forum

PP!Balance: A proposal for a new PP system

posted
Total Posts
40
show more
winber1
i've heard china is getting extremely close to a facial password, assuming you give it a very particular kind of photo. i'm not exactly assuming that you can detect someone from all angles, literally just at a particular frontal view with very similar lighting

it's true you can basically become a self-proclaimed data scientist by just taking online courses, but you still have to spend tons of time to really understand the capabiliies and pitfalls of whatever you're using. it's almost not too different than just getting a higher degree (though of course it will take less time since you will basically be specially in literally just one thing I guess)
chainpullz
If you give Tom's pp algorithm a very particular kind of map it does a very good job of identifying pp values. I'm not assuming it can detect all sorts of different patterns, overlaps, and angles, literally just at a particular bpm with simple spacing, patterns, and a rhythm like a metronome.

Grad courses teach you how to derive and prove things. Most data science is done by brute force testing different initial conditions and following basic rules of thumb. A Masters dosent even require research and a PHD is mainly a way to obtain experience as well as publications and a professional network.

They offer machine learning, vision, etc. as upper level undergrad courses minus the theory because CS majors can't do proofs to save their lives and while possibly useful it definitely isn't necessary. Applied machine learning isn't rocket science.
Edgar_Figaro

B1rd wrote:

The only real solution is to dispense with the PP system altogether. Then instead of people scrambling to get the top PP plays, they will be scrambling to get the best plays as defined by themselves and their peers. In other words, an algorithm challenges the human mind to exploit it. Having no PP system uses the human mind itself as a judge of performance, which is vastly superior to any PP system.

#downwithPP
What’s even the point of this being an online game if there is no ranking/scoring system? Might as well just play offline 100% of the time as my plays won’t even matter and I’ll get slightly better performance (game stability) from playing offline.

I know “removing ranking system” sounds like a good idea to people but this only stems from long periods of disgruntlement from the flaws in the systems. Osu is a competitive game and many players (myself included) find the competitive aspect to be a massive draw. If there was no ranking system I tbh wouldn’t play half the songs on Osu as they aren’t my preferred music choice, Also I’d care a lot less about getting better as “what’s the point” and I’d play this game a lot less than I currently do.

TL:DR: Ranking systems are a major aspect of players desire to play games and shouldn’t be removed just to appease the hipsters
B1rd

Edgar_Figaro wrote:

B1rd wrote:

The only real solution is to dispense with the PP system altogether. Then instead of people scrambling to get the top PP plays, they will be scrambling to get the best plays as defined by themselves and their peers. In other words, an algorithm challenges the human mind to exploit it. Having no PP system uses the human mind itself as a judge of performance, which is vastly superior to any PP system.

#downwithPP
What’s even the point of this being an online game if there is no ranking/scoring system? Might as well just play offline 100% of the time as my plays won’t even matter and I’ll get slightly better performance (game stability) from playing offline.

I know “removing ranking system” sounds like a good idea to people but this only stems from long periods of disgruntlement from the flaws in the systems. Osu is a competitive game and many players (myself included) find the competitive aspect to be a massive draw. If there was no ranking system I tbh wouldn’t play half the songs on Osu as they aren’t my preferred music choice, Also I’d care a lot less about getting better as “what’s the point” and I’d play this game a lot less than I currently do.

TL:DR: Ranking systems are a major aspect of players desire to play games and shouldn’t be removed just to appease the hipsters
If only there was a way to judge the skill or players and plays without an arbitrary number...
repr1se

B1rd wrote:

If only there was a way to judge the skill or players and plays without an arbitrary number...
there isn't... LoL uses MMR, Dota2 uses ELO, CS:GO uses Glicko2, ESEA uses RWS, CEVO uses efficacy, Hearthstone uses MMR, SC2 uses ELO
ELO was designed for chess

you get the point
there's no way to rank people without using numbers
B1rd
There is.
Tamako Lumisade
Remove pp.
pls enjoy game
my name is hi

B1rd wrote:

There is.
Then please enlighten me, because I don't have a single clue of what it could be while you seem to have a pretty good idea
repr1se

B1rd wrote:

There is.

