Hello everyone,

I've got some initial difficulty lists to show off here: https://docs.google.com/spreadsheet/ccc ... 0VGc#gid=1
This spreadsheet does not contain every single beatmap, only the ones I selected for testing. Please give me feedback and let me know which version of the algorithm works the best.

Why is that ?

Edit: I get it, you're trying to place a new taiko ranking system. However, the calculation of difficulty is yet to be defined as it's not perfect, which would be a factor of performance (patterning + speed), accuracy too. LunaticP started on such a calculator, but it's not 100% finished.

Three parameters (at least, for me) to take into account into the difficulty of a map:

Speed, Accuracy: Nothing to say about these ones, you know'em too well.

Patterning: The core of the difficulty of the map. A longer a pattern is, the tougher it is to hit. As complexity goes, difficulty drastically increases.

_Gezo_ wrote:

Why is that ?

This thread is mainly for the Taiko people to leave suggestions on how the future system should work. I already have a solid plan of what I plan on doing, but I'd like to hear the community's voice.

Posted (edited actually) what I would see on the new pp system.

Taiko maps should weight much more than a converted map does, though.

Converted maps should not be counted as we have enough taiko specific map IMO, or at least it should give a strong diff. malus comparing to taiko specific ones.

For the difficulty calculation itself, the main parameter is the density of the map (note/length), but you should also adapt this regarding the number and the lengths of streams each map contains.

EDIT : Also DT mod seems to give A LOT of bonus in std mode atm, but in taiko, I personnaly think it should give a little less, as it's way easier to DT an easy Oni even at 180+ BPM than FCing a nomod 180BPM map with a lot of streams.

Converted maps should count as there are some maps which are incredibly hard to even pass nomod.

About difficulty:

1) Note/s - Full map of 240 1/4 stream and map with clusters 1/4 300 bpm is about the same same difficulty at least for me, assuming I practiced both streams and high speed maps.

2) Maximum density of a certain amount of notes, for example if a map have an 1/6 1/8 parts of 240 bpm well, a person who can stream that is incredible so even though it's a small part of a song it should count, maybe not much but should.

3) SV should count as well - less time on screen - harder to react

4) DT>HR>HD>FL I think there is no need to explain this?

5) For example DT'ing Big Black requires much more skill than playing any other 180 bpm stream song even with HR. So don't overrate streams please.

6) Best taiko players in the world in no particular order: Orukaa, -applerss-, PM____, -[ ix Ishida xi ]-, zx_baka_0502442, Xay, kusaannanora
Please take a look at those people, they are the best of the best so if they are not in the top, well, you system sucks balls

Nashmun wrote:

Converted maps should not be counted as we have enough taiko specific map IMO, or at least it should give a strong diff. malus comparing to taiko specific ones.

For the difficulty calculation itself, the main parameter is the density of the map (note/length), but you should also adapt this regarding the number and the lengths of streams each map contains.

EDIT : Also DT mod seems to give A LOT of bonus in std mode atm, but in taiko, I personnaly think it should give a little less, as it's way easier to DT an easy Oni even at 180+ BPM than FCing a nomod 180BPM map with a lot of streams.

The difficulty spike in standard from DT exists due to differentiating between singles in streams. Since in Taiko you can use both hands in all circumstances, DoubleTime will indeed be less of a factor.

EBAWER123 wrote:

Converted maps should count as there are some maps which are incredibly hard to even pass nomod.

How do you tell which one is hard ? If the standard map has a lot of ddddddddd streams and OD8 then it will be considered hard even if it's actually quite easy ? The risk of having easy-mid converted map giving too many pps is too high considering the number of really interesting converted maps. Also we do have enough taiko specific maps to base a ranking system on it.

EBAWER123 wrote:

4) DT>HR>HD>FL I think there is no need to explain this?

Yes pleas explain. HD is actually harder on slow BPM than HR. (And I'm an HR-only player)

EBAWER123 wrote:

5) For example DT'ing Big Black requires much more skill than playing any other 180 bpm stream song even with HR. So don't overrate streams please.

You're overrating speed. There are a lot of players capable to DT regulars oni @180BPM+ who cannot even pass 14-15* Firce's maps at regular BPM.

EBAWER123 wrote:

6) Best taiko players in the world in no particular order: Orukaa, -applerss-, PM____, -[ ix Ishida xi ]-, zx_baka_0502442, Xay, kusaannanora
Please take a look at those people, they are the best of the best so if they are not in the top, well, you system sucks balls

Ishida is indeed one of the best players here. But she doesn't make incredible ranks on ranked maps, so I don't know how she could actually be placed in top #10, even with the most accurate system in the world ? If a players doesn't rank online, I don't know how you can expect him to be on top of the ranking and judging its accuracy by that is not a good way.

Tom94 wrote:

The difficulty spike in standard from DT exists due to differentiating between singles in streams. Since in Taiko you can use both hands in all circumstances, DoubleTime will indeed be less of a factor.

Ok, thanks for the explanation.

Generally, the difficulty of a Taiko map can be determined by two basic factors

- Note density
- Color changes

The note density can be easily termined by considering the time between notes in milliseconds. (this ensures that 1/4 snapped objects on low BPM are not considered as really hard). As higher the note density, as harder the map.

