SF4 Combo Engine Investigation

[desk]

just tried to make a hugely complimentary comment on one of the blog posts but it wouldn't let me without signing up for something or other.

Basically, it's great. I've already taken notes on the rufus post (very useful) plus it's just awesome to hear from someone who knows what they're talking about (god knows I'm sick of some of the shit that get's posted on SRK) and has access to the game. So yeah, keep the posts coming.

[gilley]

Yeah, don't worry about not being able to post comments. I'm really not looking for any If you really want to discuss anything I talk about in my blog, I guess this forum might be as good a place as any. SRK is garbage.

I'm just writing down anything I find, or want to try. If you see any insane Guile combos that don't have "confirmed as working" next to them, don't freak out. I'm trying to come up with ideas for the most craziest combo I can think of. Some of them might not work at all.

Speaking of combos...does anyone have any ideas for a 100% stun combo with any character?(besides fuerte's infinite). I've been trying to figure out a stun combo on Akuma and have been unsuccessful. I'm not sure, but I believe the stun damage scaling is the same as a the normal damage scaling. The damage scaling curve goes like this....

1st hit = 100%
2nd = 100%
3rd = 80%
4th = 70%
5th = 60%
6th = 50%
7th = 40%
8th = 30%
9th = 20%
10th = 10%
after this, every hit only does 10% of its base damage.

So as you can see, the damage scales FAST. If stun scaling is the same, then I don't think we're going to be seeing many(any) 100% stun combos in SF4. The closest combo I've found to being 100% stun on akuma(850 stun) is...

lvl3 FA > f+rh > c.strong > EX SB > lvl2 FA > f+rh > c.strong > SB > c.rh (795 scaled stun)

This is with Guile and assuming that stun scaling = damage scaling. There's another idea for a combo that I think will stun Akuma but it's really not the type of combo I want to put in my video...though I might have to...
have opponent throw FB and FADC it, backdash to the other side of the level, opponent goes over/under you. You do j.fierce > s.fierce > opponent's FB hits you > s.fierce > EX SB > lvl2 FA > f+rh > c.strong > SB > backfist (850 scaled stun)

If anyone has any wacky ideas for a 100% stun combo they would like me to try, please let me know. Hopefully nothing too difficult cause I don't have training mode or a 2nd player to help me.

[desk]

a really simple idea would be this...

throw a sonic boom when backed into the corner, then dash after it while the opponent jumps back. The desired result would be that you corner them wth the SB well behind you so you could land j. fierce, st. fierce [SB hits] j. fierce, st. fierce xx SB xx FADC, cr. strong xx Flash kick.

It may have to be modified so that you fadc the initial SB (to gain more ground) or even the first st. fierce (to lessen recovery and allow the 2nd j. fierce).

Probably won't work but it's the only thing I can think of right now. Also, it might just be too hard to do manually, stuff like this often requires very precise timing to space out the hit stun for everything to connect.

EDIT: also, I didn't really calculate the stun damage, lol. I'm just guessing it'll be enough.

EDIT: haha, I've just realised I've posted pretty much the same combo you did, plus your's has a much better ender. So the 2nd jump fierce wouldn't be possible?

[Mike Z]

Do CH's do more stun, or just more hitstun?
How about Zangief vertical j.Strong, land, junk? That does 600 stun by itself, with his c.Jab doing 50 stun and the green hands doing 100 to 200 it almost seems like the B&B c.Jab x2, s.Jab, link s.Short xx glove ought to do it right there.

[gilley]

@desk

unfortunately, the SB causes 22 frames of hitstun. I'm not sure this will give Guile enough time to do a j.fierce. Guile's jump is 4+36 frames. The SB has to hit immediately after the first s.fierce. I will try this later though to confirm.

@mikeZ
that combo does 830 stun, assuming stun scaling = hitscaling. Almost able to stun Akuma.

I would imagine something like u+mp > u+fp will stun if the u+mp is counterhit.

[desk]

does counter hit stun apply to every hit of a combo? also would you be averse to doing it on a cr. akuma? Might squeeze out a few more frames.

