I would actually like to see proof that the reversal window really is 10 frames. Even so-called “knowledgeable” people throw around numbers without substantiating them.
Actually, maybe I’ll test it myself later.
Yeah, good point.
My understanding is that it came from some official document that capcom released in japan and was then translated, I guess it was supposed to be the original source for the EventHubs frame data too. The people saying it was the truth are people who I basically trust to get that kind of stuff right.
The thing is…there is absolutely no denying that the reversal window in SF4 is WAYYYY bigger than in previous games like 3s and ST.
Just try it with something simple like Guile’s flash kick countering a meaty…shit is DEAD simple in SF4 and its MUCH harder in ST.
But I’m with you…I guess I would like a solid answer one way or another.
piano inputs help time the reversals
which makes them easier in some if not most cases.
EDIT: After more testing, it seems that the following values aren’t really reliable because input latency and software issues.
The results of a bit of superficial, unscientific testing (the values have been approximated):
With Akuma, :dp: (wait x number of milliseconds) :p:
I timed this on the PC and the max number of milliseconds the games allows for before “forgetting” the SRK motion is about 170, which is roughly 10 frames (noodalls once measured it to be 11 frames. YMMV).
Let’s call those 10 frames the leniency window.
The reversal window is generally understood to be the max number of frames after the final button press that will still result in a special move on the first active frame. This is not the same mechanic as the leniency window.
If this is correct, then for a reversal SRK (given a 10 frame reversal window), once you’re in block stun you can input the SRK motion, wait 10 frames, press :p:, wait 10 frames, and still get a reversal. This doesn’t make sense IMO, (at least not with SRK’s), considering that most moves cause less than 20 frames of blockstun. Players don’t input the move on the first frame of block stun, so even if it were true that the reversal window is 10 frames, it’s unlikely that you can obtain that value using blockstun as a method of measurement.
However, consider jumping:
A neutral jump with Akuma takes approximately 38 frames (633ms), meaning that a standing normal will only execute if the command is inputted 38 frames after the start of the jump. Special moves however can be buffered in the air, so given what we know about the SRK and the leniency window, you should be able to buffer it at approximately (38 - 10 = 28) frames after the start of the jump, press :p:, and still get an SRK (obviously you can’t press :p: immediately after the motion because you’ll get an air fireball/punch, depending on your inputs). Superficial testing howevers shows that you can buffer the SRK after as few as 24 frames. It appears that the game gives you a four extra frames for the SRK.
I didn’t test everything thoroughly, so everything here should be taken with a grain of salt. However, it seems to me that the 10 frames reversal window that everyone refers to is actually the leniency window for certain special moves, and the actual reversal window is only 4 frames (or at least less than 10 frames) Also note that generally the game doesn’t allow for input buffering during non-cancellable moves, so it’s possible that you may arrive at different values if you perform an aerial attack before attempting to buffer.
edit: I’m going to try to setup a test to calculate the reversal window during blockstun/recovery later. This should be able to determine if there are differences. It’s going to take me a while though…
Results of wakeup test:
http://forums.shoryuken.com/showpost.php?p=7436563&postcount=46
thanks for the hard work. im looking forward to seeing the results of your blockstun / reversal testing.
I started testing the reversal window on knockdowns (which is actually quite hard to do with software).
(Ryu vs Ryu in training mode)
From the start of the test, the earliest executable lp can occur after 4741ms. Anything before this won’t execute. So if the reversal window is 10 frames, you should be able to get a reversal SRK if you input the command at 4741 - 167 = 4574ms.
Testing however indicates consistent reversals require 8 frames (somewhere between 4600 and 4607ms on my system), which is less than 10, but double the amount I suspected. I didn’t take into account leniency in this test though. Just a straight SRK motion followed immediately by P.
Note that the testing procedure isn’t 100% reliable, but I doubt these numbers are far off.
