One of the least discussed technical topics in fighting games is the existence of impact freeze, also known as hitstop. Someone asked about it earlier this week and i couldn’t find any articles addressing it, so i thought i’d talk about it today.
Frame data usually gives you startup, active period, and recovery for each move in the game. When an attack connects, the remainder of its active period is converted to recovery (assuming it’s not a multi-hit attack).
Impact freeze definitely exists in most fighting games and the easiest way to spot it is by mashing jabs. It’s impossible to chain a jab into another jab before it reaches an active frame. You can stand next to someone and double-tap jabs all day and you’ll never get one to whiff chain into another before it connects.
Now that we’ve established that mashed whiff jabs reach at least their first active frame, you can compare whiff jabs against connecting jabs. It’s easy to tell that whiffed jabs come out three or four times faster than connected ones. That’s because when they whiff, you’re looking at 3 frames of startup followed by 1 or 2 active frames and then canceled back to startup. When they connect, it’s like 3 frames startup, followed by around 10 frames of impact freeze and then back to startup. Therefore whiffed repetitions take around 4 frames, whereas connected reps last around 14.
Take a look at the Rufus/Viper Combo Challenge video for a few examples. First of all, Rufus’ j.DF+MK divekick connects almost on the last frame, right before he lands. Yet you can still see the impact of the attack. If there was no impact freeze then he’d basically shoot right through Viper’s leg and continue the combo with no pause whatsoever. The same thing is true of his s.LK -> s.HK chain, which was executed with only a 5 frame delay between the buttons, yet you can clearly see the s.LK connect.
As it turns out, Ultimate Mortal Kombat 3 has no impact freeze. On the ground, most punches are cancelable and most kicks aren’t. Shang Tsung chains s.LP -> s.HP then cancels into the Kung Lao transformation. Yet you only see s.HP animate for one frame before it’s canceled. Furthermore, Kung Lao cancels vertical j.HK into divekick and then vertical j.HP into divekick. In both cases the divekick actually whiffs, so it’s almost impossible to see the move that actually connects. You barely see the startup of j.HK and j.HP before they disappear on impact. That’s what it looks like when a game has no impact freeze.
Going back to the SF4 example, check out Viper’s c.HP xx HP Seismo Hammer. You can clearly see her c.HP animate before she cancels it. That command was completed quickly as well, so if there was no impact freeze then it would look like some phantom attack connected then immediately canceled into ground pound.
Impact freeze is important is because it stabilizes cancel timing. For example, Viper’s EX Seismo Hammer is a very slow move which always combos when she cancels c.HP into it. You have that entire 12-frame impact freeze window to complete the input. That’s what makes it easy. As long as you do it before impact freeze ends, she automatically cancels into the special move as soon as c.HP hitstop ends (but never sooner). If there was no impact freeze, then the timing on it would vary dramatically. For slow specials like that, you’d have to complete the command very quickly in order to get them to combo. Match gameplay would be a mess.
As for its attributes, each move causes a specific amount of impact freeze regardless of how it connects (antiair, air to air, air to ground, etc). In most games, light attacks freeze around 9 frames, medium attacks around 11, hard attacks around 13. In the oldschool SF2 series they were all around 14 frames regardless of button strength. Some specials have very long impact freeze while others have relatively short impact freeze.
Projectiles operate by slightly different rules. As expected, they cause impact freeze to the victim but not to the attacker. When a fireball connects, the victim freezes in place for around 12 frames, then starts getting pushed back. Since pushback has no effect in the corner, there’s no practical difference between one-sided impact freeze and pure hit stun if the victim is cornered. Lastly, impact freeze caused by other attacks has no effect on projectiles. For example, if Dhalsim’s DF+LP hits Ryu by reaching under his LP Hadoken, Ryu’s fireball will continue traveling forward without pause.
Knowing all this information isn’t particularly important for understanding gameplay or dealing with frame data, since most of it works fine behind the scenes. However, impact freeze is crucial for advanced charge character combos because it provides additional recharge time.
In a more practical setting, it’s also important for avoiding projectiles in matches using slides and jump attacks. For example, ST Dee Jay’s c.HK slide can go under certain projectiles only when it connects, because it freezes both characters without affecting the projectile. This works for invincibility as well. There’s a famous clip of Fei Long uppercutting through a Dee Jay fireball which wouldn’t have worked if the uppercut hadn’t connected and frozen during an invincible frame.
Make sense?
Credit goes to ano for providing SF4 hitstop numbers.
My apologies to everyone who was expecting the SF4 Cammy TACV today, but it’ll have to wait until next Tuesday. Over the weekend, i got owned up by this cold/flu from hell – sinus pain, watery eyes, the works. I can type with my eyes closed but i can’t test combos without looking at the screen. Sorry guys.
Nice Article Maj!
Derek did a good write up of hit pause. But its slightly different to what you’re talking about I think. He was talking from a game design perspective too, rather than a gameplay one.
http://lowfierce.blogspot.com/2007/03/why-some-games-feel-better-than-others.html
We hope you get better.
yeah i got the flu too a week ago or so :/ hope you get better man
Final Atomic Buster: Yeah that was a great article. It’s funny how omni caught it right away even without having to test it out. He might have been the first person to explain it to me, years before he wrote that piece. Guess he was always meant to be a combat designer. Though maybe he should have disabled ratings on those u2b clips.
Nice article Maj. I like how you explored the game play elements that hitstop affects. Do you think it’s common knowledge though? Not the frame data element, but just the existence of it is easy to notice. Get better, though by the time you read this, you’ll be able to open your eyes. Meaning that you’ll be feeling better.
Impact freeze has a very important application in option selects. For example, in SF4 when Ryu does a jump HK on a waking opponent and immediately inputs a DP. If the j.HK makes contact (blocked or hit), the DP input occurs during the impact freeze so the DP won’t execute (because you can’t cancel j.HK into DP). If the j.HK whiffs (for example the opponent backdashes), there is no impact freeze so Ryu immediately lands. The DP input occurs while Ryu is on the ground so he executes a DP. I don’t think there’s a name for this sort of os, but a good one would be impact freeze option select. These are widely used in SF4.
Impact freeze to stabilize cancel timing is very nice. It’s slightly counteracted though in newer games because you can input the cancel even after the impact freeze, as long as you are still on a cancellable frame in the animation. This has its pros, but makes for inconsistent combos that can fail if your input is too slow. In older games, either it only allowed you to input during the impact freeze, or the cancel window on the moves themselves were much smaller.
So regarding fireballs, the victim undergoes impact freeze, plus hit stun immediately after that, right? So with projectile frame data, it’s important to know the hitstun as well as the impact freeze. Unlike with other attacks, since both characters undergo the same impact freeze, creating a net effect of 0 in frame advantage.
Actually, you don’t need any of that to determine frame advantage at point blank range. You just need to calculate when one character recovers compared to the other – same as always. The listed frame advantages for fireball attacks are accurate at point blank range, and obviously they increase the longer the projectile travels before connecting.
As for attacks being cancelable after impact freeze, that’s relatively rare. Cancel windows being bigger in newer games isn’t true either, apart from certain older games simply running at faster turbo speeds. If you compare base speed SF2CE/ST/SFA2 to SF4, the older games might even have bigger cancel windows. But turbo speed HF/ST/A2 got blazing fast all around.
And you’re right about the option select stuff. We kinda had that discussion in the comment section of the Rufus/Viper Combo Challenge entry. Thanks for bringing it up again here though, cuz it’s totally relevant and i probably should’ve mentioned it.
I have some questions on this topic:
Just to confirm: Impact freeze/hit stop applies to both hit AND blocked attacks, right? If hitstop for an attack is 14F, it’s 14F if that attack is hit or blocked, right? Or is there a separate “block stop” value? Or does it differ by game?
Secondly, what’s the deal with stationary projectiles like Dhalsim’s Yoga Flame and Benimaru’s Raijiken? I’m looking their frame data in various games and trying to figure out how their frame advantage is calculated, and it’s confusing me. Either these moves have some odd behaviour, or their listed values are wrong.
Take LP Yoga Flame in SFIV. It’s data is 15/23/22/D/-6. It says it has 20F block stun. If hitstop is 12F, and assuming the data is listed for when the move is blocked on the first frame, that should be 22+22 – 20+12 = -12. Where’d the extra 6F of hitstun come from? It doesn’t say anything about hit stop being modified like with Dhalsim’s Yoga Fire (which is listed as 8F, and also confuses me, but I’m more concerned about Yoga Fire right now). It would need hitstop of 18F for that to work.
