The KOF 98 UM Frame Data Project

Was it all just Shingo's dream?
AtTheGates
Charging...23%
Charging...23%
Posts:24
Joined:Tue Jul 01, 2008 22:27
A.K.A.:Turned ON
Location:frankfurt, germany
Contact:
The KOF 98 UM Frame Data Project

Post by AtTheGates » Wed Jul 02, 2008 11:52

Hi everyone,

I started researching framedata for kof 98 um with the help of the "Command" feature in training.
With this little tool, you can not only create macros, but also effectively count frames, if you are a little creative.

All of my data is built on several assumptions, which means, it's not fool proof, and prone to error.

Assumptions:

1. a move can be cancelled into a roll or ABC in EX mode on its first active frame (exception: several special moves). Also true for jumping normals.

2. The moment you active a normal / special (the frame where the button gets pushed) counts towards the startup time.


The process of creating this data takes a very long time. Startup frames take around 20-30 minutes per char, but active time takes a lot longer, since setting up the procedure to check them takes a lot longer, see bottom of the post for details.


So i would like to ask for help from the community: if we all split up chars between us, we can easily create a full list of frame data, at least for all the normals.

Please see below the list for some guidelines.

Here is what i have so far:


GENERAL:

time a direction has to be active for the game to register:
4-5 frames.
example: down+LP, let go of down at 4-5 frames or later: c.lp comes out. let go earlier, standing lp will come out.
by the way if i could name one thing i hate the most about kof, THIS IS IT. While this feature exists in almost all fighting games, the frame window is the biggest in KOF, which makes some combos really hard for me.


========================
GEESE
========================

wakeup time 76/77 frames (tested with geese sweep, only comparative value)

5A(c) - 7f, 3f active
5A(f) - 7f, 4f active

5B(c) - 6f, 3f active
5B(f) - 9f, 4f active

5C(c) - 7f, 2f active
5C(f) - 9f, 6f active

5D(c)[1] - 7f, 3f active
5D(f) - 11f, 5f active

6B - 23f, 3f active

CD - 15f, 2f active


2A - 7f, 3f active, total duration 25f, <7f advantage on hit
2B - 7f, 4f active, total duration 25f, <7f advantage on hit
2C - 7f, 6f active,
2D - 12f, 5f active

j.A - 10f
j.B - 8f
j.C[1] - 9f
jump up, j.C - 10f (when geese jumps up straight, normal jump, no small jump, his j.C has a different animation)
j.D - 12f
j.CD - 12f

----------
236A - 7f, 9f active, 32f recovery, 48f total. meaty on block still negative static difference. (couple of frames).
236C - ??

623A - 9f
623C - 18f

63214A - 13f
63214C - 21f




===================
KYO
===================

wakeup time 80/81 frames (tested with geese sweep, comparative value only)

5A(c) - 7f
5A(f) - 7f

5B(c) - 9f
5B(f) - 11f

5C(c) - 6f
5C(f) - 16f

5D(c) - 8f
5D(f) - 14f <-- changed from KOF 98 classic! 4 frames faster.

6B - 15f

3D - 6f (!)

CD - 19f

2A - 7f
2B - 7f
2C - 8f
2D - 15f

j.A - 9f
j.B - 9f
jump up, j.B - 9f
j.C - 10f
j.2C - 10f EDITED - used to be 13f before, which is wrong.
jump up,j.D - 11f
j.D - 13f
f.CD - 17f

---------
623A - 6f
623C - 6f

236A - 12f

236C - 18f

236B - 14f
236D - 18f

421B - 19f (perfect distance), 20f (point blank)
421D - 32f (perfect distance), 33f (point blank)

63214B/D - 14f
------------------------------------------------------


so, how to find out startup frames?

you program this:

[normal],NNNNNNNNNNNNNNNN,[AB]

make sure to include a generous pause between the push of the button and the roll cancel (start with around 12 frames for quick attacks, 20 frames for slower attacks)

1. once you see your character roll canceling the hit, keep reducing pause frames until your character stops roll canceling the hit.
2. Add the last frame you took away, so that he roll cancels again

