osu!mania ScoreV2 live!

posted
Total Posts
476
show more
Loily

smoogipooo wrote:

Hi all,

You may have noticed the ScoreV2 changes in the changelogs recently, with just over 7 weeks left until MWC begins we've released ScoreV2 for osu!mania in hopes that we can perfect the score system before the tournament. You will need to be on the Cutting Edge release stream to use this for now, but we will propagate it to all release streams (excluding fallback) when it is ready, just before MWC.

You'll be please to know that there are no more hidden multipliers and rounding issues have been eradicated, but that is not all. Let's go through a list of changes in this initial version:

THIS IS NOT FINAL
Please, do not discuss Star Rating and PP here.
  1. Score is made up of 20% combo and 80% accuracy.
    1. We want to value the more accurate players (accuracy) whilst applying a small reward for consistency (combo).
  2. LN starts and ends are now judged separately.
    1. Previously LNs considered a joint timing distribution between the start press and end release. This made it unclear as to whether you'd get a MAX after an LN end as you had to take into consideration the LN start. Judging separately should feel more natural, rewarding (as you get instant feedback), and a bit more challenging.
  3. LN ends are given a 1.5x lenience to the hit windows.
    1. LN starts were previously given up to 1.2x timing window lenience and LN ends were given up to 2.4x timing window lenience. This reduces the complexity of releasing an LN whilst you're focusing on pressing other notes.
  4. If an LN is broken but re-pressed, the LN end will not award more than 50 points.
    1. Works similar to the current system depending on when you release the hold, but is lenient enough to feel rewarding even for newer players (consider that ScoreV2 will be used as the normal ranking in the future).
  5. LNs do not give combo ticks any more - only one combo tick for the start and end notes.
    1. Feels more natural rather than displaying a useless number.
  6. Mods are back! NF/EZ/HT give 0.5x score multipliers and DT/HR/HD/FI/FL give 1.06x score multipliers.


We've had some internal discussions about how LNs should work, but have not reached a definitive conclusion as there are split opinions. We are eager to hear your feedback regarding osu!mania scoring and this new scoring system!

I'll be adding here a list of changes I will consider. Please remember that we are fully intending to break the game with these changes. We will apply any changes necessary to make things work:
  1. Make DT adjust to 100%/110%/.../150% with score bonus increments of 0.05x (or something like that).
  2. Increase the bonus of HR or decrease the tightness of the timing windows.


---- Changelog:

2016-06-16:
Cutting Edge has been updated with changes to ScoreV2 that were proposed by Shoegazer here. I want to stress that the changes are not final and we are still tweaking the system to properly represent a player's skill in a competitive setting.

Please note that HD/FI/FL mod multipliers have not yet been removed. These are slated to be removed in the next iteration of changes.

Edit: Posting this because I've explained it on reddit:

There are two components to the score.
- Accuracy
Essentially as accuracy increases we want you to gain more and more score while accounting for the difficulty of maintaining a 99%+ accuracy over 90%. To do this accuracy is exponentiated so that it is not quite a linear multiplier. In the previous iteration it was raised to the 10th power, in the new iteration it is raised to a factor of the accuracy.
This has the effect of causing lower accuracies to not be so much of dead weight as they were previously, while still providing a steep curve towards 100% accuracy as seen in this graph (red = old, blue = new): https://u-gi.me/sykzM

- Combo
Combo is the harder one to talk about. We want to award holding combo, but at the same time not punish holding 4000 combo and missing once too much. To achieve this your individual hit scores are weighted by the combo you have after hitting the note. In the previous iteration this was a linear relationship, which resulted in punishing for missing after holding 4000 combo. In the new iteration it is logarithmic, with a cap at log_4(400) (meaning combo > 400 will be weighed as if your combo was 400), as shown in this graph (red = old, blue = new): https://u-gi.me/oJ6sa
Adri
Don't tell me that 1M SS is reasonable and will be implemented in the final version ... This is the only thing in ScoreV2 that I find not only really bad, but also game breaking (and not in a really good way !!).

You can still say a score is an SS when it's not 100% accuracy if you say that a 300 is only 98%, the same way that in standard you can get a 99,5% A and a 96% S ...

I see it like that :
D: 0 - 69,99%
C: 70% - 79,99%
B: 80% - 89,99%
A: 90% - 94,99%
S: Equal to or Above 95%, only if there are not only 300 and MAX (95,00% - 99,99% range)
SS: There are only 300 and MAX (98,00% - 100% range, more usually around >99,5%) - Depending on OD, all hits are under 30 to 50ms, exactly like in standard mode (which I find fair)

With current Scorev2, SS requires every hit to be under 16,5ms not depending on OD (meaning even new players require insane accuracy to achieve SS). This is harder than Taiko's OD10 (19ms).