Boaring wrote:

Then please enlighten me, because I don't have a single clue of what it could be while you seem to have a pretty good idea
"I have discovered a truly marvelous proof of this, which this margin is too narrow to contain." -Pierre de Fermat

in other words, he knows but we're too dumb to understand, so he's making us figure it out
B1rd
Well it shouldn't be that hard to figure out considering I just told you what the method was. That method being your brain. I don't need to look at the rankings to say that Cookiezi is the best player. I can look at Freedom Dive HDHR and multitude of other scores he was made and use that to determine he is the best player. Of course it's not easy to neatly rank people from 1-100,000,000, but that's how it should be, because the game is so complex with all the skills that is very difficult to make an objective ranking of overall skill. You can't look at any one player in the top 100 and say for sure that he is worse than all those above him and better than all those below him, because the PP system only measures a specific skillset. And it's not hard even without the PP system to rank players according to a few singular aspects. However we shouldn't pretend that those few aspects that are measures by the PP system create an objective ranking of skill, and then only compete in those few skills, such as we do now. Just because you get rid of the PP system, doesn't mean it gets rid of the competitive aspect of the game; it's just that instead of competing on farm maps like Daidai Genome and Haitai, they will be competing on actual good maps. As far as I know, other rhythm games like Stepmania or Beatmania don't have any rankings like osu! does, so you can't complain that it would ruin the game.
Zid

B1rd wrote:

As far as I know, other rhythm games like Stepmania or Beatmania don't have any rankings like osu! does, so you can't complain that it would ruin the game.
Not sure about Stepmania but I know Beatmania uses the Dan System, where players are required to pass a course consisting of 4 maps that each test different skillsets of the game. If they successfully pass the course they rank up and they do this up until they reach Kaiden, which is the highest rank you can acquire.

Not disagreeing with your point though, just wanted to explain the rankings for anyone that doesn't play.
chainpullz

B1rd wrote:

<snipped>
A true ranking system should be well ordered. What you described is only a partial order and that's assuming it's even transitive. I'm not saying you're wrong about removing the ranking aspect, just that what you described does not produce a ranking.

After having experienced a rhythm game arcade environment and the mentality it breeds I personally think osu would be better off without rankings. There is no mid-map retry in arcade versions and people mainly play to beat themselves as opposed to others. The mentality is never "yay my rank went up." Rather, it's something more like "holy shit I finally did it."

The Dan system would probably be alright too since, again, it's more about beating yourself than trying to rank you against everyone else.
repr1se

B1rd wrote:

Well it shouldn't be that hard to figure out considering I just told you what the method was. That method being your brain. I don't need to look at the rankings to say that Cookiezi is the best player. I can look at Freedom Dive HDHR and multitude of other scores he was made and use that to determine he is the best player.
this leads to subjectivity as to what constitutes a good player. especially with the number of players complaining that DT is not a skill, and jumps aren't skill (which... is retarded). while the pp system isn't perfect it clearly shows that cookie is better than i am

Of course it's not easy to neatly rank people from 1-100,000,000, but that's how it should be, because the game is so complex with all the skills that is very difficult to make an objective ranking of overall skill. You can't look at any one player in the top 100 and say for sure that he is worse than all those above him and better than all those below him, because the PP system only measures a specific skillset. And it's not hard even without the PP system to rank players according to a few singular aspects. However we shouldn't pretend that those few aspects that are measures by the PP system create an objective ranking of skill, and then only compete in those few skills, such as we do now.
you just wrote that you determined cookiezi to be the best player
i beg to differ. the top ranks have their variety of skillsets, especially rohulk and idke
do i think that pp overvalues certain aspects of the game? yes. do i think that pp measures ONLY one skillset? no.

Just because you get rid of the PP system, doesn't mean it gets rid of the competitive aspect of the game; it's just that instead of competing on farm maps like Daidai Genome and Haitai, they will be competing on actual good maps. As far as I know, other rhythm games like Stepmania or Beatmania don't have any rankings like osu! does, so you can't complain that it would ruin the game.
then what constitutes a good map? whatever answer you give is subjective and everyone's opinions vary, you can't just redo the pp system to have "actual good maps" and satisfy everyone. undoubtedly someone will say "your 'good maps' are shit, give me the old pp system"

i tried searching for beatmania and stepmania leaderboards and didn't find anything. osu is the only game i know of with a leaderboard this hardcore and that's what sets it apart.
B1rd

chainpullz wrote:

B1rd wrote:

<snipped>
A true ranking system should be well ordered. What you described is only a partial order and that's assuming it's even transitive. I'm not saying you're wrong about removing the ranking aspect, just that what you described does not produce a ranking.

