forum

Fixing Taiko - Megathread (SR and PP)

posted
Total Posts
15
Topic Starter
Jaye
Dear Taiko Community,
This thread is going to discuss A LOT of things about Taiko’s Star Rating, Performance Points, and Modifiers. Most of what I present is just going to be thoughts on the current situation, issues presented, and possible solutions (for some things). I’m not going to talk about Taikomania because we all already know that it’s not Taiko and discussion of this is not what I want to become of the thread. Of course I’m not an authority on this, and whatever I say does not necessarily go. All comments and ideas welcome (except those of a defeatist nature); it’s time we sorted this shit out properly.

Star Rating

Alchyr’s attempt is good but needs more fleshing out. Whilst we almost all agree to nerf 1/4 + 1/6 and doublets (of the 1/8 variety), the challenge is managing to do this without sacrificing SR for technical maps (also nobody even knows what the definition of a technical map is so /shrug.) All I can say on this is keep up the good work, keep testing the algorithm, and keep breaking it so that we can develop better solutions.

Another issue to mention is that no pattern is inherently more difficult than another, yet we still see things like kddkddk awarding more. This of course does not factor in rhythm or snapping complexity (which is a whole other story, mentioned above). Making a system that does not encourage particular patterns of notes is ideal. Remember that we players adapt to the patterns that are ranked, which are often influenced by what the SR system favours. That being said, longer patterns in general are more difficult (probably) just because of density.

There’s also been a small debate about whether or not we should buff speed maps. Personally, I don’t have any problems with this; take a look at the amount of people capable of FCing the end of Galaxy Collapse or TLDNE. As I don’t really have an argument for this, I welcome any in the comments. My only other piece of information is that whilst speed may be the argument, stamina is important too; 300bpm is significantly easier if it’s bursts versus streams (now to just argue for how a stamina value would work, if it would even. The short answer is probably No because you can pause sometimes without consequence; long answer is Maybe for reason you guys can discuss if interested.) Alternatively, concepts like speed and stamina are irrelevant because we can just look at density, but then we’ve got a 1/6 and 1/8 problem again (how fun.)

ARGENTINE DREAM has a list of all ranked 7*+ maps and arranged them into appropriate-looking categories based on their actual difficulty (https://osu.ppy.sh/forum/t/747516); I can agree with it for the most part and I think it would come off as a useful point in showing what we as a community value as difficult.

Performance Points

So PP scaling seems to be a bit weird; it’s easy to get high pp on lower stars, but harder to achieve that on high stars (sort of like x=y2 for x≥0). Please make this more balanced? Not particularly sure what the reasoning for this is but I have an idea; later on I’ll discuss OD and why it needs a PP change along the lines of that lower diffs generally have lower OD, which is kind of overweighted.
Until some SR changes are made, it feel that it would be otherwise hasty to offer suggestions for PP changes, if any are needed.

Converts
They’re not Taiko maps; we have a ranking criteria with guidelines and a team of dedicated nominators for a reason. Converts are good for when a mode has no maps, but Taiko has existed for years and we have a heap of maps now. Please do not make these worth anything. I’m fine with them existing to have leaderboards, but like the Loved category that is all they should have (not even give bonus PP). As for other gamemodes, they can feel free to decide this for themselves.

Modifiers

A couple of things to say about mods, including the state of them as is, whether we need/want them or not, and what ones are impacting the game the most and how to solve the issues they create.

Readability Dilemma
On one hand, HR is almost entirely useless because it changes reading but awards no pp for it (the only PP increase awarded is based on the OD increase); similarly, there is nothing in the PP or SR algorithm that factors in “readability”, yet HD is a mod that purely affects readability and it does award a PP bonus, showing a lack of consistency. One method is to continue as is, with readability having no importance, and make HD yield no bonus; however, doing so would also make HR useless as the game might as well just have adjustable scroll (like Mania) and maybe adjustable OD (like Malody, or maybe HR isn’t totally useless by just increasing OD, again like Mania, but that’s kind of boring.) The downside to this is that, for one, less PP, and two, reduced motivation for the playerbase to use modifiers; this sort of ruins how we structure tournaments. Only other option then is to make readability matter in the PP algorithm; how that is achieved, beats me. It is possible to award a sort of bonus for HD versus low scroll and HR versus high scroll, but the exact amounts are subjective and up-to-debate ie. how much does the bonus give, when do we reach the point of memorisation being required (which ruins the whole idea of the bonus), and how much extra will be awarded versus how long the map has this reading difficulty (since SV changes). Opinions welcome. You may also wonder what such a change would do to gimmick maps; all I have to add is that they’re not rankable so it probably doesn’t matter (but would be nice if it worked for them too).