Aside from this I really like the whole Gaussian thing, but I think that this would only work for a totally different game. I believe that the four main modes should remain as close as possible to make them more accessible and keep some integrity (that is also why I think that a 1M cap is a bad idea compared to what is in place with Taiko and CtB for example ; Std too but combo is too broken).
If this would be implemented tho, I would be far from complaining (just sad to drift away from fellow gamemodes :D).
Ditroon

FrenzyLi wrote:

Out of curiosity, is it possible to post multiple score of different mod combinations on the same beatmap?

e.g. AIAE MX nomod pass (counts as pp), AIAE MX DT pass (counts as pp) ....

and pls nerf HT pp . w .
Kiranto -
aaa
Fifkee
I'm really scared.
Adri
Has anyone listened to my proposition ever since ? :x
abraker

Adri wrote:

Aside from this I really like the whole Gaussian thing, but I think that this would only work for a totally different game. I believe that the four main modes should remain as close as possible to make them more accessible and keep some integrity
Apply Gaussian to all gamemodes. I don't see the issue. If you argue it will screw up scores, there are plenty of changes the went into effect before that made some scores impossible to beat on older maps already.
Cuber
I'm still of the opinion that the best scoring system would be one that takes into account the instantaneous difficulty of a specific note and the (gaussian) accuracy when finding the total score for that object. I totally agree that all the gamemodes should be similar, which is why I think this should be used for every gamemode (dunno enough about ctb to propose anything for that).

Of course that would need an accurate way to measure difficulty, but apparently, the score is a bigger problem than the star rating...
Adri
I take your side about changing the four gamemodes at once, but this would be a total meta change ! I don't know how all the players would react :/
The way star rating and judge works are bound to osu! and I'm afraid we will have to deal with it, that is why I was focusing on improving score
Kiranto -
-
abraker

Adri wrote:

I don't know how all the players would react :/
Just include it as a continuous judgement mod then and let the devs play with whatever discrete scoring method they believe is best while not bothering the rest who see continuous judgement as better
Adri
For some time I have been trying to get some information regarding what you guys thought about the score cap in osu!mania, and I finally found this reddit post where smoogipooo tells the scorecap should stay in place (those comments are 7month old so I don't know if they are still valid, but i'd still like to say things about this).

I will only be talking about the score cap in this post

First of all, i'd like to say that a lot of players are waiting for the score cap to be removed ; I don't know if this is true but in the mind of some people this was just a temporary setting, in the waiting of a proper scoring system to not break the game. If this is, well it might be time to seek for a balance ?

What I understand is that currently, the points in mania scorev2 (and scorev1) reflects "How well you did on a 1,000,000 base". From what I know, there is already what we call accuracy ranging from 0,00 to 100,00 and that we can nearly always expect to correlate along with the ranks (D,C,B,A,S,SS).

For example, from 650k to 700k we will be expecting an A and from 800k to 850k an S. The SS will usually stand from 985k (low 300/300g ratio) to 1M (perfectly insane score). This is basic meta and anyone learns it when reaching 2 stars difficulty, if not before.

This system in itself is very good for comparing scores on a map, like you had better accuracy but a lower score so maybe you had some combo breaks etc ..
The problem is that when comparing two scores on different beatmaps, this is quite unfair. Doing 985k FC on a 30sec 0.8* map does really not have the same value as 985k FC on a 4*+ 18minutes map, and still only pp differs. Why would we call "score" something that doesn't reflect how much effort we put on something ? This is more of a "scale" ...

But anyways, I have a solution to put on the table ;
We do not need the logarithmic scoring we see on standard or ctb, this is causing problems, i totally agree. We do not want to lose what is great in the current scoring system, a.k.a comparing scores on a map, having it tied to accuracy somewhat. We just need to give a better meaning to the scores. Peppy said himself in the coffee hour that in lazer, it wouldn't be about compatibility, so why bother keeping this cap that unbalances the scores ?

So, for the score to not be an exponential, we should just keep the combo cap as it is : It is balanced and works well, the meta is in place. But instead of calculating the score on a 1,000,000 scale, why not just give a fixed amount of score for each note depending on the judge and then apply a small combo bonus like there already is?

Each difficulty would still have a fixed maximum score, accuracy would still have a lot of impact and combo wouldn't be a game breaker in tournaments. Nothing great from v1 would be removed but the meaning would be given back to the score. Doing an FC on a 0.8* would give a small amount of points, like maybe 90 000 like in other gamemodes, and the FC of the 18 minutes map would be like 75 millions maybe ? I'm not doing the math, but i'm taking it as a reference since it has one of the (or the?) highest hitobject count in mania, showing that this doesn't go out of control like in standard. After all, this can be balanced to fit what you want to see in scores.