IIRC SF2 Dhalsim’s Yoga Flame used to end instantly if the move was blocked, which should give it a frame advantage of +/- 0. That’s not the case in CvS2 or SFIV though. Or if it is, it recovers at some fixed point after block stun + hit stop. But it can’t be that, because the MP/HP Yoga Flames in SFIV have recovery times of 16/13, and their block advantage is -1/+2, which is consistent with the LP Yoga Flame having BA of -6. So… how exactly is the recovery of the move coming into play, considering that block stun doesn’t change (20F for all versions) and, presumeably, hit stop doesn’t change, since it mentions nothing about hit stop.
For this to work, LP/MP/HP Yoga Flames would have to have hitstops of 18F/21F/26F, which seems really arbitrary. Again, unless there’s some behaviour I’m missing (like the active duration aborting on block, but the block stun + hitstop values still don’t add up to the listed block advantage), or the values are wrong.
Beni’s Raijiken in CvS1/2 is just as bad. Except I have even less data to work with. CvS2 has frame advantage (+1 for all versions), but it only lists when the projectile starts and not how long the projectile remains active. CvS1 lists the move’s proper active length, but doesn’t list HA/BA. The data is different in both games so I can’t just map one to the other. That, and it seems that Raijiken behaves differently in CvS1 (looks like it can be powered up by holding down P). GAH.
Impact freeze is generally the same regardless of whether the attack hits or gets blocked, but i wouldn’t be surprised if it was different in some games. Either way it’s pretty easy to measure. Just get some 60fps footage of the game and count frame by frame. You can tell when impact freeze ends by when pushback begins. Even if your absolute measurement is one frame off, your relative measurements should be accurate.
Sorry, i don’t know whether the listed startup/active/recovery times for Yoga Flame are based on the move whiffing or hitting on the first frame. Obviously Dhalsim reacts to the flame connecting and skips the rest of the active period. Data on how many frames of recovery are added/subtracted doesn’t exist.
With a move like Yoga Flame, it’s fair to say that its block advantage is fixed. In other words, there’s no benefit to meaty Yoga Flame.
Why do you need this info? If you really care, just fire up an emulator and count frames with frame stepping. You have the entire SF2 and SFA series’ to work with.
I need this info for my SFP Project. I only recently became aware of the relatively fixed nature of hit/block stun for various games and need to tweak my Hit Advantage/Block Advantage values (HA/BA) accordingly, but I want to understand the exact nature relationship between active frames (AC), Hit Stun/Block Stun (HS/BS) and HA/BA. I have some 5000 or so moves to go over, so I want to make sure I understand things properly *now* before I go editing stuff.
I was just wondering if this info was available or known already and I just couldn’t find it. I know you’ve done a lot of digging with esoteric stuff like this, so I thought perhaps you knew. I guess I’ll have to go digging myself via emulator. The only problem is that it obviously differs from game to game, but at least I can use a couple games as reference. Thanks Maj.
Well, the reason this data doesn’t exist is largely because nobody cares. Impact freeze is transparent. All it does is make buffering into special moves easier. We were discussing examples where it would come in handy and we only came up with two or three situations above. Nobody really cares apart from charge characters, and even then it’s played by feel unless you’re doing advanced tool-assisted multi-charge combos.
Most games do have uniform levels of hit stun for light attacks, medium attacks, hard attacks, and maybe one or two extra levels for certain types of special moves. There are exceptions among normal moves, but not too many.
The relationship between those numbers is simple:
HS – (AC + RE – 1) = HA
BS – (AC + RE – 1) = BS
Like i said, impact freeze is transparent. Generally speaking, it doesn’t factor into this equation.
Yeah, those are the formulae I’ve been using. However, it doesn’t apply to projectiles. I thought projectiles used HA/BA = (AC+RE-1) + (Hitstop), but then I ran into the problem with stationary projectiles, which have their own behaviour that I can’t figure out. Also, double checking SFIV’s data, that formula doesn’t work there either, since they list HS/BS/Hitstop values, but nothing adds up right.