EZHDNC
Yeah it’s broken, but why? Well let’s look at the mods individually… We can rule out NC as being broken considering it’s the same as DT, which works fine. EZ is relatively broken because of two reasons: firstly, it makes patterns easier to hit by the abuse of timing windows, which in a sense reduces the difficulty of a chart (slightly), and secondly, EZ allows for greater accuracy (generally) than without (essentially EZ isn't broken, it's the OD that is.) HD acts as a counterbalance to the PP drop from the OD decrease. Again, HD is only broken if the algorithm does not consider readability. Two solutions to the EZ problem: one, we scale PP versus OD more harshly so that lower ODs are of a greater penalty, or two, we tie OD into SR. The issue with both of these changes is that high bpm maps start to reach OD9, which is kind of a necessity, but then we’ll just be facilitating more PP from OD literally just because “it’s fast” and not because “it’s harder to hit.” In essence, from playing maps like TLDNE I have learnt that if you’re not accurate on fast maps, you generally miss, so high bpm leads to high OD, but results in high accuracy (that’s a lot of PP boosting and not sure it’s particularly right.)
So the decision regarding OD that I present is this: should we scale OD more harshly, or should we incorporate it into SR? Arguments welcome, but the one I have settled upon for myself is as follows:
* OD scaling more harshly is good because it solves the problem of EZ making maps slightly easier, and reduces the strange PP spike that is awarded on lower difficulties (because they too have low OD). Furthermore, fast charts are hard enough and we would only be reducing the lower OD values, and not affecting the higher ones (if you were to increase the higher values slightly though, then I wouldn’t mind; it gives more weight to HR and makes speed maps less cheese-able.)

HDHR
I think it goes without saying, but make this option open to the playerbase; in its current state it is like FL and is mostly unusable (congratulations to the few that possess the insane reading ability to use this combination.) It works in v2 perfectly fine, and worked for about an hour with v1 back in 2017, so just implement it already. Furthermore, making v1 HDHR act like v2 HDHR may discourage dualscreening.

FL
As for FL, it is generally agreed upon to be unusable; only a few players can use it consistently (I’d say it’s harder than v1 HDHR). Making FL give no PP does not hinder the playerbase’s attitude towards mod use for competitive play, and also will discourage dualscreening. Whilst some people worked hard to remember an entire chart (myself included with Maniera and Black Lotus), it’s a worthy sacrifice. Best case scenario, keep the score bonus for the mod, but that’s it.

Alright so that’s my lengthy summary of issues and potential solutions. If anything above is flawed (which is likely) then feel free to tear me a new one. If there’s any good or even better ideas (I’m sure there are) then let them be known. Do remember that this thread is made to improve the Taiko ranking and competitive experience; we’re not here to nerf everything, nor are we here to give ourselves 30kpp. Again, no comments of a defeatist attitude, I know the Taiko community is better than that; we can make this better and we will.

Cheers,
Jaye
DarkStoorM
A small note on HR, when a map has base OD of 9.7 (IDK if this can actually happen on ranked maps), then Hardrock gives literally NO pp - only below this value, where Hidden gives 2.5% (affecting the strain value) AND additional 10% on top of that. Like, come on...

About converts, I agree, it would actually be good to not include them to the performance, just like TAG maps on standard.