After having experienced a rhythm game arcade environment and the mentality it breeds I personally think osu would be better off without rankings. There is no mid-map retry in arcade versions and people mainly play to beat themselves as opposed to others. The mentality is never "yay my rank went up." Rather, it's something more like "holy shit I finally did it."

The Dan system would probably be alright too since, again, it's more about beating yourself than trying to rank you against everyone else.
I really meant that you can judge player skill without a ranking system, though you can still rank players based on certain criteria, you just don't get to universalise your arbitrary ranking system. Although you could say that judging people by their scores is still judging people based on numbers, but no reason to get overly semantic.

repr1se wrote:

this leads to subjectivity as to what constitutes a good player. especially with the number of players complaining that DT is not a skill, and jumps aren't skill (which... is retarded). while the pp system isn't perfect it clearly shows that cookie is better than i am

you just wrote that you determined cookiezi to be the best player
i beg to differ. the top ranks have their variety of skillsets, especially rohulk and idke
do i think that pp overvalues certain aspects of the game? yes. do i think that pp measures ONLY one skillset? no.

then what constitutes a good map? whatever answer you give is subjective and everyone's opinions vary, you can't just redo the pp system to have "actual good maps" and satisfy everyone. undoubtedly someone will say "your 'good maps' are shit, give me the old pp system"

i tried searching for beatmania and stepmania leaderboards and didn't find anything. osu is the only game i know of with a leaderboard this hardcore and that's what sets it apart.
Yeah, because player skill is somewhat subjective. There are so many skills in the game that it's hard to say that any certain skill is better. Although you can have a in-depth discussion about that.

What you're saying is that you want the game to prioritise certain aspects of the game, like aim and speed, over others. Well if people don't want to recognise skills like DT, then so what? You shouldn't try to force other people to respect you. People should be able to make their own judgements about skill and form their own sub communities based around certain aspects of the game. You see this with the EZ community, and this is somewhat because EZ and reading maps are so far removed from the PP system that EZ players aren't pushed into competing over PP like other players are. I think people who complain about DT aren't actually saying that DT takes no skill, but rather it takes no skill relative to other skills that give the same PP, which is pretty much true. You already admitted that the PP system isn't an objective criteria of skill as you denied Cookiezi to be the best when he has the most PP. So you should be all for getting rid of the PP system so that players with unique skillsets aren't overshadowed by players with slightly more PP.

The problem with the PP system is that is is so pervasive and affects people who don't care for it. Yeah, I do believe that some maps are objectively better than others; I don't regard copy paste PP maps to be as good as maps by Lan Wings for example. And it's the result of the PP system that people who don't like PP mapping and aren't interest in competing in that specific skillset are marginalised.
You have to realise that mapping is basically like a market place. Mappers like to make maps that are popular and well received by others. Therefore, mappers generally cater to what the community likes. With the existence of the PP system, the majority of players play for PP, therefore they play maps which give lots of PP and have skillsets catered towards the skills which give PP. Therefore, there is a large demand for PP maps, so the majority of mappers will make PP maps, and even the mappers who don't explicitly map for PP will have their map design largely influenced by the meta and the skillsets of the majority of players. That's why you don't see many AR7 or 8 maps these days, and in fact its quite difficult to get a map ranked that goes against the meta. That's why it's not valid to say that if the PP system were to go away that what is currently regarded as PP maps will disappear and the players will be left wanting. My argument is, in essence, that the PP system means that players play maps that give PP rather than what they consider good or fun to play. If it were to go, then players would play and compete on maps that they and the community in general thought were good and fun, and therefore the demand and thus supply of maps would shift towards what the community regarded as good and fun. It would also create a more diverse range of maps that focused on larger degree of skillsets, and there wouldn't be a ranking system that arbitrarily chose a few skills to be the most important ones.

Really, I don't regard osu! as any more "hardcore" than Stepmania or Beatmania. Looking at your average Beatmania or Stepmania player, I wouldn't consider them not to be hardcore.

