Random running glitch from KOF '96. Possible mind games and shenanigans? *Shrugs*
http://www.youtube.com/watch?v=quM0Eosatng
KoF Engine Research
Re: Random Obscure Findings
heh, a bit old, been used in several CMV's from the days of old.oKutabareo wrote:Random running glitch from KOF '96. Possible mind games and shenanigans? *Shrugs*
http://www.youtube.com/watch?v=quM0Eosatng
But yeah, it is a good resource material for the ones who wanna find out how to do it. Actually, if you wanna use this clip and input the step-by-step explanation, we can make a small tutorial (i could find someone to edit it) and add it to the resources material for combovid.com and/or CX.com. Hit me back if interested.
-
oKutabareo
- Posts: 40
- Joined: Mon Feb 09, 2009 10:28 pm
- Location: New York
Re: Random Obscure Findings
Oh, didn't know it was an old glitch. I just happened to fiddle around with '96 and bumped into it haha.
Yeah sure, I'm interested in it.
Yeah sure, I'm interested in it.
Re: Random Obscure Findings
On the subject of random KOF glitches, I have the following video:
http://www.youtube.com/watch?v=x2HYb859-RM
Billy connects cl.C xx f+A, cr.C xx f+A is a standard combo in corners and connects on all of the cast (or so I thought!).
But today someone who was learning his BnB's told me: Hey this combo seems to only connect on the right side of the screen.
Seemed weird to me, checked it out, worked fine against Iori on both sides, so I asked what his dummy was, and it was Athena. So I tried it with Athena, and indeed it whiffed the first hit of the second f+A!
So just to make sure human error wasn't playing a role I wrote the combo to work in FBA-RR. And without fail the combo would connect on the right side, and would whiff on the left side.
But then suddenly, while trying it a couple of times, all of a sudden it connected.
So about once every 20 times it seems to connect on the left side. I've had it whiff on the right side, but even less often, and maybe it was a range dependent thing.
But that it doesn't work on the left side is definitely not range dependent.
So what does cause it? Well it's not a hitbox thing either. Unlike 3s and SF4, during idle animations hitboxes in kof2002 are completely static. So it must be a matter of rounding off by the CPU. A point where sometimes 1.9999999 is rounded down to 1 and sometimes round up to 2. Why this only happens in the left hand side is beyond me.
The rounding up and down thing is also visible in the jumps. I've been looking at acceleration and deceleration. And technically speaking it should be a perfect parabola so the jumping up steps in pixels should be the exact same as going down. But it's not, there are slight discrepancies here and there of one or two pixels.
http://www.youtube.com/watch?v=x2HYb859-RM
Billy connects cl.C xx f+A, cr.C xx f+A is a standard combo in corners and connects on all of the cast (or so I thought!).
But today someone who was learning his BnB's told me: Hey this combo seems to only connect on the right side of the screen.
Seemed weird to me, checked it out, worked fine against Iori on both sides, so I asked what his dummy was, and it was Athena. So I tried it with Athena, and indeed it whiffed the first hit of the second f+A!
So just to make sure human error wasn't playing a role I wrote the combo to work in FBA-RR. And without fail the combo would connect on the right side, and would whiff on the left side.
But then suddenly, while trying it a couple of times, all of a sudden it connected.
So about once every 20 times it seems to connect on the left side. I've had it whiff on the right side, but even less often, and maybe it was a range dependent thing.
But that it doesn't work on the left side is definitely not range dependent.
So what does cause it? Well it's not a hitbox thing either. Unlike 3s and SF4, during idle animations hitboxes in kof2002 are completely static. So it must be a matter of rounding off by the CPU. A point where sometimes 1.9999999 is rounded down to 1 and sometimes round up to 2. Why this only happens in the left hand side is beyond me.
The rounding up and down thing is also visible in the jumps. I've been looking at acceleration and deceleration. And technically speaking it should be a perfect parabola so the jumping up steps in pixels should be the exact same as going down. But it's not, there are slight discrepancies here and there of one or two pixels.
Re: Random Obscure Findings
Are you referring to Billy's move's hitbox having numbers change, or Athena's hitstun/hitbox etc/
Re: Random Obscure Findings
No, notice that the f+A is a move that actually leaves the ground. I think that the movement of this move is actually somewhat determined like a jump. That is, calculated with gravity. The gravity is doubtlessly going to yield non-full numbers that have to be noticeably rounded when put onto a low-resolution grid. I believe that that's the case here. Or at least, that's my best guess what's going on here. Maybe it's not the gravity but simply the pushback that affects airborne characters in a different way (maybe some friction playing a role as Billy lands back on the ground? Not sure if the engine has friction it seems like landing with the ground makes you stop in your tracks immediately).onReload wrote:Are you referring to Billy's move's hitbox having numbers change, or Athena's hitstun/hitbox etc/
Re: Random Obscure Findings
reminds me of a rainbow edition, where a Hurricane kick will hit twice as much in the left corner as it does in the right
Re: Random Obscure Findings
Actually, the 2002 engine it's sometimes a little bit... strange in certain things. There is certain particular combos or glitches where only work by facing certain characters. Sometimes also has something to do with the positioning of the hitbox and/or with the active/passive stance of the sprites (yup, that's a very deep analysis that can trigger certain bugs).
I'm not sure exactly what triggers this exact bug, i'll try to run it with hitbox detection and find the "problem". But yeah, it's most likely the hitbox positioning.
Actually (talking about other topic but still relevant), there's some curious cases where the hitbox gets displaced by the movement of the character or viceversa, the sprite gets displaced by the original hitbox. An interesting case is Benimaru's taunt, where he looks like stepping back while beckoning the opponent. Usually, every hitboxs follow the movement of the sprite wherever he moves, right? but in this case, the hitbox remains static in its original position.
Here's an example in video:
http://www.megaupload.com/?d=N6VE5MMJ
I'm not sure exactly what triggers this exact bug, i'll try to run it with hitbox detection and find the "problem". But yeah, it's most likely the hitbox positioning.
Actually (talking about other topic but still relevant), there's some curious cases where the hitbox gets displaced by the movement of the character or viceversa, the sprite gets displaced by the original hitbox. An interesting case is Benimaru's taunt, where he looks like stepping back while beckoning the opponent. Usually, every hitboxs follow the movement of the sprite wherever he moves, right? but in this case, the hitbox remains static in its original position.
Here's an example in video:
http://www.megaupload.com/?d=N6VE5MMJ
Re: Random Obscure Findings
I had a closer look at the combo and 'where' it goes wrong.
Counting from the first frame of the close C, it goes wrong on the 92nd frame. That is 2 frames afer cr.C is canceled into f+A, the first frame where Billy is considered airborne. Form some reason Billy is placed but a pixel further back when it misses, and a pixel more to the front when it hits.
I have no clue why it does that, and why it seems to do it randomly. But it does. If you're interested in getting the video files for it (including hitboxes) I'll upload them.
Problem with 2k2 hitbox display is that the hitbox disappears if it becomes active so if it's connects on the first possible frame you sadly don't see the hitbox.
Counting from the first frame of the close C, it goes wrong on the 92nd frame. That is 2 frames afer cr.C is canceled into f+A, the first frame where Billy is considered airborne. Form some reason Billy is placed but a pixel further back when it misses, and a pixel more to the front when it hits.
I have no clue why it does that, and why it seems to do it randomly. But it does. If you're interested in getting the video files for it (including hitboxes) I'll upload them.
Problem with 2k2 hitbox display is that the hitbox disappears if it becomes active so if it's connects on the first possible frame you sadly don't see the hitbox.
Re: Random Obscure Findings
Hmmm okay, I was also doing some research to see if the problem is the dummy and the left/right positioning. So I did some tryouts with Ramon and Kensou with their famous endless strings (Ramon's dC, dfB, qcb C and Kensou's D, fB) and seems there's no difference from the left and right positioning BUT.... for example, Kensou's string doesn't land on Athena, but in other dummies like Mai lands with no problem.
Diagnosis: Athena's hitbox it's smaller and some attacks won't land inside a combo, however, there's no apparent difference from left/right.
Now after what you found out about the starting/landing thing of Kane's fA, now I'm running other test with Kane and see differences with left/right.
BTW, Maj, think you could split this thread into KOF engine research thread or something?
Diagnosis: Athena's hitbox it's smaller and some attacks won't land inside a combo, however, there's no apparent difference from left/right.
Now after what you found out about the starting/landing thing of Kane's fA, now I'm running other test with Kane and see differences with left/right.
BTW, Maj, think you could split this thread into KOF engine research thread or something?
Re: Random Obscure Findings
Does the hitbox dispaly dip setting work properly for you guys? Every time I try using it the colours are all fucked up, with every box being nearly the same shade (that changes as you move about the screen).
Re: Random Obscure Findings
Have the same problem, though it's slightly better viewable in kof98 than it is in kof2002.Remxi wrote:Does the hitbox dispaly dip setting work properly for you guys? Every time I try using it the colours are all fucked up, with every box being nearly the same shade (that changes as you move about the screen).
