that's why it doesn't count eitherThen we have a solution! use Hardrock or double time for the problem.
More score, more people happy. Also you'll gain more skill this way!
Genius.
forum
Hyperdash fruits don't always spawn correctly in CTB [confirmed] 4jx2m
posted
Total Posts
211
you retrograd person, stop trolling these kind of threads and gtfo.Manchineel wrote: 546y5t
that's why it doesn't count eitherThen we have a solution! use Hardrock or double time for the problem.
More score, more people happy. Also you'll gain more skill this way!
Genius.
you did the same with the ctb droplets thingy topic.
Seph
actually running on vsync 60fps makes my ryuuta move faster than it should be (or its just me) but i tried doing it on some pixel jump and i apparently caught that pixel jump on rainbow tylenol when i pause and switched from unlimited to 60fps. though playing on it makes ryuuta control hard as fuck
60fps is not faster, there is something different with the hyperdash I think.Seph wrote: e4q1c
actually running on vsync 60fps makes my ryuuta move faster than it should be (or its just me) but i tried doing it on some pixel jump and i apparently caught that pixel jump on rainbow tylenol when i pause and switched from unlimited to 60fps. though playing on it makes ryuuta control hard as fuck
I'm bumping this. I'd really like to see it fixed since CtB is the most broken osu! mode as of now.
By the way, here's one simple example of an impossible jump: http://puu.sh/1jxet
By the way, here's one simple example of an impossible jump: http://puu.sh/1jxet
Here is an updated version of the Watashi map with three scenarios where hyperfruit fail to spawn:
*3-Note Jump
*Spaced Stream (same broken mechanic as 3-Note Jump, but in a more extreme scenario)
*Fast Slider
Note that the 3-Note Jump is catchable under some strange lag-based circumstances, but the others are definitely not (and the 3-Note Jump can't be caught without lag either.) Also included is a tick rate 2 version of Fast Slider, to show that hyperfruits spawn correctly when there are no small droplets involved.
*3-Note Jump
*Spaced Stream (same broken mechanic as 3-Note Jump, but in a more extreme scenario)
*Fast Slider
Note that the 3-Note Jump is catchable under some strange lag-based circumstances, but the others are definitely not (and the 3-Note Jump can't be caught without lag either.) Also included is a tick rate 2 version of Fast Slider, to show that hyperfruits spawn correctly when there are no small droplets involved.
This isn't high-priority because it's been a problem with CTB since CTB was a game. If it wasn't high priority then, it's not just going to become a focal point now. Old bugs are low priority compared to new bugs and trust me there are a lot of new bugs that can and should be fixed before this bug which everyone is already used to gets fixed.
One can thing of this from an accessibility standpoint. How many players does this bug affect compared to other bugs? There aren't that many "impossible" CTB maps, and there are far less players who can play these CTB maps anyways. Fixing this would only please a small group of players, while gamewide bugs/ bugs affecting wider ranges of people take higher priority.
Also this will require peppy to just drop everything he's doing and focus on nothing but this bug for a whole development day.
One can thing of this from an accessibility standpoint. How many players does this bug affect compared to other bugs? There aren't that many "impossible" CTB maps, and there are far less players who can play these CTB maps anyways. Fixing this would only please a small group of players, while gamewide bugs/ bugs affecting wider ranges of people take higher priority.
Also this will require peppy to just drop everything he's doing and focus on nothing but this bug for a whole development day.
What about XAT trying to mod CtB diff wich can't even determine if the maps are FCable ?TheVileOne wrote: 4cy22
Fixing this would only please a small group of players, while gamewide bugs/ bugs affecting wider ranges of people take higher priority.
I personnally see it as high priority for the sake of CtB mapping, modding and gaming, but I understand this bug is pretty tricky to fix cause determining a maximum non hyper jump too short could make a lot of maps too easy to FC, and using a too long one could not entierly fix the problem (well the problem is a bit more complicated than that but this is an easy way to explain why it's tricky to fix).
All the cases in MillhioreF's example map should now be fixed on the test build. Please test all cases (including other maps, as hyperdash may now be spawning in places it is not required, which may or may not need to be fixed).
I also added an alert when watching a replay which moves faster than humanely possible, which only displays on the test build. Might be handy for checking any replays you think may not be legit, and pointing them out to me in a PM for confirmation. Keep in mind this may be a result of lag, too.
I also added an alert when watching a replay which moves faster than humanely possible, which only displays on the test build. Might be handy for checking any replays you think may not be legit, and pointing them out to me in a PM for confirmation. Keep in mind this may be a result of lag, too.
Detected a little problem with many sliders :
Taking this map as example :
http://osu-ppy-sh.jeuxcrack.net/p/beatmap?b=46477&m=2
00:22:537 (1,2,3) - those sliders have now hypers in the first beat and the middle droplet but the hyper isn't triggered when the player catch them. For this map I catched the sliders sometimes without the fix so basically they was catchable (with dash ofc) but I think they still need this fix since they was really really really hard to catch.
http://osu-ppy-sh.jeuxcrack.net/p/beatmap?b=27737&m=2
00:45:257 (1,1,1) - Kinda same problem some elements of sliders seems to trigger the hyper showed in the fruit and some not.
http://osu-ppy-sh.jeuxcrack.net/s/65294
01:28:530 (7,8) - Here it's a mapping problem. Randomness of the position of the droplet could be kinda annoying to map cause for the same slider speed/direction it's going to pop or not the hyper droplet (you can compare with previous sliders). Anyway the hyper isn't trigger when playing as in the others maps.