And yes I said that Cookiezi is the best player - because he is. No one can match Freedom Dive HDHR, no one can match a plethora of his other plays either. He has amazing aim, the best streaming consistency in the game - not only that, he has great reading, and an exceptional ability to play difficult patterns and has very good acc. You might find players that are better than him in one or two aspects, but there is no one who is so exceptionally good in so many areas. He is in a league of his own.
Edgar_Figaro
I don't feel like quoting all of B1rd, but he is pretty much saying get rid of ranking altogether and just go based on individual map scores but not have that affect any sort of ranking system.

Here's the thing, people LIKE the PP system (for the most part). Before the PP system even existed there was PPV1 which was a ranking system just based on map ranks which is pretty similar to what OP suggested. People wanted a system that more accuratrely measured skill between players and thus came around Tom's TP system.

So even BEFORE the current PPV2 was in-place there was a demand for a system to rank players with more quantifiable numbers. Also we have other things currently like OsuSkills and the sort.

In-short, most people want a ranking system of some sort and not just to arbitrarily decide that "oh this perosn is good". That only really works for about 50 or so players who people have even heard of and leaves everyone into obscurity. It just would lead to a system where the only people who are ever seen as "good" have to start streaming themselves on twitch and gather a following so people recognize their skills.

PP is a nice way for people to see their improvement in the game and lets them know how much they've improved. Is PP perfectly balanced? Well no. Will there always be maps made to try and get more play based on popularity through PP? Probably yes. Do I think these problems means that "no ranking system" would be a better alternative? Absoloutely not.

If PP is removed, all it'll do is cause people to start using a system similar to TP which wasn't official and start ranking that way as trying to just go based on a "word of mouth" system to decide who is good at what skills would be extremely obnoxious.

Just to use myself as an example. In Taiko I am a primarily Hard Rock Specialized player. I am pretty good at high bpm hard rock play and have very good high scroll reading. Now am I the very best Taiko Hard Rock player to ever play this game? No not by a long shot. Now if we simply by a "word of mouth" system. The only people who would ever be recognized would be the very best of the best. While I do have a few impressive HR scores on maps that most people don't HR, it still wouldn't be enough to get my name recognized as being a well known HR player as there are people out there much better at the skill. Basically removing ranking systems means you're either the best or your not. People would have no incentive to improve as they'd feel like the top players were out of reach and until they reached the pinnacle they'd never get recognized for their achievements.
B1rd
Speak for yourself.

How does the current ranking system take people out of obscurity? People don't even know all of the players in the top 50, let along random players in the top 1000. People don't go searching through the rank list to find players, it makes zero difference to the publicity you get. You can easily judge the skill of players. Just replace the ranks in your profile by a showcase of your plays. People can look at a random sample of your best plays that you choose to show and use their knowledge of those maps as a benchmark to judge your skill. And you can judge your own skill and your own improvement by the plays you get, just as easily as using an arbitrary number. If we used this system, it would be much more conducive to publicizing your best plays, because currently, you can make amazing plays but no one will know if those plays aren't worth PP.
repr1se

B1rd wrote:

What you're saying is that you want the game to prioritise certain aspects of the game, like aim and speed, over others. Well if people don't want to recognise skills like DT, then so what? You shouldn't try to force other people to respect you.
where have i made that claim? i only mentioned that too many bitches whine about DT without recognizing that it actually does take skill to play. and i'm not even a DT player.
I think people who complain about DT aren't actually saying that DT takes no skill, but rather it takes no skill relative to other skills that give the same PP, which is pretty much true. You already admitted that the PP system isn't an objective criteria of skill as you denied Cookiezi to be the best when he has the most PP. So you should be all for getting rid of the PP system so that players with unique skillsets aren't overshadowed by players with slightly more PP.
where have i made that claim? i was pointing out that you wrote that you can't determine who's the best, but you also write that cookie was the best
rohulk bro

The problem with the PP system is that is is so pervasive and affects people who don't care for it. Yeah, I do believe that some maps are objectively better than others; I don't regard copy paste PP maps to be as good as maps by Lan Wings for example. And it's the result of the PP system that people who don't like PP mapping and aren't interest in competing in that specific skillset are marginalised.
just about all of the top players don't care for pp. they play for fun and pp is gained on the side.
lul maps are subjective, you can say that hollow wings is a great mapper and i could say he's shit, but there's no objectivity behind either statement
You have to realise that mapping is basically like a market place. Mappers like to make maps that are popular and well received by others. Therefore, mappers generally cater to what the community likes. With the existence of the PP system, the majority of players play for PP, therefore they play maps which give lots of PP and have skillsets catered towards the skills which give PP. Therefore, there is a large demand for PP maps, so the majority of mappers will make PP maps, and even the mappers who don't explicitly map for PP will have their map design largely influenced by the meta and the skillsets of the majority of players.
that's simply not true... there are mods and ranking criteria for a reason. i could pp map for this map but it won't get ranked because it doesn't meet criteria simply because it doesn't fit the song