Re: Random Obscure Findings
In Kawaks the colors are fine. I don't use FBARR.
Re: Random Obscure Findings
What version of Kawaks are you using? With me they most definitely are not.Don Vecta wrote:In Kawaks the colors are fine. I don't use FBARR.
Re: KoF Engine Research
Probably not as random and obscure than a combo that is side dependant and only works half of the time. But during the #snkplaymore battle gauntlet this happened:
http://www.youtube.com/watch?v=yMgv5JE6bgg#t=20s
Dandy J does a hop CD with kusanagi, and does another one, but ends up with the move coming out backward? How did that happen?
http://www.youtube.com/watch?v=yMgv5JE6bgg#t=20s
Dandy J does a hop CD with kusanagi, and does another one, but ends up with the move coming out backward? How did that happen?
Re: KoF Engine Research
Apparently some moves in KOF2002 can cancel active frames rather than just their recovery.
That's why K9999's cl.A xx Moon DM works. Moon DM is 9 frames startup, cl.A is startup 4, active 3, recovery 10.
Light attacks do 11 frames of hitstun.
If you could only cancel after the active frames, the 2 remaining active frames would eat 2 of the 11 frames of hitstun, which would leave with 9 frames, which isn't enough because Moon DM only becomes active on the 10th frames.
So instead you have to cancel on the 2nd or 3rd active frames for it to connect.
Does anyone know of any other moves that are cancellable during their active frames?
That's why K9999's cl.A xx Moon DM works. Moon DM is 9 frames startup, cl.A is startup 4, active 3, recovery 10.
Light attacks do 11 frames of hitstun.
If you could only cancel after the active frames, the 2 remaining active frames would eat 2 of the 11 frames of hitstun, which would leave with 9 frames, which isn't enough because Moon DM only becomes active on the 10th frames.
So instead you have to cancel on the 2nd or 3rd active frames for it to connect.
Does anyone know of any other moves that are cancellable during their active frames?
Re: KoF Engine Research
that's how special canceling always works in capcom fighting games, do all moves not work that way?
Re: KoF Engine Research
I was always under the impression that, specifically in capcom games, moves only cancel after their active frames. But just having a quick look that seems to not be the case. I wonder what gave me that idea now.error1 wrote:that's how special canceling always works in capcom fighting games, do all moves not work that way?
Re: KoF Engine Research
http://www.megaupload.com/?d=860XFMQM
http://www.youtube.com/watch?v=HVRSj1Fj8XQ
rock2no once again has discovered a couple of interesting bugs from classic KOF, 95, 97 and 98 to be precise. All of them allowing to disable any sort of grabs in between the characters.
Transcription:
In 95, seems it's possible to negate both grabs and Daimon's earthquake effect by the use of this bug.
1) Both characters should be in MAX mode.
2) Daimon hits Takuma with a CD (could be any attack) and Takuma must block it and alpha cancel it with qcf C. IMPORTANT: Takuma should NOT hit Daimon whatsoever, in this case, the fb bypasses Daimon's sprite. If everything's correct, hald of the bug it's triggered.
3) Now it's the turn of Takuma to pull a CD that Daimon should block and alpha counter with qcb D, then again, Takuma should not be harmed when this happens.
4) If everything's done as explained, now none of them would be able to perform grabs, the Shoran Kyaku of Takuma hcb D will whiff and Daimon's earthquake dp A also won't have any effect in Takuma. Also Daimon won't be able to perform his command grabs either (dp K, hcb f P, hcb x2 C DM).
5) Please notice something curious, in the displayed example Takuma's Shoran Kyaku will rush towards Daimon and stay next to him, but in the second time, he will pass through him, ignoring him completely. To cause this small bug, Takuma should go completely back to the edge of the screen and immediately perform hcb D, when he reaches Daimon he'll pass through him.
In 97/98 is as follows (credit to KOFplayer for helping with the transcription):
The principle of this bug it's quite simple.
1P as any other character, 2P as either Kyo or Athena.
When 2P is performing SDM, 1P has to perform a special move (or command move) which comes into contact with the SDM but not get hit by the SDM. 2P cannot be hit by 1P's move as well.
Upon success, 1P can no longer throw/grab 2P.
In the example shown we have Orochi vs Kyo and Athena in 97.
1) Kyo performs his qcb hcf A SDM and Athena her hcb x2 A SDM.
2) Orochi performs his lighting attack (note: forgot the motion of this command), which touches the SDM attack frames, but do NOT harm either Kyo or Athena.