Taking this map as example :
http://osu-ppy-sh.jeuxcrack.net/p/beatmap?b=46477&m=2
00:22:537 (1,2,3) - those sliders have now hypers in the first beat and the middle droplet but the hyper isn't triggered when the player catch them. For this map I catched the sliders sometimes without the fix so basically they was catchable (with dash ofc) but I think they still need this fix since they was really really really hard to catch.
http://osu-ppy-sh.jeuxcrack.net/p/beatmap?b=27737&m=2
00:45:257 (1,1,1) - Kinda same problem some elements of sliders seems to trigger the hyper showed in the fruit and some not.
http://osu-ppy-sh.jeuxcrack.net/s/65294
01:28:530 (7,8) - Here it's a mapping problem. Randomness of the position of the droplet could be kinda annoying to map cause for the same slider speed/direction it's going to pop or not the hyper droplet (you can compare with previous sliders). Anyway the hyper isn't trigger when playing as in the others maps.
Regarding this: Even with the hyper-droplets, they seem to be a bit unstable and those sliders are very hard to catch. Can someone confirm this?Drafura wrote: 3pc1b
http://osu-ppy-sh.jeuxcrack.net/p/beatmap?b=27737&m=2
00:45:257 (1,1,1) - Kinda same problem some elements of sliders seems to trigger the hyper showed in the fruit and some not.
About the generation of hypers, yes it seems to trigger a little bit too easily for example :
http://osu-ppy-sh.jeuxcrack.net/p/beatmap?b=127363&m=2
02:27:853 (2,3,4) - Example of unrequired hyperdashes in a 2 fruits jump. The fruits are allready catchables and not that hard.
http://osu-ppy-sh.jeuxcrack.net/b/85550&m=2
00:29:714 (1,2,3) - Example of unrequired hyperdashes in a 3 fruits jump. The fruits are allready catchables and not that hard. I just realize with this map that with little circle size the hypers are triggered way to often.
More example wich shouldn't get hyperdash :
http://osu-ppy-sh.jeuxcrack.net/p/beatmap?b=93842&m=2
All extra hyperdashes added in this map are not needed.
http://osu-ppy-sh.jeuxcrack.net/p/beatmap?b=128872&m=2
00:16:303 (2) - Not required.
I hope this will be helpfull to adjust the triggering values.
http://osu-ppy-sh.jeuxcrack.net/p/beatmap?b=127363&m=2
02:27:853 (2,3,4) - Example of unrequired hyperdashes in a 2 fruits jump. The fruits are allready catchables and not that hard.
http://osu-ppy-sh.jeuxcrack.net/b/85550&m=2
00:29:714 (1,2,3) - Example of unrequired hyperdashes in a 3 fruits jump. The fruits are allready catchables and not that hard. I just realize with this map that with little circle size the hypers are triggered way to often.
More example wich shouldn't get hyperdash :
http://osu-ppy-sh.jeuxcrack.net/p/beatmap?b=93842&m=2
All extra hyperdashes added in this map are not needed.
http://osu-ppy-sh.jeuxcrack.net/p/beatmap?b=128872&m=2
00:16:303 (2) - Not required.
I hope this will be helpfull to adjust the triggering values.
Droplet hyperdashes still don't seem to work correctly. The first fruit of a slider now gives hyperdash up to the droplet as it should, but the droplet doesn't renew the hyperdash status and Ryuuta only moves at normal speed (which isn't always enough.) Fast Slider on my example map seems to still be uncatchable because of this.
Seph
Here's an example of hyperfruits spawning where they aren't needed.
This map on TAG4 difficulty:
00:12:042 (1,1,1) - these three notes are just barely catchable (you have about 1/10 second of leeway) but now they're much easier because hyperfruit spawn anyway.
This map on TAG4 difficulty:
00:12:042 (1,1,1) - these three notes are just barely catchable (you have about 1/10 second of leeway) but now they're much easier because hyperfruit spawn anyway.
This has extra hyperfruits and it was possible to FC before. They are unneeded.
00:19:900 (12,1) - in this map is perfectly catchable without a hyperfruit.
http://osu-ppy-sh.jeuxcrack.net/b/90192
I really don't like thinking about making already possible maps easier just to fix some bug that affect 0.01% of maps.
00:19:900 (12,1) - in this map is perfectly catchable without a hyperfruit.
http://osu-ppy-sh.jeuxcrack.net/b/90192
I really don't like thinking about making already possible maps easier just to fix some bug that affect 0.01% of maps.
Then peppy obviously set the catchable distance rate wayy too short, because the hyperfruit change to fix these issues would have been a tiny fraction of a change in the spacing differential. This is a large, way overdoing it change. It fixes it, but it's breaking other things.
There shouldn't be a limit that's considered too difficult to catch. If it's catchable, then it should be allowed. the hyperfruits shouldn't present a threshold for too difficult, unless that threshold is uncatchable.
There shouldn't be a limit that's considered too difficult to catch. If it's catchable, then it should be allowed. the hyperfruits shouldn't present a threshold for too difficult, unless that threshold is uncatchable.
Slider behavior has improved somewhat - now all fruits in it are catchable. However, the hyperdash now ignores droplets completely, and as a result it speeds past most of them, making SS impossible. Additionally, the (1,1,1) - in the map I previously linked still generates hyperfruit when it does not need to.
The example I posted earlier is now resolved. I will look for other maps though. Kind of a pain that in order to properly test I keep having to switch from test build to regular build. Shouldn't include this update until everything is thoroughly tested or we might not be able to figure out the full effect of these changes.
Yup catching droplets depend on the slider shape/consistency, for some sliders they seems to be totally ignored.
Antoher wierd behavior is about very very long sliders like this one :
http://osu-ppy-sh.jeuxcrack.net/s/19156 @ 01:40:987 (3) - The hyper is here but the speed up doesn't seems to work if you aren't in the very good timing.
About hyperdash spawning on lot of maps :
It's a bit better now but not enough. Too many hyperdashes seems to spawn on a lot of maps. My previous post contain maps wich still have this problem.
Antoher wierd behavior is about very very long sliders like this one :
http://osu-ppy-sh.jeuxcrack.net/s/19156 @ 01:40:987 (3) - The hyper is here but the speed up doesn't seems to work if you aren't in the very good timing.
About hyperdash spawning on lot of maps :
It's a bit better now but not enough. Too many hyperdashes seems to spawn on a lot of maps. My previous post contain maps wich still have this problem.
