forum

Performance Points feedback and suggestions (Taiko)

posted
Total Posts
645
show more
unknowwiiplayer

Dolphin wrote:

unknowwiiplayer wrote:

Oh, sure, sure, go play Sacrifice (Taikosaki) and then play it with HT and tell us if it was easier or not.
Bad example, then you have the same density but at a lower BPM.
A map thats 280 bpm can surely have lower density than one that is of lower BPM.
I didn't say that it's not possible The Big Black is easier than sacrifice, and the BPM are 360 vs 270 ewe
but you're taking the density considering the BPM as base or real seconds as base? ewe they're different
OnosakiHito
In Taiko BPM is beside SV the most important factor. BPM also discribes the note density by notes per seconds, which is different when comparing a low BPM song with a high one.

So does have a BPM 250 song on 2 minutes more notes than a BPM 100 song on 2 minutes.
And a monotonically pattern plays also differently on a higher BPM than on a lower one.

So wiiplayer isn't actually wrong. The ND changes. So those the difficultie of the map.
(Of course, it depends of what difficultie you are actually mapping lol)
Dolphin
Nah, don't get me wrong, I'm not neglecting that BPM should be accounted for, but it's truly not the main factor for difficulty.
OnosakiHito
I'm the opinion it is the main factor in Taiko. Every placed pattern is effected by the BPM. Same goes to the SV, which effects the patterns directly(its density) due to the BPM.
lolcubes
No, since the use of 1/8 and double SV is allowed, the BPM is no longer the main factor here. It should be the timing between the notes, the amount of notes per second and the SV.
Example: https://osu.ppy.sh/s/45205

Those things make the BPM difficulty, but it's not BPM.

As for my opinion on the difficulty, I'd say we're mixing apples and pears here. Both complexity and "bpm" create difficulty, each one on it's own way. Just think of it as 2 separate things which when combined raise the difficulty exponentially.
DarkStoorM
Hmmmm... I have a question to... maybe I will ask Taiko BATs at the moment:

Can you guys take a look at two maps?
1) S.S.H.Holy Orders [Sin] (270 BPM - 3:38 drain length, 5.98 star difficulty, 1429 notes)
2) t+pazolite feat. Rizna - Distorted Lovesong [Taikocalypse DX] (240 BPM - 6:28 drain length, 6.23 star difficulty, 3026 notes)

At the moment, SS nomod on Holy Orders gives 374pp, where the Lovesong SS nomod gives 326pp.
Can you guys check if it's accurate? I am just wondering.

I just tried the Holy Orders and I was able to FC it on 3rd try, but I still can't FC the Lovesong.
t+pazolite has lots of different patterns, where Holy Orders is just about single tapping (I know, depends on player speed)+ one don stream and some kddkdd~~~ (1/3? didn't check :D )

Just wondering what is your opinion guys about this one.
lolcubes
Well, I'd say, logically, that the pp system got those wrong, cause the star rating is actually higher on distorted lovesong.
o_o

No idea how pp works though.
roufou
could be because the pp system calculates the standard difficulty for converts >_>
TimmyAkmed
As I said previously the maps that have a lots of patterns repetitions seems overrated alot.
I think holy order weight lots of pp only because of the kdd kdd kdd kdd kdd part at bpm 270
The maps that have this "issues"
Megpoid GUMI - Sorry For Being a Closet Otaku MM's Taiko
Bring Me The Horizon - Anthem [31's Taiko]
Renard - TU4AR [Nogard]

All those maps contains the repetitions thingy so it must be what increase the pp value alot.
silmarilen
holy orders is rated 6.86 ingame, while distorted lovesong is 6.23. holy orders also has a higher od on top of that.
Kurokotei
Is it me or did I just get a 100 ranks drop?
Yuzeyun

OnosakiHito wrote:

I'm the opinion it is the main factor in Taiko. Every placed pattern is effected by the BPM. Same goes to the SV, which effects the patterns directly(its density) due to the BPM.
https://osu.ppy.sh/s/138886 Oni (bpm 151 3.12nps) is less dense than:

https://osu.ppy.sh/b/358122&m=1 Muzukashii bpm 127 3.3nps
https://osu.ppy.sh/b/131806?m=1 Oni-like bpm 95 3.2nps
https://osu.ppy.sh/b/91802&m=1 Muzukashii bpm 142 3.57nps
https://osu.ppy.sh/b/153417&m=1 Muzukashii bpm 128 4.24nps

BPM doesn't affect the density directly at all. Going with Dolphin on that one.
OnosakiHito
Hm, probably I misunderstood something. Nvm then.
Kuro
Just a quick question for Tom:

Let's assume I played a map with no mods, let's call this "score A"
Now let's say I played that same map again but with HD, let's call this "score B"

Between the two scores, it came out to be: score A < score B

Now then, after obtaining score A my rank goes up +11 ranks
But after playing again and obtaining score B my rank goes down -6 ranks

Is this a result of HD and FL not being weighted? And if so, how long till it will be weighted, if you don't mind me asking?
Just a time frame would be fine.
lolcubes
Was actually expecting people to post that, but since nobody did, I will.

https://osu.ppy.sh/b/120132

The above map is worth extreme amounts of PP, not to mention it's star rating looks really weird too. The star rating is above 6 stars for some reason, while almost any more difficult map is worth around 5.2~5.4.

Take for example: https://osu.ppy.sh/b/174345

This map not only is faster and much more complicated in patterns, but it's much more dense too, yet it's star rating is 5.52.

https://osu.ppy.sh/b/170038 - Same thing, star rating 5.85.
https://osu.ppy.sh/b/98825 - 5.44
https://osu.ppy.sh/b/101732 - 5.46
etc



The only thing that comes to mind why the star rating is that high is that at the end there are 2 really big streams that are very difficult to perform, other than that, idk.
Dolphin

lolcubes wrote:

Was actually expecting people to post that, but since nobody did, I will.

https://osu.ppy.sh/b/120132

The above map is worth extreme amounts of PP, not to mention it's star rating looks really weird too. The star rating is above 6 stars for some reason, while almost any more difficult map is worth around 5.2~5.4.

