Guest
home
beatmaps
rankings
community
store
help
forum
  1. Forums
  2. osu!
  3. Gameplay & Rankings

How should score work?

posted
Total Posts
48
GhostFrog
416 posts
Joined March 2013
Topic Starter
GhostFrog
It's generally agreed that it's too late to make any changes to osu!'s scoring system, but this seems like an interesting exercise anyway. If you could go back in time to the creation of osu! and redesign its scoring system given what you know now, what would it look like? You don't have to limit yourself to something as simple (or as complicated) as the current scoring system, but using something like tom's star ratings for beatmaps in order to determine per-map mod bonuses is probably going too far to be reasonable.

If you don't feel like doing that, just use this thread to complain about the current scoring system or something.
crud41
33 posts
Joined May 2014
crud41
#1 roundabout post on forums.
uzzi
786 posts
Joined September 2012
uzzi
I feel like score should revolve more around accuracy than combo.
Xayne
240 posts
Joined August 2013
Xayne

- [ U z z I ] - wrote:

I feel like score should revolve more around accuracy than combo.
That's more of an issue with the PP system than the actual score. It's pretty common for combo to be the overwhelming factor in rhythm games in terms of score because that's what a multiplier does. I do agree that accuracy needs a greater impact on PP gain, though. I feel like it takes a greater understanding of a map to have 98% accuracy with a break in the middle than it does to have a full combo with 95% accuracy.
Insyni
281 posts
Joined July 2012
Insyni

Xayne wrote:

- [ U z z I ] - wrote:

I feel like score should revolve more around accuracy than combo.
That's more of an issue with the PP system than the actual score. It's pretty common for combo to be the overwhelming factor in rhythm games in terms of score because that's what a multiplier does. I do agree that accuracy needs a greater impact on PP gain, though. I feel like it takes a greater understanding of a map to have 98% accuracy with a break in the middle than it does to have a full combo with 95% accuracy.
Though you also have to take in consideration why the break was made.

1. Was the break simply due to a poor movement and they could normally pass that part with perfect accuracy? In this case, I agree accuracy is more important.
2. Was the break due to the inability to complete a new pattern without spamming? If they can't do it without spamming that one tiny section, then combo is more important.
winber1
4,520 posts
Joined February 2010
winber1
combo should stop giving extra points after 420
RaneFire
1,181 posts
Joined May 2011
RaneFire
Without the foreknowledge from today, probably just something like this:

winber1 wrote:

combo should stop giving extra points after 420
I would also redesign the Easy and Hard Rock mods to affect higher map settings in a less drastic way.
E m i
E m i
More based on accuracy. With a curve that makes the score more or less skyrocket around 98-99+%.
Less based on the MOMENT you got a 300/50/miss - possibly calculated after setting the score.
Less based on combo.
In the end, directly or indirectly based on PP.

RaneFire wrote:

I would also redesign the Easy and Hard Rock mods to affect higher map settings in a less drastic way.
This. There should be more options to HR than ar10. ar9.8 is a coincidence, and I'm not counting decimal approach rates.
theLiminator
68 posts
Joined October 2011
theLiminator
But decimal approach rates have a massive difference on the playability. There are tons of people that can play ar10, but only a tiny fraction of people can really play ar10.3.
E m i
E m i

theLiminator wrote:

But decimal approach rates have a massive difference on the playability. There are tons of people that can play ar10, but only a tiny fraction of people can really play ar10.3.
I'm talking about HR since it never goes above 10, and it stays at 10 from 7.2 to 10.
DT is pretty... Balanced? It keeps the same density, at least.
DahplA
1,099 posts
Joined March 2014
DahplA

- [ U z z I ] - wrote:

I feel like score should revolve more around accuracy than combo.
This. I hate it when I see HD+DT 100% get beaten by HD+DT+HR 95%.
Bauxe
1,624 posts
Joined August 2012
Bauxe

DahplA wrote:

- [ U z z I ] - wrote:

I feel like score should revolve more around accuracy than combo.
This. I hate it when I see HD+DT 100% get beaten by HD+DT+HR 95%.
You don't understand the difficulty in AR11 and OD11 clearly.
Xayne
240 posts
Joined August 2013
Xayne

Insyni wrote:

2. Was the break due to the inability to complete a new pattern without spamming? If they can't do it without spamming that one tiny section, then combo is more important.
If you're mindlessly spamming, I highly doubt you get 98% accuracy anyway.
E m i
E m i
b-but acc is overrated :oops:
7djames7
17 posts
Joined February 2013
7djames7
Just a quick question. Why is it to late to change the scoring system? Just wipe the online stats and leaderboards clean.

I also agree that it should be less based off of combo and more on accuracy. Not that hard to just plug the current combo length into a logarithmic function so that as the combo gets higher and higher you get more diminishing returns.

Second maybe make the scoring for accuracy scale smoothly from "perfect accuracy' to just enough. ie rather than it jump from 300>100>50 you might get a 295 because you were just barely off or a 1 for really horrible accuracy. This would require a slightly higher accuracy to achieve a 300 and would severely punish people for spamming.

Third maybe give mods a bigger multiplier than the measly 6-12% They definitely make the game a lot harder most of the time.

One last thing not really related to scoring. I would make it so that sliders don't rape your HP for even the slightest mistake.
E m i
E m i

7djames7 wrote:

Just a quick question. Why is it to late to change the scoring system? Just wipe the online stats and leaderboards clean.

I also agree that it should be less based off of combo and more on accuracy. Not that hard to just plug the current combo length into a logarithmic function so that as the combo gets higher and higher you get more diminishing returns.

Second maybe make the scoring for accuracy scale smoothly from "perfect accuracy' to just enough. ie rather than it jump from 300>100>50 you might get a 295 because you were just barely off or a 1 for really horrible accuracy. This would require a slightly higher accuracy to achieve a 300 and would severely punish people for spamming.

Third maybe give mods a bigger multiplier than the measly 6-12% They definitely make the game a lot harder most of the time.

One last thing not really related to scoring. I would make it so that sliders don't rape your HP for even the slightest mistake.
It is not late, but it will come at a price. A lot of scores will be lost, and if this decision will be made, they must change and fix as many things as possible, at once.
Your idea on acc is pretty good, the current acc system is based more on consistency.
The mods are planned to be reworked too, so the bonus score reflects pp more.
And yeah, sliders really do rape HP. >:( - Disconnected Hardkore
Saphirshroom
627 posts
Joined May 2013
Saphirshroom

7djames7 wrote:

Just a quick question. Why is it to late to change the scoring system? Just wipe the online stats and leaderboards clean.

I also agree that it should be less based off of combo and more on accuracy. Not that hard to just plug the current combo length into a logarithmic function so that as the combo gets higher and higher you get more diminishing returns.

Second maybe make the scoring for accuracy scale smoothly from "perfect accuracy' to just enough. ie rather than it jump from 300>100>50 you might get a 295 because you were just barely off or a 1 for really horrible accuracy. This would require a slightly higher accuracy to achieve a 300 and would severely punish people for spamming.

Third maybe give mods a bigger multiplier than the measly 6-12% They definitely make the game a lot harder most of the time.

One last thing not really related to scoring. I would make it so that sliders don't rape your HP for even the slightest mistake.
I agree on most of the parts, except why would you put combo length into a logarithmic function when getting to a certain combo gets harder exponentially? That just makes no sense.
Usually you lose combo because you're unable to do a pattern correctly, simply missing happens not very frequently - this at least applies to me. Combo shouldn't have as much of an impact but the exponential function is definitely justified.

On a side note, your general idea on accuracy is good but it comes to the point where it doesn't have anything to do with rhythm anymore. If you get punished for tiny fractions of a second you might as well blame it on your keyboard's input lag instead of on yourself. OD 10 already punishes you for small mistakes and songs have 1000+ notes, so taking the average accuracy out of all of those is already a very good representation of your actual accuracy.
7djames7
17 posts
Joined February 2013
7djames7

Saphirshroom wrote:

I agree on most of the parts, except why would you put combo length into a logarithmic function when getting to a certain combo gets harder exponentially? That just makes no sense.
Usually you lose combo because you're unable to do a pattern correctly, simply missing happens not very frequently - this at least applies to me. Combo shouldn't have as much of an impact but the exponential function is definitely justified.

On a side note, your general idea on accuracy is good but it comes to the point where it doesn't have anything to do with rhythm anymore. If you get punished for tiny fractions of a second you might as well blame it on your keyboard's input lag instead of on yourself. OD 10 already punishes you for small mistakes and songs have 1000+ notes, so taking the average accuracy out of all of those is already a very good representation of your actual accuracy.
The reason for the logarithmic function is so that for extremely long maps FCing the map becomes less and less rewarding with more of a focus being put on accuracy. If you add a offset based on the length of the map you actually wont notice it for shorter maps. It basically puts more emphasis on being consistent through the entire song rather than getting lucky and happening to get an absurdly long combo. For people who can consistently get long combos they will still be able to edge out others on the leaderboard but you also won't see people ranking higher than they should be as often. Also PP farming is still a thing and this would probably cut down on it.

As for the accuracy change most people would actually benefit from it as you would be getting 250-200 points for being slightly off rather than an immediate 100. It would only punish newer players and people that spam through entire sections.

You still get punished for small errors with the system as it currently is. So I don't know what your talking about there. This does however make the scores at the top of the leader board less based on who spun that one spinner the quickest and more who messed up less and by how much.

Edit: Rhythm and accuracy go hand in hand so I don't know how my system discourages good rhythm. Maybe for people that can't sight read? Or maybe for people without a formal education in music? I don't know.
TL;DR : All in the name of competitive gameplay. :D
Saphirshroom
627 posts
Joined May 2013
Saphirshroom

7djames7 wrote:

It basically puts more emphasis on being consistent through the entire song rather than getting lucky and happening to get an absurdly long combo.
Getting an "absurdly long combo" is not possible through sheer luck. And if you want to measure consistency throughout the entire song then just disable the combo feature entirely and measure the ratio of notes missed to the total amount of notes. I still don't know why it would make sense to feed combo length into a logarithmic function. This punishes players who can get a high combo.

7djames7 wrote:

You still get punished for small errors with the system as it currently is. So I don't know what your talking about there. This does however make the scores at the top of the leader board less based on who spun that one spinner the quickest and more who messed up less and by how much.
Leader board is irrelevant. Of course it looks cool to be #1 but you don't gain anything other than that. And I'm talking about the fact that small errors already get punished enough, like you said. With your system, even smaller errors would be punished.

Edit: Rhythm and accuracy go hand in hand so I don't know how my system discourages good rhythm. Maybe for people that can't sight read? Or maybe for people without a formal education in music? I don't know.
That's not what I meant. It's more the fact that with such tiny hit errors blaming the equipment becomes a thing. Notice that the hit window for getting a 300 on OD 10 is 18ms. If we, say, make score gain more subtle, for example divide it into steps of 5, realistically we get a window of like 1ms that makes the difference between getting a 300 and a 295. That makes it nigh impossible to get a 300, you could even say it was luck if someone gets a 300. Of course you could adjust the roughness of those values but I don't see the benefit of that compared to how it is now.
Right now you have a certain error margin within which you get a perfect score. The systems looks at your accuracy throughout the entire song and takes the average, there's no need for a finer scale. Like I said, your actual accuracy gets reflected very well by the system. Making in-between score values just adds fake-depth that is not actually relevant for the game.
nrl
1,402 posts
Joined October 2011
nrl

7djames7 wrote:

Just wipe the online stats and leaderboards clean.
You realize this literally means throwing out every single score set in the past 7+ years, right? I get that the very roots of the scoring system would have to be changed to get it on the right track, but let's not throw around the idea of a total wipe so lightly.

7djames7 wrote:

It basically puts more emphasis on being consistent through the entire song rather than getting lucky and happening to get an absurdly long combo.
This statement is completely backwards. Long combos don't happen by chance.

7djames7 wrote:

This does however make the scores at the top of the leader board less based on who spun that one spinner the quickest and more who messed up less and by how much.
This is also totally incorrect. Currently you need an FC just to get on the leaderboard, and after that the deciding factor is accuracy. On really close leaderboards it also matters where your 100s were, but on tougher maps simply having higher accuracy is enough.

Contrary to your opinion, this system actually does put the most consistent players near the top. The two sources of error are the fact that each note is worth more as the map goes on and the totally arbitrary weightings applied by mods. If those two things were fixed the leaderboards would almost perfectly reflect pp distribution.

Saphirshroom wrote:

Right now you have a certain error margin within which you get a perfect score.
The easy solution is to retain that margin and implement scaling outside it. That is to say, you have an 18ms window for a 300, but starting at 19ms the point value will begin to scale smoothly downwards. This margin would have to be smaller than it is now (maybe cut it in half?), and it could be controlled by OD. There's really no downside to this, it would introduce more variation in scores, helping rank spread on heavily saturated leaderboards, and it would essentially introduce OD10+, which is something a lot of players have been asking for.
7djames7
17 posts
Joined February 2013
7djames7

Narrill wrote:

Saphirshroom wrote:

Right now you have a certain error margin within which you get a perfect score.
The easy solution is to retain that margin and implement scaling outside it. That is to say, you have an 18ms window for a 300, but starting at 19ms the point value will begin to scale smoothly downwards. This margin would have to be smaller than it is now (maybe cut it in half?), and it could be controlled by OD. There's really no downside to this, it would introduce more variation in scores, helping rank spread on heavily saturated leaderboards, and it would essentially introduce OD10+, which is something a lot of players have been asking for.
This is exactly what I was trying to say.