I checked With a dance number.
Here's what I found
Before:
Hyperfruits
00:07:889 (3) -
00:35:557 -
00:39:104 -
00:42:551 (3,4,5) -
00:45:794 -
00:53:294 (7) -
01:02:112 (4) -
01:11:537 -
01:15:186 (4,5,6) -
01:17:416 (8) -
01:21:267 (2,3,4) -
01:45:185 -
02:34:746 -
02:36:368 -
02:56:739 (1,2,3) -
03:11:132 (4) -
03:11:435 -
03:15:996 -
03:17:821 (1) -
03:28:970 (7) -
03:29:984 (6,7,8) -
03:36:064 -
03:36:571 (5) -
03:43:260 -
03:45:592 (2) -
03:47:314 -
03:48:632 (1,2) -
03:55:929 (1) -
03:56:335 (3) -
04:01:807 -
04:06:673 (2) -
04:14:784 -
04:16:504 (3) -
04:19:138 -
After:
check and new indicate hyper fruits that I believe are newly spawned. I might have missed one or two in the original though.
00:07:889 (3) -
00:34:746 -
00:35:557 -
00:39:104 -
00:42:551 (3,4,5) -
00:45:794 -
00:48:835 (2) - (check)
00:53:294 (7) -
00:59:983 (1,2) - (check)
01:01:706 (2,3,4,5) - (check)
01:09:915 -
01:11:537 -
01:13:969 - (check)
01:15:186 (4,5,6) -
01:17:416 (8) -
01:21:267 (2,3,4) -
01:29:173 (1,2) - (check)
01:34:038 (1) - (check) 1.5
01:38:700 (8) - (check) 1.6 snap
01:44:172 - (new)
01:44:679 -
01:45:185 -
01:45:591 - (check)
01:45:794 - ^
01:46:199 - ^
01:58:564 - ^
01:59:578 (2) - ^ 1.5
02:34:341 - (check)
02:34:746 -
02:35:962 - (check)
02:36:368 -
02:42:652 - Why you should make your spacing consistent....
02:44:273 -
02:54:003 - (check)
02:54:611 - (check)
02:54:814 - (check)
02:55:016 - (check)
02:55:827 - (check)
02:56:739 (1,2,3) -
03:11:132 (4) -
03:11:435 -
03:15:996 -
03:17:821 (1) -
03:19:746 - (check) 1.7 snap
03:22:888 - (check)
03:24:510 - (check)
03:28:970 (7) -
03:29:984 (6,7,8,9) - (check 9)
03:32:720 (2,3) - (check)
03:34:645 - (check) -doesn't seem like a proper hyper to me
03:36:064 -
03:36:571 (5) -
03:43:260 -
03:45:592 (2) -
03:47:314 -
03:48:632 (1,2) -
03:50:558 (3) - (check) 1.51
03:55:929 (1) -
03:56:335 (3) -
04:00:591 - (check)
04:01:807 -
04:03:328 (2) - (check) 1.5 suspicious hyper
04:06:673 (2) -
04:07:889 (7) - (check)
04:10:321 - (check)
04:13:969 - (check)
04:14:578 (1,2,3,4) - (check)
04:16:504 (3) -
04:18:327 - (check)
04:18:733 - (check)
04:19:138 -
04:19:341 - (check)
Here's what I found
Before:
Hyperfruits
00:07:889 (3) -
00:35:557 -
00:39:104 -
00:42:551 (3,4,5) -
00:45:794 -
00:53:294 (7) -
01:02:112 (4) -
01:11:537 -
01:15:186 (4,5,6) -
01:17:416 (8) -
01:21:267 (2,3,4) -
01:45:185 -
02:34:746 -
02:36:368 -
02:56:739 (1,2,3) -
03:11:132 (4) -
03:11:435 -
03:15:996 -
03:17:821 (1) -
03:28:970 (7) -
03:29:984 (6,7,8) -
03:36:064 -
03:36:571 (5) -
03:43:260 -
03:45:592 (2) -
03:47:314 -
03:48:632 (1,2) -
03:55:929 (1) -
03:56:335 (3) -
04:01:807 -
04:06:673 (2) -
04:14:784 -
04:16:504 (3) -
04:19:138 -
After:
check and new indicate hyper fruits that I believe are newly spawned. I might have missed one or two in the original though.
00:07:889 (3) -
00:34:746 -
00:35:557 -
00:39:104 -
00:42:551 (3,4,5) -
00:45:794 -
00:48:835 (2) - (check)
00:53:294 (7) -
00:59:983 (1,2) - (check)
01:01:706 (2,3,4,5) - (check)
01:09:915 -
01:11:537 -
01:13:969 - (check)
01:15:186 (4,5,6) -
01:17:416 (8) -
01:21:267 (2,3,4) -
01:29:173 (1,2) - (check)
01:34:038 (1) - (check) 1.5
01:38:700 (8) - (check) 1.6 snap
01:44:172 - (new)
01:44:679 -
01:45:185 -
01:45:591 - (check)
01:45:794 - ^
01:46:199 - ^
01:58:564 - ^
01:59:578 (2) - ^ 1.5
02:34:341 - (check)
02:34:746 -
02:35:962 - (check)
02:36:368 -
02:42:652 - Why you should make your spacing consistent....
02:44:273 -
02:54:003 - (check)
02:54:611 - (check)
02:54:814 - (check)
02:55:016 - (check)
02:55:827 - (check)
02:56:739 (1,2,3) -
03:11:132 (4) -
03:11:435 -
03:15:996 -
03:17:821 (1) -
03:19:746 - (check) 1.7 snap
03:22:888 - (check)
03:24:510 - (check)
03:28:970 (7) -
03:29:984 (6,7,8,9) - (check 9)
03:32:720 (2,3) - (check)
03:34:645 - (check) -doesn't seem like a proper hyper to me
03:36:064 -
03:36:571 (5) -
03:43:260 -
03:45:592 (2) -
03:47:314 -
03:48:632 (1,2) -
03:50:558 (3) - (check) 1.51
03:55:929 (1) -
03:56:335 (3) -
04:00:591 - (check)
04:01:807 -
04:03:328 (2) - (check) 1.5 suspicious hyper
04:06:673 (2) -
04:07:889 (7) - (check)
04:10:321 - (check)
04:13:969 - (check)
04:14:578 (1,2,3,4) - (check)
04:16:504 (3) -
04:18:327 - (check)
04:18:733 - (check)
04:19:138 -
04:19:341 - (check)
Are you sure you really wanted this?