Take for example: https://osu.ppy.sh/b/174345

This map not only is faster and much more complicated in patterns, but it's much more dense too, yet it's star rating is 5.52.

https://osu.ppy.sh/b/170038 - Same thing, star rating 5.85.
https://osu.ppy.sh/b/98825 - 5.44
https://osu.ppy.sh/b/101732 - 5.46
etc



The only thing that comes to mind why the star rating is that high is that at the end there are 2 really big streams that are very difficult to perform, other than that, idk.
I thought it was already mentioned but yeah, that map is my "best performance" ._.
Yuzeyun
HD weighting is too arbitrary imo... And too high to be honest. I think it should be about the same as HR at the moment; see thie example:



It's a pretty large difference in accuracy yet that's 5pp difference to the benefit of HD players. I honestly believe that HD and HR should be, generally speaking, equally treated. At some thresholds these mods need to be separated: Higher BPMs will benefit more to HR players (anything 230+ is difficult to HR) and lower BPMs should benefit to HD players (135 and less starts to be real tough to HD; as well as very high BPMs - I can read HD easily past 250bpm). That said, you know what happens, SV calculation kicks in and that should be the "core scrolling speed value" which will dictate whether chart A should weigh more with HR or chart B should weigh more with HD.

I don't know what you guys think.

that said: https://osu.ppy.sh/b/204956?m=1
it gave me 259pp LOL
Chiswell
ppv1 rollback plz.......................
too confusing
Topic Starter
Tom94

_Gezo_ wrote:

HD weighting is too arbitrary imo... And too high to be honest. I think it should be about the same as HR at the moment; see thie example:



It's a pretty large difference in accuracy yet that's 5pp difference to the benefit of HD players. I honestly believe that HD and HR should be, generally speaking, equally treated. At some thresholds these mods need to be separated: Higher BPMs will benefit more to HR players (anything 230+ is difficult to HR) and lower BPMs should benefit to HD players (135 and less starts to be real tough to HD; as well as very high BPMs - I can read HD easily past 250bpm). That said, you know what happens, SV calculation kicks in and that should be the "core scrolling speed value" which will dictate whether chart A should weigh more with HR or chart B should weigh more with HD.

I don't know what you guys think.

that said: https://osu.ppy.sh/b/204956?m=1
it gave me 259pp LOL
Keep in mind, that for HR scrolling speed is not yet considered. With that the weightings will swing in the favor of HR again.
I'll wait for further feedback to see whether the HD bonus needs to be scaled down again.
Yuzeyun
as I said, scrolling speed should be one of the core values; imagine for example 244 BPM at 0.7 SV compared to 110 BPM at 2.8 SV. The hardest to read will be the latter due to it being 308 SV1 (the other is 170.8 SV1) thus making it more rewarding for HR. The other will be rewarding for HD due to its lower scrolling speed.
TKS
is this system possible to give a bonus for OD? in that case, HR will get better.
XK2238

TKS wrote:

is this system possible to give a bonus for OD? in that case, HR will get better.
inb4 DT
Yuzeyun
gotta skyrocket fast
lolcubes

XK2238 wrote:

TKS wrote:

is this system possible to give a bonus for OD? in that case, HR will get better.
inb4 DT
DT technically doesn't increase OD, but the relative timing window itself (decreases).
By this logic, OD shouldn't matter in pp calculations, but the timing window. HR does increase the OD, which affects the timing window, so both are covered this way.
Luna
Hidden should scale with low SV, high OD, and high pattern complexity.

I'd mostly like to address the difficulty algorithm though. Some examples of inconsistencies:
Falcom Sound Team jdk - Over Drive - Pheon's Oni is rated at 5.36 stars while 52's gets only 5.17. This makes absolutely no sense, as 52's Oni is like a gazillion times harder. Pheon's doesn't really deserve an X-tier difficulty icon either, it's pretty repetitive.
Nami Nakagawa - DON'T CUT - The Inner Oni is underrated; it should definitely be higher than the regular Oni (harder and longer patterns, more rhythmically complex), yet it has almost .2 stars less.
Rohi - Kakuzetsu Thanatos - qoot's Diff is underrated
Yousei Teikoku - Hades: The Rise - Ridiculously overrated
Bring Me The Horizon - Anthem - ^
DJ YOSHITAKA - Evans - ^
Megpoid GUMI - Sorry For Being a Closet Otaku - ^ (with DT only)
Shiraishi - Shinsekai - Underrated (HR SS only gives 302pp, wtf)

And a very big (but unranked) example:
REDALiCE ft. Ayumi Nomiya - Foughten Field (t+pazolite Remix) - 6.68 stars is higher than any ranked map, but I can FC this without too much trouble. Wat.

There are many more, but these are some of the biggest offenders/weirdest cases. Most of the overrated maps are moderately high BPM (220-240) with lots of simple and repetitive note clusters (mostly triples, few longer patterns).
DarkStoorM
Just a question about HT:
I just found this little thing as my 100th record


Isn't it a bit too high for Half Time? No mod gives 301pp with 21x100.
There is only one... weird pattern, but with HT it's a piece of cake.

Also, how does EZ affect the pp drop?





or maybe I'm just missing something
Kuro

DarkStoorM wrote:

Also, how does EZ affect the pp drop?





or maybe I'm just missing something
pfff... o3o
Nicely done...

Using DT on converts like those almost guarantees high pp payout and EZ couldn't do anything to deter the difficulty from rising because it only affects HP and OD.
I've done lots of testing to see what does and doesn't give high pp payout and that looks like one of the cases I tested. Although I did it with taiko-specific maps, the overall results should still be the same.