[Mike Z]

Well U+Strong, U+Strong should stun everyone then, no? Z has the most stun of anyone, which is 1200. (^.^)

[desk]

oh yeah, forgot to ask. What about fadc-ing the first st. fierce? is that quicker than letting it recover? also, even if his jump is that long, you'd be hitting fierce well before the jump is actually complete. I think it could be possible... good luck testing it on your own though, lol.

[Maj]

How about the originally proposed Guile combo starting with the oldschool TZW setup of Dhalsim jumping over and whiffing s.HP to reaching into an LP Sonic Boom?

[gilley]

@desk
That's a good idea, I will have to test it out this weekend, it could possibly work. If it does the combo will definately dizzy akuma (890 stun)

@maj
I dunno if that will dizzy dhalsim, he has more stun than akuma and from what I've been calculating, you really need those first 2 hits to be fierce/rh attacks in order to have the slightest chance of dizzying your opponent. After about 5-6 hits, the scaling is just retarded. It's worth a shot though.


Also, I was playing against the cpu tonight and something happened that gave me a great idea. I was fighting Blanka and he was winding up his lvl3 FA. My Guile was just mashing on far s.fierce. The s.fierce traded with Blanka's FA when it came out. This sent me into crumple stun and knocked Blanka out of the recovery of his FA. Normally you have to dash cancel to get out of FA recovery. After he was hit out of his FA, he jumped at me and hit roundhouse while I was falling over from FA crumple. 2-hit combo flashed on screen. So this is the idea...

lvl3 FA (trade) > j.fierce > s.fierce > EX SB > lvl2 FA > f+rh > c.strong > SB > backfist (850 stun)

This combo should be ALOT easier to setup than the others.

[Maj]

Rufus' EX spinny move acts like a vortex right? Does that pulling effect work from across the screen? If so, you might be able to set up a rejump combo using the same setup concept as shown against Shadow at 2:52 in this SFEX2+ combo video.

[Maj]

The newest issue of Hardcore Gamer Magazine has a whole mini-strategy-guide devoted to SF4.

HGM #34 Available to Download!

It's got a ton of random info including special/super move properties along with an impressive selection of miscellaneous frame data. Guess it was written by Arlieth of Denjin-Video fame. Definitely worth a look.

[ZenFire]

Probably not the right thread to be asking but I'll just go ahead and do it anyway:

Is there an off-by-one error on the startup frames of every move in the frame date available to us (SRK wiki)?

I played a bit of console sf4 yesterday at a friend's. I don't have the game myself so I can't test. I swear I was able to do 0-frame links which shouldn't be possible, which makes me think the startup frames of moves were measured at the point of button presses and not the number of actual animation frames. According to http://www.shoryuken.com/wiki/index.php/Ryu_%28SFIV%29 C.LP > C.MK shouldn't link, but after messing around in training mode I'm pretty sure I pulled it off, though I'm starting to doubt myself now.

[Xenozip.]

Well you know there's the Japanese and Western way of writing frames.

A 3F move could mean that is has 3 frames of startup animation and then hit on the 4th. Or a 3F move hit on the 3rd frame.

According to the wiki it's +5 and then the startup is 5. If they wrote advantage like the opponent is still in hitstun for 5 frames while you animate for exactly 5 frames, and they wrote startup like 5F hits on the 5th frame then it'd be possible, but a just frame link.

Which, by the way, is how I would write it. But you know, the other way to write it make just as much sense because it's easier to calculate total move duration by just adding startup active and recovery.

[ZenFire]

I always assumed the difference in notation was due to where you start timing/measuring and not a different interpretation of 'startup'.

See illustration This is what I think they're saying.
According to roman numerals: "takes 4/60th of a second to hit starting from input" = "4 frame startup"
According to numbers in the squares: "3 frames of animation, hits on 4th" = "3 frame startup"

Doesn't matter why there's a discrepancy, I'm just wanting to know if I should subtract 1 from every startup value in the frame data on wiki for it to make sense in my head.

[Maj]

I've never heard of a distinction between "the Japanese and Western way of writing frames." The standard way of writing startup is that 5 ST means it hits on the 6th frame. Nobody i know prefers to write it the other way around. It would work if the other way was the standard but it's not and having a standard is important. If Prima or whoever got it wrong, then i'm sure all the numbers will get adjusted eventually.

