forum

[Statistics] Average performance.

posted
Total Posts
26
This is a feature request. Feature requests can be voted up by supporters.
Current Priority: +1
Topic Starter
Archangel Tirael
Requirements: pp statistics for osu!mania.
I often find that many players play well in all four modes simultaneously, and some very good play in the same mode, but very bad in others. When I noticed it, I immediately got the idea: "Why not do the last performance for all modes simultaneously? Because it immediately displays the high places of the players that are playing really well in the rhythm of the game in general and not in what-or discipline. "

How to calculate the average performance? It is easier than you think! .. Probably. ;)

(osu!standart pp + taiko pp + CatchTheBeat + osu!mania pp) / 4 = your average pp!
DJKero
It would be really cool to see the people that actually fits us to have some challenge... ^_^
deadbeat
what if someone doesn't play all the modes?
theowest
Most people doesn't have time to play every single game mode professionally.
LoGo

theowest wrote:

Most people doesn't have time to play every single game mode professionally.
Kurokami
I want to be good in all game mods but I suck in Taiko (only Normals), not too good in Standard (only Hards) and also suck in Mania too. Probably more players out there like me. This request although it sounds awesome but don't think it will work this way.
Saten
I suck at every modes except osu! standard

I don't think it would work too well.
cirno-_old
Don't like the idea. If you are great all modes, good for you. Wish I had that kind of time to spare for just Osu! standard.
winber1
It shows absolutely nothing imo.

I don't get any sort of useful information.
bwross
I think people are getting the wrong idea about this... it clearly isn't interesting for everyone. If you don't play all modes to some degree, then you'd probably ignore this, just like you ignore the modes you don't play, and the stats tab and achievements for those modes on your profile. Nobody is going to fault someone with a handful of plays of CtB for having low CtB pp (unless they're doing something like bragging about their CtB skills)... and they're not going to fault them for having a poor average either, because they're clearly not a jack of all trades player. Which most people aren't, and so, yeah, most people won't do well on this stat, but that's just because it represents something they don't care about anyway... although, it might generate some interest for some people in moving outside their current mode into playing a bit of the other ones more often.

I think the top list of such a stat would feature people that are very strong in one or two modes, and have some competent (but not overly remarkable) play of the others. It's a bit like the decathlon, no competitor is actually at the top of all ten events... they all have their strengths and weaknesses, and the one that comes out on top is the one that performs perfectly on their strengths and manages to not completely bomb out on their weaknesses (where they're probably in the bottom half... but that doesn't matter so long as their time/distance gets them enough points to stay in the running).
Lno
It seems like a waste of time to add this when there are other more useful things that could be done instead. :\
bwross
I don't know... this is pretty simple and doesn't require much in the way of work at all. It's the sort of thing that I'd put off as a developer until I suddenly found myself with a tiny bit of time and wanting to do something, but not enough time to properly start on anything of those "more useful things". Because those often require work.
lolcubes

theowest wrote:

Most people doesn't have time to play every single game mode professionally.
Or at all.
Pegram
The average pp could just exclude the modes that the player doesn't play (which would of course show somewhere, as then players who play only one game mode would easily beat the ones who play all 3 or 4), or probably rather game modes that he/she has less than, say... 500 or 1000 playcount on?

And about the average pp itself; if the system works like this...

Archangel Tirael wrote:

(osu!standart pp + taiko pp + CatchTheBeat + osu!mania pp) / 4 = your average pp!
...then who cares if it's added or not? You can count that yourself in 3 seconds with a calculator, and in 10 seconds without one. Then again, with that said, it probably wouldn't cause much stress to the server, but the actual/possible average rankings would require some thinking.

BUT the main point in the OP (I suppose) was something like showing the person's average rank in the client instead of the mode he/she last played, and I doubt that would be a good idea. 1-mode pros (especially standard) would fall pretty down right away, as standard is completely different from taiko and mania with just its 1 key. I don't even know what to say about CtB - I play all the rhythm games I can, but Catch the Beat doesn't really require that much sense of rhythm AFAIK: it just helps a little.

Anyway, the game modes are all so different that this would probably be something to add as an optional choice. You could choose which game modes to include in your average rank which would show in the client. The game mode icon could be half mode A and half mode B, or if you want to include three game modes, it could be 1/3 osu! logo, 1/3 taiko icon and 1/3 fruit or whatever. Or, of course 1/4 all game mode icons. That might work out, IMO.