Quick recap of my proposition in (very) short (because it doesn't seem hard to apply), without all the argumentation. I know this might have been thought of but i wanted to express myself as you recommend players with an experience to reach out to you

- Give a fixed amount of score for each note depending on the judge (LN still separated as 2 notes), instead of doing a 1,000,000 scale
- Keep the current combo cap to avoid exponential scoring

What it takes from the game :
- Farming only easy maps to get a lot of ranked score
- Frustration of doing relatively small scores even when FCing a marathon
- Incomprehension of new players "Ahah he did better than you look at the score" ; When you hard cleared a 6* and the other picked some 2* (exagerating)
- A fixed scale from 0 to 1,000,000 that indicates the quality of a score, whatever difficulty it may be

What it adds to the game :
- Rewarding high score when doing a quality score on a hard and long map
- Sense of progression for new players "My scores are starting to weight a lot more, I am doing so much better now !"
- No more fixed scale, but we will still be able to know the quality of a score judging by how other people are clearing it. 90% will roughly have the same score values, as for 95%, 97%, 85% ...
- Some consistency regarding other gamemodes as well, who don't have caps !

I might not be exhaustive and this is a subjective post, but I think that many people think like me (Or closely). If you have anything to add or say, please do it, the more we talk the better it gets ...
abraker
^ What you don't realize is that the score cap MUST be there because for long enough maps we run into this technical problem called an integer overflow. Exhibit A:



In fact this map would be unrankable, and quite sad that we can't have multi hour long ranked maps in std and ctb just because of this.
Adri
I am aware about this : I talked about it and this is why I proposed to keep the combo bonus cap as it is. If the scoring is not exponential an integer overflow would only happen on maps lasting hours at consistent speed. The only problem is the combo bonus going nuts ^^
Adri
I made this spreadsheet showing basically how I would see this system without cap, that restrains to be lower than 2bil points : https://docs.google.com/spreadsheets/d/ ... sp=sharing

I set the values to something that seems logical to me, but it can be changed by anyone who wants to try some different balances.

on the bottom right i set the maximum scores for some amounts of hitcircles. Keep in mind that the largest map has 9250 hitcircles and that something like 65000 hitobjects crashes the game currently. So if the value for 65000 hitcircles done perfectly is still below the integer value, we should be fine for a looong run ...
Bobbias
Uncapped maximum score still doesn't allow for useful comparisons between songs. In standard If someone gets an SS on something easy and gets 3m points, and someone plays a much harder song and gets a low B with 3m points, you can't compare those scores in any useful manner.

The whole idea of comparing scores between songs is silly and pointless to begin with. If there were an objective way to ACTUALLY measure difficulty between 2 songs, then it might make sense because you could use that value to help compare scores, but the current star rating system is woefully inadequate for this. And your solution of effectively setting the maximum number of points based on the number of hitobjects in a song will result in wildly different maximum scores for songs of similar overall difficulty making it useless even over a short difficulty range, let alone a large one. The number of notes in a song can vary greatly depending on the kind of patterns in a song, meaning that an unbound maximum score which only really depends on the number of notes is going to make comparing even scores on similarly difficult maps absolutely painful.

LN heavy maps will tend to have a lower overall density, and thus tend to be underrated in the current system (with exceptions when using mini LNs). BMS style patterns tend to be overrated due to higher average density even though I would argue that they are much easier to hit with a few exceptions.

Score itself is almost meaningless if you actually want to compare how well 2 people did on a map. You ignore the score and look at the actual number of judgements they got. Every rhythm game I've played has ignored the ingame score calculation when comparing actual plays because ingame scoring simply does not do a good job of simplifying a score down to a single overall value.

At least with the 1m score cap you can look at your score at a glance and have some idea of how well you did overall. The moment you introduce a variable maximum score, the ability to gauge how well you did on a song at a glance disappears. I'd say this is way more useful than trying to create a system that allows some way to "compare" scores between songs while removing what small benefits the current system has. It might suck (like every score system I've ever seen in a rhythm game), but the current system is at least useful for something.
Adri
Well I cannot disagree ! Seems right to me :)
DaddyCoolVipper
Comparing scores between songs is a very silly thing to do, yeah. The only leaderboards that matter on a particular chart are, well, that chart's leaderboards. Getting an SS on some mega-easy chart is of course a lot easier than getting an SS on a super hard chart, anyone can understand this.

With that in mind, "an uncapped score lets you compare between different charts more fairly" isn't a good argument at all. Capped scores have a lot of benefits (consistency in terms of being able to tell how well you're doing on any chart simply by looking at your score) with no arguable downsides.
sarabugs72
ha. again. someone hit me with a brick plez
Tohka-Kun
[quote="smoogipoo"]Hi all,

You may have noticed the ScoreV2 changes in the changelogs recently, with just over 7 weeks left until MWC begins we've released ScoreV2 for osu!mania in hopes that we can perfect the score system before the tournament. You will need to be on the Cutting Edge release stream to use this for now, but we will propagate it to all release streams (excluding fallback) when it is ready, just before MWC.

You'll be please to know that there are no more hidden multipliers and rounding issues have been eradicated, but that is not all. Let's go through a list of changes in this initial version:

not gud
Please sign in to reply.

New reply