Hi guys!

You may know me as the guy who made this weird game. In the last 48 hours I've been been told osu! has been broken since its inception and to "fix pp" many times, so I figure it probably best to centralise my thoughts and replies in one place (even though in most cases I did respond locally on twitter, forums, reddit or other).

Recently, a score was set using a touchscreen device which achieved a new high 900+ pp score. The performance point algorithm designed by Tom94 was made with many factors considered, one of which is "aim". Touchscreens greatly reduce the difficulty of aim as players using them now have up to 10 localised input points (their fingers) which both need to travel less distance and can preempt where they need to be for future hit objects.

Many people have proposed fixes so I'll quickly explain why each one has flaws or inherent issues that are not easy to solve:

"The new play had low 80-90% accuracy! Just de-weight the accuracy in pp!"

There's no reason touchscreen players can't improve their accuracy to a higher level. This would only delay the inevitable -Tom94

"Make touchscreen only leaderboards!"

This is a HUGE infrastructure change. It needs database, osu-web (new and old), bancho, client, team and community structural changes. It would take me (and other team members) off any other task for a few weeks. All of this new infrastructure would be made for a currently very small group of touchscreen players.

I do not see this as a good solution when at the end of the day the majority mostly care about making sure scores like this do not appear higher than tablet scores. There's definitely a time and place for splitting out ranking systems, but now is not that time.

"Lower the amount of pp weighting for aim!"

This would effectively place a skill cap on non-touchscreen plays and is not a true solution.

"Ban touchscreen players!"

No user should be banned for how they choose to play a game (within limits of course).

_______________________________


verto does a good job summing up quite well why this isn't a simple algorithm-change issue and has some strong opinions on mapping which i can generally agree with to boot.



We are actively looking into a temporary or semi-permanent solution until we have more flexibility in changing infrastrcuture (this will need to come after lazer/new web goes live) and will keep you guys posted as we figure things out. A proposed solution of a touchscreen mod which devalues touchscreen pp values across the board may be something we apply as a stop-gap measure, for instance.

The pp calculation engine and infrastructure has always been designed to allow for online changes, which propagate out to all existing scores, so making changes retroactively is very possible, which means there is not an imminent danger from delaying any changes we push out.

In the mean time, if you are a programming/maths/statistics guru and want to help improve the very-stale pp algorithm – not necessarily to fix touchscreen because that is very much not possible with a single algorithm, but to make ongoing balancing changes – we have opened a new channel specifically for technical discussion of the performance points algorithm on the osu!dev discord network. Please note that a certain level of competence will be expected; this is not a place to meme, rage or bitch.

Keep playing and please enjoy osu! as a game, in a fun way, not a grim WORLD-IS-GOING-TO-END-DUE-TO-ONE-HIGH-TOUCHSCREEN-SCORE way <3

Keep in mind everything that most everything I've said here is my opinion. Other people on the team may have differing opinions as many community members. We won't be making any decisions without taking all feedback on board, so do feel free to add your own below!

Nice

No reason to change the way pp values are calculated or even change the types of people we allow to play this game..

Why not just rerank tag4 maps since touchscreen was the reason why it was removed in the first place?

One step closer to a PP-less future <3

edit for a constructive thought: I haven't seen any high level touch screen players (other than "sorry did it for the memes" from freedomdiver) comment on this situation. I would like to see what someone like CatBagasm thinks about the touchscreen situation so we could have a less one-sided conversation.

Great to see some communication on the problem, thanks pepper.

Glad to hear that change is being looked into, regardless of whether or not the decision is what I would prefer

-Vanilla wrote:

No reason to change the way pp values are calculated or even change the types of people we allow to play this game..

Why not just rerank tag4 maps since touchscreen was the reason why it was removed in the first place?

because it's just going to add to the current problem. reranking tag4s would only make the top ranks even more dominated by touchscreen (for example, the new top score would be around 1050pp on exgon's megamix score).

i think once the touchscreen problem is fixed(?) then tag4s should be re-ranked. people who can get pp from those maps with tablet/mouse deserve it imo

Maybe there could be an option on your profile where u select touchscreen and it will automatically be sent to another leaderboard?

poggers
if touchscreen leaderboards happened, will that include touch kb players?

k

Nice

Thanks for getting to addressing this so fast, peppy!