Color changes is the amount of changes between don and kat. Generally said, this is hard to describe. Patterns like kkk k kkk k kkk k are way harder than ddk d k ddk d k, because of exhaustion effects, whereas a ddk triplet is easier than a kdk triplet. When we come to longer streams (above 6 notes or higher), a very high color change increases difficulty, but also a very low one. (e.g. kdddkdddkdddk is easier than kddkdkkdkkdkk or ddddddddddk)

Finishers can be also hard, especially when the previous and upfollowing notes are of the same color. ddd D ddd is harder than ddk D kkd. Finishers at the end of a stream are easier, when most of the previous notes were of the opposite color (e.g. ddddK is easier than ddkdK).

Doubles are hard to get if mixed together with triplets or quadruplets.

Every pattern that is a combination of xxy/xyy patterns is easier to play than patterns of a xyx/xxx combination. (whereas x is color 1 and y is color 2) (kddkddkddk is easy. ddkkddkkddkk is easy. ddk is easy. dk kd is easy. kd kd is harder. kdkdkkkd is harder. kdk is harder. dddk is harder than ddk

Let me give out a few examples with maps working in various ways to create difficulty (it could be that many of these maps are my own, but I just know them the best, so sorry to appear biased):

https://osu.ppy.sh/b/171025&m=1 a quite simple Oni, mainly working with triplets (easiest "hard pattern" in Taiko) that are working in a xxY fashion. (ddk oder kkd are easier than dkd and kdk)
https://osu.ppy.sh/b/262458 finish overusing map. Please consider that the finishers at the streamends are painful and also the 1/2 quicktapping part with finishers inbetween is really difficult. It also provides the above example of kkk k kkk k kkk stuff that is really really really painful.
https://osu.ppy.sh/b/284037&m=1 Working with longstreams (many color changes here), being an eggpain to play. If you drop once you are out. Out of everything.
https://osu.ppy.sh/b/263683 Using doubles in a fancy fashion, a decent Oni level (Insane level). The hardest part for me is the dddddddk dddddk ddk thing, because its hard to keep up streaming the same color without switching to kat inbetween.
https://osu.ppy.sh/b/169371 the monotony of this patterns makes your hand complain. I assure you.

Anyways, Taiko's skill is basically based on tapping speed and reading skill. BPM is highly correlated to skill.

Moreover, HD and HR are equal, DT increases OD massively. HD players are having more problems on lower BPMs (due to low scrolling speed), whereas HR players are having more problems with higher BPM (due to HR causing a too fast scrolling speed). DT doesnt change scrollspeed based on SV but on BPM, but it decreases the hitwindow massively. HD+HR is impossible to read. FL+HD is memorization (you wont see ANY note at all), FL+HR is also memorization, but objectively easier (since you can at least see the notes, even if its just a short while).

Basically I see it like (vision based mods. HD/HR/FL only affect vision): (HD = HR) < HD+HR < FL < FL+HR < FL+HD < FL+HD+HR
DT is an extra factor, increasing OD and BPM and therefore should be calculated into the difficulty where note density calculations happen.

Please comment.

Some factors I can think of:

1. Pattern type, e.g. ddk vs ddkddkddk
2. Pattern repetition, e.g. ddk ddk ddk ddk
3. Pattern consistency,e.g. ddk kkd dkd dkk
4. Pattern speed, more or less falls under BPM
5. Mods
6. BPM
7. Accuracy
8. etc.

As others before me already said, speed and accuracy are pretty basic factors that you should have a decent understanding of even ig you don't play the mode (it may be relevant that the taiko hit judgement formula is a bit more strict than in standard, not sure if your algorithm needs the specifics. Link to the table in any case: here).
Pattern/reading skill is probably the hardest part to measure, so I'll give all the input I can give on this topic.

1) SV - Extremely high/low SV is hard to read, but should probably only give a small bonus since DSing is a thing and not provable =/

2) Mix streams - Streams get a lot more complicated if you mix in clusters of 1/3 or 1/6. Not only are they more rhythmically challenging, but they also require advanced hand switching if you fully alternate (which almost all highly ranked players do).
In case you are not entirely sure what I mean, a ddddkkkkdddd 1/4 stream is really easy since the leading hand hits every white tick while the secondary hand just plays blue tick beats; a dddd[kkkk]dddd stream (square brackets signifying 1/6) however requires a leading hand switch on the second dddd part since it starts on a blue tick.

3) Pattern complexity - As a rule of thumb, the more repetitive a stream, the easier it is. Streams are especially easy if they only really change on white ticks. For example. dkkkdkkkdkkkdkkkd... is very easy, so are ddkkddkkddkkd... and dkdkdkdkdkdkd...
dkkkddkkdkkkddkkd... is a bit more advanced, but still very basic.
More complicated streams implement kddkddkddk or kkkdddkkkddd style patterns - the difficulty here stems, again, from hand switching. While you don't need to switch leading hands offbeat like in the 1/6 example, the "sub-patterns" still force minor leading hand switches (both kddkddk... and kkkdddk... are actually non-repeating 6-note clusters even though they look like repetitive 3-note patterns).
All that being said, a pure kddkddkddkddkddkddk... stream or similar structure isn't extremely hard due to its repetitive nature, it only really gets tricky once you mix it with easier patterns, thus forcing irregular hand switches.

