What's new

MK9 Frame Data Project

DrDogg

Noob
So is the jump method good enough to estimate adv on hit/block? That's what I tried for Stryker but I'd like to know what you guys think..
I'm not a fan of that method because there's a lot of room for error. Yes, it's possible to do it on your own, so you don't need a second person, but I'd rather make sure it's as accurate as possible as opposed to being done faster.

Are you guys working on Smoke's?
The first post has this to say about your question. ;)

I ask that you do not request specific characters. All of the characters will be done, but Lomyn and I are going in the order we prefer.
 

doomfarmer

unorthodox
I don't even know how to read frame data or what to do with it to put it to good use (i'm a noob). BUT, i'm sure when I understand what to do with it i'll LOVE this thread.

I can only imagine how much work you've put into this. Thanks DrDogg and thanks Lomyn!
 

Vulcan Hades

Warrior
Please can you give us Quan Chi & Mileena frame data?
You clearly haven't read the post above you lol.


I don't even know how to read frame data or what to do with it to put it to good use (i'm a noob). BUT, i'm sure when I understand what to do with it i'll LOVE this thread.

I can only imagine how much work you've put into this. Thanks DrDogg and thanks Lomyn!
Yeah I remember when I was new to SF4 I had no idea what frame data was. It looked just like boring maths and didn't understand how it could be useful. But then I started to play more competitively in SSF4 and I wanted to get better and reach a higher level so one day I decided to study it.

Man I'm telling you, there's nothing complicated with frame data. You just need to understand how to read it.

Here's the gist of it:

Start up/Execution frames = Speed of the attack
Active frames = Duration of the attack
Recovery frames = Cool down before you can move, attack or block again
Stun frames = Duration in which you cannot move or respond (but you can still block in MK9 unless it's a chain)

All attacks inflict hitstun / blockstun which leave you either at +frames or at -frames. And players with frame advantage can attack/respond sooner.

ex: An attack that leaves you at -7 disadvantage on block can be punished by Kung Lao's Spin (which has 6 start up).

That's really all you need to understand for starters. :)
 

DrDogg

Noob
Your skipping non-DLC chars? and even the first DLc char well wht ever
Yeah... I'm doing them in whatever order I want to do them in. I'm going to download Kenshi, so I figure I may as well do his frame data while I'm checking him out. I'd do Skarlet, but I haven't downloaded her yet.

In the end, every character will be done. There's no specific order, I just do the characters I decide I want to do... as I've stated numerous times... >_>
 
Yeah... I'm doing them in whatever order I want to do them in. I'm going to download Kenshi, so I figure I may as well do his frame data while I'm checking him out. I'd do Skarlet, but I haven't downloaded her yet.

In the end, every character will be done. There's no specific order, I just do the characters I decide I want to do... as I've stated numerous times... >_>
And we appreciate it

::give stern look at people complaining when they have no right to at all because he is doing a lot of work he doesn't have to do!
 

Somberness

Lights
I'm not a fan of that method because there's a lot of room for error. Yes, it's possible to do it on your own, so you don't need a second person, but I'd rather make sure it's as accurate as possible as opposed to being done faster.
The jump method does not always work, yes, but it can work. You just have to repeat it so the cpu actually jumps as soon as possible. I don't know if they are experiencing extra block stun or what when they don't jump fast. The duck method seems to always work but you can't use that for low attacks. I know the jump method does work because of Raiden's 334 and b312 strings. We know 334 is 0 on block and that b312 is -1 and you can actually visually confirm that there is a slight delay for the Raiden who uses b312 but not for 334. It's also looks the same if you do it when using 2 controllers. What method did you think was more reliable?
 

DrDogg

Noob
The jump method does not always work, yes, but it can work. You just have to repeat it so the cpu actually jumps as soon as possible. I don't know if they are experiencing extra block stun or what when they don't jump fast. The duck method seems to always work but you can't use that for low attacks. I know the jump method does work because of Raiden's 334 and b312 strings. We know 334 is 0 on block and that b312 is -1 and you can actually visually confirm that there is a slight delay for the Raiden who uses b312 but not for 334. It's also looks the same if you do it when using 2 controllers. What method did you think was more reliable?
The main problem with the jump method is that it's not always clear what frame the jump animation begins. This is especially true for the attacking character because the last few frames of animation on the attack are canceled into the jump.

I was working on hit/block frame data a couple days ago. Unfortunately, as Brady pointed out awhile ago, the hit/block frames of MK9 are all kinds of messed up. Distance, stance and when the opponent releases the block button all seem to factor into hit/block frames. It's going to take a lot more work than I expected to do the hit/block frames, and even then it will be difficult to ensure they're 100% accurate.

What I was originally going to do was get all of the execution frame data, then use that to calculate hit/block data based on which attacks beats out another attack. For example, Kitana's d+1 is at least -7 on block. I figured this out by using a Kitana mirror match and attacking with d+1 after a blocked d+1. Since d+1 is 6 frames, the d+1 hit, which confirmed that Kitana is at least -6. I then switched to punish with d+3, which is 7 frames. That also connected, confirming that the attack is at least -7. From here I was going to move to an 8 frame situation, but then you start to run into distance and stance issues.

What I will have to do now is record every hit/block situation, then play it back frame by frame. I'll have to do this multiple times because the results seem to frequently vary when you get down to a 2-3 frame window. It will probably take a full day to record hit/block situations for every character, then at least several hours per character to analyze the captured video frame-by-frame.