Narrill wrote:

You realize this literally means throwing out every single score set in the past 7+ years, right? I get that the very roots of the scoring system would have to be changed to get it on the right track, but let's not throw around the idea of a total wipe so lightly.
If the people are as consistent as you seem to think they are the leaderboard should easily sort itself out after a month or two. The only problem with wiping the stats clean is that if done too frequently the leaderboard becomes kinda moot. If anything a wipe every 2 years or so would be good since the PP system still isn't farm proof(its only harder to farm).

Narrill wrote:

This statement is completely backwards. Long combos don't happen by chance.
It was merely a suggestion. Definitely worth a shot to try out. And yes sometimes it is kinda up to luck. Especially if you have trouble jumping or aren't the most consistent person with a mouse. I have scores from a year or two ago that I still can't beat without retrying the map over and over. My first 100% was almost entirely luck. As with any skill based game luck can play a considerable part some times. Even if it isn't all up to a dice roll or something similar.

Narrill wrote:

This is also totally incorrect. Currently you need an FC just to get on the leaderboard, and after that the deciding factor is accuracy. On really close leaderboards it also matters where your 100s were, but on tougher maps simply having higher accuracy is enough.

Contrary to your opinion, this system actually does put the most consistent players near the top. The two sources of error are the fact that each note is worth more as the map goes on and the totally arbitrary weightings applied by mods. If those two things were fixed the leaderboards would almost perfectly reflect pp distribution.
I understand this. My point was that they aren't that consistent. Getting some of those scores takes a considerable number of retries even if they already know the beatmap. However I guess that is one of the flaws with having a leaderboard.
Saphirshroom
627 posts
Joined May 2013
Saphirshroom

Narrill wrote:

The easy solution is to retain that margin and implement scaling outside it. That is to say, you have an 18ms window for a 300, but starting at 19ms the point value will begin to scale smoothly downwards. This margin would have to be smaller than it is now (maybe cut it in half?), and it could be controlled by OD. There's really no downside to this, it would introduce more variation in scores, helping rank spread on heavily saturated leaderboards, and it would essentially introduce OD10+, which is something a lot of players have been asking for.
I agree on that completely. There is one small downside, actually: Skinning will get slightly less flexible. You'd have to introduce fonts in every skin so you can portray the numbers according to your score. There are some skins now which use coloured dots as a representation of hit-accuracy which wouldn't be possible anymore. We (read: peppy) could just introduce a toggle to switch between these representations, of course.
Also,

7djames7 wrote:

This is exactly what I was trying to say.
Never mind my argument against your system then. When you said you'd be getting 295 for just a tiny bit off, I thought you meant a tiny bit off the perfect hit, not a tiny bit off from getting a 300. Sorry about that.
DahplA
1,099 posts
Joined March 2014
DahplA

Bauxe wrote:

You don't understand the difficulty in AR11 and OD11 clearly.
I don't really care. It's going to be them with crap accuracy on their profiles.
George Lopez
53 posts
Joined May 2013
George Lopez
I just simply wish combo wasn't as important as it is :( Maybe im just bad at this game, but i wish that accuracy would be valued more than it currently is.
SomeLoli
SomeLoli

DahplA wrote:

Bauxe wrote:

You don't understand the difficulty in AR11 and OD11 clearly.
I don't really care. It's going to be them with crap accuracy on their profiles.
Shit just got real. Bad accuracy on profile? omg!!!
xGx
173 posts
Joined December 2013
xGx
　
Last edited by xGx , edited 2 times in total.
eeezzzeee
979 posts
Joined August 2013
eeezzzeee
Make HD worth less than HR
nrl
1,402 posts
Joined October 2011
nrl

7djames7 wrote:

If the people are as consistent as you seem to think they are the leaderboard should easily sort itself out after a month or two.
It has nothing to do with overall rankings. You'd be wiping thousands of legendary scores, many of which were set by players that no longer play.

7djames7 wrote:

And yes sometimes it is kinda up to luck.
Sometimes people have particularly good plays, sure. That's not luck, and claiming that it is reveals a fundamentally misguided understanding of skill application. Monster plays come from the combination of a proper mental state, physical skill, and unwavering concentration, and while the human element introduces some measure of unpredictability, it does an injustice to the player to call it luck.

7djames7 wrote:

Getting some of those scores takes a considerable number of retries even if they already know the beatmap. However I guess that is one of the flaws with having a leaderboard.
Variation introduced by the human element increases as the demands increase relative to skill, that's unavoidable. Honestly, I don't even know why a scoring system that attempts to eliminate that variation is even desirable. 99% of maps can be and are FC'd within hours of being qualified.
show more
Please sign in to reply.

New reply

0
1 / 48