4) dk kd kd kk dk dk dd dk kd doubles spam and similar shenanigans can be really hard to read and play, at least at high BPM. Again, the more irregular the composition, the harder.

5) ddddddddddddddddkdddddd... (even number of d, then k) is intuitive and easy to read/play.
dddddddddddddddkddddddd... (uneven number of d, then k; k on blue tick) is really unintuitive and most people will probably miss here on a sightread attempt.

EBAWER already listed some of the best players, so I'll go ahead and give you a few of the hardest ranked taiko maps to check your diff calculator with (also no particular order):
http://osu.ppy.sh/b/179231&m=1 I'm Your Daddy [Fatal Oni] - Unrelenting high speed map with tricky 1/6 clusters
http://osu.ppy.sh/b/92643&m=1 Shinsekai [Taiko] - These streams make absolutely no farking sense, and they last for a long time
http://osu.ppy.sh/b/261205&m=1 IMAGE -MATERIAL- <Version 0> [Firce777's Taiko Oni] - Several long, somewhat complex 260 BPM streams
http://osu.ppy.sh/b/192573&m=1 Because Maybe! pt. 3 [Taiko Collab] - Stamina Killer #1. Patterns are pretty basic, but the speed and length make it incredibly hard to FC/get a good combo
http://osu.ppy.sh/b/245124&m=1 Distorted Lovesong [Taikocalypse DX] - A few of the note clusters towards the end of the map are just stupid at this BPM

If you have a good score system (algorithm) you do not need to exclude converted maps because they give points relative to their real difficulty, so there is nothing to fear about i think.
Personally i would miss something if converted maps don't give any points because compared to the whole osu mappool the taiko mappool is pretty small and many converted maps are still fun.
If a taiko specific map is actually harder to play it should give more points of course, but not more points for "no reason" just because it's taiko specific.

However i'm very excited about the new Taiko Ranking System and want to thank you, Tom94, for your awesome work so far.

Greetings,
NoYzE

I'm rather neutral to this but have to agree with Nashmun when considering the facts we have. Our gamemode is well established with its specific maps and could work even without the converted maps. Moreover would it be pretty hard to state what converted map is hard or easy. The times when Taiko was only a mod for farming points is over, so should be probably the pp for converted maps, I think. We can keep it for fun, but have it for ranking is rather questionable... again, we have specific maps and even sets now. Kantan/Futsuu can take over the job of the converted osu-taiko maps.

Currently the ranking is better than before I would say. It seems that taiko specific difficulties give a higher amount of pp than converted osu-taiko maps, which is good. In most cases specific diffs are harder anyway. - It is obvious that we have people who would deserv higher ranks in the ranking, but these people mostly play offline maps and have only here and there some awesome ranks which can't conver the amount of ranks of a player who played mostly all taiko maps in this community.
That's the reason why I would say that current rank from kanopu #1, Gezo #6, Nashmun #8 or me #16 is -somewhat- right, since probably all of us played all these years mostly specific taiko difficulties, not converted. I can't tell about the others for sure, but in my case it would apply in this way.




Also I have something else to add here which could help in some way. I'm not sure how you want to make the new pp system for taiko, but I always thought that it is important to know what difficulty patterns have, and how their diff. rise when putting them together to a stream or something else. As gezo said, Lunatic made a programm which contains these factors and more. Maybe you wanna give it a read (t/177867), but for now have these links which contain my ideas from the past (Rokodo was so nice to write it down):

http://puu.sh/4e0d9.png (graph)
http://puu.sh/4e0hg.png (graph)
http://puu.sh/4dZLK.png (Star System pic)
http://puu.sh/4eB03.pdf (Star System explanation)


@EBAWER123: HD is mostly equal to HR. It only depends of the map's BPM which makes one of the mods harder/easier. You can say that HR makes OD harder, but for that you gain more points for spinners.

*moved topic to Announcements

@Nash Just by default even if we do not consider reading problems with HR/HD, ie HD ultra slow SV and HR ultra fast SV, HR requires more acc, that's all.

Converted maps which actually require skill:

https://osu.ppy.sh/p/beatmap?b=199304&m=1
https://osu.ppy.sh/s/55560 both diffs
https://osu.ppy.sh/b/95382?m=1

etc...

Thank you for assuming I'm stupid. I know how to recognize a hard converted, but it's hard/too risky to let an algorithm make the difference.

OnosakiHito wrote:

Flashlight
Should be disabled or give no points. Too often there was a problem with this mod.
Most people do not play it and it has been often abused (HD+FL+HR). That's my opinion.

This, flashlight may be an challenge to the player but shouldn't influence in the ladder at all, people are just abusing of it to raise pp.

imo, most of converted stuff is "hard" just because the high SV, OD and maybe the HP, many of them a're basically d d k d d.d k d d d k ... or a d k d k spam.
I haven't played converted stuff for a while, so I may be wrong, but I do agree specific difficulties should have more weight.

OnosakiHito wrote:

Currently the ranking is better than before I would say. It seems that taiko specific difficulties give a higher amount of pp than converted osu-taiko maps, which is good.

Wat.
Are we looking at the same stuff?

Let's play a guessing game, okay? Guess whose top performance list this is:

He didn't play since the update, his scores are not updated yet I guess.