I'm curious to hear about your stance on disabling mouse buttons, aiming with a touchscreen using a single finger and hitting notes with two keys, as one would with a tablet or mouse. Do you believe this would potentially fall into the touchscreen mod category?

Mits wrote:

Thanks for getting to addressing this so fast, peppy!

I'm curious to hear about your stance on disabling mouse buttons, aiming with a touchscreen using a single finger and hitting notes with two keys, as one would with a tablet or mouse. Do you believe this would potentially fall into the touchscreen mod category?

it'd be based on cursor "warping", and automatically applied. so it depends on how the user decides to play.

peppy doing good job

I think its a long awaited time for a meticulous review of how PP score is calculated.

There is a very tangible difference between 4 min 7* HDHR map fc and 20 seconds HDHRDT haitai FC.

This thing HAS to be addressed at some point, I truly hope that the solution will be finded and we gonna shift from TV-size farming meta to something a bit different.

I am not a math-guru, but I can say that the formula of PP calculation is struggling when it comes to sliders.

Maps with lots of sliders gives less score, doesnt matter if the actual difficulty is higher than the common TV-size anime maps.

So the main thing that comes to my mind is that PP system doesnt care how you hit sliders, how much input sliders are having on the player during the gameplay, how it affects reading and such.

But for temporary solution I can suggest to nerf how much PP is coming from short maps, it is just not ok that 20 second map gives 900 pp. Its just wrong.

It's good that u brought this up so quick peppy.

As technology is getting more advanced by time maybe there are other "tools" we should consider an fix on as early as possible.

There are certain companies out there that are currently developing eye-tracking software. What are you thought on this seeing as this may be an issue?

Wow the world is gonna end because oa f 900 pp play omg!!! Please panic erase all the pp

for real though I'm sure many people (myself included) would be willing to make a more balanced pp system, I doubt separating the leaderboards will make any difference in terms of people's satisfaction with both the pp system and the touchscreen issue

peppy wrote:

I do not see this as a good solution when at the end of the day the majority mostly care about making sure scores like this do not appear higher than tablet scores.

I really doubt there exists a solution that would ensure this never happens without completely separating the two, due to the way the current pp system deals with jumps. Having to consider touchscreen when balancing for mouse/tablet spacing and vice versa would just throw everything off. (we've seen this already, back when TAG4 maps were unranked btw please rerank)

Also, just a suggestion, but if touchscreen had their own leaderboards, they could have different mechanics as well. Maybe taking the streaming mechanic of osu!droid(aiming with one finger and tapping elsewhere with the other hand) and putting it into osu!std could make touchscreen more approachable. What's your opinion on that?

lolarisan wrote:

Maybe there could be an option on your profile where u select touchscreen and it will automatically be sent to another leaderboard?

Yeah but people could just not select that option and have it submitted regularly, and to prevent that there would always have to be people looking at the leaderboards and all plays for that, which is kinda impossible imo

lolarisan wrote:

There are certain companies out there that are currently developing eye-tracking software. What are you thought on this seeing as this may be an issue?

I felt sick for three days after trying to play osu! with eye-tracking. Two other people who tried almost vomited on my desk. Keep in mind that when playing osu! your eyes tend to read ahead. Trying to target hitcircles directly is hard. Sliders is puke (eyes are not made to smoothly track).

I completely agree with basically everything mentioned in the OP. The best time to solve this issue isn't now, but after the release of lazer. Knowing that it WILL be addressed is more than enough for now, while the issue isn't hugely damaging. It's better to inform touchscreen players of proposals sooner rather than later to help them decide whether the peripheral is something they'd still enjoy using.

Your temporary fix of applying a mod will work in the short term, but I think the only true solution to the issue is creating a separate gamemode for touchscreen when lazer drops. Touchscreen is a completely different game that needs different maps and different gameplay features and by extent, it's own leaderboard (like taiko/ctb/mania).

hi-mei wrote:

But for temporary solution I can suggest to nerf how much PP is coming from short maps, it is just not ok that 20 second map gives 900 pp. Its just wrong.