That's why you don't see many AR7 or 8 maps these days, and in fact its quite difficult to get a map ranked that goes against the meta.
because ar7 and ar8 are where players still learn the game mechanics, and ar9 is where reading actually begins
That's why it's not valid to say that if the PP system were to go away that what is currently regarded as PP maps will disappear and the players will be left wanting. My argument is, in essence, that the PP system means that players play maps that give PP rather than what they consider good or fun to play. If it were to go, then players would play and compete on maps that they and the community in general thought were good and fun, and therefore the demand and thus supply of maps would shift towards what the community regarded as good and fun. It would also create a more diverse range of maps that focused on larger degree of skillsets, and there wouldn't be a ranking system that arbitrarily chose a few skills to be the most important ones.
and if the community thinks that copypasta jump maps are fun?
before you pounce, i actually enjoy technical maps, but a majority is a majority

And yes I said that Cookiezi is the best player - because he is. No one can match Freedom Dive HDHR, no one can match a plethora of his other plays either. He has amazing aim, the best streaming consistency in the game - not only that, he has great reading, and an exceptional ability to play difficult patterns and has very good acc. You might find players that are better than him in one or two aspects, but there is no one who is so exceptionally good in so many areas. He is in a league of his own.
again, i didn't confirm or deny cookie being good or not.
B1rd
Yeah... I'm not gonna respond to that monstrosity of post.
repr1se
i'm sorry. i have been enlightened. i now realize that the pp system is completely broken and should be removed as a whole for the benefit of us that enjoy maps that don't get much pp.

jk later bitch
B1rd

repr1se wrote:

i'm sorry
Apology accepted.

repr1se wrote:

i have been enlightened.
Are we talking about the Buddhist type of enlightenment? As a Christian I'd advise you not to let your soul be corrupted by heathen religions.

repr1se wrote:

i now realize that
Realise what?


repr1se wrote:

the pp system is completely broken and should be removed as a whole for the benefit of us that enjoy maps that don't get much pp.
Agreed, that was my main point.

repr1se wrote:

jk later
Okay bye.

repr1se wrote:

bitch
Wait, who is this directed at?
Topic Starter
Mio Winter
Sorry for not replying earlier. I've been busy and then it took a while to compose my reply. I edited the main post to include a section called "Solving the differential popularity problem".

Full Tablet wrote:

Using the average score is not a good indicator of the difficulty of the map, easier maps tend to have relatively few good players attempting plays (mostly people who hunt for score ranks only) and a lot of players who aren't good at the game in general.
Yeah, I tried solving that problem in my edit. Would really like to hear what you think of it since you have experience in trying to create a statistical PP system.

Full Tablet wrote:

Here is an example of another ranking system (for osu!mania) based on statistical data only:
t/329678
It is based on calculating "Difficulty Curves" for each beatmap ("Player Skill" required to achieve certain score in the beatmap), and simultaneously fitting and those curves and the "Player Skill" values each player has (based on the scores they have set). It requires an enormous amount of computing power to calculate, though (here it took several months to calculate the rankings with only ~2000 players, with the time required to calculate increasing approximately quadratically with the amount of players and linearly with the amount of beatmaps).
Wow! Really nice. Have you made a post explaining this algorithm anywhere, or do you have the code on github or something?

repr1se wrote:

there are other things that the pp system leaves out, like unstable rate and overall stability of the player. i'm going to propose some generic equations where specific numbers can be filled in later yada yada yada

this proposition encourages stability in accuracy and increases the amount of pp you get from high OD (like HR):
https://gyazo.com/ab41ade7e2796cafc99fd3c5b2f6276b

this proposition reduces the amount of pp awarded in jump maps, to align with stream and technical maps, as well as reducing the pp of slider-only maps:
https://gyazo.com/96a13706bf510199adfae034a234e471

this proposition increases the pp value of technical maps.
https://gyazo.com/591ea23d810d681155da6e7d93d88b71

do i think these changes have to be implemented? no. but i think it does address the concerns of people whining about the current pp system.
Cool! But I think statistical measures will do better at capturing what "difficulty" is than direct measures because the direct measures will have to deal with the near-infinite complexity of the real world.

repr1se wrote:

k_1 and k_2 are constants that determine the range of PP values.
-how will this be determined?
Just by picking some arbitrary values and judge whether they look about right. You can fiddle with the numbers here.