Also..
Simplified: DT and HT are like night and day. If DT gives out a higher payout then HT gives a lower payout.

But there's no reason for me to spill all of my secrets so I'll stop here.
Two Fruit Cakes

Kuro wrote:

DarkStoorM wrote:

Also, how does EZ affect the pp drop?





or maybe I'm just missing something
pfff... o3o
Nicely done...

Using DT on converts like those almost guarantees high pp payout and EZ couldn't do anything to deter the difficulty from rising because it only affects HP and OD.
I've done lots of testing to see what does and doesn't give high pp payout and that looks like one of the cases I tested. Although I did it with taiko-specific maps, the overall results should still be the same.

Also..
Simplified: DT and HT are like night and day. If DT gives out a higher payout then HT gives a lower payout.

But there's no reason for me to spill all of my secrets so I'll stop here.

You should be posting your test results, not trying to profit off of them.
Kuro
Woooow bro, I only answered his question
There's no need for the hostility

I'll be sure not to hang around here anymore
I bid you all a good day~
karterfreak

Kuro wrote:

Woooow bro, I only answered his question
There's no need for the hostility

I'll be sure not to hang around here anymore
I bid you all a good day~

I saw no hostility there, only him telling you to post results as the point of this thread is to improve the performance point system for taiko, not to hide results that don't make sense for your own gain. Making a comment like "there's no reason for me to spill all my secrets" is the complete opposite of what this thread should be used for and won't help the system be improved at all.
Topic Starter
Tom94

TKS wrote:

is this system possible to give a bonus for OD? in that case, HR will get better.
The timing window already accounts for the accuracy portion of the pp that a score gives.

DarkStoorM wrote:

Just a question about HT:
I just found this little thing as my 100th record


Isn't it a bit too high for Half Time? No mod gives 301pp with 21x100.
There is only one... weird pattern, but with HT it's a piece of cake.

Also, how does EZ affect the pp drop?





or maybe I'm just missing something
I think the HT score seems semi-alright. It might be affected by the same problem as I will describe in the following, though. EZ internally only changes the timing window and HP drain, so the non-accuracy portion of the pp stays the same. I'll eventually have to revisit the accuracy formulas for all modes, though.

Luna wrote:

Hidden should scale with low SV, high OD, and high pattern complexity.

I'd mostly like to address the difficulty algorithm though. Some examples of inconsistencies:
Falcom Sound Team jdk - Over Drive - Pheon's Oni is rated at 5.36 stars while 52's gets only 5.17. This makes absolutely no sense, as 52's Oni is like a gazillion times harder. Pheon's doesn't really deserve an X-tier difficulty icon either, it's pretty repetitive.
Nami Nakagawa - DON'T CUT - The Inner Oni is underrated; it should definitely be higher than the regular Oni (harder and longer patterns, more rhythmically complex), yet it has almost .2 stars less.
Rohi - Kakuzetsu Thanatos - qoot's Diff is underrated
Yousei Teikoku - Hades: The Rise - Ridiculously overrated
Bring Me The Horizon - Anthem - ^
DJ YOSHITAKA - Evans - ^
Megpoid GUMI - Sorry For Being a Closet Otaku - ^ (with DT only)
Shiraishi - Shinsekai - Underrated (HR SS only gives 302pp, wtf)

And a very big (but unranked) example:
REDALiCE ft. Ayumi Nomiya - Foughten Field (t+pazolite Remix) - 6.68 stars is higher than any ranked map, but I can FC this without too much trouble. Wat.

There are many more, but these are some of the biggest offenders/weirdest cases. Most of the overrated maps are moderately high BPM (220-240) with lots of simple and repetitive note clusters (mostly triples, few longer patterns).
When I have the pattern algorithms for mania done which are supposed to detect repetetive things, then most likely I can also apply them to Taiko and fix stuff like that. Thanks for compiling that list - I will add it to my references when adjusting the algorithm further.
Bara-
Maybe add comboheight as a small factor?
I think a play with combo 150 and 97% should get less than a 900combo with 92%
If the combo is higher, but with lower accuracy, you still get the same amount of pp maybe?
This will make not-so-accurate people who can get high combos on maps get more pp than people who can't keep their combo up, yet still being very accurate.
Idk if this will fit in pp good or not.
Dolphin
Combo doesn't mean anything about how accurate your timing is. All combo is there for is to tell how you long you can play, and what you can play, without fucking up. If people miss a lot their accuracy will get worse either way, so making combo weighed is a pretty stupid idea.
Coro
It's pretty much a carry over from the TnT scoring system, standard also has it due to Ouendan/EBA also being combo oriented
(I know Shinta mode is a thing but still)
Nothing much you can do about it I guess orz
Dolphin
It can stay in the scoring system (it's way too late to change it now anyway) but I'd prefer for it to not be in the pp system as well.
roufou
just noting that long monocolor streams are pretty overrated still, not talking about 300 bpm ones in general but monocolor streams in general like for example https://osu.ppy.sh/b/48098 and https://osu.ppy.sh/p/beatmap?b=34211&m=1


I also noticed this in one of my own maps wherein one of my maps with a decently long 300 bpm stream became higher than some of my real hard maps, if you make them worth less maybe hades would also get a more fitting star difficulty
Luna
Some more feedback in addition to everything that has been mentioned so far:

1) The most overrated (ranked) map in the entire game has to be Kyoai with DT.
6.99 (essentially 7) stars is completely ridiculous. I got a 1 miss sightread play, and I wasn't even warmed up. This map deserves 6 stars max, probably even less than that. It is super easy for 90% of its playtime and has only 2 slightly tricky parts (which are short and repetitive) and a very basic 1/3 stream.
I feel like the straight-up kddkddkddk / dkkdkkdkkd difficulty assessment is too heavily skewed in favor of those specific patterns and not enough towards actually difficult, confusing streams. Something like kddkdddkdkkdkkddk seems to be rated as easier than the basic stream pattern despite being significantly harder.

2) Low BPM is underrated. That is, low BPM with difficult patterns. I have a bunch of deathstream maps in the ~130/140 BPM range (many mapped by well known "overmappers" like Firce and Ishida) that are rated below 5 stars even though they are easily as hard as most ~5.8* maps. While pattern difficulty mostly becomes a nonfactor at super low BPMs (like 100 BPM), there is a pretty wide range of speeds where star difficulty simply does not reflect the complexity of the map to an appropriate extent at the moment. This might also be related to the limitations of stream complexity calculations mentioned before.

3) Something wonky is going on with speed scaling I guess. Take a look at this map. Both [Last Taiko] and [Nineteenth Taiko] are rated at 9.23 stars, with 19th being listed as harder by some 1/100 of a star or something. However, the difficulties are not even comparable in practice - while a seriously dedicated top level player could absolutely practice 19th enough to FC it, I haven't seen anyone even pass Last yet despite it having a life drain difficulty of 0 (seriously)! Last Taiko is essentially a true 400 BPM map. I'm assuming this is a side effect of whatever you did to nerf 1/8 mashfest conversions, but it's not working as intended since this map has absolutely nothing in common with those types of maps. This actually has patterns (and lots of them) rather than completely monotonous dddddddddd spam. You cannot mash this. You will miss 90% of the map if you do. As long as Last Taiko is assigned the same rating as 19th, there is something seriously wrong.

4) Hidden weighting is really messed up, but I guess you already know this. Just to illustrate, some scores from the Kyoai map:
HR DT 100% = 410pp
HD DT 99.35% = 409pp
HD DT 99.86% = 421pp
The HR DT score is ridiculously difficult, even though the map itself may be overrated. Yet it can be tied and even surpassed by way easier HD plays.
On the other hand, there are also lots of maps where HD is underrated (mostly high OD maps I think).
Yuzeyun
I once made a very difficult 7:15 map with absolutely ridiculous streams (starting with 2,050 notes stream) to have in the end a whopping 5.04 rating.
It's like a lot harder than any other map ever done despite it being 179 BPM, just that it doesn't feature a lot of handswitchers.
Luna
Another example I just came across of a map with kinda tricky streams @200 BPM that is rated below 5 stars for some reason: https://osu.ppy.sh/b/123834
Thought I'd mention this one because it uses a different kind of "difficult" stream (dddkkkdddkkk-based).
Nwolf
can't give any real examples but in lower diffs 3/2 notes are overrated a lot
MMzz
You realize pp doesn't calculate color changes, right?

It just takes the amount of notes played in the time of the song. Plus some acc and something called "strain".
Luna

MMzz wrote:

You realize pp doesn't calculate color changes, right?

It just takes the amount of notes played in the time of the song. Plus some acc and something called "strain".
That is not true. If you want to check it for yourself, make a map that just consists of a single don-only stream and look at the star rating. Then take the same map, but make it a kddkddkddk... stream. The second map will have a significantly higher star rating.
Rhythm complexity is also taken into consideration.
MMzz
First off, pp does not look at star rating. According to the wiki and what I've found through farming that is not the case at all.

Maps like https://osu.ppy.sh/b/114207?m=1, https://osu.ppy.sh/b/177271?m=1, and https://osu.ppy.sh/b/113409?m=1. Give boat loads of pp because they are dense, and with double time are pretty short. (making the ratio of notes per song length even higher) Mind you there is no pp bonus for having DT active. All the patterns are very straight forward and have little to none in rhythmic challenge compared to maps like I'm your daddy or Shinsekai. (These are great examples because the rhythm is random and all over the place.)

It is pretty obvious that there is no weighting for color changes/star rating.
OnosakiHito
We need updates for this thread and its current state. Especially about the system itself.
There are a lot of bugs in the Taiko rating. People are daily talking about it, yet very few post into this thread.

First step would be to have some data in OP from current stance.
Nashmun
Giving a pp boost for HD mod was a great idea, but isn't it too much now ?


Yuzeyun
In some maps the HD boost is lower than the HR one that's odd

Example : https://osu.ppy.sh/b/166598&m=1 - HDSS gives 185pp and HRSS gives 204.
TimmyAkmed

_Gezo_ wrote:

In some maps the HD boost is lower than the HR one that's odd

Example : https://osu.ppy.sh/b/166598&m=1 - HDSS gives 185pp and HRSS gives 204.
It must be because of the OD, Hades is OD 5 and Naotyu OD6, HR gives more when OD is 6+ and HD gives more if OD is 5-
That would also explain why DTHD SS gives more than DTHR SS on Kyoai.
Yuzeyun
This seems to be a valid point as Kyoai is OD4
roufou
https://osu.ppy.sh/b/95102?m=1 jesus christ this map is underrated, it gives 171 pp which is nothing by my standards while it deserves to give at least 200

compare it to this map https://osu.ppy.sh/b/135609?m=1 which is like 130 bpm and so stupidly easy which gives like 207 pp

30 pp difference is pretty huge

admitedly the one that is 207 pp is quite a bit harder in the accuracy spectre, and accuracy is actually kinda underrated in my opinion, the fact that the first one is worth less is pretty dumb
verto
http://puu.sh/dCmm4/989b1353f1.png

Just some EZ vs no EZ comparison
Full Tablet
I made a difficulty calculation algorithm for taiko maps (based on the rhythm complexity algorithm I was working on for osu!, but also taking strain into account; the value represents both the strain and the rhythm-pattern complexity, since checking the algorithm by playing maps is less straightforward if the 2 values are separated. I was testing with this mode since it is the simplest one to analize); the scale of the values is made similar to the scale used in the current star difficulty (Tom Stars).

Finishers are considered the same as regular notes for patterns (since hitting them with both keys is optional). Drumrolls aren't considered in the algorithm at all.