3) Orochi won't be able to either grab the opponent neither connect his grab DM. Athena also won't be able to pull her hcf C command grab.
In 98 we have Yamazaki vs/ Athena. In this special case, to make both characters nullify their own grabs, it's necessary to use Ryuji Yamazaki.
1) Athena performs her hcb x2 A SDM
2) Ryuji then should use his qcf C attack, which is supposed to absorb energy attacks but not bouncing back like with qcf A. Please notice the positioning of Yamazaki: he's placed in a position where the SDM won't damage him and the punch motion of the qcf C won't connect either BUT still absorbs Athena's energy aura.
3) Upon completion, Ryuji can't connect neither hcb f P grabs nor his hcb x2 P DM/SDM. Athena also has her hcf P grab disabled.
http://www.youtube.com/watch?v=HVRSj1Fj8XQ
rock2no once again has discovered a couple of interesting bugs from classic KOF, 95, 97 and 98 to be precise. All of them allowing to disable any sort of grabs in between the characters.
Transcription:
In 95, seems it's possible to negate both grabs and Daimon's earthquake effect by the use of this bug.
1) Both characters should be in MAX mode.
2) Daimon hits Takuma with a CD (could be any attack) and Takuma must block it and alpha cancel it with qcf C. IMPORTANT: Takuma should NOT hit Daimon whatsoever, in this case, the fb bypasses Daimon's sprite. If everything's correct, hald of the bug it's triggered.
3) Now it's the turn of Takuma to pull a CD that Daimon should block and alpha counter with qcb D, then again, Takuma should not be harmed when this happens.
4) If everything's done as explained, now none of them would be able to perform grabs, the Shoran Kyaku of Takuma hcb D will whiff and Daimon's earthquake dp A also won't have any effect in Takuma. Also Daimon won't be able to perform his command grabs either (dp K, hcb f P, hcb x2 C DM).
5) Please notice something curious, in the displayed example Takuma's Shoran Kyaku will rush towards Daimon and stay next to him, but in the second time, he will pass through him, ignoring him completely. To cause this small bug, Takuma should go completely back to the edge of the screen and immediately perform hcb D, when he reaches Daimon he'll pass through him.
In 97/98 is as follows (credit to KOFplayer for helping with the transcription):
The principle of this bug it's quite simple.
1P as any other character, 2P as either Kyo or Athena.
When 2P is performing SDM, 1P has to perform a special move (or command move) which comes into contact with the SDM but not get hit by the SDM. 2P cannot be hit by 1P's move as well.
Upon success, 1P can no longer throw/grab 2P.
In the example shown we have Orochi vs Kyo and Athena in 97.
1) Kyo performs his qcb hcf A SDM and Athena her hcb x2 A SDM.
2) Orochi performs his lighting attack (note: forgot the motion of this command), which touches the SDM attack frames, but do NOT harm either Kyo or Athena.
3) Orochi won't be able to either grab the opponent neither connect his grab DM. Athena also won't be able to pull her hcf C command grab.
In 98 we have Yamazaki vs/ Athena. In this special case, to make both characters nullify their own grabs, it's necessary to use Ryuji Yamazaki.
1) Athena performs her hcb x2 A SDM
2) Ryuji then should use his qcf C attack, which is supposed to absorb energy attacks but not bouncing back like with qcf A. Please notice the positioning of Yamazaki: he's placed in a position where the SDM won't damage him and the punch motion of the qcf C won't connect either BUT still absorbs Athena's energy aura.
3) Upon completion, Ryuji can't connect neither hcb f P grabs nor his hcb x2 P DM/SDM. Athena also has her hcf P grab disabled.
Re: KoF Engine Research
Special canceling in the Street Fighter 2 series is timing based. I get the impression that later games have a 'move properties' approach instead.error1 wrote:that's how special canceling always works in capcom fighting games, do all moves not work that way?
Re: KoF Engine Research
Moves are generally cancelable (or not) based on startup being shorter than a certain number, and only cancelable up to that standardized frame. It's kind of a messy convention, as with everything in SF2.
Anyway the end result is the same. Canceling stuff makes you do the followup immediately after impact freeze ends, eliminating the remaining active frames. You can manually delay it though, by a frame or two in most cases.
Anyway the end result is the same. Canceling stuff makes you do the followup immediately after impact freeze ends, eliminating the remaining active frames. You can manually delay it though, by a frame or two in most cases.
Re: KoF Engine Research
SRK is down, but I did some googling and I don't know if anybody is working on compiling a list of changes made to KoFXIII; seems like the update is "ver. 1.1", and I think they modified Raiden's dropkick charge times, and supposedly the bugs will be gone, but...who knows.