->> count the frames, INCLUDING the normal, EXCLUDING the roll cancel. That is the startup time of the normal.

for crouching normals, make sure the setup looks like this:

2222[2+normal]2222NNNNNN and so on. If you don't do this, your char will stand up to do the normal.


for special moves, here are some examples (i will use numblock notations for this one):
qcf
22336[6+normal],6666NNNNNNNN [ABC]
hcf
41122336[6+normal] see above


for jumping moves, program it like this:

6x up or 9x up, (depending whether you want a smalljump or normal jump),NNNNNNNN(just program a short pause), normal, NNNNNNNN, ABC.

1. keep reducing the first pause if the char is too high when he does the normal, increase it if he is too low.
2. keep reducing the second pause like you do for standing normals, until the ABC doesnt come out anymore, then add 1 frame, count.


finding out ACTIVE time:

this takes quite a while.

1. program this:

2222[2+D]2222, then a LONG pause (around 75-80 Ns), then the normal, after the normal, add a pause that equals his startup time, then [AB] or EX mode [ABC].

an example would be:

geese vs. kyo:

2222[2+D]2222, 80xN (roughly), HP,6xN (since HP has 7 frames startup, and the HP counts as frame 1), then [BC].

then keep reducing the first pause. EVERY FRAME THAT HITS WITHOUT TRIGGERING THE ROLL IS AN ACTIVE FRAME. Count them one by one and keep decreasing the first pause until the move does not hit anymore.
Last edited by AtTheGates on Fri Jul 04, 2008 13:20, edited 2 times in total.

AtTheGates
Charging...23%
Charging...23%
Posts:24
Joined:Tue Jul 01, 2008 22:27
A.K.A.:Turned ON
Location:frankfurt, germany
Contact:

Post by AtTheGates » Wed Jul 02, 2008 12:32

==============
EIJI
==============

5A(c) - 7f
5A(f) - 7f

5B(c) - 9f
5B(f) - 11f

5C(c) - 7f
5C(f) - 16f

5D(c)[1] - 7f
5D(f) - 12f

CD - 13f

2A - 7f
2B - 7f
2C - 8f
2D - 10f

j.A - 9f
j.B - 9f
jump up,j.B - 9f
j.C - 10f
j.D - 12f
j.CD - 14f

-------------
236A/C - 9f




==============
HEIDERN
==============

5A - 7f

5B(c) - 9f
5B(f) - 9f

5C(c)[1] - 9f
5C(f) - 12f

5D(c) - 9f
5D(f) - 16f

CD - 25f

6B - 22f

2A - 7f
2B - 6f
2C - 10f
2D - 10f

j.A - 10f
jump up, j.A - 7f
j.B - 7f
jump up, j.B - 12f
j.C - 8f (!)
j.D - 10f
jump up, j.D - 15f
j.CD - 9f

-----------------------
all charge moves have a charge time of 30f (half a second).

[2],8+A - total duration 47f on hit
[2],8+C - total duration 50f on hit

[4],6+A - total duration 48f
[4],6+C - total duration 43f




==========
YAMAZAKI
==========


5A(c) - 7f
5A(f) - 10f

5B(c) - 8f
5B(f) - 9f

5C(c) - 6f (wtf)
5C(f) - 14f

5D(c)[1] - 9f
5D(f) - 13f

6A - 16f

CD - 16f

2A - 9f
2B - 9f
2C - 8f
2D - 12f

j.A - 7f
j.B - 10f
jump up, j.B - 7f
j.C - 9f
jump up,j.C - 16f
j.D - 12f
jump up, j.D - 12f

j.CD - 12f

---------------------------
623A - 12f
623C - 14f

236A/C - 9f

623B[1] - 15f
623D - 13f



===========
CLARK
===========

5A - 8f

5B(c) - 6f
5B(f) - 7f

5C(c) - 7f
5C(f) - 15f

5D(c) - 13f
5D(f) - 12f

6B - 17f

CD - 15f

2A - 8f
2B - 8f
2C - 9f
2D - 8f (wtf)

j.A - 7f
j.B - 7f
jump up, j.B - 7f
j.C - 10f
j.D - 8f
jump up, j.D - 15f
j.CD - 13f




