The King of Fighters XIII Hyper Guide Wiki


First of all I would like to say that the wiki guide is awesome. I was able to find some good HD combos and general strats in it for some of my characters.

Second, I have a question and I apologize if I should ask this in a different thread. I notice that you have frame data for only some characters right now (such as Kyo). I was wondering if you guys are getting this information from personal testing and if so are the other characters being worked on for frame data or are we waiting for an official release on this information form Atlus/SNKPlaymore?

Thanks again guys.


I think Phoenixnl measured Kyo and a few other characters’ frame data manually, but it’s based on the arcade information so it’s not exactly accurate and the measuring process is a chore. There might have been a Japanese resource like an issue of Arcadia Magazine that had Arcade XIII’s frame data, but as far as the console version goes there’s nothing yet.

I’ve heard that there’s supposed to be something in the works with all the new data so we’ll see. In the mean time you can already get a feel for most attack speeds and recoveries from links and juggle speeds, or for specifics you can always set up a training dummy and then testing how safe an attack or tactic is. For most basic punishes, you can assume most Close Normals are 2-3 frame startup, all instant command grabs are 1 frame startup, and most EX Ranbus are 2~4 frame startup which help them punish just about any negative move on block.

crimzonflame, the Yuri combo section needs to be edited, but you should only use dp+C to initially launch the opponent and from then on in the juggle do dp+A like you were doing. In the corner you can go straight into her cl.C f+A xx qcf+K loop unless they had to be juggled into the corner in which case you’d do the usual corner juggle loop.


Good guide, though some character pages are still kinda barren, Raiden’s for instance.


Hello guys, I’ve got some frame data here and would like to throw it up on the wiki. Right now I’ve got Saiki’s done, and plan to do a few more in the next little while. Is there a list of data that’s already completed so that I know what I shouldn’t waste time on doing?


From what I’ve been hearing, “something” will be coming out soon. I don’t know how much data, who’s compiling it, nor when it’ll come out.

So I say, go for broke and do everyone please and upload it to the wiki. o4L
No one else I know from #SNKPlaymore or my other compatriots are working on frame data so yeah please contribute.


I’m working on some framedata, but I’ve been rather busy. But I can share some of the stuff I have, either pop in #SNKPlaymore and PM me, or PM me on the forum here. I have pretty complete data for Kyo, and some start for Benimaru and Daimon.

I updated some of Kyo’s data to console already.


I’m not sure where to put this since I haven’t got access to the wiki yet, but here’s some info on the hitstop and hitstun/blockstun of attacks:

All light attacks on the ground have 8F hitstop.
All air attacks (light and heavy and CD) as well as all ground heavy and CD attacks have 12F hitstop.


Grounded lights: 15F/13F
eg. Saiki 2A is 4F active, 6F recovery, +5 on hit, so 15F hitstun. Links to 2C which goes active in 5F.
Grounded heavies: 21F/19F
Grounded CD: KD/23F

Jump/hop lights: 13F/13F
Jump/hop heavies: 13F/16F
Jump CD: KD/23F
Hop CD: KD/19F


Hmm, that’s odd. When I looked at hitstop I arrived at 7Frames and 11Frames (same as KOF2002)

The way you count blockstun is odd, or at least, different from the way it is traditionally done with KOF hitstun/blockstun.

A hit is only registered after the first hit, so the first active frame is over. So after hitting with Saiki cr.A, there are 3 active frames remaining and 6 recovery

That’s 9 in total. 15-9=5 so +6. In other words, you have a frame of blockstun too many. It should be:
Grounded lights do 14/12 frames of hitstun/blockstun
Heavies do 20/18
Grounded CD I guess does 22 then (when I looked at it, I arrived at 21, but I might have done it wrong, and maybe they changed it in console).

jump lights = 12/12 (used to be 14/14 in arcade)
jump heavy = 12/15 (used to be 14/20 in arcade)
Jump CD = 22 (used to be 24 in arcade)
I’m really surprised that hop CD does less blockstun than jump CD. It’s never been like that in previous KOF’s But haven’t tested that myself, exactly because I didn’t expect it, you might be right.

Older KOF’s had:
Lights: 11/9
Heavy: 19/17
CD: 21
Jump light: 11/9
Jump heavy: 11/17
Jump CD: 21

And that’s counting frame advantage in the way I did it, and it is actually the game processes the hitstun, so I think it’s worth writing it as such rather than 1-frame higher than ‘real’ hitstun the way you did.

Also for the wiki, keep in mind that startup frames are usually counted without the first active frame unlike the tradition in SF4 and GG. So a 3 frame startup sf4 move is equal to a 2 frame startup KOF move.