The algorithm doesn't include factors that affect reading difficulty (such as scrolling speed, scrolling speed changes, finishers blocking other notes).

Here are some results (for most ranked taiko maps):
https://www.dropbox.com/s/e4l17ig7vtwat ... .xlsx?dl=1

a) What do you think about those results in general?

b) An earlier version of the algorithm used to detect color-inversed repeated patterns (for example, ddkkd kkddk), but I removed that detection (making the algorithm consider both 5-note patterns as completely different) since when I was playing I noticed that color-inversed repetitions confused me more than helped (I tended to play the 2nd pattern the same as the first instead of inverted), but that could have been only because I am pretty bad at reading Taiko (in osu!mania, the equivalent concept of mirroring patterns actually makes thing easier for me). Should I re-add color-inversed pattern detection?

c) Some maps that players have said to be pretty hard, but are rated considerably lower in this algorithm compared to the current Star Difficulty, are Shinsekai and Taiko Time. They are rated lower in this algorithm mainly because the algorithm finds many long repeated patterns:
In Shinsekai, the algorithm finds many patterns that are 8-notes long or higher next to each other (up to 16-notes long).
Should I make the algorithm less sensitive to long patterns repetitions (longer than the ones the average person can keep in their short term memory)? If I did, then people who are good at memorizing patterns in their long term memory would have an extra advantage. On the other hand, if a pattern is long (and not composed of sub-patterns), there is a higher chance the player doesn't have any experience playing that specific pattern.
roufou
if you'd like you could compare your stats with a human-made list I did here: p/3563986

just as a warning, I do not claim that I'm the one person to make an objective list like that as I'm not THE MOST EXPERIENCED TAIKO PLAYER, but I have a decent idea for what maps are harder I guess (and some people also tell me if they really disagree with something I did, which I sometimes realise is true)

also my list is pretty unfinished, so that's something worth taking note of

as for critic on your list, disconnect hardcore is overrated if I were to say something, like it sure is really hard for a ranked map, but I wouldn't consider it 2nd, as it's not near compared to distorted, never mind and because maybe, image material is kinda overrated.
speaking of never mind it's really underrated, ztrout's visit to the proctologist is also very underrated imo, like seriously, bellow team plasma appears?! I think it's because patterns like [dddd]dd[dddd] which are really weird and complicated to play are highly underrated
Luna
With a little cheating kiddie running rampant today, I noticed something.

Take a quick look at these scores:



Is there a specific reason why the pp values are so ridiculously low? Equivalently difficult scores in any of the other modes would be worth at least 1k pp each, probably way more (judging from the huge scaling at the top level, they might even be worth multiple thousands of pp).
Topic Starter
Tom94

Luna wrote:

With a little cheating kiddie running rampant today, I noticed something.

Take a quick look at these scores:



Is there a specific reason why the pp values are so ridiculously low? Equivalently difficult scores in any of the other modes would be worth at least 1k pp each, probably way more (judging from the huge scaling at the top level, they might even be worth multiple thousands of pp).
I'll take a look at the scalings when I get the time to work on Taiko pp again. For now I can't really do much about pp at all. :/
Thanks for reporting!


Full Tablet wrote:

I made a difficulty calculation algorithm for taiko maps (based on the rhythm complexity algorithm I was working on for osu!, but also taking strain into account; the value represents both the strain and the rhythm-pattern complexity, since checking the algorithm by playing maps is less straightforward if the 2 values are separated. I was testing with this mode since it is the simplest one to analize); the scale of the values is made similar to the scale used in the current star difficulty (Tom Stars).

Finishers are considered the same as regular notes for patterns (since hitting them with both keys is optional). Drumrolls aren't considered in the algorithm at all.

The algorithm doesn't include factors that affect reading difficulty (such as scrolling speed, scrolling speed changes, finishers blocking other notes).

Here are some results (for most ranked taiko maps):
https://www.dropbox.com/s/e4l17ig7vtwat ... .xlsx?dl=1

a) What do you think about those results in general?

b) An earlier version of the algorithm used to detect color-inversed repeated patterns (for example, ddkkd kkddk), but I removed that detection (making the algorithm consider both 5-note patterns as completely different) since when I was playing I noticed that color-inversed repetitions confused me more than helped (I tended to play the 2nd pattern the same as the first instead of inverted), but that could have been only because I am pretty bad at reading Taiko (in osu!mania, the equivalent concept of mirroring patterns actually makes thing easier for me). Should I re-add color-inversed pattern detection?

c) Some maps that players have said to be pretty hard, but are rated considerably lower in this algorithm compared to the current Star Difficulty, are Shinsekai and Taiko Time. They are rated lower in this algorithm mainly because the algorithm finds many long repeated patterns:
In Shinsekai, the algorithm finds many patterns that are 8-notes long or higher next to each other (up to 16-notes long).
Should I make the algorithm less sensitive to long patterns repetitions (longer than the ones the average person can keep in their short term memory)? If I did, then people who are good at memorizing patterns in their long term memory would have an extra advantage. On the other hand, if a pattern is long (and not composed of sub-patterns), there is a higher chance the player doesn't have any experience playing that specific pattern.
Awesome! If the response to your system is generally better than to the current system in place I'd love to use your system (or at least incorporate it in some way - there need to be certain guarantees fulfilled). That is if you'd be willing to share it of course. :)
Nwolf
Don't wanna be rude but Full Tablet's system overrates and underrates maps too much. Busy modding atm so I can't go in further detail.
Luna
Yeah, Full Tablet's system in its current state is actually worse than what we have in my opinion. I'll write up a more detailed critique with reasons, examples etc when I find some more time (probably early-mid January)
Full Tablet
New version of the algorithm. It probably needs some further adjustments, but currently it should be better than the previous version.

https://www.dropbox.com/s/bug2etc301ymb ... .xlsx?dl=1

Tom94 wrote:

Awesome! If the response to your system is generally better than to the current system in place I'd love to use your system (or at least incorporate it in some way - there need to be certain guarantees fulfilled). That is if you'd be willing to share it of course. :)
I am willing to share it when the response is positive enough.
Kuro
This newer version is looking a lot better than the previous

Anthem isn't too op anymore +1
Forgotten got the boost it truly deserves +1
Hades got nerfed big time +2
Jehovah's YaHVeH is overrated now even though it's nothing but ddk everywhere... lol, it's basically the new Hades -1
Intersect Thunderbolt dropped a bit, not cool bro -1
Pavor Nocturnus took a small drop, i think it's pretty hard though +-0
o'er the flood took a small drop too, i think it's a bit underrated now +-0
Atama no Taisou - Love's taiko got a big boost, which is cool, and it's still lower than inaba's taiko which is also good. A lot of people say inaba's taiko is definitely harder but the gap shouldn't have been that big between the two so this works out well. +2

A lot of other changes were for the better. Overall, I'm liking where this is going. It's not in shape to be integrated yet but it's making progress.
Nashmun
Speed seems to be a little overrated compared to pattern difficulty, but with some adjustment it could end up in something great. I couldn't take a long look at it yet since I'm not at home, but I'll dig further into it by the end of the week. Also some maps which were overrated due to dkkdkkd like patterns + 1/6 seems to be rated better here (Ohai evans)
Yuzeyun
Anjuu Tianlan is above Anjuu Zhuhong now -514

What would be nice is to have it tested with non-ranked maps. I can provide you some example maps that are deemed overrated / underrated by people.


small edit : wtf how come Anjuu Kantan DT is rated that much
Full Tablet

_Gezo_ wrote:

Anjuu Tianlan is above Anjuu Zhuhong now -514

What would be nice is to have it tested with non-ranked maps. I can provide you some example maps that are deemed overrated / underrated by people.


small edit : wtf how come Anjuu Kantan DT is rated that much
I had an older version of that mapset in my beatmap list. Anjuu Kantan DT is rated 1.77 actually (1.52 No-mod)
Tianlan's is rated 4.08157 (5.45859 with DT)
Zhuhong's is rated 4.08052 (5.45962 with DT)
Yuzeyun
Definitely a problem, it took me forever to FC DT the Zhuhong diff while the Tianlan one was already DTHRed at that time. Zhuhong is really tricky and focused more on pattern complexity rather than density; while Tianlan was much more on density. I think because of these different focuses, it ended up being nearly the same for a totally different field.

Also I maintain that Anjuu Kantan is overrated
roufou
SPOILER
just a few oddities in the pp system that are pretty weird in general and some opinions on overrated maps/underrated maps:

first off lets start with my opinion on evans, a lot of people think evans is VERY overrated in the pp system, (I can agree that the SR is a bit overrated) however nerfing stuff like dkkdkkd patterns which evans features a lot of would not be the right move in my opinion, and 1/6 (which I think is kinda underrated, at least if they're in 1/4 streams), why do I think so? well first off if we nerf evans the pp system will probably become less pp in general, I guess, and I think that would make it a bit harder to judge (this is not the major point though), it's the fact that I personally don't think evans is overrated, it's a tough map, however I do think evans overshadows a lot of underrated maps, I'll give an example: unhappy refrain which gives 240 pp, nerfing dkkdkkd would probably make evans around that pp range or lesser, and I doubt unhappy refrain will be affected. However, I don't believe anyone should think unhappy refrain is harder than evans, not near. Evans is a pain to fc mainly because of the end and it's pretty tough to read overall, while unhappy refrain is WAY easier at those aspects, unhappy refrain may be slightly more stamina craving, but I think evans readability outweights that by far. Unhappy refrain is not the only map, however, these are more argueable but the following maps aren't far from evans but are way easier in my opinion: cruel clocks, daybreak's bell (with DT, the monocolor streams can be kinda tough to some people however and accuracy isn't that easy), tu4ar (the convert). There are even maps that are easier than evans which gives more pp, hades and augoeides are good examples (I'd even argue that both of them are examples, but I don't think everyone will agree ono's around the difficulty level of Evans)

here's two maps that are about as hard as evans (if not harder) but are rated lower, there are probably more but these maps are really underrated: Etude Op. 10-4 and MEPHISTO (etude should probably give roughly 250 pp and mephisto more than 270, imo)

other underrated maps and a pp estimate of how much I think they should be worth and roughly what they are worth (I know these numbers don't mean a lot but it gives an estimate of how underrated they are, and again, this list will change drastically after changes):
schrodinger's cat (180~ pp to about 200 pp+)
No way back (I have no idea how much pp this gives, most people who can fc this are too good to have this in their best perf, 250+ sounds reasonable to me)
Music Revolver (not sure how much pp this gives but I think it's like 230, 260 sounds more like it)
Tenjiku2000 (not sure how much pp this gives but I swear a HRSS should be 280 pp)
ztrout's visit to the proctologist and similiar (maybe not underrated in pp but maps with 1/4 and 1/6 in them should be like 5 stars imo, it is like 4,43 stars)

tl;dr: don't work on nerfing too many aspects, buff the aspects that are underrated: complicated or tricky 1/6 streams in 1/4 streams and longer more complicated streams

these are the following maps that probably deserve a direct nerf in what makes them rated highly: anthem (not sure how this is rated high), hades: the rise (gives way too much pp no matter how you see it), holy orders (not too overrated, but XK's score should be more in the 370 range imo), rising hope (if you get lucky on the "hard" part this map rewards you just too much, and with HT it's even worse), kyoai (rewards people who can do the kddkddkddk stuff with DT a bit too much imo), lepi's augoeides (eh, pretty easy)
I might've forgotten like two or three maps but I think that's mostly everything, other maps should just decline to a suitable level by other maps being buffed

bonus: no idea how much pp this map would give https://www.dropbox.com/s/3lrz0pvozt8s1 ... e.osz?dl=0 but it's like 5,88 stars...6,5 would be a lot more suitable thanks to being a tough map playing at around 255 bpm
Yuzeyun
HD is way too op, now.
WAY too op.

