Input Delay in Brawl

[#HBC | Joker]

Based on what we've just learned, I doubt it would affect it very much, if at all ever. Every input you did for the DACUS would have the same amount of delay to it, so it would sync up just fine.

 

[dettadeus]

Does anyone know if this negatively impacts DACUS? Especially the tighter ones like for Sheik. It would certainly help explain why there's actually nobody in the world who can do Sheik's DACUS at 100% success rate.

As far as my testing has shown, within that 4-frame window it'd be almost impossible for the inputs to have different amounts of lag.
It might explain why you might miss a GR > DACUS every once in a while, but probably not missing the DACUS itself.

It especially wouldn't affect BDACUS because everything is buffered anyways.

goddangit ninja'd with my own answer :T

 

[Zankoku]

So the delay is constant among a large enough period of time to contain quick sequences of inputs like DACUS? Good to hear I guess.

And yeah I suppose I'm referring to DACUS post-grab-release.

 

[dettadeus]

Well I lost my own bet... by one frame.
15/16 spotdodges caught with the same frame setup. The one time I missed by one frame (14th repetition), I caught him out of his buffered spot dodge at the end without even counting the frames.

Here's the frame setup I used exactly.

Frame | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | 14 | 15 | 16 | 17 | 18 | 19 | 20 | 21 | 22 | 23 | 24 | 25 | 26 | 27 | 28 | 29
Dedede | 1st SD Frame | 2nd (Invinc. start) | 3rd | 4th | etc | | | | | | | | | | | | | | Buffer SD | | Invinc. end | | | | | | | 1st SD Frame |
MK | Crouch | | | | | | | | | | | | | | | | | | | | | | Input Dtilt* | | | | | Hit D3's Foot

*This is the first frame of the input; in frame-advance the input would be held while frozen on frame 22 and about to advance to 23.
**This entire table assumes Dolphin's default 2-frame delay, which is why there's 2 extra frames between MK's Dtilt input and it hitting.

I input MK's Dtilt on the same frame of D3's spotdodge every single time and hit him with 94% accuracy.

Also I tested a couple of times, watching for the input delay, and got the same delay for both D3 and MK twice in a row.


Hypothesis:

When an input is made, the "random" amount of lag becomes set for all subsequent inputs made for the duration of the first input's animation. This way, even though the amount of lag at any given moment may seem random, someone reacting to an animation before it ends will experience the same lag that the animation first experienced.
(Ignoring the 2 frame delay within Dolphin)
Example: D3 inputs a spot-dodge and it comes out after 1 frame of lag. The number of frames from the input to the end of the animation is 28 (input frame > no input frame 1 animation starts > 27 frame animation). For that entire 28 frame window, all inputs made experience the same 1 frame of lag. MK starts his Dtilt on frame 24 (input frame > no input frame 1, animation starts > no input frame 2 > no input frame 3 > no input frame 4, dtilt hitbox) and it connects on frame 28; D3 is hit.

This is probably incorrect and would need more testing but the current data upholds this hypothesis fairly well.

 

[Biz_R_0]

So it is or isn't the same for both players?

 

[dettadeus]

So it is or isn't the same for both players?

Read the edit I just made.

At any given moment (or frame) it's likely that both players are experiencing the same lag, although the lag itself isn't always a constant value.

Obviously this disadvantages the person reacting, but presumably they get the same advantage against their opponent in an opposite situation.

In a way this would almost encourage more aggressive/offensive play because chances are the person defending (who has to react to your move) will have to deal with at least one frame of lag while reacting to your action, which potentially puts them at a massive disadvantage, seeing as the difference between Powershielding and getting hit is just one frame.

 

[Biz_R_0]

Someone tell Reflex.

 

[Vkrm]

It's still unfair. Leaning the game towards aggressive play is only good for aesthetics. It does not change that in some case the lag will negatively impact a player independent of his or the opponents skill.

 

[dettadeus]

It's still unfair. Leaning the game towards aggressive play is only good for aesthetics. It does not change that in some case the lag will negatively impact a player independent of his or the opponents skill.

1. will you please get out of the brawl boards omf

2. Reflex was complaining about how this would make the game even more overly-defensive than it already is, and I suggested a way in that it actually makes aggressive play harder to punish and therefore more favorable. I never said it wasn't unfair. I just said it might not have the effect Reflex suggested it would.

 

[Zankoku]

It makes reactionary defensive actions in shielding an incoming attack on reaction, and using shield offensively by advancing with it into an expected defensive attack equally unreliable, by a margin of two frames.

 

[Vkrm]

Even if it did help aggressive play, it would doing so in the worst way possible. Need them codes from the pmbr.

 

[Biz_R_0]

Why not just wait until they release 2.5? You won't be playing regular Brawl ever anyway.

 

[dettadeus]

Even if it did help aggressive play, it would doing so in the worst way possible. Need them codes from the pmbr.

Again, I never said in a good way or a bad way. I simply said that it could encourage aggressive over defensive play.

Also, we've gone 4 years without needing a fix for it, and the likelihood of every Wii in a tournament being hacked and running a no-frame-delay code (and in that case, also a no-tripping code) is near to nothing. I'm not saying it's not worth fixing, but I am saying we can live without it being fixed.

Try running an entire Melee tournament using only one version of Melee (in the US). Not going to be easy.

 

[Ishiey]

Question, dettadeus. This lag consistency that occurs with normals, does it also occur with specials and movement inputs? I'm mostly concerned with the movement inputs, just for the sake of seeing if the analog stick is affected in the same way that button presses are.

 

[Zankoku]

My previous statement was not adequately complete. I suppose I should say that any plays involving responding to the opponent in a frame-tight situation by reaction or timing, rather than a pre-determined sequence of inputs, will be adversely affected by the random delay. Perhaps the reason this doesn't seem to have as great of an effect as it would in, say, Street Fighter, is because a large part of our current generation of top players developed on Wifi, where a definitive delay of at least 10 frames would heavily discourage relying on such situations in the first place.

 

[bubbaking]

I feel this thread was dumb, tbh. I could have summed up the whole debate to, "Well, now that we know our game has lag, will you continue to keep playing or not?" We can't really do much other than seethe at Sakurai for screwing up, which doesn't help.

Edit: A large part of our current top players developed on Wifi? O.o

 

[Zankoku]

Off the top of my head, Nairo, Ally, ZeRo, Salem, Kain, and many others enjoyed much success on AiB's ladder before becoming superstars offline.

 

[teluoborg]

This would be like if, when you trip, your opponent trips as well no matter what they're doing, even while in the air.

inb4 core mechanic of Smash 4.

Thanks for all the details dettadeus, makes this whole input lag story a lot less frightening.

 

[shanus]

it makes sense that there is short-term consistency in lag. I could imagine it's process related in the game and the processing time could lead to short periods of consistency. The question is why it was even there (and so easily removed with a few lines of ASM) in the first place.

 

[Supermodel From Paris]

These days I'm believing more and more that they accidentally pressed some beta revision to DVDs.

 

[Magus420]

It doesn't actually affect menus btw. The (usually) delayed set of inputs are only used in gameplay to control your character.

Also, if anyone is familiar with using the frame advance code before I updated it to disable Z button input while the game is stopped, this variable input delay is the cause of the Z button sometimes inputting what Z is set to (like a grab) while also advancing the frame. During the times it's advancing + inputting Z action the input delay is at 0 frames. Any other time the input delay is 1 or greater. The code looks at the most recent inputs in memory to check the start/Z/X/d-pad up buttons that it uses, and to advance a frame it unfreezes gameplay for 1 frame when you press Z. While there is input lag, however, the code will have stopped the gameplay again by the time the game would try to use the Z input for stuff so it does nothing.

In training if you switch to 2/3 speed then change back to 1x speed (a trick to temporarily get rid of the above quirk in the old FA code), it resets the input delay timing to be 2 frames.

 

[bubbaking]

Wait a sec, why are people saying that the lag affects everything in and out of versus matches and Magus is saying otherwise?

 

[dettadeus]

Because they believe most people have a sense of humor.

 

[AmKhokar]

IMO, compile your data over a large sample size and calculate percentages of 0-1-2 frame lag if you're willing to spend that time ^_^ <3

 

[dettadeus]

IMO, compile your data over a large sample size and calculate percentages of 0-1-2 frame lag if you're willing to spend that time ^_^ <3

I might do that at some point, but right now I have a bunch of other stuff higher on my list.

What I would like to do at some point is possibly recreate/emulate a match between two players and check if the input lag changes ever if there is always an input being made or an animation being played.

Might also be interesting to test with four players but I don't have enough controllers to make four sets of inputs every frame.

 

[Vkrm]

Wonder what m2k thinks about all this....

 

[SpiderMad]

Off the top of my head, Nairo, Ally, ZeRo, Salem, Kain, and many others enjoyed much success on AiB's ladder before becoming superstars offline.

Was ADHD one too?

 

[RaptorTEC]

So basically lets say someone makes an attack and it comes out on 1 frame of lag then the second player inputs the shield button. You're saying the guy who shielded will also experience one frame of lag?

 

[bubbaking]

Well if I'm understanding that data correctly, I think that's only guaranteed if both players input their actions on the safe frame. If that isn't the case, then it seems very possible for the 'lag threshold' to change in between the two players' actions.

 

[RaptorTEC]

No he didn't say the same frame. He mentioned if the second input is made before the animation of the first one ends, then they experience the same lag. I'm just trying to make sure with him. This honestly isn't a big deal if that's the case.

 

[Vkrm]

Hey raptor, could you quote that bit from detta's post because I missed it entirely.

 

[dettadeus]

Hey raptor, could you quote that bit from detta's post because I missed it entirely.

Then I tried having Sheik Dsmash (hitbox out Frame 4) and Falco Powershield with the first shielding frame. They still experienced the same number of lag frames, despite Falco making his input on Sheik's third frame of input, and Falco did indeed powershield the Dsmash on the first frame. I repeated this successfully two more times.