============
LEONA
============

5A(c) - 7f
5A(f) - 7f

5B(c) - 9f
5B(f) - 8f

5C(c)[1] - 7f
5C(f) - 15f

5D(c)[1] - 8f
5D(f) - 16f

6B - 30f (thats a lot)

CD - 18f

2A - 7f
2B - 7f
2C - 6f (wtf)
2D - 11f

j.A - 10f
j.B - 7f
j.C - 7f
j.D - 8f
j.CD - 15f

jump up j.A - 7f
jump up j.B - 13f
jump up j.C - 10f
jump up j.D - 7f

--------------------
all charge moves need to be charged at least 30f (half a second)

[2],8+A - 7f
[2],8+C - 9f

214B - total duration 61f
214D - total duration 65f

User avatar
SonicTempest
Posts too much
Posts:3409
Joined:Wed May 04, 2005 14:26
A.K.A.:>8|
Currently Playing:Dark Souls II
XBL:SonicTempest84
PSN:SonicTempest
Location:Seattle, WA
Contact:

Post by SonicTempest » Wed Jul 02, 2008 17:36

I'll go ahead and sticky this.

AcidicEnema
ON Elite Spammer
ON Elite Spammer
Posts:544
Joined:Sat Jul 09, 2005 22:10

Post by AcidicEnema » Wed Jul 02, 2008 18:01

I think this is an excellent project, but a few points that need to be considered:

- Are you sure that each input represents 1 frame? E.g., NNN = 3 frames. While it might seem to be the obvious conclusion to make, your findings so far seem to be in direct contradiction with the information on Dune's frame data page for KOF98: http://dune-enud.web.infoseek.co.jp/kouryaku/frame.html

For example, you list Kyo's close stand C as needing 6 frames before it becomes active, while Dune's webpage states it as 2, 3+9, 9 (meaning 2 frames to become active, 3 frames active for the gut punch attack frame, 9 frames for the uppercut attack frame, 9 frames to recover).

While this wouldn't be a problem for comparable attacks (I know for a fact that Kyo's close s.C, Leona's c.C and Yamazaki's close s.C are all as fast as each other so 6 frames to become active for all seems about right). Its problematic when comparing attacks of different speeds. For example, you list Kyo's close s.B as 9 frames before it becomes active, meaning it supposed to be 1.5 times slower than his close s.C. Dune's page, however, lists that Kyo's close s.B takes 5 frames to become active, suggesting that its 2.5 times slower than the close s.C.

So since its hard to imagine that the frame data was changed so drastically between KOF 98 and 98UM, there must be either a problem with one of your assumptions or a problem with the data on Dune's webpage.

Hope you understand what I'm trying to get at. Do let me know what you think!

AtTheGates
Charging...23%
Charging...23%
Posts:24
Joined:Tue Jul 01, 2008 22:27
A.K.A.:Turned ON
Location:frankfurt, germany
Contact:

Post by AtTheGates » Wed Jul 02, 2008 19:53

true, i made several asumptions in the process. If a move cannot be cancelled on its first frame into dodge,roll or EX mode ABC, then all the data is wrong :(

about kyo's close C: yeah 2 frames of startup is definitely wrong, no move in any fighting game is that fast except for throws and supers (ok so someone just told me that 4s chun has a 2 framer, but hey, that's 3s, doesnt make sense anyway)

also, i have a feeling that there can be no normal faster than 5 frames, because of the 5 frame window you need to hold an input for it to register. This is just a gut feeling, we'll have to see how the data develops.

about the frame data page: what language pack do i need to display the characters? i thought installed everything there was for asian fonts :(
is there a translation?

about the scans on the site:

shouldn't the 2nd, empty column be the startup time? cause that one is missing.
do you have a translation for all the columns?