Comment to Loctav

1. SV is also an elementary factor here, especially when playing with mods. The scrollbar is after all the playfield.
2. Well, you showed now some patterns which might be hard/easy, but the question is rather which play-style do we considere. Is it ooxx or xoox ? This is important to be able to determinate what patterns are hard/easy. So if you ask me, the authentic style (xoox) should be considered in the calculation, since this is after all a game based by a drum which is played in this way. So are oxxoxxo really hard patterns since the alternate way of them is really complicated.

1. As I stated above, I think FL is not really needed here.

Comment to Luna


To point 3.) You are mostly right. Just let me mention that it is easier to play patterns which have more main notes than semi notes?
Easier: kdddkddk
Harder: dkkkdkkkd

Everything else I would personally agree to. The more irregular a pattern becomes, the harder it is to play.

Nashmun wrote:

Thank you for assuming I'm stupid. I know how to recognize a hard converted, but it's hard/too risky to let an algorithm make the difference.

I wrote in my post that the algorithm should take into account clusters of high density notes, like in forgotten: 250 1/8 streams, if it thinks that's an easy stream than that algorithm...

EBAWER123 wrote:

Nashmun wrote:

Thank you for assuming I'm stupid. I know how to recognize a hard converted, but it's hard/too risky to let an algorithm make the difference.

I wrote in my post that the algorithm should take into account clusters of high density notes, like in forgotten: 250 1/8 streams, if it thinks that's an easy stream than that algorithm...

The problem isn't considering a hard converted hard, but an easy-mid converted not hard. Because the settings will obviously gives EVERY converted map a boost, plus they generally have high density so once again boost. What's the big deal with disabling it ? Because chipscape would not count anymore ? That's really a big deal...

EDIT : Moreover the maps your are linking now are just stupid maps full of spam, so it's even better not including them in the ranking anyway

Nashmun wrote:

Because the settings will obviously gives EVERY converted map a boost, plus they generally have high density so once again boost.

EDIT : Moreover the maps your are linking now are just stupid maps full of spam, so it's even better not including them in the ranking anyway

First of all converts have less notes than taiko-specific map on the same song, for example Wind God Girl has 2367/1424 notes, etc... And assuming we take a map without a taiko diff and make an Oni there, well, 99% chance it will have more?

Maps I linked are not stupid - they require skill to play, what you just said is that it's better not to include hard maps in the ranking?

Think what you want but fact is, if you include converted maps the ranking will be biaised.

And if you think that the converted maps you linked are well made, there is no point to argue anymore...

I do believe a good point was brought up:
the way people play Taiko.

In Taiko and osu!mania, the style of the game is heavily dictated by tapping the keyboard in an entirely different fashion from the regular and Catch the Beat modes. Some rhythmical patterns may become harder if a certain playstyle is adopted.

For example, take oxoxoxoxoxox...
There are two main ways you could do this: you could alternate your hands (default czczczcz...) or you could alternate your fingers (default cvcvcvcv...).

Although it's probably very minor, the stamina of hitting these notes over a long duration may become very difficult.
I should probably flesh this out more, but I lack the patience to test each rhythm over many methods, especially since I have adopted a playstyle, and bias is always present in that case.

I definitely believe that this should be incorporated into the pp in some way. Perhaps we can put an option on how we perform certain patterns into a pp calibrator? Just ideas.




As for flashlight, I think it should be worth some pp at the very least, as it takes quite a bit to memorize something. Definitely not as much as other mods!

Dodging the answer is not a very good way to have a fruitful discussion, don't you think? Just because you do not wish to accept those are very hard maps and you think that they are not supposed to be played in taiko does not make them any easier. The whole ranking is about becoming better, not hiding in the corner and playing only in your comfort zone. Playing higher BPM, increasing your stamina, your reading skills, trying new mods, increasing your accuracy; striving to be better - isn't this the whole point of having a higher rank than someone else? Of course it's easy to say that this map is too fast and it's a convert to boot but we have to acknowledge that pm___ passed forgotten with nomods and it took skill not just random spamming of buttons =) And that Orukaa has 98% on another diff, if it was just stupid spam people would not play it.

I'm in agreement with not including converted maps or weighing them FAR less than taiko specific maps.

There is a fairly large pool of taiko maps now, and considering how 95% of converts end up being completely garbage in taiko due to high SV / 1/8 spam / boring patterns, there's no real reason to include them in the ranking metric anymore.

As for flashlight, I'm still of the opinion that it should be considered a mod that doesn't give any multiplier and is more for showing off... and yes, I think this solely because of previously rampant issues regarding Dual Screening and people using it to boost with these methods.

As for patterning and whatnot, don't consider ddkk / kkdd playstyles into how hard a pattern is. Most of the patterns that are harder for kddk are hard for those playstyles as well (ddddddddddddd / kkkkkkkkkkkkk as examples). Considering that kddk is the default playstyle, difficulty should be considered around it, not other playstyles.

That said, most of what I would have said is already covered here by others for patterning and such.

I didn't say it doesn't require skill to play them, even if it's indeed mainly spamming, some score are just really impressive, and I do know those players are good. My main point was about easy-mid difficulty converted maps and you pushed the discussion on the hardest one because you think it's unfair not to include them. I personnally think it would be more unfair to make the ranking biaised because of 99% of converted maps rather than ignoring the 1% that would indeed worth being added.

Oh wow, that's some overwhelming feedback.