So I actually support this, if my points in that last paragraph are seriously taken to consideration.

EDIT: like this, perhaps? (not with a rank that good, obviously)
theowest
used to play or actively playing?

this still doesn't sound like a good idea.
Pegram

theowest wrote:

used to play or actively playing?

this still doesn't sound like a good idea.
It does.

And I don't get your question.
theowest
I've played some taiko and ctb (just trying it out), but that was ages ago. My rank for those game modes are terrible.
Pegram

theowest wrote:

I've played some taiko and ctb (just trying it out), but that was ages ago. My rank for those game modes are terrible.
Like I said in my tl;dr post, I think it might be good to let people decide which game modes are included in his/her average rank/pp.
benguin
I think the point behind this is to encourage people to try out the other modes and to actually give them more of a chance. If you saw some ugly number on your profile, wouldn't you be a bit more inclined to fix it?

The issue is the formula makes sense when you think about it intuitively (taking the average that is), but there is some little issues that can be seen if looking at the issue closely:
-Trying to compare raw pp from different game modes isn't an exact science, it's hard to compare what is "equally good" between modes. For example, I'm primarily a standard player and a play CtB casually and dab in Taiko from time to time, and at some point I had more CtB pp than standard which made my profile default to the CtB tab which annoyed me a bit since I felt like I was better at standard (but then again, who knows, the two modes are hard to compare in terms of how good one is)

-Another alternative would be to average the rankings themselves instead of raw pp since at least rankings are more objective and more comparable between modes. The only issue with this is that there is a disproportionate amount of people that play each of the different modes which creates some sort of natural "weight" on the rankings. One way to solve that would be to have "counterweight" constants to balance the "weights", like:
weighted average rank = standard + 3*ctb + 3*taiko + 5*mania (these are random numbers I chose just for the sake of example)
The issue with this though is the determining which counterweight constants are fair and what not, it's not going to be an exact science to determine one.

-Probably the most accurate way of doing this would be this idea: for each member, first add up all their pp from each of the four game modes. Then sort each member in descending order by their sum. Create a ranking based on that order. (In other words, the person with the most combined pp will be rank 1) As for showing this ranking and making it tangible/sensible, instead of having "global performance rank" above the pp graph, put something like:
---------------------------------------------------------
[current tab] performance rank: [rank]
Overall performance rank: [rank]
---------------------------------------------------------
and perhaps put all the lines for each mode and overall ranking on the graph, with the current tab mode being the boldest line, the overall ranking bold enough to be visible, and the other modes as faint lines

Anyways, I support, if this can be done well :)
Pegram
Dunnolol

Rather than talk about how it would work... why would we really need this feature? Some sort of an average rank could be nice, yeah, but would it need to show in-game to other people? IMO it could be enough if it showed just in your profile.
benguin

ReZero wrote:

Dunnolol

Rather than talk about how it would work... why would we really need this feature? Some sort of an average rank could be nice, yeah, but would it need to show in-game to other people? IMO it could be enough if it showed just in your profile.
I was mostly talking about that information being displayed on the profile. Let me make a quick mock up in paint or something of what it could be...
benguin
Here we go, something like this perhaps? And something similar for each game mode when you switch to that mode's tab?
theowest

benguin wrote:

Here we go, something like this perhaps? And something similar for each game mode when you switch to that mode's tab?
How's that better than viewing it separately?

Ranks can differ a lot between game modes and you do want detailed information about your rank.




See how I'm rank 159009? That would make my osu! graph insanely hard to view. My average rank would be around 150675, this means you should only be able to get the average performance if you are around the same rank in all game modes.
This would also leave the ability to select which game modes you want to view (since most people doesn't have around the same rank in all game modes), making it extremely hard to compare with other people. It wouldn't create a new ranking list or anything, making it impossible to implement in-game. Some people might even get the same average rank.

I'd be much easier to view the average accuracy, combo or score in the different game modes. Accuracy would obviously be excluded in ctb, score in osu!mania, leaving combo the most useful average information between all game modes.
Pegram

benguin wrote:

Here we go, something like this perhaps? And something similar for each game mode when you switch to that mode's tab?
Well damn.

Before saying anything regarding average performance, I'd like to point out that the graph actually lacks a lot of potential at the moment, and that picture made me realise that. I mean, it COULD show more than just one stat, right?
benguin

theowest wrote:

benguin wrote:

Here we go, something like this perhaps? And something similar for each game mode when you switch to that mode's tab?
How's that better than viewing it separately?