Re: KoF Engine Research
http://cyberfanatix.com/forums/index.ph ... l#msg33981onReload wrote:SRK is down, but I did some googling and I don't know if anybody is working on compiling a list of changes made to KoFXIII; seems like the update is "ver. 1.1", and I think they modified Raiden's dropkick charge times, and supposedly the bugs will be gone, but...who knows.
Re: KoF Engine Research
So I've been working on a website with all the hitbox data of KOF2002... but I've run into a problem which I haven't been able to solve.
If you switch on the hitbox dipswitches in kof2002, first of all the colors are pretty horrible to look at compared to kof98, but also the hitboxes are a lot more ambiguous. With almost every attackbox a second 'ghost' attackbox appears around it.
Usually these ghostboxes stay out just as long as the actual attackboxes, but some moves this is not the case. An example of this is Maxima's jump A which has the ghost hitbox for the first 6 active frames and then disappears on the 7th till 9th.
Here are pictures of the hitboxes in question:
Does anyone have an idea what these ghostboxes mean?
They are usually found a couple of pixels around the actual attackbox, but this isn't always the case, sometimes they take up a slightly different box, like Iori's jump C.
It almost seems as if it is irrelevant and should be ignored but I don't feel comfortable ignoring something in the hitbox data that I do not understand.
If you switch on the hitbox dipswitches in kof2002, first of all the colors are pretty horrible to look at compared to kof98, but also the hitboxes are a lot more ambiguous. With almost every attackbox a second 'ghost' attackbox appears around it.
Usually these ghostboxes stay out just as long as the actual attackboxes, but some moves this is not the case. An example of this is Maxima's jump A which has the ghost hitbox for the first 6 active frames and then disappears on the 7th till 9th.
Here are pictures of the hitboxes in question:
Does anyone have an idea what these ghostboxes mean?
They are usually found a couple of pixels around the actual attackbox, but this isn't always the case, sometimes they take up a slightly different box, like Iori's jump C.
It almost seems as if it is irrelevant and should be ignored but I don't feel comfortable ignoring something in the hitbox data that I do not understand.
Re: KoF Engine Research
For the hitbox colors me and some buddies are working in decipher the colors of it. We are working in a full investigation out of it. KOF engine (especially 2002) has very complex hitboxes and require tons of research. We're on that.
Re: KoF Engine Research
Is it possible that the hitbox properties of the move are changing while the graphics stay the same? I know it happens in other games. I'm not familiar with KoF, but my inclination would be to look at throwability and pushback properties.phoenix wrote: Usually these ghostboxes stay out just as long as the actual attackboxes, but some moves this is not the case. An example of this is Maxima's jump A which has the ghost hitbox for the first 6 active frames and then disappears on the 7th till 9th.
...
It almost seems as if it is irrelevant and should be ignored but I don't feel comfortable ignoring something in the hitbox data that I do not understand.
Re: KoF Engine Research
It could be, but the thing is the 'ghostboxes' don't actually like do anything, they never collide with anything, you'd think if they'd mark certain properties you'd expect them to 'do' something right, even when a move hits only the 'real' box disappear, not the ghostbox.Rufus wrote:Is it possible that the hitbox properties of the move are changing while the graphics stay the same? I know it happens in other games. I'm not familiar with KoF, but my inclination would be to look at throwability and pushback properties.
It might be interesting to mention that also kof2000 has these ghostboxes, by extension I'd guess also 2k1 and probably 99 has them but I haven't checked those yet.
Re: KoF Engine Research
Is there anyplace I could get framedata and/or hitbox images for 98 Mai? I noticed both the JP wiki and SRK wiki have identical data sets, which lacks anything on Mai.
Looks like Jolly Ranchers & Baskin's Sherbet.
Re: KoF Engine Research
Only stupid assumptions...
It may have something to do with the strikers? Like a possibility to use it, or something like striker start up?... Sounds really stupid...
Maybe a basic position for the game engine purpposes? That ghostboxes do not seem "very accurates"...
It may have something to do with the strikers? Like a possibility to use it, or something like striker start up?... Sounds really stupid...
Maybe a basic position for the game engine purpposes? That ghostboxes do not seem "very accurates"...
So what happen exactly from the 7th frame to the 9th? Properties? And what happen with the enemy´s ghostboxes in that moment?phoenix wrote: An example of this is Maxima's jump A which has the ghost hitbox for the first 6 active frames and then disappears on the 7th till 9th.