There wont be super difficult CTb maps as you know it now, because anything past a certain degree of catchability is now hypered. So anything that is 1.9 plus horizontal is now a hyper in With a dance Number, when before only some of them were. Fixing this is going to drastically reduce the difficulty of maps and its not like you can just bring that difficulty back.
There wont be super difficult CTb maps as you know it now, because anything past a certain degree of catchability is now hypered. So anything that is 1.9 plus horizontal is now a hyper in With a dance Number, when before only some of them were. Fixing this is going to drastically reduce the difficulty of maps and its not like you can just bring that difficulty back.
If you take a look at my profile: 220 plays - Hatsune Miku - With a Dance Number [0108 style] (Probably +400, counting offline tries and before-ranking tries).
That means it's a pretty well memorised map. I'm quite sure somebody who hasn't played it so much won't be able to do a nice performance there. It's challenging enough even with the new HDashes in my opinion, but I'm pretty sure most pros will rage about the "simplicity" of such a popular map.
That means it's a pretty well memorised map. I'm quite sure somebody who hasn't played it so much won't be able to do a nice performance there. It's challenging enough even with the new HDashes in my opinion, but I'm pretty sure most pros will rage about the "simplicity" of such a popular map.
I'm most worried about getting huge limitation in term non hyper jump difficulty for ctb mapping than for converted maps. Non hyperdashed jumps are part of CtB, think about Zhsteven's maps, he don't use that many hyperdashes, and it is the intetended behavior he wanted while mapping.Deif wrote: 6h6j6d
That means it's a pretty well memorised map. I'm quite sure somebody who hasn't played it so much won't be able to do a nice performance there. It's challenging enough even with the new HDashes in my opinion, but I'm pretty sure most pros will rage about the "simplicity" of such a popular map.
Imo jumps wich are already catchable by most experienced players shouldn't change. If some jumps have to became hyper to fix the bug i'm ok, but i'm totally against a huge change wich will affect a lot of maps (inculding ctb maps, ranked or not).
http://osu-ppy-sh.jeuxcrack.net/p/beatmap?b=93842&m=2
I've tried test build but this map still has problem. 210x~220x part(without hyper) was the best part of that beatmap T_T
I don't know what is the current triggering mechanism but I think the best way is just stick to previous triggering method but add some situations
1) on multiple jumps to same direction: add hyperdash where a jump can't be catched if previous jumps were catched
2) add hyperdash on very fast sliders
I'm pretty sure that current one is not like this. I didn't check case 2 yet though..
I've tried test build but this map still has problem. 210x~220x part(without hyper) was the best part of that beatmap T_T
I don't know what is the current triggering mechanism but I think the best way is just stick to previous triggering method but add some situations
1) on multiple jumps to same direction: add hyperdash where a jump can't be catched if previous jumps were catched
2) add hyperdash on very fast sliders
I'm pretty sure that current one is not like this. I didn't check case 2 yet though..
I'm thinking that this would also help find out cheaters, to some degree. People would take advantage of this limitation and cheat their catcher with some boosted speed.peppy wrote: 6w4l42
I also added an alert when watching a replay which moves faster than humanely possible, which only displays on the test build. Might be handy for checking any replays you think may not be legit, and pointing them out to me in a PM for confirmation.
I think is not the correct way to fix this, the spacing are fine as it is, the only problem are the "pixel jumps" and there should be a way to fix them without changing anything else.TheVileOne wrote: 4cy22
Are you sure you really wanted this?
There wont be super difficult CTb maps as you know it now, because anything past a certain degree of catchability is now hypered. So anything that is 1.9 plus horizontal is now a hyper in With a dance Number, when before only some of them were. Fixing this is going to drastically reduce the difficulty of maps and its not like you can just bring that difficulty back.
peppy, I'm not sure it should be "fixed" in the sense that all fruits that meet a certain criteria should be hypered. Because if that is the case, it will alter maps that were previously possible without said hypers. I'm sure there will be hate if people start playing their uber pro map and find these hypers that weren't there before, because the algorithm was broken before, and now the fixed result is easier than how people knew it.
I am not a super pro or regularly play CTB. So I cannot speak on behalf of the pro community without the pros assessing their favorite maps and potential changes to the patterns. Really we need people checking difficulties for pattern consistencies and checking how far one should reach for a pattern.
I am not a super pro or regularly play CTB. So I cannot speak on behalf of the pro community without the pros assessing their favorite maps and potential changes to the patterns. Really we need people checking difficulties for pattern consistencies and checking how far one should reach for a pattern.
The idea is that only fruits which were impossible before (or possible only with certain FPS that gives you faster speed than intended) will have hyperdashes added, and all other hyperdashes will stay as they were. If it wasn't important that this was the case, the hyperdash fix would already be complete and have gone live.
Pseudo code (hopefully better than the current one)
For each jump in the 3-note jump situation alone, we can catch the two notes with dash starting from the right side of the first fruit (and ends up hitting the left side of the second fruit). But considering the 3-note jump as a whole, this algorithm should be able to detect that the third fruit is without reach, spawning hyperdash on the second fruit.
