Stealth Raptor
Smash Legend
Me and DeLux did some testing of SDI today and i would like to post what we found.
first off im going to explain some of what i know from my own personal testing. i would like to start off with a brief mention of vector analysis. nothing fancy its more for a visual aid to some of the mechanics. if you know what it is feel free to skip this, im gonna assume a lack of knowledge of it
when it comes to vectors, the best way to deal with them is to break them down into vertical and horizontal components that can be used for vector addition and such. as you can see in the picture, the red vector can be broken down into a horizontal (Blue) component and a vertical (Green) component. there are formulas that allow you to get the magnitude and angle and such from this, but thats unimportant for SDI.
NOW when you SDI, the game seems to look at the components to determine if to SDI or not. if there is a new component in any direction, it will input the SDI in the direction of the control stick.
lets say someone is going to HCDI a hit, starting straight left and going clockwise to straight right. first left is hit, and a new vector to the left is seen by the game. it says yes to SDI, and it goes to the left. lets say on the next frame, the control stick is up-left. the game sees two components, a left and up component. the left component is old so it ignores it, but the up input is new, so it triggers SDI, and you SDI to the up-left (note it isnt up but up-left. the components that i talk about do not influence SDI, they just sorta serve as a yes/no to whether or not to do it). one frame later, your control stick is straight up, leaving only an up vector. as there are no new component vectors, the game triggers a no, and no SDI is visible. on the next frame, the control stick is up-right. the game sees a new component to the right, and triggers a SDI to the up right. this continues down to right, where as no new components are made there is no SDI. this behavior was seen by me in frame advance and by what me and lux did tonight. i would also like to point out that when it goes from straight up/down/left/right to a diagnol, that diagnol input could be in any direction between the cardinal directions, it doesnt have to be straight in the diagnol, which means more or less that even though there are a lot of possible SDI directions, the game really only registers eight seperate directions, the four main directions and each quadrant between the four main directions
this is more or less how it sees SDI. note if SDI is done by the tap method, there is at least one frame of no input so any tap SDI is still registered as it goes from no vectors to one vector in whatever direction you tap.
now lux looked at the data K Prime collected in regards to holding the control stick in a direction and tapping the cstick (ISSDI and grab mashing) and was thinking it could be used for SDI. what we found is that if you hold a direction and tap the cstick once, it inputs two very rapid SDI inputs, seperated presumably by one frame (wasnt frame tested, didnt have the codeset + inconsistancy of cstick in frame advance) both in the direction that the cstick was hit. note these must be at least ninety degrees apart ie both must have at least one vector component different from each other to trigger the SDI. if the cstick is tapped in the same direction as the control stick is held nothing will happen.
a quick summary of various SDI:
single tap- one SDI per input of tapping the control stick or cstick, probably wont get more then one off except large hitlag
double tap- two SDI inputs per tapping of the cstick + control stick slightly offset. probably wont get off more then a single double tap except in large hitlag
QCDI- rotating the control stick in a quarter circle. as its only nintey degress there will only ever be two new vectors so only two SDI inputs are technically possible (if you happen to go slightly beyond ninety degrees you may get a third input depending on starting point)
HCDI- rotating the control stick in a half circle. This will only give you a max of three SDI inputs (if you go slightly beyond a half circle you may get four inputs depending on starting location but at that point you are SDI'ing the opposite direction so it wont help)
DSDI- holding a direction on the control stick and tapping the cstick. this gives two SDI inputs seperated by a frame. can be done a little faster then double tap as its a bit easier to tap one hand really fast then coordinate both
the DSDI has a lot of potential, after we looked into this DeLux was doing some crazy SDI shenanigans almost immediately in our matches
if you have any questions please ask, i can ramble at times so i may have been a bit confusing.
first off im going to explain some of what i know from my own personal testing. i would like to start off with a brief mention of vector analysis. nothing fancy its more for a visual aid to some of the mechanics. if you know what it is feel free to skip this, im gonna assume a lack of knowledge of it
when it comes to vectors, the best way to deal with them is to break them down into vertical and horizontal components that can be used for vector addition and such. as you can see in the picture, the red vector can be broken down into a horizontal (Blue) component and a vertical (Green) component. there are formulas that allow you to get the magnitude and angle and such from this, but thats unimportant for SDI.
NOW when you SDI, the game seems to look at the components to determine if to SDI or not. if there is a new component in any direction, it will input the SDI in the direction of the control stick.
lets say someone is going to HCDI a hit, starting straight left and going clockwise to straight right. first left is hit, and a new vector to the left is seen by the game. it says yes to SDI, and it goes to the left. lets say on the next frame, the control stick is up-left. the game sees two components, a left and up component. the left component is old so it ignores it, but the up input is new, so it triggers SDI, and you SDI to the up-left (note it isnt up but up-left. the components that i talk about do not influence SDI, they just sorta serve as a yes/no to whether or not to do it). one frame later, your control stick is straight up, leaving only an up vector. as there are no new component vectors, the game triggers a no, and no SDI is visible. on the next frame, the control stick is up-right. the game sees a new component to the right, and triggers a SDI to the up right. this continues down to right, where as no new components are made there is no SDI. this behavior was seen by me in frame advance and by what me and lux did tonight. i would also like to point out that when it goes from straight up/down/left/right to a diagnol, that diagnol input could be in any direction between the cardinal directions, it doesnt have to be straight in the diagnol, which means more or less that even though there are a lot of possible SDI directions, the game really only registers eight seperate directions, the four main directions and each quadrant between the four main directions
this is more or less how it sees SDI. note if SDI is done by the tap method, there is at least one frame of no input so any tap SDI is still registered as it goes from no vectors to one vector in whatever direction you tap.
now lux looked at the data K Prime collected in regards to holding the control stick in a direction and tapping the cstick (ISSDI and grab mashing) and was thinking it could be used for SDI. what we found is that if you hold a direction and tap the cstick once, it inputs two very rapid SDI inputs, seperated presumably by one frame (wasnt frame tested, didnt have the codeset + inconsistancy of cstick in frame advance) both in the direction that the cstick was hit. note these must be at least ninety degrees apart ie both must have at least one vector component different from each other to trigger the SDI. if the cstick is tapped in the same direction as the control stick is held nothing will happen.
a quick summary of various SDI:
single tap- one SDI per input of tapping the control stick or cstick, probably wont get more then one off except large hitlag
double tap- two SDI inputs per tapping of the cstick + control stick slightly offset. probably wont get off more then a single double tap except in large hitlag
QCDI- rotating the control stick in a quarter circle. as its only nintey degress there will only ever be two new vectors so only two SDI inputs are technically possible (if you happen to go slightly beyond ninety degrees you may get a third input depending on starting point)
HCDI- rotating the control stick in a half circle. This will only give you a max of three SDI inputs (if you go slightly beyond a half circle you may get four inputs depending on starting location but at that point you are SDI'ing the opposite direction so it wont help)
DSDI- holding a direction on the control stick and tapping the cstick. this gives two SDI inputs seperated by a frame. can be done a little faster then double tap as its a bit easier to tap one hand really fast then coordinate both
the DSDI has a lot of potential, after we looked into this DeLux was doing some crazy SDI shenanigans almost immediately in our matches
if you have any questions please ask, i can ramble at times so i may have been a bit confusing.