Completely incorrect. IF that map was a 20 second 300bpm full screen spaced deathstream, it definitely should give 1000s of pp. This shouldn't even have to be said.

fieryrage wrote:

I doubt separating the leaderboards will make any difference in terms of people's satisfaction with both the pp system and the touchscreen issue

Well I mean, it would completely solve the issue people have with touchscreen. PP issues are an unrelated topic.

What I don't really get is that changing the pp system doesn't work.
By nerfing difficultyspikes(Example: 5* map with one or two short 7* jumpspikes)
and improving pp gained from streammaps or technical maps,
the pp mappers, that would have mapped short longjump difficultyspiking maps would start to map different styles, effectively nerfing touchscreen indirectly.

Letting length, average difficulty, and accuracy play a greater role in the current system would probably shift the meta to a more balanced set of skills required for pp.

A few friends of mine who reached the top 1000 fairly fast, all have only one skillset in common, and that's long fast jumps.

No streams, barely any technical ability, no low AR reading, no complicated tap variations, barely any consistency on non-jumpmaps.

It's sad to see those playstyle slowly go away, they add so much depth to the game, but in the new mapping meta it's very rare to see stuff that isn't 1:28 jumps with the odd triple in there.

thank you peppy for listening to the community

Tune wrote:

What I don't really get is that changing the pp system doesn't work.
By nerfing difficultyspikes(Example: 5* map with one or two short 7* jumpspikes)
and improving pp gained from streammaps or technical maps,
the pp mappers, that would have mapped short longjump difficultyspiking maps would start to map different styles, effectively nerfing touchscreen indirectly.

Letting length, average difficulty, and accuracy play a greater role in the current system would probably shift the meta to a more balanced set of skills required for pp.

A few friends of mine who reached the top 1000 fairly fast, all have only one skillset in common, and that's long fast jumps.

No streams, barely any technical ability, no low AR reading, no complicated tap variations, barely any consistency on non-jumpmaps.

It's sad to see those playstyle slowly go away, they add so much depth to the game, but in the new mapping meta it's very rare to see stuff that isn't 1:28 jumps with the odd triple in there.

I completely agree.

hi-mei wrote:

I think its a long awaited time for a meticulous review of how PP score is calculated.

There is a very tangible difference between 4 min 7* HDHR map fc and 20 seconds HDHRDT haitai FC.

This thing HAS to be addressed at some point, I truly hope that the solution will be finded and we gonna shift from TV-size farming meta to something a bit different.

I am not a math-guru, but I can say that the formula of PP calculation is struggling when it comes to sliders.

Maps with lots of sliders gives less score, doesnt matter if the actual difficulty is higher than the common TV-size anime maps.

So the main thing that comes to my mind is that PP system doesnt care how you hit sliders, how much input sliders are having on the player during the gameplay, how it affects reading and such.

But for temporary solution I can suggest to nerf how much PP is coming from short maps, it is just not ok that 20 second map gives 900 pp. Its just wrong.

I think this is more or less what people should be focusing on. The problem is not strictly a "touch screen problem" because stupid plays like this can and will happen on any input device if a map is short enough.

Trying to separate the playerbase with different leaderboards is both unnecessary and abusable. The bigger problem lies in the system not weighing consistency properly which creates a "spam retry meta". The touch screen players can thrive on this since their accuracy varies a lot and all it takes is "one good run" to set the score.

The only real solution I can think of is introducing a soft limit on how much pp can be earned before a certain map length is reached. Anything below 60 seconds or so should start seeing some diminishing returns that scale exponentially with how short the map is.

You might say that this is a "band-aid fix" because "what if freedomdiver just gets better at playing longer maps" but let's be real, if you can sustain 85%+ acc on OD11 with touchscreen for over a minute then you probably deserve to set the score.

happening or not, im glad to see some action going on

peppy wrote:

lolarisan wrote:

There are certain companies out there that are currently developing eye-tracking software. What are you thought on this seeing as this may be an issue?

I felt sick for three days after trying to play osu! with eye-tracking. Two other people who tried almost vomited on my desk. Keep in mind that when playing osu! your eyes tend to read ahead. Trying to target hitcircles directly is hard. Sliders is puke (eyes are not made to smoothly track).

That is true. But this sickness might just be temporary like motion sickness. People might play enough to get used to it like changing from mouse to tablet.