EDIT: various fixes and some comments
EDIT2: brief explanation
SPOILER
The point is that we work out the reasonable range the catcher is currently able to reach, and base hyperdash spawning on this.prevFruit.x = 256; // Arbitrary initial values
prevFruit.t = -Inf;
minX = 0; // Leftmost catcher position
maxX = 512; // Rightmost catcher position
prevMove = Normal; // Bit flags, Normal=0
// SpaceLeniency and TimeLeniency may (or may not) depend on CircleSize and OverallDifficulty respectively
// Old (public build) behavior is roughly equivalent to SpaceLeniency = CatcherWidth * 0.1875 and TimeLeniency = 0
// when successive hyperdashes are not encountered with
foreach (fruit in fruits)
{
if (fruit is spinner or something like this)
{
continue;
}
fruitLeft = Max(0, fruit.x - CatcherHalfWidth + SpaceLeniency);
fruitRight = Min(512, fruit.x + CatcherHalfWidth - SpaceLeniency);
minX -= Max(0, fruit.t - prevFruit.t - ((prevMove & LeftFullDash) != 0 ? 0 : TimeLeniency)) * DashSpeed;
maxX += Max(0, fruit.t - prevFruit.t - ((prevMove & RightFullDash) != 0 ? 0 : TimeLeniency)) * DashSpeed;
prevMove = Normal;
if (minX > fruitRight)
{
prevFruit.isHyperDash = true;
prevMove |= LeftFullDash;
minX = fruit.x;
maxX = fruit.x;
}
else if (minX >= fruitLeft)
{
prevMove |= LeftFullDash;
}
else
{
minX = fruitLeft;
}
if (maxX < fruitLeft)
{
prevFruit.isHyperDash = true;
prevMove |= RightFullDash;
minX = fruit.x;
maxX = fruit.x;
}
else if (maxX <= fruitRight)
{
prevMove |= RightFullDash;
}
else
{
maxX = fruitRight;
}
prevFruit = fruit;
}
For each jump in the 3-note jump situation alone, we can catch the two notes with dash starting from the right side of the first fruit (and ends up hitting the left side of the second fruit). But considering the 3-note jump as a whole, this algorithm should be able to detect that the third fruit is without reach, spawning hyperdash on the second fruit.
EDIT: various fixes and some comments
EDIT2: brief explanation
Last Remnant_old
Simply put,
Let's say that there is a FOR loop that goes through all fruits and that we are currently at some fruit $i. Current algorithm will take a look at $i+1 fruit's time and X coordinate and decide whether fruit $i will become hyper or not.
This needs just a small tweak:
Let's say that at past example we concluded that fruit $i doesn't need to be hyper (in other words, jump $i - $i+1 is catchable), now we also need to check fruit $i+2 (and maybe $i+3, although cases with 4 fruits in a row are very rare, I know only 1 map). If we conclude that jump $i - $i+2 isn't catchable, we must make both fruits $i and $i+1 hypers.
Here jumps 1 - 2 and 2 - 3 are catchable, while jump 1 - 3 isn't, which is shown on the next picture.
Let's say that there is a FOR loop that goes through all fruits and that we are currently at some fruit $i. Current algorithm will take a look at $i+1 fruit's time and X coordinate and decide whether fruit $i will become hyper or not.
This needs just a small tweak:
Let's say that at past example we concluded that fruit $i doesn't need to be hyper (in other words, jump $i - $i+1 is catchable), now we also need to check fruit $i+2 (and maybe $i+3, although cases with 4 fruits in a row are very rare, I know only 1 map). If we conclude that jump $i - $i+2 isn't catchable, we must make both fruits $i and $i+1 hypers.
Here jumps 1 - 2 and 2 - 3 are catchable, while jump 1 - 3 isn't, which is shown on the next picture.
That isn't the only case, but it is one of them. I have already allowed for this. Thanks for your input with proposed solutions, but please wait until I make further changes as I already know exactly what is wrong and how I will fix it.
Keep in mind my current fix fixes not only the main case you specify, but others which were mentioned by Millhiore. We are not dealing with a single point of failure here.
Keep in mind my current fix fixes not only the main case you specify, but others which were mentioned by Millhiore. We are not dealing with a single point of failure here.
Last Remnant_old
Yeah, the smallest droplets in middle of slider also causes break (as if in my example fruit 2 was droplet) and other things mentioned by MillhioreF.
There is one more thing I would like to ask, not much related to this, but it is connected to "traversing the distance" with catcher :
In past weeks / months we could see that lags / pauses and other tricks (even not tricks, just some PCs having natural lags that are helpful) could be used to make your catcher cover more distance for the same time (thus many players used this to get these same impossible jumps we are trying to fix). This is because catcher's movement and song aren't synchronized. I'm not exactly sure why, since song's timer is updated on 10ms intervals. Isn't it possible to tie catcher's movement to song, same as hitObjects are synced (in other words, update catcher's position by the fixed distance each time the song timer is updated)? This means that in each update catcher may either stay put, or move left or right by that same distance.
This is also related to your recent implementation of ALERT feature, which is triggered for almost any replay (probably because each PC has some kind of lag that makes catcher's speed uncontrollable and variable)
PS. Since we want equal conditions for everyone, I think something could be done about this. (already present check that won't submit your score if catcher is faster or slower isn't good enough, maybe make it more strict? )
There is one more thing I would like to ask, not much related to this, but it is connected to "traversing the distance" with catcher :
In past weeks / months we could see that lags / pauses and other tricks (even not tricks, just some PCs having natural lags that are helpful) could be used to make your catcher cover more distance for the same time (thus many players used this to get these same impossible jumps we are trying to fix). This is because catcher's movement and song aren't synchronized. I'm not exactly sure why, since song's timer is updated on 10ms intervals. Isn't it possible to tie catcher's movement to song, same as hitObjects are synced (in other words, update catcher's position by the fixed distance each time the song timer is updated)? This means that in each update catcher may either stay put, or move left or right by that same distance.
This is also related to your recent implementation of ALERT feature, which is triggered for almost any replay (probably because each PC has some kind of lag that makes catcher's speed uncontrollable and variable)
PS. Since we want equal conditions for everyone, I think something could be done about this. (already present check that won't submit your score if catcher is faster or slower isn't good enough, maybe make it more strict? )
The behavior seems to have gotten worse - none of the scenarios in my example map work correctly any longer.
