k some people use left thumb, some right, some both and some none. I don't think a variable that depends on playstyle would be a smart thing to do. And it all depends on the patterns anyways, the value would be pretty irrelevant anyways.

Drace wrote:

Also @tom, checking for single finger strain simply covers for jacks and hand strain is pretty much irrelevant. I mean there's so many different playstyles out there. I personally play 4k with one hand. In 7k some people use left thumb, some right, some both and some none. I don't think a variable that depends on playstyle would be a smart thing to do. And it all depends on the patterns anyways, the value would be pretty irrelevant anyways.

Limiting yourself to one hand if you can use both should not be considered by such a system imho. That's like assuming people would use one finger, even for streams, in standard, thus giving people who actually use both an unrealistic bonus.
Sure, algorithm should be as much as possible independent of playstyle, but it should orient itself at the most successful ones. As for space-bar jacks in 7K - it should be common sense to use both thumbs at least in that case.

The algorithm is supposed to automatically determine pattern difficulty, I don't understand why it would be irrelevant. :p

Karuta-Roromiya wrote:

how about rank of osu!mania separate become 4 (4k player only,5k,6k and 7k) ?
and every player can only choose 1 or 2 of them

That's the stupidest idea ever. I play all K's and I want to keep playing all of them without having "to choose" any in particular.

Tom94 wrote:

Limiting yourself to one hand if you can use both should not be considered by such a system imho. That's like assuming people would use one finger, even for streams, in standard, thus giving people who actually use both an unrealistic bonus.
Sure, algorithm should be as much as possible independent of playstyle, but it should orient itself at the most successful ones. As for space-bar jacks in 7K - it should be common sense to use both thumbs at least in that case.

how is that different from jacks on any other column? nobody is stopping you from using two fingers or even more for other keys, in fact I've seen people who specifically learn to play certain charts like that with the iidx controller.

you should assume by default all columns are even. this is not djmax

what about a ranking system like EZ2ON? I think it would be good.

For those who don't know how EZ2ON ranking works i'll try to explain. In EZ2ON u have two kind of level, one is the level we use to know (with the exp) and the other one is your "clear level" that increase when u clear a song (for example if u clear a song which has difficulty level 5 ur clear level will be 5 and it will get higher once u clear a harder difficulty level song, which mean that u can skip from lv 1 to 10 if u're just able to clear a lv 10 difficulty song), and the clear level could be used to block the more skilled people from farming pp in easy maps by not giving pp if they clear a song which has a lower difficulty than 10 (just to make the ranking fair, that's my point of view).

Please don't hate me for this post

Or we can just play osu!mania and talk about mania without constantly bringing o2jam, synthesia, whatever games that aren't mania; since not everyone played those in the past before coming here.

I also think you should assume all columns are even.

Tom94 wrote:

Limiting yourself to one hand if you can use both should not be considered by such a system imho.

I'm not limiting myself at all, I actually play better that way due to my 8K background. It's an example to how per hand strain values aren't ideal for determining map difficulty. Actually, one of the best 4k players in the world plays with 3 keys one hand and one key on the other (7k+1 background I suppose).

Tom94 wrote:

The algorithm is supposed to automatically determine pattern difficulty, I don't understand why it would be irrelevant. :p

Sorry I believe you misunderstood, what i was referring to as irrelevant is per-hand strain, since it all depends on the pattern difficulties anyways.

Aqo wrote:

how is that different from jacks on any other column? nobody is stopping you from using two fingers or even more for other keys, in fact I've seen people who specifically learn to play certain charts like that with the iidx controller.

you should assume by default all columns are even. this is not djmax

I completely agree. Far too many people play with different styles. An algorithm forged around one specific style would be unfair.

Tom94 wrote:

Limiting yourself to one hand if you can use both should not be considered by such a system imho. That's like assuming people would use one finger, even for streams, in standard, thus giving people who actually use both an unrealistic bonus.
Sure, algorithm should be as much as possible independent of playstyle, but it should orient itself at the most successful ones. As for space-bar jacks in 7K - it should be common sense to use both thumbs at least in that case.

The algorithm is supposed to automatically determine pattern difficulty, I don't understand why it would be irrelevant. :p

I believe like Drace said, you have misunderstood something, or you actually have your words contradict each other within your post!

