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

Accuracy calculation

posted
Total Posts
13
casmith789
45 posts
Joined June 2013
Topic Starter
casmith789
I read on the wiki that your accuracy is the average of all the top ~600 ranks you have. "pp's accuracy is dependent on top 600~ submitted online score's accuracy." This doesn't seem to be the case. For instance, I can set a 100% on any easy map (as I just did on S.A.T.O.R.A.D.E. Easy +DT +HR #349) and it doesn't get counted.

So, how is it counted?

Presumably it's the average accuracy of all maps you have gained at least 1pp? or some other arbitrary cutoff? The average accuracy of your top 100 maps in pp?

If it is all the maps I have got pp on, how do I check? I've got a spreadsheet of maps which I know gave me pp in the past, and pp now, but it doesn't match with my accuracy changes from other maps. Likewise, I went SSing tons of maps recently, and some surprised me in counting towards my accuracy score, even though they were no mod SS ranks of 2.5 star maps. Most didn't, of course.

I know I can improve accuracy by playing maps I know have bad accuracy that give me pp, like my A ranks in my top performance list. I'm just interested in the underlying formula.
silmarilen
5,579 posts
Joined February 2010
silmarilen
the acc difference between before and after you played it just wasnt big enough to show
casmith789
45 posts
Joined June 2013
Topic Starter
casmith789
Pretty sure that's not the case. For instance, after SSing a new map that did count, I get a gain of more than +0.01%. It's around +0.10%, actually.

If I drop accuracy on a map I know is in my PP list (my top 10), say if I improved a +HD +HR to a +HD +DT with lower accuracy, I drop some accuracy, of around -0.03% (for a recent map).

So, I know I have less than the number of scores for that to happen, and as such I know it isn't counting these scores.
RaneFire
1,181 posts
Joined May 2011
RaneFire
Yes it weights accuracy similar to pp, I'm not sure if it is linked directly though, but generally doing "well" (according to pp) and "badly" (acc) at the same time will net you the most accuracy drop. Do "well" and "well" and get most accuracy gain.
casmith789
45 posts
Joined June 2013
Topic Starter
casmith789
Not sure it does weight accuracy similarly to PP.

I don't have concrete evidence, but I have a bit.

I worked out how many songs would be required if each counted equally towards accuracy when adding a new, low PP song to my list

This was around 40-90 (error bars of +/- 0.01)

Similarly, I did the same for a song that was already in my list, and modified accuracy

This was around 50-90 (error bars of +/- 0.01)

I haven't done extensive testing but I am inclined to think every song that counts counts the same. I could be wrong, though.

EDIT: Conclusions: I would imagine that if it counted the same way as PP, the second song would count way more than 10x as much towards PP as the first, whereas if this were the case for accuracy too you would expect to see around 500 songs required if each counted equally towards accuracy. Thus I conclude that it is more likely that they do count equally towards accuracy.
RaneFire
1,181 posts
Joined May 2011
RaneFire
Problem with testing and "low pp" songs is that pp is relative. When I said similar, the weightings would not be according to pp's calculations of course, they would be mostly linear.

I wouldn't waste my time testing though... I'm fine not knowing.

Just play.
jesse1412
1,457 posts
Joined January 2010
jesse1412
Accuracy is bullshit anyway. You can be the worst player in the world and have better accuracy than cookiezi, lewa, mugio etc.
casmith789
45 posts
Joined June 2013
Topic Starter
casmith789
That makes sense, it could be using the raw unweighted pp values.

The reason I ask is because I find a game much more fun if I incorporate these kind of things. It helps me to enjoy the game, and measure levels of improvement, etc.

I don't have a problem with motivation for playing, as you can see if you check how much I have played recently, and I find this kind of mathematics fun. I don't consider this kind of thing timewasting, more a goal in itself.
uzzi
786 posts
Joined September 2012
uzzi

casmith789 wrote:

That makes sense, it could be using the raw unweighted pp values.

The reason I ask is because I find a game much more fun if I incorporate these kind of things. It helps me to enjoy the game, and measure levels of improvement, etc.

I don't have a problem with motivation for playing, as you can see if you check how much I have played recently, and I find this kind of mathematics fun. I don't consider this kind of thing timewasting, more a goal in itself.
Even if it may not be accurate measurements, I think the same way as you do. Best way I've seen to improve accuracy is just to fix up your top ranks, and finish songs that get you a decent pp value so it'll effect your accuracy.
silmarilen
5,579 posts
Joined February 2010
silmarilen

casmith789 wrote:

Not sure it does weight accuracy similarly to PP.
yes it does

casmith789 wrote:

I don't have concrete evidence, but I have a bit.
peppy said so himself

casmith789 wrote:

I worked out how many songs would be required if each counted equally towards accuracy when adding a new, low PP song to my list

This was around 40-90 (error bars of +/- 0.01)

Similarly, I did the same for a song that was already in my list, and modified accuracy

This was around 50-90 (error bars of +/- 0.01)
yeah nice and all, but it doesnt work like that.

casmith789 wrote:

I haven't done extensive testing but I am inclined to think every song that counts counts the same. I could be wrong, though.
you're wrong, as said above

casmith789 wrote:

EDIT: Conclusions: I would imagine that if it counted the same way as PP, the second song would count way more than 10x as much towards PP as the first, whereas if this were the case for accuracy too you would expect to see around 500 songs required if each counted equally towards accuracy. Thus I conclude that it is more likely that they do count equally towards accuracy.
what?
casmith789
45 posts
Joined June 2013
Topic Starter
casmith789

silmarilen wrote:

casmith789 wrote:

Not sure it does weight accuracy similarly to PP.
yes it does

casmith789 wrote:

I don't have concrete evidence, but I have a bit.
peppy said so himself
I'd be interested to read that, if you know where the thread is?
silmarilen
5,579 posts
Joined February 2010
silmarilen
t/97850
t/92503
casmith789
45 posts
Joined June 2013
Topic Starter
casmith789

silmarilen wrote:

https://osu.ppy.sh/forum/t/97850
t/92503
Thankyou :)

Certainly makes it interesting.

I was wondering about using accuracy of a measure of determining what pp my songs are worth, as it is based on pp this makes it more likely this could work.

Of course, this is unlikely to give exact results but I'll have a play around with the numbers and see what they say.
Please sign in to reply.

New reply

0
1 / 13