Let me at first clarify how the current situation is:

1. By far not all scores, have been processed, more than half is still more or less random.
2. Difficulty of taiko diffs is based purely on star rating.
3. Taiko-specific maps are worth 3 times more than converted maps.

This obviously is not a very good system as it is. The star system is arguably broken and mods aren't even taking into account.


What I plan on doing:

I'll make a difficulty calculator similar to osu!tp incorporating what you are suggesting me here. I plan on using strain values, just like in tp (see the info page for a graph) to account for note density. For color changes I am not yet sure how to tackle the problem, but I'll experiment a lot with it and let you know when I have some results.


Converted vs. Taiko-specific

I plan to test including them and present you all with a list of beatmaps together with their difficulty. If it endss up working well I don't see a reason to exclude converted maps. If it won't work they won't be included (or by a small margin).


I can't give any in-depth answers yet, since I don't quite have the time to work on anything during the next 2 weeks, but more information will follow.
Thanks again for the huge amount of information!

Oh, and I'd like to add that patterns with even numbers of notes tend to be a lot harder to play than those with an odd number of notes.

Loctav wrote:

Every pattern that is a combination of xxy/xyy patterns is easier to play than patterns of a xyx/xxx combination. (whereas x is color 1 and y is color 2) (kddkddkddk is easy. ddkkddkkddkk is easy. ddk is easy. dk kd is easy. kd kd is harder. kdkdkkkd is harder. kdk is harder. dddk is harder than ddk I don't agree with the dk kd/kd kd case when we're hitting higher BPM. I have a hard time FCing Sakura Sunrise on DT due to dk kd, but if it were dk dk I'd easily do it.

Moreover, HD and HR are equal, DT increases OD massively. HD players are having more problems on lower BPMs (due to low scrolling speed), whereas HR players are having more problems with higher BPM (due to HR causing a too fast scrolling speed). DT doesnt change scrollspeed based on SV but on BPM, but it decreases the hitwindow massively. HD+HR is impossible to read. FL+HD is memorization (you wont see ANY note at all), FL+HR is also memorization, but objectively easier (since you can at least see the notes, even if its just a short while).

Basically I see it like (vision based mods. HD/HR/FL only affect vision): (HD = HR) < HD+HR < FL < FL+HR < FL+HD < FL+HD+HR
DT is an extra factor, increasing OD and BPM and therefore should be calculated into the difficulty where note density calculations happen.

HR affects OD as well. Else, well you'd see many more HRDT scores on average BPM songs (that is 110-140 bpm range no more)
HRHD is not impossible to read, it's very tough and playable at a very limited range !
FL+HD+anything is "memorization", FL+HR is also memorization (except when BPM gets abusively low) as you have a ridiculous reading timespan.

My comments are in blue, Loctav.

I do not agree with your comment on HDHR @Gezo, it's only readable if you play with a widescreen resolution.

"very tough" and "very limited range"
You got room for literally one note even on wide (half a note on 4:3)

So, one thing I have to say about the taiko vs. converted argument.
If this is going to be a proper difficulty rating system, it shouldn't weight certain taiko maps based on whether they are converted or not, but on how difficult they are. Put your stance of taiko vs. converted away, because we are trying to sort them as whole, not two separate things. Whether it's full of 1/8 spam like https://osu.ppy.sh/p/beatmap?b=107875&m=1 , or complex patterns like most taiko maps, it's still gotta weigh as if it's still one mode.

capeseverywhere, if we try to sort them as a whole like you say, then converted maps are going to end up being weighted way heavier than they should be due to bad conversions leading into 1/8 spam (Mad Machine [Champion] in taiko is a good example of this).

I'm gonna let tom do his thing and try to make it work with both, but at the same time, if he can't, I'd rather he just focused on a good metric for taiko specific maps over trying to manage something that works for both.

capeseverywhere wrote:

So, one thing I have to say about the taiko vs. converted argument.
If this is going to be a proper difficulty rating system, it shouldn't weight certain taiko maps based on whether they are converted or not, but on how difficult they are. Put your stance of taiko vs. converted away, because we are trying to sort them as whole, not two separate things. Whether it's full of 1/8 spam like https://osu.ppy.sh/p/beatmap?b=107875&m=1 , or complex patterns like most taiko maps, it's still gotta weigh as if it's still one mode which it is.

I agree with you here.

Another point I'd like to make is regarding certain patterns being ranked as harder/easier. Let's say we have two players, both ranked around #10000. Where Player X might have problems with ddkdd at 180 BPM, Player Y might not be able to do kkdkk at the same BPM. Neither pattern is necessarily harder than the other, but obviously Player X and Player Y are scoring well on beatmaps the other can't. The players who can do both patterns however, are the ones who should be ranked higher overall.

I believe it would be fairly easy to group certain patterns together based on the logic used in the above example. These groups of patterns would be initially ranked by how tricky they are to play at certain BPMs. After this, we can use a plethora of options to see which ones within the group should be worth more. Some possible options could include popularity of certain patterns over others in the group, overall accuracy/fail~pass rate of songs that include these patterns, etc.

The groups themselves would be something like
(ddd, ddk, dkk, kkk, kkd, kdd)
(ddddd, ddddk, dddkk, ddkkk,dkkkk, kkkkk, kkkkd, kkkdd, kkddd, kdddd)
(dd dd, dd dk, dd kk, dk kk, kk kk, kk kd, kk dd, dk dd, dk kd)
and on and on.