I'm certainly open to other methods, but given how MK9 frame data works, I don't know if there's any better way other than NRS just giving us the frame data.
 

Somberness

Lights
Hmm, does it actually matter if you can't tell where the jump animation begins? What I'm saying is pick the same character and just see where their highest point is and calculate how many frames difference it is between them.
 

DrDogg

Noob
Hmm, does it actually matter if you can't tell where the jump animation begins? What I'm saying is pick the same character and just see where their highest point is and calculate how many frames difference it is between them.
I didn't think of calculating from the top of the jump. However, there's still the issue of can you block before you can jump or vice-versa? You also may be able to attack before you can jump. Doing it this way I'll still have to test the same attack multiple times, but I may give it a shot.

Do Kano. I want to know his knife speed and his upball shenanigans.
I have no intentions of doing Kano... ever.
 

Somberness

Lights
I noticed Sprint used jump too, but did it by measuring total jump duration (basically the same thing). Camera movement might actually prevent you from identifying the highest frame.
Cyrax's neutral jump has a 41 frame total duration (including startup and recovery).
The first frame can be hard to spot after certain other animations, so I used the easily identifiable 41st frame as a reference point and subtracted 40 to verify the location of the first frame in each case.
As far as being to attack before you can jump, Sprint actually checked that but used a crouching attack as an example, with the need to add a frame in that case.
There appears to be a 1 frame transition between crouching and standing that I overlooked.

Kitana's d+1 is -1 on hit.
According to the jump method it's -2 (apparently the "animation fluff" made you misidentify it)

More specifically, taking the hitting frame as frame 0:
- The opponent can jump and attack on frame 21
- Kitana can attack on frame 22
- Kitana can jump on frame 23

From what I've checked so far, this discrepancy only seems to apply to crouching moves;
- Cyrax's d+1, d+3 and d+4 do indeed appear off by a frame.
- So far his standing moves seem correct, though I've only verified a few so far.

I'll look into it a bit more later on, but from what I've seen you just need to add a frame to the advantage for crouching moves when testing using the jump method.


p.s. The jumping input is actually to hold up for 3 frames. This doesn't make a lot of difference when buffering it after something (just make sure you buffer it early), but when jumping from a neutral state you'll see that it doesn't happen instantly. Pressing up for less than three frames causes you to not jump at all.
I'd like to believe that you can always attack, jump or block on the same frame unless of course you are staggered. Paulo already said that blocking is instant so I'm sure he would clarify this issue or anything else if you were to shoot him a PM. I hope for your sake that a non-attacking method will be good to go.
 

PeeJaeys

Sabi
If you could tell me how to do it, I could probably calculate frame data for smoke. Would it be enough to record the gameplay, and then view it using video editor frame by frame? PM me if you want/need help. I do have a HD DVR.
 

DrDogg

Noob
I noticed Sprint used jump too, but did it by measuring total jump duration (basically the same thing). Camera movement might actually prevent you from identifying the highest frame.

As far as being to attack before you can jump, Sprint actually checked that but used a crouching attack as an example, with the need to add a frame in that case.


I'd like to believe that you can always attack, jump or block on the same frame unless of course you are staggered. Paulo already said that blocking is instant so I'm sure he would clarify this issue or anything else if you were to shoot him a PM. I hope for your sake that a non-attacking method will be good to go.
I haven't had a chance to test the jump method thoroughly, but going by what you just posted, I can already see it's wrong. Kitana is my main and d+1 is absolutely not -1 on hit. At the very least it's +1 on hit. It was difficult to test beyond +1, but it could easily be +2 or +3 as well.

If you could tell me how to do it, I could probably calculate frame data for smoke. Would it be enough to record the gameplay, and then view it using video editor frame by frame? PM me if you want/need help. I do have a HD DVR.
As long as you can record in 720p or 1080p at 60 fps and playback at the same resolution, frame by frame, you can help. You need to record each attack hitting an opponent (preferably Sub-Zero in costume one), with button inputs, damage and collision area on. The first frame of the attack is when the red collision area lights up. For attacks where you can't see this or it doesn't occur (projectiles, etc.) the button input appears on the second frame of the attack.

The first active frame is the frame just before the damage shows up. So you count from the first collision area frame and stop counting at the frame just before the damage appears.
 

Somberness

Lights
Caught some of your play on stream, awesome Kitana, D1 for the win! And great commentary as well, lots of humor but not overdone. Also good to see NRS' shenanigans not deterring you from finishing this.
 

Vulcan Hades

Warrior
For those that wanna try out DrDogg's calculation method for frame advantage on hit/block, here's some moves you can use:

-Sub Zero's D,B,1 (2 start up)
-Kung Lao's D,F,1, Reptile's B,F,2 or Kitana's D1 (6 start up)
-Johnny Cage's uppercut or Kung Lao's D1 (7 start up)
-Raiden's B,F,3 (8 start up)
-Johnny Cage's D,B,3 (9 start up)
-Throw or Nightwolf's F,F,4 (10 start up)*
-Reptile's 2 or Kabal's uppercut (11 start up) *
-Reptile's B,F,4 or Kabal's D4 (12 start up)
-Kenshi's 3 or Scorpion's B,B,1 (13 start up)
-Kenshi's D,B,1 (14 start up)

*Might need something better here..
 

Somberness

Lights
Spin would be 5 frames as well but the first active frame is being counted in the total. Don't be confused, elbow dash should be considered as 6 frames, just like spin.