-In the 3-note jump, the first fruit hyperdashes while the second doesn't, which is probably intentional. However, due to the catcher slowing down once you're under the next note while still in hyperfruit status, it's impossible to get over far enough to catch the third fruit. Both should still grow hyperdashes.
-The spaced stream doesn't grow hyperfruits at all, and so is still impossible.
-The slider has reverted to the state it was in after the first fix:
-In the 3-note jump, the first fruit hyperdashes while the second doesn't, which is probably intentional. However, due to the catcher slowing down once you're under the next note while still in hyperfruit status, it's impossible to get over far enough to catch the third fruit. Both should still grow hyperdashes.
-The spaced stream doesn't grow hyperfruits at all, and so is still impossible.
-The slider has reverted to the state it was in after the first fix:
MillhioreF wrote: 6w3d4u
The first fruit of a slider now gives hyperdash up to the droplet as it should, but the droplet doesn't renew the hyperdash status and Ryuuta only moves at normal speed (which isn't always enough.)
The second fruit absolutely needs hyperdash while the first doesn't really require that (getting hyperdash on both fruits makes sense though).MillhioreF wrote: 6w3d4u
-In the 3-note jump, the first fruit hyperdashes while the second doesn't, which is probably intentional. However, due to the catcher slowing down once you're under the next note while still in hyperfruit status, it's impossible to get over far enough to catch the third fruit. Both should still grow hyperdashes.
Sliders :
Hyperdash on sliders doesn't works ryuuta seems to stop at the droplet, even if the droplet is hyperdashed the ryuuta stops on it.
1-1 jumps :
Seen too many extra hyperdashes appearing on many maps.
1-1-1 jumps :
Doesn't seems to spawn in the good placements. Most of the time only the first fruit is hyperdash'd and the behavior is kinda same as sliders, the ryuuta stops in the second fruit.
Hyperdash on sliders doesn't works ryuuta seems to stop at the droplet, even if the droplet is hyperdashed the ryuuta stops on it.
1-1 jumps :
Seen too many extra hyperdashes appearing on many maps.
1-1-1 jumps :
Doesn't seems to spawn in the good placements. Most of the time only the first fruit is hyperdash'd and the behavior is kinda same as sliders, the ryuuta stops in the second fruit.
If there's the same spacing between the 3 fruits the 2 first fruits should be hyperdash'd. Makes much more sense for mapping, and also in term of readability of the map.statementreply wrote: 4c3764
The second fruit absolutely needs hyperdash while the first doesn't really require that (getting hyperdash on both fruits makes sense though).
http://ha.ppy.sh/ss/1273 It should be exactly the opposite, only the center one requires hyper.
All right, here's one for the tutorial. (Watashi mp3 was too short) The first and third fruits become hyper, making the pattern uncatchable. If the second and fourth fruits were hyper though, it would be.
http://puu.sh/1IMkE
http://puu.sh/1IMkE
http://ha.ppy.sh/ss/1277 The pixel jumps are again uncatchable, no hyper dash appears x_x
There are some unnecesary HDashes as well, taking this map as reference, which contains extreme patterns: http://osu-ppy-sh.jeuxcrack.net/s/19156 [Sp's Darkness]
In the described cases, those notes had a not-so-long distance with the next ones, thus they shouldn't be HDashed.
In the described cases, those notes had a not-so-long distance with the next ones, thus they shouldn't be HDashed.
I've noticed that droplets are totally ignored during sliders making them quite impossible to catch for most cases.
Here's a map to easily test play them : http://osu-ppy-sh.jeuxcrack.net/p/beatmap?b=46477&m=2 @ 00:22:537 (1,2,3) -
I've succeeded to catch sometimes the droplets adjusting my dash during the hyper, in other words it's greatly luck based. For other maps it seems like they're just impossible to catch.
Here's a map to easily test play them : http://osu-ppy-sh.jeuxcrack.net/p/beatmap?b=46477&m=2 @ 00:22:537 (1,2,3) -
I've succeeded to catch sometimes the droplets adjusting my dash during the hyper, in other words it's greatly luck based. For other maps it seems like they're just impossible to catch.
If so, that can be discussed in the CTB rules thread to avoid putting sliders at such high speeds that they need to be HDashed. Otherwise, they'll be incopatible with:
- Your map must theoretically be SSable. This generally means you must be able to catch absolutely all fruits, including droplets. This ties in the previous point, but is fundamentally for different reasons.
http://puu.sh/1Kntd
First three patterns: There should be hyperdashes somewhere for all the fruits to be catchable, preferably on all fruits except the last one of each pattern. (3 note jumps)
The last pattern: The 3rd fruit should be spawn hyperdash. (jumps after hyperdash)
First three patterns: There should be hyperdashes somewhere for all the fruits to be catchable, preferably on all fruits except the last one of each pattern. (3 note jumps)
The last pattern: The 3rd fruit should be spawn hyperdash. (jumps after hyperdash)
.
3-note jump is working beautifully!
Spaced stream is a bit awkward since every 3rd fruit isn't hypered, but it should be fine, since it's still possible to catch everything. CTB players might even welcome this, since it's harder to catch the stream than it would be if every fruit was hypered.
Sliders with big droplets are still perfect.
Aside from missing the little droplets, which may not be such a big deal, sliders with only small droplets in between sliderends are a bit strange. If you aren't already moving in the direction you need to go when you catch the hyperslider fruit, the screen will darken and Ryuuta will turn red, but he won't actually get a speedboost and will miss the next fruit, making these unintuitive to catch. Normal fruit and sliders with only big droplets aren't affected by this - you can catch the fruit, wait a few milliseconds, then start moving and still make it to the next fruit.
Gonna edit this post once I analyze a map that's very hard but SS-able on the public build, making sure there aren't any hyperfruit where they aren't needed.