For those who are interested, here’s how to replicate the test using AutoIT (I didn’t use AutoHotKey because of timing issues):
option
Spoiler
Create a file called “test.au3” containing this code:
HotKeySet("+q", "Test") ; binds the test to shift+d
Func Test()
WinActivate ( "STREET FIGHTER IV")
Sleep(20)
; This next block resets stage
Press("ESC")
Sleep(400)
Send("{DOWN down}")
Sleep(1)
Send("{DOWN up}")
Sleep(100)
Press("LP")
Press("LP")
; You'll need to adjust the value in the next line depending on the
; timing of your recording
Sleep(4741)
Press("LP")
; Once you've determined the earlier timing for a normal move,
; remove the previous 2 lines of code and uncomment the next
; few lines
;Sleep(4607)
;SRK_Right()
;Sleep(1)
;Press("LP")
EndFunc
While(1)
Sleep(8000)
WEnd
Func Press($move)
$move = StringUpper($move)
Switch $move
Case "L"
$move = "LEFT"
Case "R"
$move = "RIGHT"
Case "U"
$move = "UP"
Case "D"
$move = "DOWN"
Case "LP"
$move = "A"
Case "MP"
$move = "S"
Case "HP"
$move = "D"
Case "PPP"
$move = "F"
Case "LK"
$move = "Z"
Case "MK"
$move = "X"
Case "HK"
$move = "C"
Case "KKK"
$move = "V"
EndSwitch
Local $btn_dwn = "{" & $move & " down}"
Local $btn_up = "{" & $move & " up}"
Send($btn_dwn)
Sleep(1)
Send($btn_up)
EndFunc
Func SRK_Right()
Send("{RIGHT down}")
Sleep(1)
Send("{RIGHT up}")
Sleep(1)
Send("{DOWN down}")
Sleep(1)
Send("{RIGHT down}")
Sleep(1)
Send("{DOWN up}")
Send("{RIGHT up}")
EndFunc
This script assumes the default keyboard bindings. While the game is running, double-click the test.au3 file to launch it. You should see the AutoIT icon in the notification area.
Start SFIV in windowed mode and go to to training mode. Select Ryu vs Ryu, turn on Input Data, select “Record”, and then press Escape again and select “Restart”. This resets the stage and give a chance to input the knockdown move without delay. Record a crouching roundhouse or any knockdown move against yourself. Now press escape and select “Playback” and then “Restart” to playback the command.
Press shift-d to run the test. Because of the nature of the software, it sometimes drops inputs so you may have to run it several times. Also, the SRK command is fairly ghetto so you may bet better results if you tweak it. You’ll need to play with the numbers a bit, but if you’ve gotten this far without problems then I’m sure you’ll figure it out easily.
I encourage people to verify the results, and maybe even produce better/more reliable tests.
Generally fighting game companies don’t release frame data. Most of the time it comes from Arcadia doing the leg work themselves. I don’t know for SF4, but I’d assume this comes from the Arcadia master’s guide, and the more recent softbank release. Not to criticise the translators either, Arcadia often makes mistakes. Also, sometimes they will release corrections but then it’s up to the person who translated the original document to be aware of these as well.
Anyway, having just done some testing in another thread I found this one.
My test there was to do Ryu d+HK on F1, and see when up would be read as an input first. The result was on F88 for most characters. I modified my sketch to test what’s of interest here.
2P programmable stick. Frame accurate.
My results were
P2 Ryu does d+HK on F1.
P1 Ryu can press HP on F88 or later to give HP. This is the same time as pressing up for the jump. F87 gives nothing.
P2 Ryu does d+HK on F1.
P1 Ryu can do fireball as early as d on whenever, df on F62 and f on F74 then P on F84 and still have it display reversal. (Also, if you do the FB motion a little faster, P on F84-88 will work as reversals.)
The 84-88 bit is the important one I think, as these are the frames when the input finishes. The rest is just a reflection of the leniency of special move inputs.
Tested with same setup.
Ryu 2P d+HK on F1.
Ryu 1P can input normals from F88. bd can be input as early as b on F72 then b again on F81.
If this is correct, then it seems my original assumption of a 4 frame reversal window was right, and it’s exactly the same as the window during jumping. Of course, my results during that test could also be suspect because of software timing issues but I’ll run it again to confirm.
edit: I can’t confirm a 4 frame reversal window on PC. I still get a reversal DP even if :p: is pushed as early as 8 frames before wakeup.
What value do you get for the window during neutral jumping?
Nice, noodalls. So reversal backdash is actually easier than reversal special/super by 3 frames.
And it looks like normals get a 1-frame reversal window, so specials get a 5-frame window and backdashes get 8 frames.
What stick did you use, I’d like one!
This might break your heart, but, it’s a Mayflash. Check out the programmable stick thread to see it.
http://forums.shoryuken.com/showthread.php?t=109169&page=6
When do you release the P? I wonder if you might not be negative edging it? When I said that I press the P for the FB motion on F84, I neglected to mention that I release it on F85. Retesting now, FB motion with the P pressed on F85 (and released on F109 just to put it way out of the way) it will still release.
I suspect then that F85-88 all are reversal frames, whether by pushing or by negative edge.