Quick comparison between two same-acc scores on Kyoai
For the record, wzxxzw's SS score doesn't even come close to HDNC.
Stefan
mfw I stuck since months between 2000 and 2100 because of updates and doing nothing.
Fudgy
Since when did HD give not enough pp? I don't understand that buff.
lolcubes
HD DT EZ spam everywhere now, for so much pp as well lol.
Vuelo Eluko
I believe somewhere along the line the intent was to buff hidden on slow songs not fast ones
it seems to have gone wrong, i'd just wait it out.
Kuro

Fudgyking wrote:

Since when did HD give not enough pp? I don't understand that buff.
I don't know about you but to me this isn't brand new... back then it wasn't even worth the effort to even try to HD maps because you wouldn't get anything anyway or you'd just get little to nothing unless you played something stupid hard and HD'd it...
OnosakiHito

Riince wrote:

I believe somewhere along the line the intent was to buff hidden on slow songs not fast ones
it seems to have gone wrong, i'd just wait it out.
Exactly. That was the plan in the past. Giving some boost for HR on faster songs and a boost for HD on slower songs. At least we discussed about it.
But as you said, let's just wait for it.
roufou
as a sidenote, I haven't noticed a single pp increase on any of my FL scores, I could be wrong though
Yuzeyun
didou's hrhdfl on whatiforgotlol was 275 before the buff, it's now 305. @agu


something made real quick about SV scroll speeds


note:
ASS = absolute scroll speed. yeah. not arse.
despite having a huge difference in speed, 180 BPM SV 1.4, HR 16:9 (448) is easier to perform than 210 BPM HR SV 1.6 (448). it's due to the fact that 16:9 and 4:3 have the same time between notes' apparition and supposed hit. the ASS is like "units", and each "unit" does not have resolution considered.

I'm pretty sure the best thing to use is the upper table with them values that represent the ASS (yeah it's just SV*bpm) to consider HD and HR. HD will use the ASS tp tell whether it's hard or easy. Fairly sure 150-170 and below is where it gets harder, 225-230 is comfort for most, and hitting 420 it starts to get crazy.

HR on the other hand, is harder to tell. let's say it gets easier from 100ish (below you will have to resort on OD), and the starting point for HR to truly fade out is 280-300, using nomod ASS.

if you have anything to add, either on the HR side or the HD one, please use the first table in light blue to show where it hits a point. if you need more SV values tell me. I think [1.2;2.0] is a fairly large range for most maps. however I don't consider SV changes for now.
Topic Starter
Tom94
Unfortunately the infrastructure of taiko pp still doesn't really have a measure for scrolling speed, so I can't easily add a HR or HD bonus relative to scroll speed. The general consensus seems to be, that HD is worth too much right now, so I'll move it downwards a bit.
Yuzeyun
can't even manage to fetch SV and BPM at all? kinda odd, considering they're a bit like AR (not completely the same) so yeah. x_x
TKS
since we can add any scroll speed by using green line, isnt like AR. it seems really hard to manage for that indeed.
Yuzeyun
they're not exactly like AR, I know it. we don't even have "fixed values" but rather the scroll speed thing I talked about earlier.
but first get into the sole thing of ASS which is for 300 set pp to this, ASS 400 pp bonus to that blablah

then about SV changes it'll depend on the chart definitely. if there's a very hard section but SV 0.5 it won't be treated the same way as SV 2. for ther same thing
Topic Starter
Tom94

_Gezo_ wrote:

can't even manage to fetch SV and BPM at all? kinda odd, considering they're a bit like AR (not completely the same) so yeah. x_x
I can fetch those, but since they depend on timings within the map I'll have to conclude a reasonable "global" value for the pp calculations from within the difficulty calculations.

We don't know whether misses were made in parts of the map with fast SV or not after all.

That stuff needs additional balancing and implementation, hence it's not there yet. :p


PS: HD and FL boni should be a fair bit lower now. Does it look better?
FLANKs
I think HD is fine how it is right now. The first update might have been a bit much but this is fine.
TimmyAkmed
imo the system is still overrating HD alot, just look at the recent DTHR of Sobatsuyu, his performances weights way less than some DTHD that have 0,50 accu below.

HR beats HD only when base OD is 7. Now there is just no point in doing DTHR on maps since it's just way harder and it will reward you way less.
TKS
HD is fine. as it is, HR is too underrated. has to buff before nerf HD :)
ikin5050
Maybe for like 2000-3500 pp players buff how much pp a 93% oni play with a few misses?
I personally cannot properly play onis (proper meaning fc 97%) and right now i always get some misses, look at my top plays :P
I feel that a lot of people in my country for example have a lot of pp from convert insanes, when oni maps are much harder.

http://puu.sh/gj3WM/41b3ba129a.png
http://puu.sh/gj3W2/6fdfc0c8c3.png
Redon
MMzz
While I don't like the current state of lower difficulties for taiko either. It is better to make easier diffs that can be universally accepted, rather than what YOU think provides the proper challenge for a noobie. Easy/Normal serve the purpose of basic rhythms and learning the difference in color. That is it. There isn't really a reason for them to be more complex than they already are. Muzukashii serves as a good platform for applying your rhythm and reading abilities learned from Kantan/Futsuu into full fledged rhythms.

I used to think how you do, so I know where you are coming from. But I do agree that the star rating scale for kantan/futsuu is ridiculous. You can't even add a few simple 1/2 patterns without the star rating skyrocketing.
Yuzeyun
Let me correct you a bit MMzz on your last sentence ;P

You can't even *USE 1/2 patterns without the star rating skyrocketing (1/2, 3/2, 5/2, 7/2...)