Just to make sure, I’ve gone back and checked these numbers (thanks Irysa for helping out).

Stand Lights: Hitstun: 14 Blockstun: 12 Hitstop: 7
Stand Heavies: Hitstun: 20 Blockstun: 18 Hitstop: 11
Stand CD: Blockstun 23 Hitstop: 10
Jump/Hop lights: Hitstun: 12 Blockstun: 12 Hitstop: 11
Jump/Hop Heavies: Hitstun: 12 Blockstun: 15 Hitstop: 11
Jump CD: Blockstun: 22 Hitstop: 11
Hop CD: Blockstun: 18 Hitstop: 11


This is incorrect, based on frame by frame review of how the engine registers hitstun/blockstun. A hit or block is registered on active frame 1. The defender is put into hitstun or blockstun immediately on this frame, and hitstop begins here as well. So on AF1, all of the following is occurring: AF1, Hitstop1, Hitstun1 (or Blockstun1). The animation for hitstun or blockstun is clearly present during hitstop so you can see here it doesn’t wait until the first AF is over.

Hitstop is the total freeze time for a move on hit or block. This time is 8F for lights and 12F for heavies - this time includes the initial active frame and an additional 7 or 11 frames of additional freeze.

I’m aware of how KoF convention is to not include the first AF in startup, so that’s how I’ll be posting my data up.


SNKP will release soon a frame data.

The Kyo frame data is not the same as our theorical frame data in France, we believe (because it’s all about , or you have a method i don’t have to test, like 60fps records) Kyo’s s.C is 3 and some moves are +1frames startup etc. But the differences between the frames are same (It’s Always +1 or -2 frames etc for all moves).


I took the time to make a nice overview of how hitstun/hitstop is counted:

I’m fairly confident that it works like this. But there’s an easy way to check it. According to my theory, a heavy attack can only combo into a 19-frame startup move. While according to your hypothesis it can even combo into a 20-frame startup move.

So to test it, all we have to do is find a 19 and 20 frame startup move respectively

(or a 13/14 frame startup move to test it off lights).

Let’s find moves like that.

[edit:] Just to add: The way hits are processed is exactly the same as the way it was in older KOF’s, and my hitstun numbers for older KOF’s are definitely correct. So it’d actually be quite surprising if the actual hitstun registration turns out to be different.

What makes you say that?


As the difference we (myself and phoenixnl) have been discussing relates to whether we count the first active frame or not, it’s kind of arbitrary which way we count the frames. Since it’s KoF convention to not include the hitting frame I’ll leave it off and go with the KoF standard.

That being said, I’ve added complete frame data for Saiki on the wiki, can be found here:

I don’t think I’ll be going into this much detail for other characters though (it’s a lot of work!).

King is up next and should be up tomorrow.


Being completely aware just how much work this stuff is, I want to say: fantastic job!

But: The grab supers really should be “0+0” startup. Since you can punish Kyo’s qcf+B with it which is -1 on block.

And I think it would be more correct to say the super freeze is 29 frames for all non-instant supers. The first invincible startup frame on every super is pretty real. Like Kyo’s qcb,hcf+AC has 1 frame invincible startup 29 frames of freeze and after that it’s active
So Startup: 1+(29)+0
The opponent can move in that 1 frame still, as a result the move is 1 frame startup and therefore cannot punish qcf+B but can punish qcf+D (which is -3 on block)

The exception to the rule are in fact grab supers that can punish Kyo qcf+B on block. These moves have their 1st invincible startup frame included in the freeze so moves like Saiki’s grab, but also Daimon’s and Clarks grab become 0+(30)+0 So with these grabs the opponent is frozen the first frame of ‘startup’.


I’m annoyed to see that for some reason Special blockstun seems to break universality and some undoubtedly do 17 frames of blockstun instead of 18. I’d casts the ones that knockdown in doubt since you don’t have much to compare to relatively, but the +2 vs -1 of qcf+A pretty much proves it’s possible. Ugh. Oh well.

The only other way to explain that would be if somehow hitstop is different on hit than on block on the fireball. Which would be equally ugly, and in essence doesn’t make much of a difference.

Also, I see that for the recovery of Saiki’s fireball you’ve given the total animation frames. Why not do the recovery frames instead? simply 44-14 = 30 recovery would be fine right?