How we would rank the groups would be up for debate.

Tasha wrote:

I'm in agreement with not including converted maps or weighing them FAR less than taiko specific maps.

This.

Everyone already put many reasons already, however I just think they are broken, especially a certain convert with 2 circles and sliders for the rest, where halftime actually makes you rank higher (unless you are a damn beast hitting all the rolls at that bpm). As Nashmun said, it's risky. Just because it's hard or even fun as a convert, it's not meant to be played in taiko, so it shouldn't really be counted that much.

As for the patterns, to make it more simple, during streams, whenever you see an odd number of same color notes between even number of notes (kdddkddkdddkddk etc), I'd say it's making it really difficult. Hand switching is a pain and cheesing the pattern requires high level reading skills. Been said before, but just putting it here since I feel the same way.

As for the SV changes, that's gonna be a bitch to calculate. There are times where SV increases actually reduce difficulty. You could get a certain speed X by having BPM Y and SV of Z. The problem here is everyone is different, and some people will have it easier even if it's super slow, but you will have people who will miss everything cause it's slow. All depends where people look at the screen.
In my opinion SV should be just left out from the really complex calculation and depending on bpm just add some minor difficulty (unless the multiplier is very much different or something, I think fcing HR on this map is near impossible without DS or memorization).

But then again let's see how LunaticP's thing works after some fixing/feedback, if the star rating actually works properly without us arguing here on whats harder, that could be a good difficulty estimate as a base.

How about making rolls and spins not count in the pp system?
From what I'm gathering, the converted maps feature far more of these.

I think the main factors of determining stream difficulties are:
length+
speed+
color changes+
rhythm changes+ (1/6mixed in 1/4s etc)
big notes+
repetition-

Using linear prediction filters(-2 -1, 0, 1, 2 for big blue, blue, none, red, big red respectively) in 1/12 rhythm intervals applied over overlapping 1/2/4 bars of map then record the error as entropy and include it in the estimation for difficulty may be worth a try (similar to Linear predictive coding).
^^really I have no idea what I'm talking about, just thought it may be fun to do.

My take on Taiko (was around 1900 before removal)

Similar to mania, I believe that mode-specific maps should be weighed a lot more than auto-converts. Auto-converts change the map into a rhythm map, with extremely awkward timings, abrupt tempo changes, and unrealistic big notes in the middle of a stream. And again, much like mania, auto-converts don't have certain patterns which you see when you get into maps that get more difficult, for example the long streams in Shinsekai (http://osu.ppy.sh/s/24634), MEPHISTO, or variations in streams like duckling (http://osu.ppy.sh/s/28893) or Intersect Thunderbolt (http://osu.ppy.sh/s/38316).

Accuracy could and should be used as a good determinant in skill, as there is no further differentiation between 良 and 可.
Will edit with time.

Standard maps aren't made in the mindset to be challenging, worthwhile, or played in taiko, except for the handful of taikosu! maps. They are made to be played in Standard. Not to mention most standard conversions are considered unrankable because of SV and finisher usage. So they don't even fit with the criteria a taiko map should have. Essentially, they are a completely different style of play from what the community considers taiko gameplay. (as shown from the long discussion of making the taiko ranking criteria.)

So, there is no reason they should be included in the taiko ranking system.

Anyways on terms of mod scaling, DT is obviously the hardest mod. Not just because of speed, but because it adds more OD to the map as well. In my opinion HR and HD are equal. HR provides slight vision impairment from higher speeds, and a big boost in the OD. While HD provides a substantially greater vision impairment that can severely effect your accuracy, even though there is no OD boost. This is based off a theoretical player who hasn't played either mod enough to get used to the differences these mods make. We all know after a while these mods become second hand.

I can't really say anything about FL because there hasn't been a breakthrough in how to play it efficiently without having a good memory or playing the song a million times to memorize it. Something that doesn't equate to raw skill in my mind.

As far as note density goes: I would like to think a ratio of notes to song length would be accurate, but every song is different and just isn't possible without over/undermapping. So basing it off something like BPM would be more accurate, in theory.

Factors affect the overall difficulty of a taiko map

1. OD, higher the harder you can hit

2. HP, I don't know but I think this should be included. Since you need to get a pass to submit your score....but it does not affect too much

3. Stream, or just call this the note patten.

4. BPM, yes even if you have very simple note but in high bpm they should be difficult

Example : http://osu.ppy.sh/b/179225 280 BPM but notes patten is not that difficult

5. Timing points, or the bpm change or the SV multiplier. Can affect a lot if you fill the map with green line like grass...

Example : http://osu.ppy.sh/u/242910 Trust me the link is correct.

6. Mods, I think so many big guys have explained a lot on how mods affect the map so I am not going to explain

How to deal with all these factors?