Spaced stream is a bit awkward since every 3rd fruit isn't hypered, but it should be fine, since it's still possible to catch everything. CTB players might even welcome this, since it's harder to catch the stream than it would be if every fruit was hypered.
Sliders with big droplets are still perfect.
Aside from missing the little droplets, which may not be such a big deal, sliders with only small droplets in between sliderends are a bit strange. If you aren't already moving in the direction you need to go when you catch the hyperslider fruit, the screen will darken and Ryuuta will turn red, but he won't actually get a speedboost and will miss the next fruit, making these unintuitive to catch. Normal fruit and sliders with only big droplets aren't affected by this - you can catch the fruit, wait a few milliseconds, then start moving and still make it to the next fruit.
Gonna edit this post once I analyze a map that's very hard but SS-able on the public build, making sure there aren't any hyperfruit where they aren't needed.
http://puu.sh/1KH7M (on tutorial)
00:03:808 (4) - Hyperdash isn't necessary
00:03:808 (4) - Hyperdash isn't necessary
I'll drop a few more .osu files for Watashi here with bugs that need fixing.
http://puu.sh/1KH0l - a triplet before some back-and-forths makes an unnecessary hyperdash spawn.
http://puu.sh/1KH54 - Ryuuta seems to overshoot the end of this slider if you keep holding right!
http://puu.sh/1KHbb - This is catchable, but suffers from the slider-with-small-droplets issue.
http://puu.sh/1KH0l - a triplet before some back-and-forths makes an unnecessary hyperdash spawn.
http://puu.sh/1KH54 - Ryuuta seems to overshoot the end of this slider if you keep holding right!
http://puu.sh/1KHbb - This is catchable, but suffers from the slider-with-small-droplets issue.
Fixed all these issues. Please check for any regressions and/or new cases.
I've written up exactly what was wrong with the previous calculations, for anyone interested enough: http://blog.ppy.sh/post/39917651337/get ... dash-right
I've written up exactly what was wrong with the previous calculations, for anyone interested enough: http://blog.ppy.sh/post/39917651337/get ... dash-right
It doesn't appear that there were any regressions. The triplet with back-and-forth was fixed, but Ryuuta still overshoots the slider in my second example difficulty. The slider-with-small-droplets issue seems to be somewhat better, but it's not completely fixed as far as I can tell. I might have to record it and slow down the recording to take a closer look at what's going on.
There are still a few misplaced hyperdashes, too. I've put up a difficulty with 3 examples from Talent Shredder of fruit that are hyperdashed, but don't actually need hyperdashes to be caught. (Sorry if they would have been better if they were each isolated, but I think they might be related, and also it's simpler to import this way)
Click here!
There are still a few misplaced hyperdashes, too. I've put up a difficulty with 3 examples from Talent Shredder of fruit that are hyperdashed, but don't actually need hyperdashes to be caught. (Sorry if they would have been better if they were each isolated, but I think they might be related, and also it's simpler to import this way)
Click here!
These cases involve calculations at the turn of single calculation frames. I have tweaked it slightly, so please test all previous cases for regressions. If you are still seeing hyperfruit being made where they weren't before, I think these are best left to for edge cases where lag could cause a miss/hit.
somethimes the catcher is crazy,
this bug appears when you fail the "end hdash fruit"
this bug appears when you fail the "end hdash fruit"
:3
scarlet rose
scarlet rose
1. Trigger hyperdashlineqtxz wrote: t4b5e
somethimes the catcher is crazy,
this bug appears when you fail the "end hdash fruit"
:3
scarlet rose
2. Change direction before hyperdash ends while holding dash key
3. Ryuuta moves in the opposite direction at hyperdash speed (should be normal dash speed)
It seems that hyperdash exits ~4ms before accurate timing now (~22ms before, with an additional 1-frame slowdown), which could make it harder to control on exiting. If you stop moving or change direction a bit earlier, the catcher could be quite far away. I'm not quite sure about the outcome of such changes though, let's see how pro ctb players feel about the changepeppy wrote: 6w4l42
I find the "change direction" hyperdash speed to be correct. You shouldn't be changing direction until it ends.
Another bug is that when you miss the target fruit of a hyperdash, the graphics is stuck in hyperdash mode (red catcher, dimmed background).
Red: change direction 10ms early
Green: change direction at the accurate timing
Blue: change direction 10ms late
Not to the scale
It's quite tricky to handle the ryuuta now, specially in over-HDashed songs, where you now need to do a lot of fine movements to control those HDashes. The only song I can think that could be benefited with this new movement algorythm is Cirno's Perfect Math Class, imo...statementreply wrote: 4c3764
It seems that hyperdash exits ~4ms before accurate timing now (~22ms before, with an additional 1-frame slowdown), which could make it harder to control on exiting. If you stop moving or change direction a bit earlier, the catcher could be quite far away. I'm not quite sure about the outcome of such changes though, let's see how pro ctb players feel about the change
From my quick tests, hyperdash spawning is now just about perfect, and droplets are even catchable due to the "tighter" hyperdash controls.
However, it's become much harder to actually hit hyperdashes because of this, and I probably speak for most of the CTB community when I say that they'd rather have FC-able maps with the old hyperdash controlling than SS-able maps with the much tighter, harder controlling. Maybe you should save a better droplet solution for a future patch. Perhaps adjust droplet location only on sliders that are hypered to fall into the natural hyperdash range? Not high priority though.
However, it's become much harder to actually hit hyperdashes because of this, and I probably speak for most of the CTB community when I say that they'd rather have FC-able maps with the old hyperdash controlling than SS-able maps with the much tighter, harder controlling. Maybe you should save a better droplet solution for a future patch. Perhaps adjust droplet location only on sliders that are hypered to fall into the natural hyperdash range? Not high priority though.
statementreply wrote: 4c3764
1. Trigger hyperdashlineqtxz wrote: t4b5e
somethimes the catcher is crazy,
this bug appears when you fail the "end hdash fruit"
:3
scarlet rose
2. Change direction before hyperdash ends while holding dash key
3. Ryuuta moves in the opposite direction at hyperdash speed (should be normal dash speed)
I think I got it wrong. It's definitely a bug.peppy wrote: 6w4l42
I find the "change direction" hyperdash speed to be correct. You shouldn't be changing direction until it ends
1. Catch the starting fruit of a hyperdash (but do not complete the hyperdash)
2. Try moving ryuuta around
Now ryuuta keeps at hyperdash speed until he moves onto the position (x) of the target fruit of that hyperdash.