AcidicEnema
ON Elite Spammer
ON Elite Spammer
Posts:544
Joined:Sat Jul 09, 2005 22:10

Post by AcidicEnema » Wed Jul 02, 2008 20:42

Yeah, I know 2 frames sounds wrong, but I always assumed that it was something to do with the frame rate that the games run at.

According to another source, Kyo's close s.C in KOF XI takes 4 frames to be active, so I'm thinking that 2 frames on that site really means 4 frames, at 60fps.

- About the scans on the site: They're all related to damage/push back etc. Nothing to do with frames. The frame data is seen when you scroll lower. Its displayed above all the hitbox displays.

However, the page is far from complete, and only a few characters (Kyo, Beni, Goro and a few more) have completed pages which show both the frame data and the hitbox data. Some, like Iori's only show the hitbox data, and some, like Andy's, only show the frame data.

AtTheGates
Charging...23%
Charging...23%
Posts:24
Joined:Tue Jul 01, 2008 22:27
A.K.A.:Turned ON
Location:frankfurt, germany
Contact:

Post by AtTheGates » Wed Jul 02, 2008 20:51

if anyone volunteers he could test in kof xi how quick kyo's close C is.

program:

C,N,N,N,[shift or roll cancel]

if that doesn't work,

C,N,N,N,N, [shift or roll]

and so on, until he rolls / shifts.

i'm really curious whats gonna happen in xi =)
unfortunately, my copy gives me problems on my ps2.

AtTheGates
Charging...23%
Charging...23%
Posts:24
Joined:Tue Jul 01, 2008 22:27
A.K.A.:Turned ON
Location:frankfurt, germany
Contact:

Post by AtTheGates » Wed Jul 02, 2008 21:07

=======
YASHIRO
=======

5A(c) - 7f
5A(f) - 7f

5B(c) - 7f
5B(f) - 8f

5C(c) - 7f
5C(f) - 14f

5D(c) - 7f
5D(f) - 12f

6A - 22f
6B - 15f

CD - 15f

2A - 7f
2B - 10f
2C - 7f
2D - 13f

j.A - 10f
j.B - 11f
j.C - 12f
j.D - 11f
j.CD - 11f

------------------
41236A - 17f
41236C -25f

623A - 10f
623C - 12f

63214A - 13f
63214C - 15f

63214B - 23f
63214D - 28f

AcidicEnema
ON Elite Spammer
ON Elite Spammer
Posts:544
Joined:Sat Jul 09, 2005 22:10

Post by AcidicEnema » Wed Jul 02, 2008 21:35

Okay, you're not going to like this, but I tested it in KOF XI.

For Kyo inputting this command:

C,[Roll Cancel]

Will already get you a roll cancel of Kyo's close stand C. This will work, up till 19 'N' commands have been added in the middle, meaning that

C,N,N,N,N,N,N,N,N,N,N,N,N,N,N,N,N,N,N,[Roll Cancel]

Will still get you a roll cancel off a close stand C.

On the other hand, Kyo's far stand C (which is supposed to take 5 frames to activate), will get a roll cancel from

C,N,N,N,[Roll Cancel]

Up till

C,N,N,N,N,N,N,N,N,N,N,N,N,N,N,N,N,N,N,N,N,N[Roll Cancel]

Meaning that with 3 to 21 'N's in between, you will get a roll cancel off a far stand C.

So this method definately doesn't work for XI. That doesn't mean that it wouldn't work for 98, but I think your assumptions will need to be further tested.

AtTheGates
Charging...23%
Charging...23%
Posts:24
Joined:Tue Jul 01, 2008 22:27
A.K.A.:Turned ON
Location:frankfurt, germany
Contact:

Post by AtTheGates » Wed Jul 02, 2008 21:43

hm. the thing about the far C does concern me.

about the other thing:
adding too much "N"s is no problem, it's the same in kof 98 um - you can cancel a normal with a roll or dodge ultra-late.

AcidicEnema
ON Elite Spammer
ON Elite Spammer
Posts:544
Joined:Sat Jul 09, 2005 22:10

Post by AcidicEnema » Wed Jul 02, 2008 22:09

Okay I think I figured out the difference between your data and the data on the Dune page.