EZ should also affect a flat pp drop. Currently, only NoFail drops pp by 10%, EZ is only reducing the OD, which is not enough. The greater the OD, the bigger penalty for EZ. If OD 0 ever happens on a ranked map, it will be completely destroyed by EZHDDT since EZ has no penalty on OD0... that's really bad.
An SS on 9 stars map, with OD 0 and 1700 combo gives 500pp, adding HDDT gives 606pp - EZ has literally no effect on pp, it just gives readability and accuracy boost. I hope that will NEVER happen.

Let's just make mods have no multipliers, haha.
--------

Doublets are really weird. Monocolor 1/8 doublets are extremely easy to hit for DDKK/KKDD, it's a free SR boost, and I guess the same applies for KDDK when maps have alter color doublets. I feel like the Star Rating boost for doublets should only be applied when the spacing between them is really small. Otherwise 1/8 doublets feel like playing 1/4. You know it's true.

Shame on people forcing 1/8 to push boosted SR to the rank.
Prinny God
I do not know what it might be or how it might be implemented, but might I recommend we add a unique modifier to the gamemode? We have a long history of TnT games to draw ideas from. (Disclaimer: never played a TnT game)

Back to the debate as listed, I think that a good start would be to take readability into account somehow in the SR system. This would necessarily imply that SV changes throughout the map need a set of draconian rules to help prevent this new system from being exploited, but I think it might be able to work with enough patience in making sure it stays balanced. Furthermore, going this route might provide an opportunity for a new modifier. Say, someone wanted to add SV gimmicks to their map. If SV gimmicks are banned completely from maps as a result of them being "free pp", then one might consider adding a unranked-category mod that allows someone to play the map as the creator intended it to be. The SV changes would be removed from the map by default, and one would play the map at the base speed only. Conversely, if SV changes are retained, probably in a relatively nerfed format, for ranking criteria, one could add a negative mod that removes SV changes and has the player playing the map at the base value, with a fixed pp % loss on top of the changes to the SR that result. I hope all of what I just said makes sense...

Some of the bugs in making SV factor into SR, aside from SV gimmicks that can inflate SR, would be the way mods altar the SR. It would help further nerf pp yield from EZ but buff HR pp. No matter how balanced we try to make the change to the SR calculations, someone's gonna have an issue with it. I wanna make one more suggestion here that can be taken with or independently of everything discussed above. Nerf HR to 1.5x SV instead of 2x. A SV nerf for HR might have already happened, plz don't kill me if it has. I only suggest this because of how insane the SV gets with HR at high speeds.
BabySnakes
I don't think you can account readability into the difficulty since it's extremely subjective. I can't even imagine what people would come up to inflate the system with ninjas, overlaps and other bullshit. And if scorev1 HDHR is getting fixed, it should be scorev2 but slightly easier since Scorev1 HDHR is way easier than scorev2 HDHR if you use custom resolutions.
goheegy
yeah i dont think there's a world where sv readability can be well accounted for in sr/pp without causing problems or being easy to abuse.
for hdhr, even if it were made slightly easier and limited to 16:9 (or 4:3 like v2) like how hd is limited to 4:3 i think that would be an improvement over essentially having to config edit or find some workaround to get a custom widescreen res just for hdhr to be playable.
as for od affecting sr, there would be no reason map with anything <7 and i think it's fine the way it is now. to nerf ez(hddt) it should either have an additional inherent nerf to pp or scale the od to pp more harshly as you said imo.

also yeah there is no reason for converts to be in a ranked/pp-giving state at this point when they don't follow rc and a lot are just broken for this mode.
Beat43210

Ney wrote:

just remove the entire gamemode

boom fixed


thanks for wasting your own time and contributing nothing to the conversation.

Anyways on to the REAL issue which is the ideas Jaye brings up.

Alchyr will continue to stress test and break his own algorithm to see how its going. I've talked to him every once in a while hopefully, he might be able to address some of the more tech heavy sides of this conversation.

Why the hell some patterns are classified more difficult than others? what might be difficult for one playstyle probably is easy for another. Why is that even taken to account. All i see out of that is a system where it'll create double standards.

As much as i personally hate speed maps, they shouldnt be nerfed. i hold the view that speed maps should be left untouched since it isnt really a pressing issue.