And as for your eyes reading ahead, people might change a slight delay in the software making is go a bit slower when u look and then learn to hit after the circle.

Lets make osu! Great again together

I'm glad there will be a solution in the future. A touchscreen mod sounds like it might be a patchwork solution for now that doesn't break everything..

Grae wrote:

I think this is more or less what people should be focusing on. The problem is not strictly a "touch screen problem" because stupid plays like this can and will happen on any input device if a map is short enough.

Trying to separate the playerbase with different leaderboards is both unnecessary and abusable. The bigger problem lies in the system not weighing consistency properly which creates a "spam retry meta". The touch screen players can thrive on this since their accuracy varies a lot and all it takes is "one good run" to set the score.

The only real solution I can think of is introducing a soft limit on how much pp can be earned before a certain map length is reached. Anything below 60 seconds or so should start seeing some diminishing returns that scale exponentially with how short the map is.

You might say that this is a "band-aid fix" because "what if freedomdiver just gets better at playing longer maps" but let's be real, if you can sustain 85%+ acc on OD11 with touchscreen for over a minute then you probably deserve to set the score.

This is an emotionally fueled solution. Long maps are widely considered overrated, it's not an issue with map length. It's not an issue with low accuracy scores. It's an issue with touchscreen.

To imply that a touchscreen player can't do insanely hard maps for a minute straight is ludicrous. If a normal player can manage it, so can a touchscreen player.

This thread should be about the issues with touchscreen IMO, it's hard to discuss the standard PP system when so many people are trying to shoehorn in a fix for touchscreen. There's no system that can compare touchscreen scores with normal scores because they're completely different games.

Hey peppy, I guess this is a good opportunity to ask you about this:
What are your thoughts on moving the osu! pp system towards a multi-category kind of system that allows for different objectives (such as aim, control, speed) like the guys at http://osuskills.tk/ did, rather than a single leaderboard?

I think it would be a great refresher for both playing and mapping. Instead of mapping for ''pp'' like we usually see nowadays, mappers would have a lot of diversity to do so and maps that focus on other abilities such as reading would be far more valued as well.

Perhaps you've considered something like this already but I'm just curious on your thoughts. Personally I enjoy trying to improve my scores on that website from time to time, even if it's far from finalized. Thanks for looking to improve the pp algorithm!

please dont nerf touchscreen i was gonna get myself one for christmas ;(

MariahCarey wrote:

Hey peppy, I guess this is a good opportunity to ask you about this:
What are your thoughts on moving the osu! pp system towards a multi-category kind of system that allows for different objectives (such as aim, control, speed) like the guys at http://osuskills.tk/ did, rather than a single leaderboard?

Perhaps you've considered something like this already but I'm just curious on your thoughts. Personally I enjoy trying to improve my scores on that website from time to time, even if it's far from finalized. Thanks for looking to improve the pp algorithm!

pp already works like this under the hood, but aim/acc/speed are accumulated together for each score and presented as "pp". If they can't be separated for the current system, I don't know if we can expect them to be for the new system. If they can be separated, that should be done before implementing a new system to check which pp system actually performs the best in optimal conditions.

nibs wrote:

Touchscreen players can only exploit a small selection of maps as far as I know (ones that are almost entirely jumps/no streams, like yuima-ru world and granat), so maybe just have a thing in the ranking criteria that prevents maps that are:

very short and have very high star ratings (yuima-ru world is 43 seconds and it's 6 stars)

maps that are easy until the last 10 or 15 seconds and then it's just jumps (use something like this to see https://rarelyupset.com/oppai/)

from being ranked? Sorry if this isn't helpful, just my thoughts on this entire situation.

These criteria aren't set in stone. Touchscreen players can do longer maps, it's just easier to do the shorter ones. The only things really required are huge jumps and a low amount of short streams. It's also unfair on mappers to disqualify their maps just because a subset of people want to use a peripheral that isn't suited the the performance system.

jesse1412 wrote:

This is an emotionally fueled solution. Long maps are widely considered overrated, it's not an issue with map length. It's not an issue with low accuracy scores. It's an issue with touchscreen.

To imply that a touchscreen player can't do insanely hard maps for a minute straight is ludicrous. If a normal player can manage it, so can a touchscreen player.