[For the record, you can complete the FB motion including pressing P on F74, then negative edge on F85. 11F negative edge? Maybe]
fp SRK beats out most moves =)
random start of thread response for people that think you shouldn’t try and SRK a reversal SRK:
doing ryu’s FP or MP SRK will beat ryu’s reversal SRK if you finish the input after the 1st but before the 5th frame or something like that. i don’t know the exact invuln frames on it, but basically you start your punch after theirs has started but before the first hitting frame. timed properly, it beats a lot of wakeup options (i think it only loses to EX srk and blocking vs shotos) and makes you look like you’re amazing.
for the reversal leniency, you have ten frames before the game forgets your f, d, df, right? so you can just do your srk motion within 14 or 18 frames of waking up and hammer on a punch button?. sorry if i restated what someone else just said, i need it in simpler terms like most of the rest of us.
1ms after pushing it. The software I use works in ms, not frames. I doubt timing issues would cause a 67ms difference, especially considering that the results are reproducible, but there could be other factors at play.
Basically, I’m doing the same test that you are doing, except I’m counting from the start of the training session.
After Ryu gets knocked down, the earliest an LP can execute is 4380ms. The earliest a reversal can be inputted is between 133.36ms (8 frames) and 140ms before wakeup.
Could input latency be the cause of the 4 frame discrepancy? I read today that SFIV on the 360 has 67 frames of input latency, and it’s presumably the same on the PS3. It’s possible that the software I’m using only polls at 30fps, which would explain the 4 frame difference (133ms effective latency). The problem is though I’ve managed to verify various other things as well (eg. Charge time for sonic booms etc) based on existing literature, so it’s a bit strange that it would work there and not here.
Lot’s of possible different explanations I suppose. Your joystick is probably much more accurate than the readings I’m getting. I think someone else needs to do a test using a different method on the PC version.
You can mash the motion and the button and it will come out.
I’m interested in anything you guys can come up with about reversal focusing.
It certainly FEELS like there is no reserval granted. I have also seen people wake up and get hit by projectiles/meaties that they could afford to block. I assume the are trying to focus.
Easy-mode reversal focus would be too risky, anyway. At least with invincible backdashing the momentum is set to nuetral, and not one character getting molested at EVERY wakeup situation.
The first page of this thread was pretty shitty. Glad I kept reading though.
Turns out the steam sale was extended by a day, so I ended up getting SF4PC. Now, if I can run it at 60fps with minimum settings on my terrible PC, we can have some cross platform testing.
From what I can tell, it’s the same F85-88. No reversal message comes up, but the timing is the same for these 4 frames. You can’t negative edge (e.g. press MP+MK prior to F85 and release it during 85-88) and get a focus.
You can press MP+MK on F85, release on F86 and still get a focus. Or, you can hold from 85-88 until 106-138 to get a level 2, and anything 139 on longer gives level 3.
If you start holding MP+MK on F89 LV3 focus won’t come out until F140.
Nice read.
So you have 4 frames to do a focus. Landing frame jump to focus looks easier now
But i will happy to know how option select his read by the engine of the game, how proximity blocking is handle and if it’s possible to cancel a move during this window.
For example, when you safe jump on Sagat, you can buffer a dragon.
If Sagat do a tiger, ryu block in time.
If sagat do backdash, the dragon come
First i was thinking it’s work because the tiger uppercut start just before ryu start
F1 Ryu is in the air - Sagat start his tiger uppercut
F2 Ryu land (he has 4 landing frames)
F3 Ryu landing frame 2
F4 ryu landing frame 3 (he can block)
F5 ryu landing frame (he can cancel with a normal or special) - Sagat hit with his tiger uppercut (5 frames to hit)
So i was thinking, it’s work because the tiger uppercut hit ryu in the perfect time just before he do a move but i was “wrong”
I tryed versus rufus (with chun li)
Same safe jump but with option select sweep like that if rufus backdash, she can hit him.
F1 chun li is in the air - rufus start is ex messiah kick
F2 chun li land (she has 4 landing frames)
F3 chun li landing frame 2
F4 chun li landing frame 3
F5 chun li landing frame 4 ( she normally start her sweep at this moment)
F6 chun li ??
F7 chun li ??
F8 chun li ??
F9 chun li ??
F10 chun li ??
F11 chun li ?? - Rufus hit with his ex messiah kick and chun li block.
Normally she would doing a sweep since F5 but it’s not the case. Perhaps i’m wrong but this same set tup in trainging mode allow me to block tiger uppercut of sagat (with chun li).
It’s look like the game cancel the sweep input by block since the Frame 5 even if rufus don’t hit at this moment. That’s strange