Momiji wrote:

Mine solution - Do that, then turn ppv2 into SCORE, and then filter ppv2 values through the ppv1 amplification
Huh, interesting. That might work better than using ScoreV2 since PPv2 seems to scale accuracy better. Correct me if I'm wrong, but my impression is that the difference between 99 % and 100 % acc is a lot more on PPv2 than on ScoreV2. This is as it should be because going from 89 % to 90 % accuracy is a lot easier than going from 99 % to 100 %.

ManuelOsuPlayer wrote:

Or a PP system made by players ratting. Where players rate maps under their skills to don't exploit it. For example my lowest toprank PP It's 89pp. I should be able to rate maps what i can get an S to give from 1pp to 88pp.
I really like the idea of basing PP values on player ratings and would like to see it developed further. It might just work.

Although I don't think you should require a certain amount of PPv2 in order to be able to rate maps since then you'd inherit some of the problems that the PPv2 system has. I'd rather suggest that players who have gained an A on a map should be allowed to rate its difficulty. It shouldn't be lower, since you have to be able to play a map decently well in order to truly understand its difficulty. For example, when I was a new player, I couldn't tell the difference in difficulty between Scarlet Rose and, say, Neuronecia, but I now know Scarlet Rose is much harder.

B1rd wrote:

#downwithPP
Meh. I like having a quick and fairly reliable way to determine how good random players on the forums are at aim. It's a pretty good lowest estimate for how good someone is at aim. If someone is 6K PP, I know they're better than me at aim; but if someone is 1K PP, I don't know whether they're better than me at aim.

Husky wrote:

Remove pp.
pls enjoy game
Hmpfh. There's no contradiction between trying to gain PP and enjoying the game. Likely there are very many players who think osu! is fun because they like trying to get more PP than their neighbour.
kricher

Mio Winter wrote:

[i]



make it (MapRank/N)*(1/ln(sameStuff))?
Topic Starter
Mio Winter

kricher wrote:

Mio Winter wrote:

[i]



make it (MapRank/N)*(1/ln(sameStuff))?


Hi. : ) This wouldn't work because it would give higher PP for plays on maps where fewer people had already played a map. Consider the PP of #1 on a map with N=1000, vs the PP of #1 on a map with N=10000. With your formula, it would give 10 times more PP to the former play.

(And then there are other reasons I wrote about why the first idea I wrote about wouldn't work.)
Lesbea
I was wondering for a few days if someone designed an adaptive system to fit the subjectivity of difficulty, and here it is. I didn't read it entirely but it looks solid, wow. At the very least, you can hopefully dump this on a programmer and they will be able to do it without asking too many questions.

I need to read again more in-depth to be sure, but it also sounds quite heavy to compute all these things, especially to solve the differential popularity. I believe peppy would gladly try this in a hypothetical universe where he's bored and doesn't have a lot of features to add, bugs to fix, and real life to live, but if the complexity is not lower than quadratic just for adding a new player score, then your ideas are basically moot. In that case, you need to either find a way to make the calculations progressively without breaking the ranking too much, or hope that everything can be refreshed in one go periodically with a good initial estimate for the new submissions. But first, you need to implement a ladder with some dataset and make sure it's both accurate and fast, like someone else said before me. But where is the dataset ? I'm asking Google, he says fuck you. :(
There is still the osu!api to request a few things like you said, but the top plays are definitely not going to be enough. You can still request the score for a combination of map + player, but it will hit the request limit pretty quickly.

Anyway, I can't give a proper opinion without trying out first.