You say you can automatically determines pattern difficulty, right?
It seems TP is automatically generated. Yes, I have actually read your TP algorithm publicized- and there are experimental values inside the model, right?
This is where subjective things come in.

You have also said that it should orient itself at the most successful ones. But what is the most successful ones? This is completely a subjective matter.
Just use the space key problem as an example. Does using both thumbs hitting space making things better?
Unlike in osu!, actually the space key hits are requiring more coordination than hand speed. So using both thumbs to play space key may cause severe early hits to happen, making people using both thumbs not having identical scores on every map.

And there are actually more examples, listed in my previous post, that it's not a good idea to put "one style" into the "better" assumption, making the other playing style have underestimated power. If you are trying to do things like "this map is hard for people using left thumb for space, so I should raise PP for people using right thumb to get better score." I don't think this is relevant - I bet you also don't.

So what I want to say here is clear. "There are 50% maps which player with left space can play better, 40% maps which right will do better, 10% with both thumbs, so I will award PP for these who use left thumbs, and subtract PP from these who use both thumb as space hitter, because left thumb space hitter get more #1 than both thumb space hitters. And because people having higher rating tend to play better, maps they have played great should be higher weighted." It DO make sense because these 50% players have better play records which means they should be higher rated, but coming to the difficulty calculation - this do nothing at all with difficulty. It is just a statistical reference to the maps, which means it is irrelevant.
This actually DO happens in star difficulty calculation, where 4 Key weighting is dramatically decreased. And star difficulty is another automatically determined difficulty calculator, right?

I gotta give in. After reading the arguments of you guys, I agree, that it'd make more sense to treat each key individuallly. Will still show experimental results with the additional strain measures just for consideration and comparison.

Tom94 wrote:

I gotta give in. After reading the arguments of you guys, I agree, that it'd make more sense to treat each key individuallly. Will still show experimental results with the additional strain measures just for consideration and comparison.

waa, i can't wait for the changes..! xD
keep up the good work...

Something quick and dirty I made that reflects my personal view on how such a system should deal with accuracy.
This is only on a per diff/map basis and it assumes an average experienced player's 300/300G distribution.
There are 96 points to be gained. More points mean better performance, this is what the vertical axis represents. The horizontal axis represents the end % result of the play.
Anything below 50% is worth 0 points, just because this can be often be achieved by spamming the fingers in some random pattern.

While this may look steep, I'm pretty sure it's far less steep than what is implemented in standard or what o!m currently has.
I'm basically saying someone who can get SS on a song deserves twice as much score as one who can A it. While in reality, the person who can SS is probably more than two times as good but like I said, stressing this too much leads to accuracy becoming 'the only viable playstyle' while I aimed to seek some kind of balance here between the different playstyles.

rate and flame! :p

The timing windows in o!m are fairly huge, so an exponential accuracy rating seems fair. Even on some difficult maps, 98% can be considered pretty low. I also agree about super low acc scores like 50% not getting calculated, mainly so mashing gets discouraged.

Also, a few posts back I remember seeing some plans to make HR mod count towards your acc score. Does this mean it's becoming a rankable mod? : o

I've always found ranked mods to be ridiculous. There are enough of different playstyles and mapping styles to keep everyone on a equal footing and the game entertaining. HR for example would be far too much of an advantage for accuracy-style players.

but that's just my opinion

Pyakura, the way mods work in standard is basically they don't help your scoring at all unless the map is too easy for you. (I'm not really experienced in std but this is what it seemed like to me: the difficulty of the mods far outweigh the score bonuses unless you already mastered the map)
HR would be great for accuracy players to which the map is too easy, and DT for hard clearers for whom the map is too ez.
So I don't think it's such a bad idea tbh. It's not necessary per se for overall rankings assuming that we have enough ranked maps hard enough for the best players, but for per map rankings it can certainly be cool.

Except at the moment we don't have a ranked HR mod, so the only way to get that boost is to do DT which mutilates the song.

Accuracy shouldn't be looked at at all. Use score.

I tested on many maps now and it seems like score is FAR more credible for how well you can play a map than accuracy% (honestly accuracies are completely random, who doesn't have a million stories of beating an A with a B and so on); it's also pretty much immune to mapper-selected OD values which is cool.