Then there’s the fact that certain attacks, like Blanka’s Horizontal Ball and Cammy’s Cannon Spike, have different (and unlisted) recovery times when blocked (and in Blanka’s case, on hit as well), resulting in BA that is independent of the BS inflicted. I guess stationary projectiles are also in this category.
Also, while I could be wrong, the CvS2 data book seems to calculate it using BS/HS-(AC+RE) = HA/BA. The only example I could find is Ryu/Evil Ryu’s F+MK, which is a ground move that hits like a jumping medium. Jumping medium attacks have HS/BS of 15/20. Ryu’s F+MK’s data is 19/4/15/-4/+1. -4/+1 only works if you use HS/BS – (AC+RE), and not HS/BS – (AC+RE+1) like in SFIV. More confusion for me, but I guess it doesn’t matter as long as I’m consistent.
And hey! I think you’ve covered far more esoteric stuff in games than this stuff! :p
Sorry, I meant to say projectiles used HA/BA = (HS/BS + Hitstop) – (AC+RE+1). Can’t edit posts.
Where do you see hitstop values in SF4 data?
Er, it’s listed for every projectile in the SFIV Master Guide? It says “Hit stop: 8F” for every single one. I don’t know why it’s not on the SF wiki, though it seems that the follow up guide for the console characters doesn’t say anything about hitstop for projectiles.
Ah okay, that’s cool. Anyway the formula does work for physical attacks. What doesn’t add up right?
Ryu’s close s.HP is 7 AC, 26 RE, 22 HS, -10 HA.
22 – (7+26-1) = 22 – 32 = -10
Chun Li’s LP Kikoken is listed as 14 ST, 44 RE, 44 TO, 22 HS, 0 HA. First off i have no idea why RE = TO. One of those numbers has to be garbage. Looking at a video clip, looks like the the RE number is the useless one. Let’s say true RE is 30 frames. How long is impact freeze?
HS + IF – RE = HA
22 + IF – 30 = 0
hit IF = 8 (roughly)
BS + IF – RE = BA
18 + IF – 30 = -4
block IF = 8 (roughly)
Holy crap Maj, you did it! That was the missing piece! The RE listed for all projectiles (except for Yoga Flame) is the total number of frames the move lasts, NOT the actual recovery! I have no idea why they would do such a thing. I merely subtract startup from the total to get the actual RE (actually that would be RE+1, since the startup is actually start up+1). Add that to hitstop (8F always it seems) to get the true hitstun and blockstun for projectiles. It all adds up now!
I knew the formula worked for most physical attacks, and I thought I had projectiles understood too until I checked the SFIV data and compared the numbers. I didn’t realise RE=TO for projectiles (WHY?), so that was throwing me off. So the formula works for *most* attacks. However, there are still a few exceptions:
– Dhalsim Yoga Fire
– Blanka Horizontal Ball
There a probably a couple others I’m forgetting. The main thing is that the formula doesn’t work on attacks that have a specific recovery that only activates on hit or block. Even though they have their own HS/BS, the recovery value listed is only if the move whiffs.
Shit, I did some more checking, and I figured out the formula for multi-hit projectiles as well. It’s HA/BA = (Hitstop x no. of hits) + (HS/BS) – RE. Works perfectly for Ryu’s Super/Ultra, Akuma’s Red FB, and all EX projectiles. Wow. Thanks a lot Maj. This helps a LOT.
They should probably make a note that RE = TO for projectiles on the SRK wiki. Cause they list RE as it’s listed in the Master Guide. I suspect that RE=TO for all the console characters (Seth, Rose, Dee Jay) as well – even though HS/BS isn’t listed for them, using RE=TO in calculations puts their HS/BS at the usual values (normally 22/18).
> I merely subtract startup from the total to get the actual RE (actually that would be RE+1, since the startup is actually start up+1). Add that to hitstop (8F always it seems) to get the true hitstun and blockstun for projectiles.
Dammit, I typed this wrong. I meant subtract ST from TO to get the true RE. Then add HS/BS to hitstop (8F), and subtract *that* from the true RE to get the HA/BA. Anyway, summary plz: Formula works! Except for those handful of moves with arbitrary recoveries on hit/block!