I also had Sheik and Falco jab with 1 frame of difference in their inputs, and consistently had them start their jab 1 frame after each other, implying equal input lag.

Well I lost my own bet... by one frame.
15/16 spotdodges caught with the same frame setup. The one time I missed by one frame (14th repetition), I caught him out of his buffered spot dodge at the end without even counting the frames.

Here's the frame setup I used exactly.

Frame | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | 14 | 15 | 16 | 17 | 18 | 19 | 20 | 21 | 22 | 23 | 24 | 25 | 26 | 27 | 28 | 29
Dedede | 1st SD Frame | 2nd (Invinc. start) | 3rd | 4th | etc | | | | | | | | | | | | | | Buffer SD | | Invinc. end | | | | | | | 1st SD Frame |
MK | Crouch | | | | | | | | | | | | | | | | | | | | | | Input Dtilt* | | | | | Hit D3's Foot

*This is the first frame of the input; in frame-advance the input would be held while frozen on frame 22 and about to advance to 23.
**This entire table assumes Dolphin's default 2-frame delay, which is why there's 2 extra frames between MK's Dtilt input and it hitting.

I input MK's Dtilt on the same frame of D3's spotdodge every single time and hit him with 94% accuracy.

Also I tested a couple of times, watching for the input delay, and got the same delay for both D3 and MK twice in a row.


Hypothesis:

When an input is made, the "random" amount of lag becomes set for all subsequent inputs made for the duration of the first input's animation. This way, even though the amount of lag at any given moment may seem random, someone reacting to an animation before it ends will experience the same lag that the animation first experienced.
(Ignoring the 2 frame delay within Dolphin)
Example: D3 inputs a spot-dodge and it comes out after 1 frame of lag. The number of frames from the input to the end of the animation is 28 (input frame > no input frame 1 animation starts > 27 frame animation). For that entire 28 frame window, all inputs made experience the same 1 frame of lag. MK starts his Dtilt on frame 24 (input frame > no input frame 1, animation starts > no input frame 2 > no input frame 3 > no input frame 4, dtilt hitbox) and it connects on frame 28; D3 is hit.

This is probably incorrect and would need more testing but the current data upholds this hypothesis fairly well.

These are probably the relevant posts.

So basically lets say someone makes an attack and it comes out on 1 frame of lag then the second player inputs the shield button. You're saying the guy who shielded will also experience one frame of lag?

As far as I can tell, yes.

No he didn't say the same frame. He mentioned if the second input is made before the animation of the first one ends, then they experience the same lag. I'm just trying to make sure with him. This honestly isn't a big deal if that's the case.

I said the same frame the first time I tested, but then tested for larger frame differences and got the same results.

 

[bubbaking]

OK, I knew I saw something about actions being done on the same frame somewhere, so I wasn't making up stuff. It's good to know that it expands to moves with larger differences in frames of execution inputs.

 

[Exdeath]

For those claiming that this is more helpful to offense than defense, Reflex' examples of punishing dodges still stands.

For example, Meta Knight Dtilting Falco's spot dodge with 0 frames of lag requires him to input it on frames 17-19. With 2 frames of lag requires him to input it on frames 15-17. In other words, the only consistent frame of the spot dodge to aim for is frame 17, with the other frames causing Meta Knight to either whiff Dtilt on the spot dodge, or have it power shielded (a.k.a. guaranteed Usmash).

Normally it's not unreasonable to expect top players to hit this frame perfect window, except for one thing: Due to the ambiguous number of input lag, it's impractical to practice such a timing; it's virtually impossible to know if you've hit on exactly frame 17 without frame advance.

I find this discovery particularly vexing because I now have to second guess recorded frame data.

 

[Biz_R_0]

He's already provided an example of exactly that scenario to show how it benefits offense.

 

[Exdeath]

It benefits both in inconsistent (a.k.a. competitively bad) ways. It reduces an emphasis on player skill and emphasizes chance. If you wish that Brawl were more like competitive Pokemon then this is a good thing. For the rest of us, it's bad.

 

[Biz_R_0]

You're not backing up anything you're saying, you just keep saying "IT'S BAD IT'S BAD IT'S BAD" instead of directly responding to the point.

 

[#HBC | Joker]

Actually he is. He's saying that the window is so tight that trying to respond to a spotdodge becomes impractical. The player has to respond to the spotdodge, not the spotdodge input. If it came out 2 frames late, he still only just now saw it, has no idea if it was lagged or not, and will not be able to respond properly if he guesses wrong.

 

[RFrizzle]

BIZZARO I SWEAR TO GOD I WILL SLAUGHTER YOU FAMILY LIKE PIGS caps

 

[Biz_R_0]

And what he was saying was already covered by dettadeus (that all inputs made after that would have the same lag), and after I said that he just resorted to "IT'S BAD IT'S BAD IT'S BAD".

BIZZARO I SWEAR TO GOD I WILL SLAUGHTER YOU FAMILY LIKE PIGS caps

Aww, thank you.