This thread should be about the issues with touchscreen IMO, it's hard to discuss the standard PP system when so many people are trying to shoehorn in a fix for touchscreen. There's no system that can compare touchscreen scores with normal scores because they're completely different games.

Its not ok that AIM-based short maps are considered more valuable than long maps that are focused on reading/stamina/memorization.

Short maps has to be nerfed, ill say it again: 20 second map can NOT give 900 pp. No matter how you set the score, it is just makes people who has one strong ability to be higher (lets say Rafis) than people with other strong ability (lets say -GN).

The rewarding system should consider all the players equally. For now we are living in a racist world where "aim players" are better than "reading-focused players" and "stream-focused players".

Talking about touchscreen players - they will NEVER get a good score on any map longer than 1 min with lots of bursts, streams and fast sliders.

The problem was never in touch screen players, the problem is that the system thinks that jumps are way harder than anything else (streams, fast sliders, low ar maps etc).

hi-mei wrote:

ill say it again: 20 second map can NOT give 900 pp.

And I'll say it again. If you have a 20 star 20 second map and someone fc's it, you'd have to be braindead to argue that it should be worth less than 900pp; thus 20 second maps CAN give 900pp.

hi-mei wrote:

Its not ok that AIM-based short maps are considered more valuable than long maps that are focused on reading/stamina/memorization.

The rewarding system should consider all the players equally. For now we are living in a racist world where "aim players" are better than "reading-focused players" and "stream-focused players".

Different argument that I completely agree with. It also, sadly, has no easily implementable solution.

hi-mei wrote:

Talking about touchscreen players - they will NEVER get a good score on any map longer than 1 min with lots of bursts, streams and fast sliders.

True but you're relying on the touchscreen player to only use touchscreen. If the player uses keyboard/mouse/tablet to do the first few stream sections of a map and then swaps to touchscreen for an ending jump spike, it becomes possible.

hi-mei wrote:

The problem was never in touch screen players, the problem is that the system thinks that jumps are way harder than anything else (streams, fast sliders, low ar maps etc).

Completely wrong, did you read the OP and watch the video provided? It explains why. If a tablet player could set the scores that freedom diver set, they'd definitely deserve 800+pp. That play is FUCKING INSANE, regardless of whether jumps are slightly overweighted. PP is mostly not the issue, touchscreen is the issue.

osu!droid

jesse1412 wrote:

And I'll say it again. If you have a 20 star 20 second map and someone fc's it, you'd have to be braindead to argue that it should be worth less than 900pp; thus 20 second maps CAN give 900pp.

you have some misconception of what i meant there, we have a clear comparison of maps with kinda same star rating, right? so lets say, we have these two freedom dive hdhr and haitai 3mod. the first one demands all the possible abilities to get an FC (reading, speed, aim, stamina, memorization), its 4 min and u get a 800 pp for it.

the second one is ONLY focused on ONE aspect of the game, which is aim, and its literally x12 times shorter than freedom dive map.

x12 times, and its only requires one ability to have (FD requires at least 5).

so even if the HAITAI was 4 min with the same jumps all over the table, it would still be less valuable than FD score, cause it requires less anyway.

I know it sounds kinda strange but... Ill never be ok with the thought that less effort gives more rewards. its just not how life works.

True but you're relying on the touchscreen player to only use touchscreen. If the player uses keyboard/mouse/tablet to do the first few stream sections of a map and then swaps to touchscreen for an ending jump spike, it becomes possible.

Cmon, we both know its impossible.

Completely wrong, did you read the OP and watch the video provided? It explains why. If a tablet player could set the scores that freedom diver set, they'd definitely deserve 800+pp. That play is FUCKING INSANE, regardless of whether jumps are slightly overweighted. PP is mostly not the issue, touchscreen is the issue.

I still disagree, if the system would work normally, there wouldnt be a 20 sec 900 pp scores on a map that was made in literally 30-40 mins with no effort at all. People are ranking short maps because of the score in first place, not cause its fun and shit.

Talking about touchscreens again in particular, its just allows AIM-aspect to dissapear from the actual diffuculty, but it struggles with other aspects of the gameplay (reading, streaming etc).

Nerfing the touchscreen will not resolve the issue because people will find other techniques to abuse JUMPS, the issue is more in the map, in my opinion.