Keep in mind that fireballs are weird. We’re assuming that their listed 14 ST means that it hits on the 14th frame, which would be consistent with their conventions of requiring a +14 attack in order to link into it.
But they could also mean that the projectile is created on the 14th frame and might not hit until the 15th frame. That’s how it works in a lot of games. Projectiles typically don’t become active until the frame after they’re “born.” Then you have some weird stuff going on in oldschool SF2 games where they’d randomly hit one frame later than they were supposed to.
Anyway it’s better not to get into that ’cause it’s a mess yet it comes up so rarely that nobody notices. It’s safe to say our ST/RE/IF figures are accurate within +/- 1 frame. For your SFP project, you can pick your preferred interpretation, set your own IF, and stick to that uniformly. As long as you’re consistent, the rest doesn’t matter.
For Blanka ball, the theoretical formula would be the same as the general physical attack formula:
HS – (AC + RE – 1) = HA
The only difference is you ignore the “AC – 1” component because its active period ends immediately upon contact. That leaves you with:
HS – RE = HA
We have HS and HA so: 26 – RE = -15
That means its bounce recovery has to be 41 frames long. That’s consistent with the block situation as well:
BS – RE = BA
17 – RE = -24
RE = 41
Hah. You would be correct again. And you know what? 41F IS listed in the data for Blanka Ball. It’s right there in the text, except it’s listed as 36+5. Of course I can’t read the text, but now I’m pretty sure that’s the recovery time if the move is blocked, since it’s consistent with the HA/BA and the listed HS/BS. This also works for the Vertical Ball (same 36+5 in the text ). Honda’s Headbutt too, has recovery values listed in the text that line up with the data. So this is great.
It doesn’t work for a few other moves, at least as far as I can tell with the numbers listed. I can’t find a match for Honda’s Buttslam, Super or his Ultra, for example. Perhaps it’s a case, like with Blanka H/V.Balls, that I simply can’t read the text sufficiently. Regardless, in any case where the numbers don’t match up, it’s easy enough to extrapolate the altered RE using HS/BS – RE = HA/BA, and call it a day. I think I have enough to continue with now. Thanks Maj.
Well, Honda’s super is a little complicated because it has two variable recovery periods, so it’s hard to tell what the published frame data actually represents. His ultra is only one hit on block though, so that should be straightforward.
BS – RE = BA
20 – RE = -15
RE = -35
His SF4 data lists it at 22+13, which i’m guessing means 22 airborne plus 13 grounded, so that’s consistent.
Buttdrop should be nothing special because it doesn’t react to whether or not it connects. But the data is weird. They list 6 AC, 18 RE, and 20 BS with -2 BA for the LK versions. What’s weirder is they list 9 AC, 18 RE, and 20 BS with -2 BA for the MK version. Maybe they went ghetto and just had the tallest character they could find blocking the descending hit, instead of measuring from the first active frame.
Unless otherwise stated, ST/AC/RE frame data is always listed if the move whiffs. I can’t read what the text says for Honda’s super, though it does say something about “on hit or guard”.
And yeah, I couldn’t figure out what the deal was for buttslam. However, I *think* it says something about both hits being used on a blocking Ryu.
On this note, I did some checking, and it seems the “RE=TO” for projectiles is a “frame data chart” thing, just like when data is presented in chart form, ST is listed as the first frame that the move hits on, rather than the actual start up. I checked frame data for GGXX:AC, and it has TO values in the RE column.
I miss the old graphical method of listing frame data. It’s a lot less confusing. :|
Okay i know the discussion is pretty old but i was discussing “hit confirm” in vega forums recently, basically there was an argument about whether you could “hit confirm” after say SFIV vega cr.lp
the move is :
4 ST 2 AC 8 RE +1BA +4HA
then we have two conflicting theories
On one side we have several Vega vet swearing they can hit confirm that into an otherwise unsafe combo reliably.
On the other side we have Neuroscience past 30 years experiments that strongly suggest it is very unlikely to be humanly possible.