1. OD, the number can be treated as a multiplier

2. HP , same

3. Stream, now this is the important point. Stream is the core part on how we rate a taiko map. We have to discuss a lot on which patten is more difficult than another patten. But I want to say taiko stream have unlimited combination, even if you say ddkkddk is more difficult than ddddkkd it still does not have meaning to ddddkkkkd. How if a map with endless stream like Shinsekai? So how to deal with this problem? I think the key is how player read the map. In Chinese there is a term that we " Crack the map ". Which we divide a long stream into small part and play them one by one. For example I divide a long stream by 2/3/4 method. Which I try to search for 2/3/4 patten in a long stream and crack it down. I don't know how others do but that's how I do. If we can come to a conclusion on how to crack the map, there is a easy way to crack down any stream and rate it. For example we all agree that we use 3/3/3/3/3 to crack down the map, we can just divide the long stream into 3/3/3/3/3 and rate each of the 3 and add them up. In the last few 3 I think we should also raise the difficulty a bit since it is a long stream. After all, if we need to talk about stream difficulty, we have to look from a player's eyes but not a mapper.

4. BPM affect a lot, but here is what I suggest. BPM 160 - 200, no change, from that 2 number go down and up, we add the difficulty index bit by bit. For example, BPM 160 = 1.0 , and BPM 150 will = 1.05 which is the same as BPM 210. Why? Because WHAT I THINK 160 -200 is usually a comfortable zone for player.

5. Timing points, am I don't really know how to deal with those grass, can someone think of a suggestion?

6. Mods, I think we have to find some expert in each mods to ask for opinion. What I know is Orukaa CROSS PM and ZX = DT Ishida = HR asgardv = PF Nardo = FL. I think every single mod should be included, HR DT HD FL, even the perfect mod but not sudden death which gives no meaning at all. Why perfect? Because of the super high pressure , just try once you will know.


Last, I am busy in last few weeks and I think I can come back on my rater program now. Thanks for waiting. However I need help the point 3, the stream patten.

I think the best method on how to rate ALL the stream by a program is we hard code a difficulty index to 2/3/4 patten and we "crack the damn map" with program. I am creating a google form for this please wait.....

Nashmun wrote:

I do not agree with your comment on HDHR @Gezo, it's only readable if you play with a widescreen resolution.

They are still readable (4:3) just like how 16:9 is. At least for me though. HRHD is much easier to manage than pure memorization just like HDFL, but you shouldn't say FL should be nerfed, since the hugest enemy FL players usually face is accuracy than the patterns itself, (also remember there's no OD change at all, and to be honest I am always amazed at people who can HRHDFL a map with over 97% accuracy) I still find HD easier than HR, maybe it's because I'm more used to HD. I have something on DT, where as long as the BPM is over <inserts BPM here, 140 for me>, extra pp should be given. I can't stand slow maps, anything below 140BPM, and I'd rather use DT with them, even with a lesser accuracy. And as for my standpoint,

DT < HR = HD < FL

As for coverted taiko vs. taiko only maps, I'd say give converted maps way lesser weights than taiko only maps. As skills increase, you can purely hit for taiko-only maps for pp instead of ranking on converted maps, as they are way easier to get ranks with. You can even get to the Top 50 even with a C rank.

As for giving weights based on patterns, how to we judge the pattern is hard or not? Different people have different playstyles, and different people have different opinions on patterns. Like me I still think dkkdkkd and kdkkdkkd is hard but dddkkkddd and kkkdddkkk is easier...

Also no matter how long the patterns are, the complexity of it still comes down by splitting them into simpler patterns.. For example, splitting a pattern like kkdddkdddkdkkkdkkkddkkd can also be split like kk dddk dddk dkkk dkkk ddkkd, which makes managing patterns easier

Tom94 wrote:

Converted vs. Taiko-specific

I plan to weight converted maps just as much as Taiko-specific maps and present you all with a list of beatmaps together with their respective difficulty. If it ends up working well I don't see a reason to exclude converted maps. If it won't work they won't be included (or by a small margin).

I like what Tom94 says here.
Everyone else is saying "no no no make taiko specific worth much more" But, in my opinion, converts should be worth ALMOST as much as taiko specific since some can be difficult. Not all converts are great but, they shouldn't be worth too little.

I just wanted to confirm lolcubes' statement that patterns with alternating groups of even and odd numbers of notes are generally hard to play (For patterns longer than 3 notes at least). That's the most specific start for an actual algorithm you could probably use. Add in something about repetition and it should already work pretty well.
Another thing to keep in mind is that repetition on blue ticks is usually harder; for example ddkkddkkd is super simple, while dkkddkk is a lot trickier to play (still somewhat basic, but definitely way harder)

I am not really sure how you want to measure difficulty of a pattern because it's silly, there are a few basic patterns to learn but not so much:

[1] d, k
[2] dd, kk, dk, kd
[3] ddd, kkk, ddk, kkd, dkd, kdk, dkk, kdd
[4] dddd, kkkk, dddk, kkkd, ddkd, kkdk, dkdd, kdkk, dkkk, kddd, dkdk, kdkd, ddkk, kkdd, dkkd, kddk
[5] all from [4] + d/k in the end

Only 62 patterns =) mania has 30 times more to say the least, not including combination of those patterns which play a much bigger role there than in taiko.

After learning to read all those patterns there is nothing more to learn about them because all streams can be divided into clusters of 1/2/3/4/5 notes, ie:

dddkkkddkkdkdkkkkdddkkdddkkkkkdkdkdkdddkdkdkkkdkdk = ddd kkk ddkk dkd kkkk ddd kk ddd kkkkk dkdkdk ddd kdkd kkk dkdk dd D

This is basic reading skill, we divide a stream into groups of usually 2/3/4, sometimes 5 or 1 if we have a finisher or a certain amount of same colour notes or we just prefer to have in mind clusters of 5 instead of 3+2/1+4/2+2+1/etc and that's all, after learning how to play each pattern, streams without BPM changes become easy to read.