It was a pain to fix Twitch Plays Pokemon because of that :C
Raiden

MMzz wrote:

While I don't like the current state of lower difficulties for taiko either. It is better to make easier diffs that can be universally accepted, rather than what YOU think provides the proper challenge for a noobie. Easy/Normal serve the purpose of basic rhythms and learning the difference in color. That is it. There isn't really a reason for them to be more complex than they already are. Muzukashii serves as a good platform for applying your rhythm and reading abilities learned from Kantan/Futsuu into full fledged rhythms.

I used to think how you do, so I know where you are coming from. But I do agree that the star rating scale for kantan/futsuu is ridiculous. You can't even add a few simple 1/2 patterns without the star rating skyrocketing.
Not even this. Even 1/1s are overrated as heck.

I've been told in my maps (by mods) that I overmapped the Kantans and Futsuus because 1/1s. And I'm like, seriously? They're not even high BPM, gosh...
Redon
eeezzzeee
To be fair, I think its mania that has underrated maps. If you look at all the other game modes, easy almost never has 1/2 notes, and normal will never have 1/4 triples, whereas in mania those are standard. So what I'm trying to say is we should be expecting "brain dead" maps at 1-2 star difficulty but somehow that's not the case for mania.

I'm not saying the taiko star rating is good right now though. No. lol
Redon
verto
I'm going a bit off here to address another issue.

As we know TAG4 maps now don't yield any pp, with the explanation that they weren't be rankable according to the ranking criteria.

Now, we have converts which kind of acts the same as TAG4 maps to a lesser extend. A lot of people argued in the thread that they shouldn't count toward the pp and now that TAG4 maps were removed from the pool I don't see why would you keep converts in it.
Full Tablet

Redon wrote:

I still think star ratings in the lower difficulty range are hilariously overrated. It takes an insultingly simple map to even make it below the 1.5 cutoff for a Kantan.

I made a post about this about a year ago when the new star rating calculation was introduced, but I didn't receive any response at the time. I'm not sure if the Taiko community is fine with dedicating half of every map's diff spread to the mentally handicapped or if it's just that nobody cares about easy Taiko diffs, but if you compare the difficulty of a 1.5 star Taiko map to that of a 1.5 osu!mania map, it shouldn't take too long to realize that something is terribly wrong.
For me, X star Taiko maps are harder than X star 4K osu!mania maps.

Taiko requires considerably more timing precision than mania, OD3 maps require about as much accuracy for a certain accuracy percentage as OD7 in mania (and if you use DT on OD3, the timing precision required goes beyond OD10 in mania; DT doesn't change the timing windows in mania beyond some rounding errors). Also, Taiko punishes misses more than mania.

Also, for me, Taiko maps are harder to read than osu!mania maps (with about 3000 plays, I still mess up randomly at reading colors even in 2 star maps; with 4K mania I never had that issue even when just starting playing).

I do think that 7K and 8K osu!mania lowest difficulties are underrated (learning to use ring fingers takes some time).
lolcubes
It's a matter of perspective. For example, when I started taiko, I could pass Muzukashii diffs with some effort. When I started mania, I was getting destroyed by easy diffs (7K).
We shouldn't really compare the two game modes, all they have in common is pressing the right button at the right time (and even that is questionable due to holds).

As for the star ratings, they were never really comparable cross game modes, so yeah, it's just different.
goatlife


k e k
MMzz
Joking aside that actually makes no sense. I'm your daddy has a larger note-length ratio and significantly harder patterns. So from what we are told about pp it should give much more than Holy Orders.

Oh wait. BPM. :idea: :idea: :idea: :idea:
Backfire
Ok i'm here with a complaint.
https://osu.ppy.sh/b/690982
this map is actually insanely difficult, it might as well be the hardest taiko map ever, but its only 5.8* (the standard is fucking 8*) and Flanks #1 (which is fucking insane) only gave him 123 pp. I do not understand this system. Sorry.

Not mad, just saying that I don't think this system weighs some things like length and bpm. All it seemingly weighs is patterns. Hades : The Rise is harder than this map according to SR.

I think it still needs some more tweaking towards bpm and SV in taiko.

I didn't read some of the recent pages, so im probably being overly redundant.
lolcubes
92% accuracy B, that's why. SV doesn't work at all when it comes to difficulty calcs.
OnosakiHito
And if tom didn't change it, converted maps wight much less than taiko diffs.
DarkStoorM
400~ x100 is an insanely HUGE amount along with 80~ misses. 131pp for these stats is actually high, trust me. Now imagine, that people may lose ~1pp for every x100 hit, FLANKs got 384 of them (depends of course).

I have no idea how the map looks like, but since it's a convert I assume it's full of monocolored 1/8s (?). I can only guess, that full combo nomod would give around 400pp, because it's less than 6 stars, but along with the length it may end up giving like 450.
verto

_verto_ wrote:

I'm going a bit off here to address another issue.

As we know TAG4 maps now don't yield any pp, with the explanation that they weren't be rankable according to the ranking criteria.

Now, we have converts which kind of acts the same as TAG4 maps to a lesser extend. A lot of people argued in the thread that they shouldn't count toward the pp and now that TAG4 maps were removed from the pool I don't see why would you keep converts in it.
What if we just let converts go?
lolcubes
Was actually suggested many times already, and I guess people didn't really like the idea, though it's probably the most logical one since there are so many things you will never see in a ranked taiko map.
verto

lolcubes wrote:

Was actually suggested many times already, and I guess people didn't really like the idea, though it's probably the most logical one since there are so many things you will never see in a ranked taiko map.
Before I thought that it's a matter of opinion, so keeping them is subjective, but after TAG4s were removed from the std pp pool it became an entirely different issue. They are literally the converts of standard in this regard.
show more
Please sign in to reply.

New reply