That is if we stick to frame data, but I can very much see “impact freeze” settling the matter, which would give impact freeze another use beside increasing “advance charge buffering”
the thing is vega cr.lp cannot be linked into a true blockstring, so in SFIV you can be SRKED between 2 cr.lp if the first one is blocked, frame data suggest you have less than 11 frames to react to the hit that is 166ms of frames,
http://img835.imageshack.us/img835/9…hitconfirm.jpg
while neuroscience says average response time for simpler and more rehersed task than street fighter (reaction to color, reaction to matching of a 2 letters, reaction to an anticipated flash) is in the 250 ms.
Simply put even not factoring input/display lag (that further handicap reaction time to a visual information) the numbers/claim don’t fit.
Now if we had say a pause of 7 frames of “impact freeze” before one character starts 8 frame of recovery and the other starts 9-12 frames of block-hit stun, then hit confirming even such a fast move before commiting into a link combo sequence becomes very doable even for a random dude providing one plays on a decent gaming platform.
basically the question is “(fireballs aside) Is impact freeze included in the RE / BS / HS frames figures or does it add up to them on both character side ?”
Well, you’re looking at 3 frames startup, hits on the 4th frame and causes 8 frames of impact freeze, then it stays cancelable for 1 or 2 more frames. That adds up to 12 or 13 frames total. The rest of recovery isn’t cancelable.
Is that enough? I don’t know. Wouldn’t it be easier just to ask the guy to film himself reacting to the Training Mode dummy set to random blocking?
Also if you add a jump attack before that jab, it becomes a lot easier to react. Maybe that’s what he meant.
If not, there’s still a chance those neuroscience experiments could be irrelevant. I mean are they set up where a color randomly flashes and you have to react? Because if so, that doesn’t take into account the fact that the Vega player is pushing the jab button.
In other words, the experiment would have to be set up so the subject gets to press the button that shows the color, then they react to the color itself. That’s much different from having the scientist press the button that shows the color.
Even then, Street Fighter players develop better than average reaction times. Not to mention, SF is a lot more absorbing than some boring color test. It’s easier to focus on SF.
Btw your image link is broken.
Well in fact i proposed the exact same experiment to those vets, as you say neuroscience experiment could be irrelevant in one way or the other and probably both, it doesn’t exactly match playing SFIV with a purpose for a reliable fit.
Sadly they didn’t show that much intrested in digging deeper into that topic, not even for the sake of discovering how exactly they could rely on what they apperently consider a basic tool in their gameplan. Only one of them actually hit training mode and random block dummy but just ended up saying he performed better than expected but did not really count the success/fail ratio.
If by anychance you know some top players willing to do the experiment I can always get in touch with them, As far as i am concerned, I m am not reliable enough yet on 1 frame links to do the experiment myself, I also am not sure I can be a reference as far as reaction time goes I am really not tournament level.
I wasn’t exactly speaking of cancelling the move btw, vega SFIV cr.lp cancels into Rolling crystal, Scarlet terror, and wall jump, in in that case if I understood correctly your message one would have 9-10 frames minus the lag to decide to cancel in the “2 in 1” or not, that does’nt feel very likely to be done but would still be worth an experiment.
The topic was about linking cr.mp after or not, cr.lp — cr.mp is an instresting combo opener for vega but these are link combos ( there is an exception).
So really the problem evolves from reacting on the 4rth frame of the move and before the last frame of recovery into a one frame linked cr.mp,
depending on whether you have :
8-9 frames of recovery alone minus the lag,
or
8 frames of impact freeze + 8-9 frames of recovery minus the lag
to decide and link the subsequent move, or stop there before doing something unsafe.
The jump attack thing really wasn’t part of the topic, there was a consensus a late jump in followed by a cr.lp is a true block string and can be used as a “hit confirm” sequence for the jump in hit.
Oh, if you want to link then it’s impact freeze plus remainder of active plus recovery, so 8+1+8=17 frames. Let’s say around 15 frames if you take out input lag.
I think that’s possible to react to if you’re focused, but the question is do you want to invest so much concentration on that? Personally i’d look for another option that doesn’t require such strict reactions. You lose track of the big picture when you focus on details.
If Vega really has no better options, then i guess it’s worth a try, but i’m a little confused – what’s the point of proving it with a neuroscience experiment? If you can hit confirm it, then use it. If you can’t, then don’t. Otherwise it’s theory fighter.