Ranks can differ a lot between game modes and you do want detailed information about your rank.




See how I'm rank 159009? That would make my osu! graph insanely hard to view. My average rank would be around 150675, this means you should only be able to get the average performance if you are around the same rank in all game modes.
This would also leave the ability to select which game modes you want to view (since most people doesn't have around the same rank in all game modes), making it extremely hard to compare with other people. It wouldn't create a new ranking list or anything, making it impossible to implement in-game. Some people might even get the same average rank.

I'd be much easier to view the average accuracy, combo or score in the different game modes. Accuracy would obviously be excluded in ctb, score in osu!mania, leaving combo the most useful average information between all game modes.
I only put the current tab's line and the average rank line because at second thought, looking at 5 lines at once it confusing (4 modes + average). With two lines, the theory it basically shows how your skill in [current tab]'s mode compares overall compared to the sum of all modes. In other words, if your [current tab]'s mode line is above the average line, then that would suggest you are more proficient in that mode compared to the others in general. If your [current tab]'s mode line is below the average line, it would suggest you don't put too much time/effort into that mode. If your [current tab]'s mode line is about at the average line, it would suggest you put an about equal effort into that mode compared with the other modes. That's the theory in my mind anyways. As for how the graph scales, I can see that being a problem. I was actually thinking of making a feature request though to allow the pp graph to be dynamic (that is with "handles" to change the viewing area) and to allow setting a default viewing area for when people visit your profile. If that can be done, that would be helpful not only to this feature, but in having a more accurate view of one's progress in general.
bwross

ReZero wrote:

The average pp could just exclude the modes that the player doesn't play (which would of course show somewhere, as then players who play only one game mode would easily beat the ones who play all 3 or 4), or probably rather game modes that he/she has less than, say... 500 or 1000 playcount on?
Nah... it should be all. If it isn't, you're essentially creating 11 new categories (once mania is in... 6 double stats, 4 triple stats, and the quadruple). If two or three of the modes aren't your game, than this probably wouldn't be either. It's that simple.

Although, you could go the judged sports route... toss out the highest and the lowest and average the remainder. That gives everyone a free mode they can ignore, but at the cost that they also lose their strongest. This works if goal is finding the most versatile players... with three or four items, it's using the median instead of the mean for the "average". The key here being that there's a set rule for a single stat and ranking... things are fragmented by allowing the players a lot of control. People will want to know how they rank for this stat, and adding 11 new ranking charts is a bit crazy.

ReZero wrote:

And about the average pp itself; if the system works like this...

Archangel Tirael wrote:

(osu!standart pp + taiko pp + CatchTheBeat + osu!mania pp) / 4 = your average pp!
...then who cares if it's added or not? You can count that yourself in 3 seconds with a calculator, and in 10 seconds without one. Then again, with that said, it probably wouldn't cause much stress to the server, but the actual/possible average rankings would require some thinking.
Yes, it's easy to work out one's own average. But you'd need to spider all the profiles if you wanted to find the best player, or where you place. Which is a lot more work and stresses the server with a lot of requests... so it's best if it's done server side in the database.

ReZero wrote:

BUT the main point in the OP (I suppose) was something like showing the person's average rank in the client instead of the mode he/she last played, and I doubt that would be a good idea. 1-mode pros (especially standard) would fall pretty down right away, as standard is completely different from taiko and mania with just its 1 key. I don't even know what to say about CtB - I play all the rhythm games I can, but Catch the Beat doesn't really require that much sense of rhythm AFAIK: it just helps a little.
I didn't read it as having to do with the client at all. It's fine just listed on the profile for the people that care about seeing it, with a ranking chart so those of us who do play all the modes can find out who our idol is at any time. The way things are currently displayed in the client is fine.


And I think that if things are going to be charted together, then they should probably be charted by raw pp... because that allows for a nice consistent 0-7000 scale. Which probably means a separate chart with the four modes + average plotted together.

And I've also though about the fact that maybe the average should be weighted (probably by analyzing the distribution of pp scores for each mode and working out how to shape them to the same curve). And then I thought, nah, it's just a fun number... straight average is probably good enough and it's easy for people to work out by and understand what's going on (unlike pp itself, which confounds people). But maybe people will take it seriously, and an appropriate curve should be calculated (and perhaps recalculated periodically).
Please sign in to reply.

New reply