Below 500,000 score = pure mash

500-600 = controlled mashing
600-700 = playing maps just outside of your reading limits. this is the main thing that should give performance.
700-800 = playing maps on the border of your reading limits. this is usually where you start to get S rank.
800-900 = the map is easy for you
900-999 = farming accuracy

Beating an A with a B is almost always because of the 300/300g distribution varied, which is not considered in my quick and dirty calculation. So yes, score may be more accurate. I just wanted to give more or less an approximation for a relation that tries to balance the performance gained for all playstyles.
'the map is too easy for you' on 800-900 that really depends on the player. Most accuracy players won't agree, but since most of the experienced players come from o2jam here that seems pretty accurate for me and probably most others lol.

To be completely honest, the Acc% is a good way too gauge just in WHICH category the player is on this map, being those categories the ones Aqo posted.

Just to be a lil' bit more precise on the matter, S ranking starts at 95,0001 acc% (YES 95 is an A) A goes from 95 to 90,0001 (the same applies to any other rank) and then from this point on it decreases in values of 10, so B is 90-80, C is 80-70 and D is -70.

With this in mind let's get into the scores, this is a lil' bit ambiguous since the factor that decides the Rank is the ACC ONLY so... yeh, pretty much a guide to just know how much you need to get in a Medium note density map, not too much, not too little. Going into the too much the scores tend to get lower, meaning that the more notes you get the more are your chances to increase ACC even if you have a shitty score (Already confirmed it with an autoconvert of 30 mins.... Trust me.... a nightmare to get a decent score while trying to get a 95% acc or so to speak A just barely missing the S) and for the lower note density it just goes the other way around you tend to need higher scores for getting a higher rank.

So, Aqo pretty much did the job on this part but i'll put it again with a more accurate value (on mid-density maps I repeat)

D: Pretty much everything that oscilates into 0 and 400K

C: +400K pressing on the 515K mark (This is pretty much an estimate since its difficult to get a C on pourpose, at least for me XD)

B: +540K is definetily a B, then it kinda stops to fade out in the early 600

A: +600K and it kinda stops at 740K

S: +750K until 980K And I'm pretty sure its quite accurate, there are exceptions of course, but its pretty difficult to find an S in the 985K mark and impossible to see one on the 990's

SS: from 980K onwards..... not much to say here.

Having these data in mind I think that's better making an estimation of quality play on the Point department, but for minor adjustments on the point giving use the ACC % factor as a second option, in other words weight more the actual Score, but for scores pretty similar (tends to happen a lot) Value the ones that have more ACC%

I mean, come on, I've seen cases like of an S having (not really but something like this) 985.575 and an SS having 985.475 And man I'm sure that's just cause the guy who SS'ed couldn't get the timing good and failed at MAXing everything It is painful to have a perfect ACC and just be beaten because of the random point giving from the 300/300g as xcrypt said.

Also, I think that the Density of the map in mania affects a lot these aspects.... ACC and scores tend to be thrown off gaphics by HYPER density maps.... its just silly XDD

PS: Just trying to put some more numbers in there and also try to get the ACC and Score on the same post just to compare them more easily

Aqo wrote:

Accuracy shouldn't be looked at at all. Use score.

I tested on many maps now and it seems like score is FAR more credible for how well you can play a map than accuracy% (honestly accuracies are completely random, who doesn't have a million stories of beating an A with a B and so on); it's also pretty much immune to mapper-selected OD values which is cool.

Below 500,000 score = pure mash

500-600 = controlled mashing
600-700 = playing maps just outside of your reading limits. this is the main thing that should give performance.
700-800 = playing maps on the border of your reading limits. this is usually where you start to get S rank.
800-900 = the map is easy for you
900-999 = farming accuracy

I would change that by lowering the first two by 100K and putting a replacement 500-600 as 'know what you're meant to be doing (so not mashing) but just not able to pull it off properly'. 'Controlled mashing' will usually end up giving a score in the mid-high 400ks.

Oh and A ranks of 800k+ are actually rather common in my experience (and usually mean you completely screwed up on one part of the map, but played the rest well).

>700k B rank: http://osu.ppy.sh/ss/1302389

Score and rank really don't correlate all that well.

It's common for me to get B's of 470k ~ 500k huehue

You cannot measure mania difficulty well with just strain values. Standard is a mode that requires next to no reading skill, while mania's way more about reading than physical ability. When a map is too hard for someone, in 90% of the cases it's because they can't comprehend all the objects, not because they fingers aren't fast enough, that's why you need to analyze patterns. Two patterns with the same note density and column strain might have a very different difficulty if one's easier to read than the other.