Only thing that most of people hate about streams, especially on 240+ bpm streams where it takes a little bit more skill to read the stream due to the sheer speed, is even amount of notes mixed with odd amount of notes, ie:

ddddkkkddkkkddddkkkddkkddddkkkddddddkkdddkkkkddd = dddd kkk dd kkk dddd kkk dd kk dddd kkk ddd ddd kk ddd kkkk ddd (Mixed hand change)

This stream is annoying due to changing hands but again, awarding this stream more pp than these 2 is kinda biased imho:

ddddkkkkddddkkkkddkkddkkddddkkkkddkkddkkddkkddkk = dddd kkkk dddd kkkk dd kk dd kk dddd kkkk dd kk dd kk dd kk dd kk (No hand change)
dddkkkdddkkkdddkkkdddkkkdddkkkddkddkddkdddkkkddd = ddd kkk ddd kkk ddd kkk ddd kkk ddd kkk ddk ddk ddk ddd kkk ddd (All patterns w/ hand change)

Another thing that people find hard are not streams but certain patterns at high speed:

dk kk kd dk dk dk kk kd kd dk dk dd kd dk dk kd kd dk

Even though it is difficult to some people but again there are people who don't see any difficulty with this, so I don't really see the point of getting more pp when it comes to this pattern.

Another pattern-based thing is a stream with bpm changes, like this:

1/4 stream with 1/8 clusters inside but this again can be easilly calculated in your algorithm as a stream of a certain BPM which lasts for example in this map it was from 00:32:241 (302) - 00:35:321 (358) , 224 BPM base, 57 notes, of which N amount was 1/4 and M was 1/8

I am not sure which and how you will calculate this but this stream has a difficulty based on speed not really a problem with patterns. All in all I don't really see any point in giving certain patterns more "weight" than others, just calculate a length of a certain part of the song, see if it has bpm changes or not and that's all. I think it will be more fair and unbiased, because we can argue about pattern difficulty forever but it's just our opinions whether a BPM increase is always hard.

Addressing the post above:

dddkkkddkkdkdkkkkdddkkdddkkkkkdkdkdkdddkdkdkkkdkdk = ddd kkk ddkk dkd kkkk ddd kk ddd kkkkk dkdkdk ddd kdkd kkk dkdk

You can't just say that streams are nothing more than a combination of these patterns and only marginally harder than the added value of all of them.
If you split it this way, you need to keep in mind that hand switches occurr. Assuming you start the stream on your right hand, that equates to:

dddkkkddkkdkdkkkkdddkkdddkkkkkdkdkdkdddkdkdkkkdkdk = right-hand-ddd left-hand-kkk right-hand-ddkk right-hand-dkd left-hand-kkkk left-hand-ddd right-hand-kk right-hand-ddd left-hand-kkkkk right-hand-dkdkdk right-hand-ddd left-hand-kdkd left-hand-kkk right-hand-dkdk

That's a LOT more complicated than you make it sound.

/E: Fixed the right/left assignments. I messed up on a stream with unlimited time to think about it, just goes to show how easy it is to get confused on these patterns...

To be honest it looks like a color blind test

anyway actually different people may divide stream in different way, hard to say who is right

Luna wrote:

Addressing the post above:

dddkkkddkkdkdkkkkdddkkdddkkkkkdkdkdkdddkdkdkkkdkdk = ddd kkk ddkk dkd kkkk ddd kk ddd kkkkk dkdkdk ddd kdkd kkk dkdk

You can't just say that streams are nothing more than a combination of these patterns and only marginally harder than the added value of all of them.
If you split it this way, you need to keep in mind that hand switches occurr. Assuming you start the stream on your right hand, that equates to:


That's a LOT more complicated than you make it sound.

I play full alternate and I encourage all other players to learn full alt so you won't get any problems with hand switching. While patterns are a difficulty based on a person playing and his style, it is still not that much of a factor compared to SV(Speed of your eyes), OD(Precision of your tapping), BPM(Speed of your hands), Length of a pattern(Your stamina), Length of a song(Your stamina) and mods which increase or decrease the difficulty.

Something I want to say:
① OD of course higher OD means higher pp
②mod I think DT>HR=HD>FL of course. Or you can just let FL≈0.(HR+HD) should not be more than( HR)+(HD).So is HDFL and HRFL and HDHRFL
③of course Taiko only maps far more than converted maps.From pp now I think 3 times is not enough.Maybe it should be 5 timers or more.
④The difficulty of the map itself. I think this is the most important.There are many points.
(1)BPM this BPM is the REAL BPM but not what the mapper give us. For example, the big black should be calculated as 180BPM but not 360.
(2)Change of the colors. I only have a theory of calculate the difficulty of a stream.I'm trying to translate it to English.
(3)Rhythm change. 1/6+1/4 is harder than 1/4 only of course.
(4)Note density. I think this one needn't be explained.
(5)Star rate. That's really means nothing in Taiko. So many wrong examples.

if you have any questions about what I say u can ask here or just PM me.