Well, to be fair, I edited your post after a few pages xD
In other news, peppy, statementreply, Deif and I have been discovering bugs in and ironing out the hyperdash on the test server the last couple of days. There's still a couple of minor bugs, (which will probably get their own threads) and hyperdashes are a bit tighter now to make up for formula changes, but all maps should now be FC-able and probably even SS-able. The changes are virtually done, so expect the patch to go public soon.
In other news, peppy, statementreply, Deif and I have been discovering bugs in and ironing out the hyperdash on the test server the last couple of days. There's still a couple of minor bugs, (which will probably get their own threads) and hyperdashes are a bit tighter now to make up for formula changes, but all maps should now be FC-able and probably even SS-able. The changes are virtually done, so expect the patch to go public soon.
Most experienced people will find the change to be very hard to adapt to, since it would seem like all the luck you needed to hit the pixels was transferred to the hyperdash controlling thing. I mean, now if I go and try some TAG4, only chance will tell if I'll be able to have total control of the Ryuuta when it's in hyperdash status so no, this is a regressive change and I suggest to fix the speed of these special fruits back to their normal state.MillhioreF wrote: 6w3d4u
From my quick tests, hyperdash spawning is now just about perfect, and droplets are even catchable due to the "tighter" hyperdash controls.
However, it's become much harder to actually hit hyperdashes because of this, and I probably speak for most of the CTB community when I say that they'd rather have FC-able maps with the old hyperdash controlling than SS-able maps with the much tighter, harder controlling. Maybe you should save a better droplet solution for a future patch. Perhaps adjust droplet location only on sliders that are hypered to fall into the natural hyperdash range? Not high priority though.
Basically I tried to hit my directional keys <- / -> the same way I did with some TAG4 like http://osu-ppy-sh.jeuxcrack.net/b/58064?m=2 and it was a total failure compared to the old hyperdash controlling state. If you want to improve the skill spectrum of CtB, you're doing it the wrong way.
P/D: I totally agree in fixing the Hyperdash-not-always-spawning bug, but try to do just that and not change other gameplay aspects of the game.
Sorry for my grammatic mistakes.
No, just no. It wasn't as much hard as you're predicting (for me) in 99.9% of the songs. I've been testing different TAG4 difficulties, as it seems to be your major preocupation, and these are the results (disregard NoFail and multiply all the scores by 2):VelperK wrote: 2v2r1m
Most experienced people will find the change to be very hard to adapt to, since it would seem like all the luck you needed to hit the pixels was transferred to the hyperdash controlling thing.
Comparing to my average scores in those songs, I must it those are only normal tries and not spectacular scores (I didn't pretend to do OMGWTFBBQ records anyway). That means the gameplay in those songs has not (or maybe slightly) changed with the last changes. StrangeProgram is probably the most benefited song of those, as there were a lot of speedy sliders that used to be impossible before.
I encourage experienced players to go into the test build and help to develop this new HDash system, instead of complaining and not offering any constructive solution.
And I encourage you to try this map 5 times in both osu! builds (the public and the private): http://osu-ppy-sh.jeuxcrack.net/b/74845Deif wrote: 6h6j6d
No, just no. It wasn't as much hard as you're predicting (for me) in 99.9% of the songs. I've been testing different TAG4 difficulties, as it seems to be your major preocupation, and these are the results (disregard NoFail and multiply all the scores by 2):VelperK wrote: 2v2r1m
Most experienced people will find the change to be very hard to adapt to, since it would seem like all the luck you needed to hit the pixels was transferred to the hyperdash controlling thing.
Comparing to my average scores in those songs, I must it those are only normal tries and not spectacular scores (I didn't pretend to do OMGWTFBBQ records anyway). That means the gameplay in those songs has not (or maybe slightly) changed with the last changes. StrangeProgram is probably the most benefited song of those, as there were a lot of speedy sliders that used to be impossible before.
I encourage experienced players to go into the test build and help to develop this new HDash system, instead of complaining and not offering any constructive solution.
Please, do post the results of each osu! build and then say the same.
Tried it with the original AR in both versions. Just 1 try was necessary:VelperK wrote: 2v2r1m
And I encourage you to try this map 5 times in both osu! builds (the public and the private): http://osu-ppy-sh.jeuxcrack.net/b/74845
Please, do post the results of each osu! build and then say the same.
Same accuracy, and 2 misses of difference. The test build fixes some pixel-jumps the original map has but on the other hand, the sudden direction changes are harder now.
Nevertheless, that's not a very good example, as it's not even following the criteria of rankability in CTB maps.
(Sorry I couldn't connect these days to test. /mylife)
Uh. I still can find extra unrequired hypers.
For example : http://osu-ppy-sh.jeuxcrack.net/s/40071 [Posthumous] @ 01:31:959 (2) -
I don't really like the new behavior of the hyperdashes... I think i'll have to rethink many of my patterns in quite all my maps, I used them to make huge jumps wich don't require an accurate timing, the update about this point is basically the oposite of how I see an hyperdash and how I use them in mapping.
Uh. I still can find extra unrequired hypers.
For example : http://osu-ppy-sh.jeuxcrack.net/s/40071 [Posthumous] @ 01:31:959 (2) -
I don't really like the new behavior of the hyperdashes... I think i'll have to rethink many of my patterns in quite all my maps, I used them to make huge jumps wich don't require an accurate timing, the update about this point is basically the oposite of how I see an hyperdash and how I use them in mapping.