It’s kind of interesting that if you look at the hitstop you indeed arrive at 12 hitstop here and not 11. (20+12)-30=+2. But I think that has to do with the way the engine hitchecks fireballs. I believe it takes a frame more than with normal hits to hitcheck a move, essentially adding another frame. If that extra frame of hitcheck doesn’t exist on block that would explain the difference of 3 frames instead of 2 between hit and block.

I’ll have a look at the way fireball hits are processed soon.


O and one final thing: I noticed that you’ve only put the j.CD in the framedata and not the hop CD despite hop CD doing a different amount of blockstun. i think it’s probably a good idea to include that too.


So many fucking numbers


Don’t you know? The only way you can be good at fighting games is to learn Math!


Very educational! Oh my queeen’s THRONE!


can i give you my children? Thanks for the framedata dude :D!!111


Fixed up a few errors in the Saiki data, it should be error-free now.


This is not actually correct, since there are two frames up until active for grab supers: 1 frame where enemy can move at the start of super freeze, and the second being the active frame of course. However the active frame occurs during super freeze so it’s really just the 1F until hit.

This is actually not the case - the super freeze functions exactly the same for grab supers as it does for all other supers, the enemy can still move during the first frame. After some testing I noticed that the real issue behind why grabs work against -1 moves is due to how “grab check” works. For normal grabs at neutral, the actual grab “spark” appears on frame 2, but the grab check itself happens on frame 1 - meaning, that throws are immediate (active on 1). If you have an active frame overtop of an enemy on the first grab frame (pre-spark), it will just pass through them and you will get grabbed - because well, you were already grabbed.

The same happens with grab supers. Except, the “grab check” occurs during the last frame of super freeze. Meaning that in effect, the first and only active frame of a grab super occurs on frame 30 of super freeze. In Saiki’s case, the extension of his hand after the freeze is actually part of his recovery frames, as that frame isn’t a grab check frame - the one before it is.

So basically, it’s 1F invul startup + 29 additional super freeze where everything is frozen - 1F at the end where your grab goes active but the enemy is still frozen. In other words, 1+29-1.

Or just 1-1.

Yeah that works too - but traditionally, frame data for all games (that I’ve seen anyway) lists projectiles with their startup and total frames only. Presumably because knowing the total time it takes to throw a fireball is the most important thing. And there are moves like Saiki’s qcf+AC fireball which would be kind of ridiculous to read without a solid total frames number.


I’m probably going to look you up on the chat for this, and I’ll try to get some recordings done myself, as this is pretty hard to understand.

So, if Saiki’s grab has 1 frame startup, and is able to punish a -1 move, how exactly does that work?

Once again the Kyo qcf+B on block example: Saiki reversals with grab super, Kyo jumps up instantly after qcf+B. Saiki reversals and Kyo still has 1 frame of recovery to go, the one invincible frame happens, Kyo recovers. The next frame superfreeze starts, and Kyo starts his first frame of his jump and is thus unthrowable. It should miss, but it doesn’t.

Doesn’t that mean that the grabcheck happens on the very first frame you input the move? Thus 0-frame startup?


Okay, I think I get it. So, 1st startup frame happens, but freeze happens before both players advance to the next frame, so Kyo can’t start his jump, since the hitcheck happens during freeze, he cannot jump out in time.

The way you have it in notation now as 1-1 is fine I guess, it’s both accurate and anyone can figure out that it functions as a 0-frame startup move.

Never seen it like that, but I agree it makes sense.

O, one more thing. I haven’t looked specifically at Saiki, but Kyo at least his st.CD actually does 23 frames of blockstun rather than 22.

This is because Kyo bestows 10 frames of hitstop on himself and 11 frames of hitstop, so the difference + 22 blockstun gives a total of 23 frames that the opponent cannot move.

I haven’t looked at other characters for this yet, is this a Kyo only thing then?


So… just noticed that all hop normals have different active frames than their jump normal counterparts. More work :confused:

Seems like the upper limit for active frames on hop normals is 8F. As King’s jump normals tend to have 10 or 11 active frames, they’re all reduced to 8. So far I’ve noticed that 9-10F AF jump normals lose 2 AF on hop. 11 AF normals lose 3 AF. 8 AF normals lose 1F, and potentially 6-7F AF normals lose 1 (haven’t come across these yet). 5 and lower AF normals don’t lose any.

Just tested this, and Kyo’s st.CD does 22F of blockstun (using traditional KoF counting, not mine), same as the others I’ve tested. There’s definitely no difference in hitstop - both are in it for a total of 12 (or 11 if you don’t count the hitting frame). Tested this on whiff too, and whether on whiff or block, it takes the same number of frames for Kyo to start a jump after doing CD, when subtracting 11F of extra hitstop.