Kamui Gakupo, KAITO & Len Kagamine - Arrest Rose [Level 22] is an example of a very well-done chart that becomes much harder with Random mod.

I know i'm not E.T. so my opinion is probably going to be thrown out the window into a large pile of dog mess that was left behind this morning, but i'm going to say this anyway.

Mods suck! All of them, and I have a perfectly good reason for this.
Double time- If the map maker wanted the map to be in double time he would have made the song double time.
Hard Rock- Doesn't even work in mania.
Sudden Death- If you have a high rank you should be getting this anyway
Fade in - Encourages you to look in the wrong place.
Hidden - You should be looking there anyway.
Flashlight - This is just annoying, I can't even tell if i'm doing bad or not.

Well that's that out of the way, now to move onto another one of my (most likely) globally unacceptable opinions. I would prefer it if PP was decided by score not accuracy. This is because when you hit a 300perfect and a 300 they are considered as the same accuracy even though they aren't.

There we go, I've strapped the steak to my legs, now time to click "post" and drop myself into the lion pit.

milky228 wrote:

I would prefer it if PP was decided by score not accuracy. This is because when you hit a 300perfect and a 300 they are considered as the same accuracy even though they aren't.

I can't agree any more.

PyaKura wrote:

It's common for me to get B's of 470k ~ 500k huehue

spamspamspamspam \:D/

milky228 wrote:

I would prefer it if PP was decided by score not accuracy. This is because when you hit a 300perfect and a 300 they are considered as the same accuracy even though they aren't.

quoting again for great justice. accuracy is practically meaningless (not really, it gives a general impression, but still random and inaccurate/inconsistent) in mania, go with score.

I agree that PP should be based on score and not accuracy.

Also, mods should not be factored into the ranking system. My biggest argument for this is that they don't give actual score bonuses in the songs like they do in standard so it would be counter-intuitive if they increased your pp. If you want mods to be calculated then you would have to make the max for songs be 1mil x modifier increase, and this is a bad idea because it takes away the beauty of having the max be a nice even number and forces people to play with stupid restrictions if they want to be competitive.

Even if the 0* modifier is unintuitive, I don't see why players shouldn't be rewarded for playing more difficult maps well. It's the same as osu!standard, where (for example) rrtyui will sometimes barely get in the top 50, but is also the only player who was capable of playing it with DT.

The only mods that can't really be counted for the ranking system are the mental ones, imo. Hidden makes pretty much no difference, and on Fade-in you can just lower the scrolling speed. I haven't really tried Flashlight in mania though, so I'm not sure about that one.

Seeing as all mods are currently 0*, it should be possible to recalculate old scores, but I doubt that will happen because of the 1000000 cap.

EDIT: -nvm, failed to read correctly-

Cozzzy wrote:

Even if the 0* modifier is unintuitive, I don't see why players shouldn't be rewarded for playing more difficult maps well. It's the same as osu!standard, where (for example) rrtyui will sometimes barely get in the top 50, but is also the only player who was capable of playing it with DT.

Except when you get a high nomod score, and then a slightly lower DT score (which obviously needs more skill), you won't get pp for it because only the highest score matters. Once you do well without mods, you're screwed.

Would be nice if the PP system changed to score instead of rank, so a high S can yield more points than a normal SS.



I strongly oppose the implementation of mods into the ranking system. First of all, there is a max score for a reason, as the max score of 1,000,000 gives you a good idea of how far you are from perfection. Second of all, mods aren't worth any points right now, meaning that the implementation of mods into the rank system would lead to a HUGE fluctuation in some players's rank. Ignoring the mode specific mods such as SpinOut or FadeIn, the current mods still don't fit into the mania gameplay at all. Mania has the most flexible mode, with the greatest amount of variations, and there is simply NO need in implementing mods to spice up the ranking system.

Specific reasons why each mod is pointless:

HardRock
First part of HR: There is no point in decreasing the hitbox of the notes, as unlike the other 3 modes which are either 300/100/0, mania offers so many different possible scores (MAX, 300, 200, 100, 50, 0). HR is a thing in other modes because the max score is easily attainable, yet for most mania maps, 1,000,000 is quite a hard task.
Second part of HR: What are you going to do, increase the scroll speed? The scroll speed has 0 influence on the score and the game, as everyone can change the speed to something that they feel comfortable with. If HR increased scroll speed, we could simply slow the scroll down manually.