In regards to PP changes, i would like to see an system where a small PP boost is added to the value of a map if the map itself contains enough velocity in the changes in scroll speed of notes. maybe you could assign changes in speed, like ninjas and speed ups and slowdowns as some value and once it hits a certain threshold it'll activate a small, small bonus into the maps final original PP value calulation, how that would work is beyond me but thats my proposition to resolving SV matters.

CONVERTS SHOULD FUCKING DIE AND JAYE'S ARGUMENT IS VALID, nuff said, if needed and questioned i'll give my own reasons on top of his.

The readability dilemma: i think a way to solve it would be to remove all current PP bonus methods in regards to HD and HR mods. Then, simply, let the PP value calulator run it's values on whatever the map is as a NM map. Afterwards assign maybe a 1.06x multiplier on that final PP value if either HR or HD is played. ie. if a PF of a map gives 300pp, a PF with HD or HR of the same map would give 318PP. I argue that this would be fairer for all players since players who dont care about mods dont feel punished for not using a mod while the top players feel rewarded by adding a mod and pushing themselves.

The EZ solution Jaye proposes (to scale OD more harshly) is sound to me. but im not too sure how to handle that situation

HDHR should work like V2, nuff said. This isnt really even a debate

FL SHOULD DIE. legit, what fucking use does it have to ANYTHING. It's not even a mod used at normal tournaments. Top professional players don't use it (and dont even fucking suggest to me that we should include FL in TWC). normal players are too shit early on to even use it properly, and not assigning any pp to it doesnt even hurt anyone. Like, bruhhhhhhhhh why is it even in the game!
vrnl
hi

as for someone who can play hdhr v1 decently, i dont think it should to be changed. i think its seen as a cool niche thing in taiko and making it v2 instead will decrease its status as something to look at and be like "wow he played this hdhr?" v2 to me is just hd with slightly increased spacing between the notes, and isnt a real challenge

and as for converts, as someone who literally started taiko from converts, i might be a LITTLE biased when it comes to pp and all that. but taking away bonus pp seems too much. if u like to SS farm Normals and that sort then whats the harm in gaining the slightest amount of bonus pp from just playing the game?

[of course im hella biased and would love to keep pp on converts because i suck so bad i cant get pp any other way lolololdfkjlf, but i see the small minority say that they shouldnt give any pp at all which sucks to me. but im the type of person who want to negotiate this instead of just getting rid of it all together]

for FL... who cares Lol theres maybe actually 2 people who even dualscreen at all (and who get any pp from it), this argument is so weak. give me real evidence of this claim and ill change my mind on it. keep FL pp because memorization is a skill that should be shown and granted where its due.

everything else idc about; i cant speed, dont care about the SR, readability is interesting but i also dont care about the skill of hr that much. i only use hd because nomod is too dense on the screen for me to read properly. i use hr in the same sense as hd sometimes when hd becomes too dense with low scroll stuff.

those are my ops but idk if it matters
Edgar_Figaro

Beat43210 wrote:

The readability dilemma: i think a way to solve it would be to remove all current PP bonus methods in regards to HD and HR mods. Then, simply, let the PP value calulator run it's values on whatever the map is as a NM map. Afterwards assign maybe a 1.06x multiplier on that final PP value if either HR or HD is played. ie. if a PF of a map gives 300pp, a PF with HD or HR of the same map would give 318PP. I argue that this would be fairer for all players since players who dont care about mods dont feel punished for not using a mod while the top players feel rewarded by adding a mod and pushing themselves.


This I don't think is the right solution to the HD & HR situation. The problem with this is that HD's difficulty is all visual while HR has a visual and OD component to it's difficulty but only awards based on it's OD component.

While I do understand that making SV matter in PP would potentially break the system I think there is other ways that we can approach this. The reason I am opposed to both mods just multipling the PP value by 1.06x is that it completely ignores how HR works. Sure a flat multiplier is fine for HD as it's a visual mod and if we ignore SV that's all that we really can do. The issue with doing this for HR is that a map with low OD is easier to ACC on HR than it is with HD while a high OD map is harder to ACC with HR than it is with HD.