I haven't completely read the thread either (honestly I don't want to dive into the community too much), but I don't agree with some other propositions. After re-reading I realized I'm repeating some stuff that was already said, but I'm too lazy to remove it now :

@ B1rd
TL;DR :
Even if you don't like rankings, can't you just let us chirp and brag with each other ? It's not like we are affecting how you play, is it ? Hmmm.
Actually there IS a way to even affect solitary snowflakes like you (assuming you still do play offline) but it's at the end of my rant, sorry.

The long post if you want to bother reading monstrosities :
You want to judge without a number, using the community opinion. Well that's great, but this system has been tested for many kinds of competition for centuries, and it has always been clearly biased towards the popularity of a player. I mean, look at how you (and others) fanboy about Cookiezi. I'm not saying he's not the best, he probably is, and he might even have double the PP compared to the #2 with the system proposed in this thread for all I care, but a lot of people are clearly rating him higher than he should be, because lolhype it's Cookiezi ffs bruh. It's difficult for a newer player to become #1 even if he actually had better skills, because the gaming communities are biased towards old players. The same shit happens in Super Smash Bros about Isai because this game has a lot of subjectivity about skills even though the guy is tired (and retired) of the game and the fans are rating him on pure speculation. Bias can also exists (for better or worse) towards black people, overweight people, girls, ugly people, asians, gays, etc. For example I am biased against USA and would severely bring down any player from there just because I hate that 2nd amendment and because their "USA! USA! USA!" chant is fucking annoying. And hey you're forgetting about all the players that aren't Cookiezi or that-super-popular-guy. What's up with MY rank ? Nobody is going to bother, and it's really frustrating to not have a way to see how much I progress. This is also why scoring/speedrunning in video games needed the internet, because you couldn't see who was the best one and how to improve before you could slap your name on a list with a score/time/whatever. We need to compare each other or to show off at some point to make a hobby interesting, it's not simply a matter of "plz enjoy the game in a corner of your little world". Even when you work seriously on some project, you have to deliver it to the world someday, and people will judge it with whatever method is available. You can't just store it in a cardboard and forget it in your attic.
By the way, even if you judge with a single person who decides who goes where without comparing scores, you're still going to need numbers. Cookiezi is the best ? Guess what. He's at the top of the list in your head, "1" is a number, whether you like it or not, and the same goes for the players below.

On another topic, actually we DO have a judging team of humans doing work on osu. Look at the beatmap submission system, and how mappers like Sotarks are garnering attention because they denature the game, yet every kind of bullshit he comes up with automagically goes to ranked. I don't have a clue about his personality and never read a single post from him, but he is clearly mocking someone with https://osu.ppy.sh/s/805224 (it's not like I can FC it but my opinion is still valid right ?). Look at that background, the diff title, the patterns. I'm actually afraid of clicking on the beatmap discussion to read QATs fawning all over him when they are actually the clowns of the story. I don't want to play a casualized game (giving out free PP isn't the right path to take) just because a small club of circle1-2jerkers decided to do memes instead of doing their job. Actually, maybe we should go the shaming route, and straight up balance the PPs according to how much a mapper is giving out to the playerbase. It's even funny how people talk about the mapping meta. The fuck they are talking about, there is nothing of the sort. You can only talk about a meta when there is a winning objective to reach against another strategy. Mapping the best PP/difficulty ratio isn't the objective of a rhythm game. They need to map the music/song correctly instead, and the only "opponent" they have to beat are the players. I'm not saying they should only map weird songs like an aspire, but hopefully you get my point.

I'm sorry for writing so much about the mapping and submission communities even though I hardly interact with them (people will probably say I don't know what I'm talking about because of that), but they still have a lot of effect on the rankings. If the system presented here can provide incentives to increase the quality of beatmaps and give them a proud reason to talk about a "mapping meta" without creating controversial dramas (because that's something humans are incredibly good at), then it's all the better. Still needs testing though.




@ winber1
Deep learning is a fancy tool. You said it was becoming reliable, but it's actually overblown by marketing shenanigans. It doesn't just work for everything, you're almost never going to hear about cases where it failed miserably (for obvious reasons, the successful applications are the ones who make the most buzz), and from a mathematical standpoint we can't prove an upper limit of the error rate.
Well if it "looks like" it works, okay, whatever. But keep in mind that it will only just "look like". You will never be sure. If it breaks down, you will not know why. This cannot be called science, really, but again we are calling "science" a lot of things that are mostly bullshit just because it makes people feel great, so I guess it's better than nothing.
Please sign in to reply.

New reply