At least when calculating active frames, Dune frame + 4= yours.

So lets looks at Kyo's close s.C again. According to the Dune website, it takes 2 frames to activate. Add 4 to that, and you get 6, which is what you got.

For c.B, the Dune website says it take 3 frames to activate. Add 4 to that, you you get 7, which is what you got.

This seems to hold true generally. But there are a few discrepencies. For example, it states that Kyo's far s.D takes 14 frames to be active. If you add 4, you should get 18. But your figure is 15. Another discrepency would be stand CD.

I'm thinking maybe you could double check your findings for those where there are discrepencies. If it turns out that they aren't different after all, then we could probably use this relationship to save you some time.

Btw, according to this relationship, Ryo's c.C should take 5 frames to activate, according to your methodology... Correct?

AtTheGates
Charging...23%
Charging...23%
Posts:24
Joined:Tue Jul 01, 2008 22:27
A.K.A.:Turned ON
Location:frankfurt, germany
Contact:

Post by AtTheGates » Wed Jul 02, 2008 22:19

excellent find, thank you for the help.

i will check this as soon as i can.


Another thought occured to me - we should be able to check framedata by using an emulator, kof 98 classic, and frame-by-frame-macros for attack,wait,roll cancel. While this still does not guarantee whether a roll can be activated on active frame 1, we can see if there is any frame difference from my data.

AtTheGates
Charging...23%
Charging...23%
Posts:24
Joined:Tue Jul 01, 2008 22:27
A.K.A.:Turned ON
Location:frankfurt, germany
Contact:

Post by AtTheGates » Wed Jul 02, 2008 22:27

using my method,

ryo's C(c) takes 6 frames (might have to do with the "direction has to be active for 5 frames or it won't count" bug, so this could be 5 actually).
kyo's far D takes 14 frames <--- made a mistake here in the original test, i will edit my first post.
kyo's CD is still 19 frames.

AcidicEnema
ON Elite Spammer
ON Elite Spammer
Posts:544
Joined:Sat Jul 09, 2005 22:10

Post by AcidicEnema » Wed Jul 02, 2008 22:45

Its possible that they might have nerfed Ryo's crouch C in UM. There's a simple way to test this.

In 98, when I map player 1 (Kyo's) C button on the same button as Player 2 (Ryo's) C button and make Ryo do his crouch C and Kyo do his close s.C together right in front of each other, Ryo's crouch C would win every single time. This is consistent with the information from Dune's page which states that Ryo's crouch C is 1 frame faster than Kyo's close s.C.

I'm not sure if its possible to test this the same way on the PS2, but if it is, you could try and see if Ryo's crouch C still beats Kyo's stand C. If it does, that means there is a problem with your method. If not, then it shows that Ryo's crouch C was indeed nerfed.

Sorry I can't help with all this since I don't have UM on the PS2 >_<

AtTheGates
Charging...23%
Charging...23%
Posts:24
Joined:Tue Jul 01, 2008 22:27
A.K.A.:Turned ON
Location:frankfurt, germany
Contact:

Post by AtTheGates » Wed Jul 02, 2008 22:52

i made some progress.


using kawaks, i programmed yamazaki to do:


2A,2,2,3,6,A

and c.lp, qcf A actually came out. 1 frame less and nothing happened. Which would mean: 5 frames. +4 = the 9 frames i counted on PS2.


I will try around a bit more, especially with moves that have more startup.


hm.

yashiro:

A(c),f,d,df,A
gives you A(c),623A. A(c) should have 3f startup (7-4 = 3), but counting the push of the button (A), the number of frames is 4. There is a chance the limit of this test is the number of motions that you have to enter - when they are too much, you cannot have exact results. will try with heidern (charge char) next.

what does the frame data you have say about yashiro's close A?


heidern:

programmed charge 2 for 30 frames, 2+A,2,2,8+A gave me a 2A into special. including the push of the button, again, 4 frames, 1 too much. The whole thing could be explained by not counting the first frame, but more testing is needed.

Post Reply