So the fact that it is based on the map OD is somewhat fair. The main issue imo with the current system is that HD bonus never stops increasing while HR does. The reason for this is that the OD scaling isn't based on the SR as Jaye Mentioned.

For instance if you have a 1000 note song at OD6 (8.4 for HR) and get SS. Just to show some values for HD & HR comparison at different SR.

5* HD = 260.91
5* HR = 276.88

6* HD = 322.42
6* HR = 330.56

7* HD = 396.04
7* HR = 395.01

8* HD = 481.72
8* HR = 470.2

See the problem? They don't scale linearly. HR is worth more than HD at low SR while HD beats HR at high SR despite no change in ACC, Note Count, or Map OD. Personally I am of the opinion that OD6 is the breakeven point where HD & HR should be identical PP for SS play. Lower OD maps should award more HD than HR while higher should award more HR than HD. But regardless of the SR at around OD6 they should always be the same. I don't really have an idea for to to do this but the fact that one mod is affected by SR and the other isn't has always bothered me.
fieryrage
why remove converts though lol
-Kazu-
Please let's keep this thread civilized.
tatatat
I agree with the convert suggestion. Converts should not be ranked, and as an extension hybrid mapsets including osu! and taiko difficulties should have a full reasonable taiko spread. No muzukashii/oni hybrid spreads for osu!/osu!taiko mapsets. The converts are bad.
fieryrage

tatatat wrote:

I agree with the convert suggestion. Converts should not be ranked, and as an extension hybrid mapsets including osu! and taiko difficulties should have a full reasonable taiko spread. No muzukashii/oni hybrid spreads for osu!/osu!taiko mapsets. The converts are bad.


you would effectively be killing 90% of taiko’s content by doing this

I really don’t see a reason to remove them since they do literally nothing at the higher levels sans like two maps which can be fixed with a pp adjustment anyways, kind of seems pointless lol
abraker

Jaye wrote:

Converts
They’re not Taiko maps; we have a ranking criteria with guidelines and a team of dedicated nominators for a reason. Converts are good for when a mode has no maps, but Taiko has existed for years and we have a heap of maps now. Please do not make these worth anything. I’m fine with them existing to have leaderboards, but like the Loved category that is all they should have (not even give bonus PP). As for other gamemodes, they can feel free to decide this for themselves.
This will likely be rejected by peppy.
Topic Starter
Jaye
Completely missed this idea but it's been proposed to fix the visibility of HD when combined with mods that change scroll.
This would mean more visibility for HDHR and less for EZHD, so that would be a fix and a slight nerf still need some OD adjustments but yeah.

Also as mentioned by Edgar, we weight HD and HR the same competitively (same scoring multiplier in v1 and v2) so it would be fair to give them equal weight in some fashion when it comes to pp worth; OD6 is a good middleground since it's probably the most common and about the same difficulty for both. Subjective stuff but gotta pick a middleground somewhere, might as well be there.
Beat43210

fieryrage wrote:

tatatat wrote:

I agree with the convert suggestion. Converts should not be ranked, and as an extension hybrid mapsets including osu! and taiko difficulties should have a full reasonable taiko spread. No muzukashii/oni hybrid spreads for osu!/osu!taiko mapsets. The converts are bad.


you would effectively be killing 90% of taiko’s content by doing this

I really don’t see a reason to remove them since they do literally nothing at the higher levels sans like two maps which can be fixed with a pp adjustment anyways, kind of seems pointless lol


Converts were not made by people - FACT
Converts were not designed by mappers for taiko gameplay - FACT
Almost nobody plays converts in tournament play - FACT
Non-convert maps are generally MUCH higher quality than convert maps - FACT
Convert content is overall garbage - FACT

Should we delete them/make them unplayable, no but to allow PP to be collected from maps that doesnt even take into consideration taiko's ranked criteria is complete bullshit. In fact i propose a new reason why converts shouldnt give PP

They can bypass taiko's ranked criteria
Please sign in to reply.

New reply