FadeIn/Hidden
FadeIn/Hidden would have been a decent mod for ranking, as unlike HR, players cannot manipulate the visual strain it adds. However, the idea that the hidden part increases after each successive note (until around 70%) is simply retarded (I really can't find another word), as the variation in how much of the screen is hidden makes the mod easily manipulable.

DoubleTime/NightCore
DESTROYS THE SONG. Jokes aside, a big part of DT is about the increase in visual strain, as the notes speed increase as well. Same argument for HR. DT is still barely viable just because of how you have to click the notes quicker now.

FlashLight
No comment.

RebelSlayer wrote:

mods should not be factored into the ranking system.

Andrew Chow wrote:

I strongly oppose the implementation of mods into the ranking system.

Me too! ^

Besides score, Combo is very important to determine players' skilll in hard chart. Although the combo system in o!m is fluctuated (Ex. Those who get #1 got less combo than others), some patern allows only high skilled players to FC, and the score system does little impact on that.

See this shit? Niclijou is this high when he is always using these key mods! This was today, and the whole time he had 4k on. This is why the key mods should be unranked.

HR is unranked, yet it only changes the accuracy for notes.
RD is unranked, but even though it changes the note placements, it still will let you end with the same combo number of the song.
And yet, Key Mods are ranked, even though they change the whole songs track and combo number!

Some people are saying "But some good songs are low key counts." YES I KNOW! I play those songs too and not always 7k! I just want the mod unranked, not the lower key songs!

Take this into consideration people, it's totally unfair for all of us skilled players.

October Scream wrote:

See this shit? Niclijou is this high when he is always using these key mods! This was today, and the whole time he had 4k on. This is why the key mods should be unranked.

HR is unranked, yet it only changes the accuracy for notes.
RD is unranked, but even though it changes the note placements, it still will let you end with the same combo number of the song.
And yet, Key Mods are ranked, even though they change the whole songs track and combo number!

Some people are saying "But some good songs are low key counts." YES I KNOW! I play those songs too and not always 7k! I just want the mod unranked, not the lower key songs!

Take this into consideration people, it's totally unfair for all of us skilled players.

I didn't even know key mods were ranked. What the hell?! That is a load of crap. That shouldn't be ranked at all.

Well keymods don't work on mania specific maps. So it won't be a problem if..... on second thought, let's not bring that up right now.

Also I completely disagree with who ever said combo is important.This isn't standard, a play with high acc and 2 misses, one at 1/3 and 2/3, is much more impressive then someone who missed 5 notes at the end of the song. Miss count is pretty much included in accuracy, making combo the most useless and broken score calculating utility.

Also, why is everyone talking about mods? No one ever said that they'd be considered in ppv2, this is greatly off topic. We should be contributing on what makes a map hard and what makes a map easy, with possible example, to help tom make his algorithm.

October Scream wrote:

See this shit? Niclijou is this high when he is always using these key mods! This was today, and the whole time he had 4k on. This is why the key mods should be unranked.

HR is unranked, yet it only changes the accuracy for notes.
RD is unranked, but even though it changes the note placements, it still will let you end with the same combo number of the song.
And yet, Key Mods are ranked, even though they change the whole songs track and combo number!

Some people are saying "But some good songs are low key counts." YES I KNOW! I play those songs too and not always 7k! I just want the mod unranked, not the lower key songs!

Take this into consideration people, it's totally unfair for all of us skilled players.

That rank is from the old & broken PP system, from farming easy maps that are low key by default - key mods have nothing to do with this.

If you're seriously complaining about someone using a mod combination in multi for fun that gives them a maximum score cap of 200,000 then wut.

Drace wrote:

We should be contributing on what makes a map hard and what makes a map easy, with possible example, to help tom make his algorithm.

This!

Imo, this topic should get cleaned until Tom can get his first algorithm out so we can give proper feedback. At the moment without a base, all we can really do is indicate which patterns are tricky, rather than actually help improve it

@Drace, I'm specifically talking about key mods, not trying toget other mods up.

Also, I bring up combo because it's a simpler amount of notes reference.

@Cozzzy, If the old system were the case, wouldn't those players still be low? Wouldn't we all be back to the ppv1 placements? Also, EZ and NF don't affect pp given (Same pp reduction is applied), I was trying to point out key mods -_-