Might be worthwhile to start a thread on SRK and take this up with whoever transcribed all that data. Sometimes people get stubborn and fall in love with the wrong way of doing things once they get off on the wrong foot. ("Rye you!!")

[ZenFire]

Glad to see I'm not the only one who thinks that notation is weird. I'm definitely not starting a thread on SRK, cus I'm likely not going to care about what the ppl that do reply have to say on the subject and it'll only aggravate me if anything. I agree with you maj, that this is the 'wrong' way to do it, but I quote that because apparently it's actually pretty common practice. Some guy told me that VF and GG frame data use the same 'japanese' notation. He didn't call it japanese, but I'm referring to it like that since Xenozip did. Japan is also the source of Game-restaurant 3S data and the awesome CVS2 guidebook, which use the western notation, so I'm not exactly sure it's an accurate naming convention. I was wrong in thinking the startup frames difference had something to do with the time of input.

I found out something to further corroberate the whole 2 notation thing. I remember ppl saying the Karathrow and 3S bible frame data was wrong because it was based on DC data. I totally accepted that as the truth because people said it with such certainty (so naive). Check these two Chun frame data sets:
A) http://home.att.ne.jp/star/GR/3rd/chun02.html
B) http://web.archive.org/web/200412080224 ... hunli.html

I'll spare you the trouble and name an example, Chun's low jab:
Game restaurant: 2/4/5
Karathrow: 3/3/5

So the first hitting frame is attributed to the startup in the latter. It's like Xenozip said, it's a different interpretation of 'startup'. We all know it's a 2-framer and hits on the 3rd, they just write the same thing in a different way. It makes a lot more sense in the first notation, cus the number of active frames actually means what it says. When you color them like this: http://www.drakenslag.com/ryu.html the second notation would be even more confusing because the red area should be hitting frames, but there would be 1 missing. Anyway, it's certainly wrong in my book, but since it's more widespread than I thought I won't go around shouting it from the rooftops.

[Xenozip.]

I'm getting really annoyed.

Anyway I don't know why other people call it by jp/western style. It probably doesn't have anything to do with it. I just got use to referring to it as that because others do.

Like I said before it probably has to do with how the word "startup" is used.

A "3F move" hits on the 3rd frame. If the data table says "startup = 3F" then there's 3 frames of nothing and the first active hitframe is on the 4th frame. Call that wrong or incorrect if you want, I don't care.

Whatever. ZenFire didn't really care why in the first place, just that if it was off by a frame or not. And frankly I don't give a shit either, because I can read it either way.

[ZenFire]

Annoyed? about this subject? Why exactly?

Anyway, I still don't know IF the startups are off by one in the table, but hopefully I'll know more by the end of today since I can finally pick up my copy of the game. If the notation is off by one, then the next question is if I should add +1 to the active frames? Maybe it's time I do some more capturing.

[Maj]

Sounds like it'd be really hard to test. Unless you had a program pad and used it to set up precise meaty situations, i don't know how you could accurately measure active frames. You can't really whiff attacks in front of opponents and judge by their block animation because those recovery reactions have gotta be delayed somewhat. Plus 3D animation is a lot more fluid which makes it hard to draw lines.

Also you can't deduce from meaty combos either, because those formulas are never as simple as you'd think. For example if a move has 5 active frames and gives +4 frame advantage, that doesn't necessarily mean you get +8 when you do it as meaty as possible. I don't know any more detail than that but i've seen a bunch of CvS2 meaty combos fail which should have worked if the math was that simple. Maybe there's some kind of built-in meaty frame advantage softener?

[ZenFire]

I wasn't thinking of anything complex like that. I was just thinking of recording an attack and jumping right after it so I can count the total number of frames. If they match up with the totals in the wiki frame data tables then I would assume the active frames need 1 added to them, if the actual frames were shorter then no. Though the thing about 3D animations being much harder to disect is true, I'm just hoping the first frame of a jump is easily recognisable. I'll try to do it friday, cus it requires me to take out my wifi pci card and put back in my capture card. (graphics card too bulky to have both in at the same time). Or I could make one of the three extra pc's laying about my appartment into a dedicated capture/video machine, though I believe none of them is very fast or has a large enough HDD.

