The_Jiggernaut
Smash Ace
- Joined
- Jul 7, 2008
- Messages
- 649
Despite claims that rest in Smash 4 was now usable to to engine changes and other factors, I had never seen the hard and fast numbers to support this. So @Linkshot and I took up the challenge and did the grueling measurements and frame-data analysis ourselves. We found out a handful of previously undiscovered details about the game and got a rather strange answer to sing's viability.
The full analysis with explanations how how things were done can be found here: https://docs.google.com/spreadsheets/d/1n5yQ2fySt8JymrjhAPEybbVBEfL0U1M7tSTuZXAZu5s/edit?usp=sharing
WARNING: It's rather formal, and my physics background is kind of obvious
If your eyes glaze over at charts, My written analysis is below. If you don't like reading my TL;DR verson is below.
Ok so first major discovery: Each of the 3 parts of sing that put people to sleep have 2 hitboxes each. One is smaller and puts enemies to sleep for longer and the second one comes out a frame later, is about the size of the sing graphic (ie larger than the first), but puts enemies to sleep shorter. If the closer one hits, the second hitbox doesn't effect the enemy.
The size difference of the hitboxes is hard to describe. It's slight enough that it was difficult to walk far enough to get hit by one and not the other. If you want to get the hang of the difference training mode is your friend in that regard.
Second major thing: We have a strong theoretical understanding of how the game reads controller input for struggling out of sleep and in what way the game enforces a minimum sleep time. I go into more detail in the document, but it seems the game only checks if there is ANY input on each frame. Pressing two buttons on a frame counts as 1 input; pressing every button on the controller in the same frame only counts as 1 input. Spinning (not flicking or mashing) the control stick counts as many spread out inputs, so this is the best way to escape grabs and wake up from sleep. Pressing any of the other buttons does not help, and in practice will likely reduce the speed at which you spin the control stick. This is a nice little tidbit on it's own, but it means that escaping sleep at the quickest the game allows has never been easier to pull off.
And lastly is the initial goal: When is it safe to sing? Obviously the less input someone does to escape the sleep, the easier you can punish them. However, since it's fairly easy for someone preform a top-level escape, let's just talk about the last of the 3 data sets. If you hit with the close hitbox, you have a frame advantage at 60%, but if you instead hit with the weaker hitbox, you're only safe at 160%. You can get a forward tilt in at these percents, but honestly, that's about it. And also, at 160% you might as well uptilt out of your roll read instead of singing.
If you can jump in and rest in less than half a second, then hitting with the close hitbox of sing at 100+% is for you. However, if you miss and only hit with the longer sing hitbox, you WILL get punished. Forget everything you knew about how hard sing was to land before, it just got MUCH harder.
it's long so TL;DR:
OMG you need to be super close to your opponent for sing to work properly, or they won't sleep at long
WTF spinning the control stick and pressing no other buttons is the fastest way to escape sleep and grabs
BBQ Sing is only safe on hit at 160+% and if half a second isn't enough for a decent punish then sing is just about never useful ever. Dreams crushed
Also, I move that we have a Jigglypuff Data thread, where the results of this and rest percentages with/without DI, and the upTilt percentages are all put in the first few posts with discussions of other data in the thread and then added to the first post afterwards. If no one wants to do that, I'm up for heading the Data thread.
The full analysis with explanations how how things were done can be found here: https://docs.google.com/spreadsheets/d/1n5yQ2fySt8JymrjhAPEybbVBEfL0U1M7tSTuZXAZu5s/edit?usp=sharing
WARNING: It's rather formal, and my physics background is kind of obvious
If your eyes glaze over at charts, My written analysis is below. If you don't like reading my TL;DR verson is below.
Ok so first major discovery: Each of the 3 parts of sing that put people to sleep have 2 hitboxes each. One is smaller and puts enemies to sleep for longer and the second one comes out a frame later, is about the size of the sing graphic (ie larger than the first), but puts enemies to sleep shorter. If the closer one hits, the second hitbox doesn't effect the enemy.
The size difference of the hitboxes is hard to describe. It's slight enough that it was difficult to walk far enough to get hit by one and not the other. If you want to get the hang of the difference training mode is your friend in that regard.
Second major thing: We have a strong theoretical understanding of how the game reads controller input for struggling out of sleep and in what way the game enforces a minimum sleep time. I go into more detail in the document, but it seems the game only checks if there is ANY input on each frame. Pressing two buttons on a frame counts as 1 input; pressing every button on the controller in the same frame only counts as 1 input. Spinning (not flicking or mashing) the control stick counts as many spread out inputs, so this is the best way to escape grabs and wake up from sleep. Pressing any of the other buttons does not help, and in practice will likely reduce the speed at which you spin the control stick. This is a nice little tidbit on it's own, but it means that escaping sleep at the quickest the game allows has never been easier to pull off.
And lastly is the initial goal: When is it safe to sing? Obviously the less input someone does to escape the sleep, the easier you can punish them. However, since it's fairly easy for someone preform a top-level escape, let's just talk about the last of the 3 data sets. If you hit with the close hitbox, you have a frame advantage at 60%, but if you instead hit with the weaker hitbox, you're only safe at 160%. You can get a forward tilt in at these percents, but honestly, that's about it. And also, at 160% you might as well uptilt out of your roll read instead of singing.
If you can jump in and rest in less than half a second, then hitting with the close hitbox of sing at 100+% is for you. However, if you miss and only hit with the longer sing hitbox, you WILL get punished. Forget everything you knew about how hard sing was to land before, it just got MUCH harder.
it's long so TL;DR:
OMG you need to be super close to your opponent for sing to work properly, or they won't sleep at long
WTF spinning the control stick and pressing no other buttons is the fastest way to escape sleep and grabs
BBQ Sing is only safe on hit at 160+% and if half a second isn't enough for a decent punish then sing is just about never useful ever. Dreams crushed
Also, I move that we have a Jigglypuff Data thread, where the results of this and rest percentages with/without DI, and the upTilt percentages are all put in the first few posts with discussions of other data in the thread and then added to the first post afterwards. If no one wants to do that, I'm up for heading the Data thread.
Last edited: