PPad 100% Technique for Determining Frameskip Patterns

[Magnetro]

New Technique found by joo to guarantee that your combo/situation/reset/whatever works 100% of the time. This should apply to every game. I haven't tested it but there is no reason it shouldn't work.

The main concept is that you program and do something that doesn't work 100% of the time twice and from there, anything after will work 100%. The technical part (not 100% sure on the numbers): 2 instances of whatever that isn't 100% happening in a row = 100% guaranteed anything as long as you stay in the same program sequence. I'm sure it has to do with the game's frame skipping, the problem is I don't know the exact numbers.

Example of the situation in marvel:

Venom: n_1/2/8_6/4pp_9/hk_11/2/8_6/4pp_9/hk11/

That means, neutral for one frame/down(1 frame)/up for 6 frames/left + both punches for 9 frames/hk for 11 frames/ repeat.

If both tri jumps work 100% of the time, anything from there as long as you start it at the right frame will work 100% of the time.

I'll try to answer any questions, but I really don't know much about it so far, I did however make a table to show all mvc2 characters' 100% sequences, http://zachd.com/magnetro/expvid/100psequences.html

Random note* after you do the sequence and you want to add 2 seconds worth of nothing in frames for recording purposes you have figure out the right frame. Once you do anything backwards or forward in 4 frame increments will work. So if I land on say frame 230 and anything works after 230 frames of nothing, I can add or subtract 4 frames and it will still work. Again, I don't know all the exact numbers or why, i just know it has to do with frame skipping.

Anyway, this is an enormous discovery and I felt obligated to share it on here since a lot of you have programmable controllers.


Mod edit: Changed thread title from "Attention People with PPADS" to something more descriptive.

[laugh]

Good stuff.

I've actually been thinking about the same thing myself recently. I know I'm like Maj when it comes to making combos on the default speed, which we all know to skip frames. Now, the thing is we also know that cvs2 at speed3 skips every 5 frames (I don't know if it's the 6th frame that gets skipped or the 5th, but that's easy to find out), so it's easy to tell if a combo is doable at speed3 by counting the number of 1-frame inputs and what Nth frame they are placed in.

So let's say the game skips every 5th frame, so we're always gonna be missing the 5th, 10th, 15th, 20th, 25th, 30th....so forth frames, so as long as none of the 1-frame inputs are placed at those frames, you're good.

It doesn't really have to be the frames that end in 5&0, as long as none of the 1-frame inputs are placed in 1 (out of 5 possible pairs) pair of frames that are 5 frames apart (1&6, 2&7, 3&8, 4&9, and 5&0) it's good.

I hope I made sense. I'm thinking joo is thinking the same, although I didn't really understand what you posted cuz I skimmed through.

Obviously, the example works best with games that skip every 5th frame to make the calculations easier, but it's still doable with games that skip more/less often.

[Magnetro]

i have nooo idea about how marvel works in the skipping frame department. But that makes a lot sense since once you get the first part, you have to START the combo at the right frame...I'm really hoping people here experiment a lot with this. I've already found all the sequences for the characters in Marvel so I don't really plan on doing much more with it unless something big happens.

[Maj]

It's awesome that he took that idea and formalized it, but it's kind of common sense, isn't it? I've been doing similar things all along - putting a few (unnecessarily) difficult one-frame timings at the beginning of long sequences so that if they work, then the rest of the sequence will probably work too. If they don't work, then i don't waste my time waiting for the rest of the sequence.

That's basically the core idea here. You use two desynched one-frame instances to make it so that your program only works from one point in the frameskip cycle. That way, when you program the rest of your sequence, you're taking the frameskip cycle into account.

It's a little tricky to find those two instances, because you have to make sure that they don't fall on consecutive cycle frames. I'm pretty sure that default speed CvS2 skips every 5th frame. So each full cycle takes 4 program pad increments, cuz the 5th one is invisible. If you do a one-frame command connecting the 1st frame (A) and the 2nd frame (B), plus you do a one-frame command connecting the 3rd frame (C) and the 4th frame (D), then the only way the pair will work is if you start your sequence on the 1st frame of the cycle. If you start it on the 2nd frame, the D command will fall on a skipped frame. If you start it on a the 3rd frame, then the C command will fall on a skpped frame. If you start it on the 4th frame, then the B command will fall on a skipped frame. And you can't start on the 5th frame cuz it's inaccessible.

However if you do a one-frame command connecting the 1st frame (A) and the 2nd frame (B), plus you do a one-frame command connecting the 2rd frame (C) and the 3th frame (D), then the pair will work if you start your sequence on the 1st frame of the cycle as well as the 2nd frame of the cycle. That's why you have to have your two test cases completely out of sync in order to ensure that they only work from one point in the cycle.

If you build your combo on top of a valid test sequence, everything you test will fall on the same portions of the cycle every time. Then once you have a working combo, you can add or subtract 4-frame slices from your waiting period because that's a full cycle. Everything still falls on the same part of the cycle.

Anyway this is cool and all, but it's not a magic solution for all situations. Sometimes you make a video where you want to start it as soon as the round begins, which doesn't give you time to run a test. Or a lot of times you're doing stuff that requires two program pads, so you have to push the buttons on the same frame which is always annoying. Or your combo is affected by randomness which is unrelated to frameskip, like superfreeze effects in CvS2.

Plus, this won't help in older games like HF or ST where there are three or four separate sources of randomness (damage, dizzy, projectile impact slowdown, charge time, etc.) and not even the frame skipping follows a regular pattern.

[Magnetro]

/

Yeah. Anyway, has anyone tested this for other games? This thing is beyond useful for crazy elaborate situations and combos that require the other guy to do something that doesn't work 100% of the time as a set-up. This information has made my week

[BB Hood]

Wow, this makes a lot of sense. I didnt even know that certain games skip frames every Nth frame.

[Magnetro]

I updated the page a little, some of the data was changed. http://zachd.com/magnetro/expvid/100psequences.html

pretty much the same thing with a few tweaks and info

[Magnetro]

Turns out tigerkneeing (d,df,f,uf) a move once is all you need to get the 100% technique working. Instead of 2 tri jumps you only need to do one tigerknee move and if the person does his/her move on teh floor you've landed on the right one

[Magnetro]

Like maj said, this 100% thing doesn't fix everything, there is still no way to control randomness from certain supers in the game - as shown here

http://zachd.com/magnetro/ryu_mag_rando ... t_x264.avi Most of the time the hypergrav will whiff, sometimes it will catch the person, and sometimes it will catch the person so low to the ground that it can't animate him towards magneto

[Maj]

Regarding the fake tigerknee method, basically you're talking about doing D,DF,F,UF,P/K and if the character stays on the ground for their special move, then you've found a consecutive sequence?

Let's think this through. When you push UF, most characters will be on the ground for 2 frames then airborne on the 3rd frame. Therefore, you want to push UF for 1 frame, wait 2 frames, then press a button. Completing the special move input on what would otherwise be the 1st airborne frame actually overrides the jump, because Capcom games tend to resolve special moves before movement.

Assume the game outputs four consecutive frames, then skips a frame. That seems to be what we're dealing with. Now if the skipped frame lands before the UF, then the next skipped frame won't appear until after the button press, so you'll get your grounded special move. If the skipped frame lands in any of the three positions inbetween your UF input and your button press, then the character will superjump forward. Thus, the only way for this script to work is if a frame is skipped immediately before the UF input. Here's a ghetto diagram showing all of the possibilities:

Code: Select all

D -- DF - F -- UF - n -- n -- P/K

fS - fD - fD - fD - fD - fS - fD
fD - fS - fD - fD - fD - fD - fS
fD - fD - fS - fD - fD - fD - fD   <-- success
fD - fD - fD - fS - fD - fD - fD
fD - fD - fD - fD - fS - fD - fD

fS = frame skipped / fD = frame displayed

Here's the test sequence for 2-frame jump characters:

Code: Select all

001) 005 - null
002) 001 - D
003) 001 - DF
004) 001 - F
005) 001 - UF
006) 002 - null
007) 001 - P or K
008) 255 - null
009) 001 - YOUR HEART'S DESIRE!!

Of course, characters with slower jumps would require a longer wait period on the 6th action. I'm assuming that Zangief and Mech Zangief get off the ground slower than Gambit.

[Magnetro]

Wow, great post. Yeah, Zangief is slower than everyone else. Again, great post, I might use that ghetto code table to explain it to joo (who was wondering about the 1 move thing, too). Great explanation! Also, I'm trying to figure out 1 for Guile and Nash/Charlie cause they don't have QCF moves that aren't snapbacks or supers. i use 2 flash kicks to find the right frame.

Edit* hah, i found it! all it is is: 4_30/12369_3/6lp
http://zachd.com/magnetro/guile100pt_x264.avi

again, really really great post.

[Maj]

Actually i think my first attempt was a little bit flawed. Here's a more accurate version:

Code: Select all

D ------ DF ----- F ------ UF ----- n ------ n ------ P/K

fD -fS-- fD ----- fD ----- fD ----- fD -fS-- fD ----- fD
fD ----- fD -fS-- fD ----- fD ----- fD ----- fD -fS-- fD
fD ----- fD ----- fD -fS-- fD ----- fD ----- fD ----- fD --fS   <-- success
fD ----- fD ----- fD ----- fD -fS-- fD ----- fD ----- fD

fS = frame skipped / fD = frame displayed

Because the whole point is that inputs can't be applied to a skipped frame.

[Magnetro]

So i think i found a way to manipulate the frame skipping (in marvel, duh). i was messing around with sentinel's unblockable using the 100% thing and it wouldn't work 100% of the time so i paused the game for a certain amount then upaused it and voila, it worked 100% of the time. seeing as how this is sh/forums i dont need to say much else. (hi maj, this means you, go try it)

[Ibostyle11]

I think he means we but yeah... anyway, this is me posting so someone will stop referring to this forum as "the forum that I've never posted in."
On another note, I am of the opinion that the frame skip issue is due to the fact that the pad engine for a match continues to go even when the game is paused. As long as you have an idea of the cycles of a game you can easily set it back on track. It's probably already known that a new thought for me...

I'd like to experiment with the details of the unblockable and getting out of it since I'm thinking there are ways to manipulate this fact.

[Rufus]

In Street Fighter HDR, frame skipping is regular - that suggests it may be regular in ST as well, but be aware that the period is 32 displayed frames - turbo 3 just happens to have period that's a divisor of that.

You don't need to use special moves to do a turbo skip alignment test. For a simple frame skipping pattern, you should also able to align using two normal moves, where the first move has a 'in game' duration equal to 1 mod the cycle time. So, for HDR, turbo 3, you'd want a move with a duration of 1,6,11,16, or so on frames since the skip pattern has a period of 5 in game frames. Then your test pattern can be:

Move A appropriate Move B (for 1 frame)
If things are set up correctly, Move B should only show up 25% (if it's on the first frame after a skip).