I can't think of any other way to test it without perfect timed button presses (ppad). I found that "looking at blocked animation from whiffing it in front of them" idea pretty ingenious. It might work if I could find anything to use as a reference like Chun's dissipating fireball (though then I'd have to find a way to measure when exactly it dissipates ).

I did have this other idea of just going brute force and recording a character walking forward and me mashing jab over and over, surely one of the times he'd walk into it on the last and 1 after last active frames and I could deduce that the active frames end at point x.

[fullmetalross]

Random thing I noticed while playing

Seth can probably be used to setup easy fireball interrupt combos ala rolento. His suction move will knock you out of your fireball and pull you towards him a bit. It is also invincible to fireballs during the suction (allowing said fireball to hit meaty) this should give you enough time to combo off it no problem. Also I believe if you do it from far enough away you could probably get a jump in still... is this our new steel rain????

[ShinjiGohan]

its worth a try but if the grab stun is longer than the fireball recovery then it might not work out too well. but still worth a few trys.

[Maj]

I did have this other idea of just going brute force and recording a character walking forward and me mashing jab over and over, surely one of the times he'd walk into it on the last and 1 after last active frames and I could deduce that the active frames end at point x.

That could work, if you just want to test one move. Cuz you'd have to do a ton of trials and then inspect them all, so it probably wouldn't be feasible for analyzing a whole character. There's no turbo speed in SF4 so there shouldn't be any frame skipping right?

Seth can probably be used to setup easy fireball interrupt combos ala rolento. His suction move will knock you out of your fireball and pull you towards him a bit.

Haha, that's pretty cool. We'll see it popping up in combo videos in no time.

Honda's ultra and Abel's super transport both characters all the way to the other side of the stage kinda like Rugal's wall slam, but they take too damn long. I tested Abel's super against Ryu's FB and Sagat's high FB and it seems Abel is invincible the entire time.

In other sad news, Dan's Taunt super doesn't fill up anyone's meter, not even his own.

By far the weirdest thing i've noticed in SF4 is the way jump attacks set up combos. One of Vega's Challenge Mode Hard Trial combos is j.HP, s.HP, s.LK, c.MP which is really odd because s.HP by itself puts Vega way too far away to link s.LK afterwards. But somehow if you start the combo with a frontal jump attack, Vega stays closer. The other crazy thing is that crossups suck. You actually land way further away from the opponent than with frontal jump attacks.

[Maj]

I noticed in the frame data that there's a lot of discrepancy between hit advantage and block advantage. Usually they're only one frame off in Street Fighter games, but everything has a three or four frame difference in SF4. It's like MvC1 all over again - Gief blocks anything and you get Piledriven.

CPU Abel is hella cheap too. He's like a mini-Seth. Rolls around all day flaunting his invincibility, then does bullshit reactionary wakeup shenanigans.

When i was trying to test Abel's super against Ryu and Sagat in training mode, those bastards wouldn't calm the fuck down either. Giving myself infinite meter automatically gave the opponent infinite meter as well and they literally did nothing but spam supers. I guess you can get around that by giving yourself infinite meter and then turning it off so that the CPU opponent only gets one shot at it. But even after that Ryu absolutely refused to throw regular fireballs.

Also people keep mentioning that second-round Seth is a jerk but i haven't been able to spot any differences between his default form and his powered up form after he loses a round. Maybe tougher AI but that's about it.

[fullmetalross]

Man they need to patch rose in some juggles. haha. Not really she plays pretty decent without them its just funny how her reflector launches them and almost nothing connects after that.

[Maj]

Button priority is kinda interesting. If you push PPP, you get fierce. If you do a special move with PPP you either get EX or the fierce version if you don't have meter. Guile has F+MP and F+HP command normals so F+PPP gets him F+HP, but Abel gets F+MK when he presses F+KKK because that's his only command normal. It's interesting that F+MK takes priority over s.HK in this case.

Guile's F+KKK gives him F+HK upside-down kick when in close range, but turns into F+MK hopkick at long range. Not too surprising. There are some weird exceptions though. For some reason, Chun gets DF+LK when she does DF+KKK even though she has a DF+HK move.

Oh yeah, and the Challenge Mode combos are pretty flexible. You can insert extra moves as long as it combos, like extra jabs or juggles or whatever. Also "EX Focus Cancel" is generally a hint that you should be dash-canceling the Focus Attack before it connects. It'll say "EX Focus Attack" when they want it to connect. But they're actually fine with the Focus Attack connecting even when it says "EX Focus Cancel."