Yes focus and big picture, happens to me all the time I say to myself next time he does the cr.mk xx hado shit I am goona scarlet terror his face, and then…. get thrown.
Thanks for the info anyway, it ‘d be simple if it was only about me and was say like how tall you are, not gonna change.
But lets consider you discarded doing something because in your previous experiment it didn’t work-out. Then you meet someone and they say they can, You ‘ll want to try again and train until you can do that too. That is when neuroscience and therory crafting came into the picture.
15 frames is a realistic figure for an automated reaction time btw, now it is all about what you said, is it enough for SFIV, how reliably can every one guy do it and thus can he integrate that into his gameplan, is getting focused on making the best of what is otherwise merely a poke worth losing track of the rest.
As for neuroscience, it is rather funny to consider using neuroscience for getting better at a fighter, but i was considering the thing the other way round in fact, fighters are intresting as far as getting better at neuroscience goes.
Neuroscience and AIs are kind of my “thing”, I liked your article on “fighter’s AI” btw, both are intresting if you are a game designer, not much if you are a player. we could go into an intresting debate from that staring point but i ll save it for another day.
As for vega he has plenty other options, even commiting into a “pseudo blocksctring” after cr.lp is not such a big deal as long as you don’t do get too predictable.
Ah okay. In that case i’d be happy to discuss AI with you under that article. Or at least share whatever i’ve observed from a player’s perspective. I can’t say i have any technical background on it or anything.
Strange Focus hitstop properties – http://www.youtube.com/watch?v=RIN1eMOXIE4
So I was messing around with some option selects, trying to make it so that if my low jab was focused or whiffed, tornado throw would come out. Ideally so that EX-karakusa wouldn’t work against this due to tornado throw being unthrowable.
It didn’t work as I intended, and tornado throw comes out too slowly. But as I was messing around I saw that if –
–HERE’S THE STRANGE PART–
the low jab was absorbed by focus and the focus was immediately backdashed, nothing came out. If the focus was held, tornado throw was cancelled from the low jab.
I don’t know anyone else who could break this down and my thread on SRK didn’t get any views or replies lol.
It’s similar to Third Strike option selects (Ken for instance) meaty fierce late buffer fierce SRK. If the fierce is parried, it cancels into SRK due to parry hitstop lasting so long.
I was curious about focus hitstop ever since original SF4 but no one knew anything.
Please take a look and help me out if you can!
It may well be that the game treats the immediate backdash as an ‘armor cancel’ which will have different properties than the normal dash cancel.
If that’s the case, then Makoto should be able to delay the backdash to cause Abel’s command grab to come out right?
Also if you have the original video, you can go frame by frame to count the armor impact freeze in each of those two cases. We can’t really do that once u2b owns up the frame rate.
Yeah I have the original video, can I electronically mail it to you?
For the record I was never able to replicate it.
Well, i don’t really have a great way of counting either. Basically you just have to estimate and it might not even tell you anything. Just watch it frame by frame starting from the first contact frame, and count how long it takes for Makoto to start animating. If she moves earlier in the backdash clip than the other one, that might give you an answer.
If you don’t know how to watch frame by frame in Windows Media Player, it’s under View -> Enhancements -> Play Speed Settings.
The problem is I don’t have a PC, just an iPhone.
Wow, crazy. Alright i guess just e-mail it to me and i’ll take a look.
Sent the video. Can you guys think of any practical application for this? Since this doesn’t work against ex-karakusa I’m kind of at a loss.
I converted it into an image sequence using Super Encoder and in the first case there are 5 blue Makoto frames before Abel’s left elbow moves, whereas the second case shows 6 blue Makoto frames before Abel’s left elbow moves.
Also in the first case, there are only 4 blue Makoto frames before her backdash begins, which does seem kind of low. So maybe these numbers indicate that dashing out of Focus Attack startup does affect impact freeze.
On the other hand, this video is only 30fps and the game runs at 60fps, so it’s possible that important frames simply show up in the video. (That’s what i meant by “it might not tell you anything” earlier.)
In any case, even if this is true, you’d probably need frame-perfect timing to create any kind of option select out of it. It’s probably not realistic. But then again, i’m not much of an SF4/SSF4 option select expert.