For example, Gouken's fouth Hard Mode combo is supposed to be j.HP, s.MP, c.LK xx HK Hurricane Kick (one hit) xx FADC, ShinSho without letting the Focus Attack come out. But i was able to do it with j.HP, s.MP, c.LK xx LK Hurricane Kick (one hit) xx lvl1 Focus Attack -> dash forward, ShinSho instead. The ShinSho only connects once and you lose most of the damage but it satisfies the challenge conditions.

[fullmetalross]

I thinks its actually st.mp cr.lp but whateva...

Alsooo computer feilong did a funny combo to me. I with his cr.fp with gouken cr.lp and then feilong did reversal ultra and it combo'd haha.


You got a pm for this Maj sorry I somehow hit the wrong button.


Edit: In other news I found another way for dan to combo ex gaodoken into ultra, outside of the throwing it from sort of far away and then just ultraing after it. when near corner you can do cr.fp xx ex gaodoken ~ ultra. I got it to work on blanka after some work, probably not useful for actual match play but maybe for combo video.

(should we make a community video for this? I can't record still cause of some money issues but I might be buying a usb card of some sort soon, maybe the usb blackmagic box depending on my tax return.)

Edit again: Gouken can full stun seth in one combo! and it only takes 1 ex, viper can do it to for her whole meter, but maybe for less if I can get the feint down better.

[ZenFire]

Finally got my recording PC all set up. Did some recording yesterday, nothing interesting yet, I'm afraid. It's a friggin Athlon 700 Mhz processor with only 128 MB of RAM and yet it records without dropping any frames, like... EVER. So I split the fields and counted some frames. Looks like jumping from crouching takes 1 extra frame than jumping from standing. This might get me banned from this forum but: that's normal, right? I mean, I remember ppl saying in CvS2 how rock can reversal HP elbow a blocked blanka ball onlyt if he's standing which made me assume standing up took time, but I never actually bothered to test it. Has it been like that in all SF's?

More questions:

Also, Chun's short>short>lightning legs is a cancel, which won't surprise anyone, but begs the question: why can she cancel off of rapid fire chains? The cancel isn't immediate on hit but looks a like it cancels after the last crouching short is retracted a bit, kind of like it was renda canceled. Chun can also do c.lk chained close s.lp and cancel that into a special.

Also, counter hits do more hitstun, but how much more? I can' search the Master Guide thread on SRK cus they put the search time limit to ~infinity and it has too many pages to read. Was it even in the guide?

Chun's DF+LK knocks down on counter hit, I have no idea why, but maybe it has some combo potential.

Also, Gen's hard trial #5 is stupid hard...

EDIT:
Is there anything you guys want me to try and count the frames for? Like, a specific normal for a console-only character?

[Maj]

(should we make a community video for this? I can't record still cause of some money issues but I might be buying a usb card of some sort soon, maybe the usb blackmagic box depending on my tax return.)

Honestly i don't have a whole lot of faith in community videos coming through at this point. We've had one fizzle out and another that's currently getting zero support. If you have combo ideas, it's probably better to find someone who's already making an SF4 video and earn some special thanks.

Looks like jumping from crouching takes 1 extra frame than jumping from standing. This might get me banned from this forum but: that's normal, right? :oops: I mean, I remember ppl saying in CvS2 how rock can reversal HP elbow a blocked blanka ball onlyt if he's standing which made me assume standing up took time, but I never actually bothered to test it. Has it been like that in all SF's?

I believe it does take time to go from crouching to standing and crouching to jumping in most Capcom games. However it doesn't take time to go from crouching to reversals. What happens in that Rock/Blanka example is that crouching characters are fatter than standing ones, so Blanka's bounce-off point happens further away than it would if Rock had been standing.

There's a lot about SF4 that has yet to be explained and rapid fire chain cancelability is definitely near the top of that list.

I don't want to give you any extra work but i am kinda curious how many impact frames a lvl3 Focus Attack causes. Connect with one of those and dash early, then count the number of frames the attacker stays frozen, from initial contact to dash startup. I'd like to have an estimate of how many frames of charge time you could get out of that.