Grandeza
Smash Master
There was absolutely no difference between the two replays.
-
Welcome to Smashboards, the world's largest Super Smash Brothers community! Over 250,000 Smash Bros. fans from around the world have come to discuss these great games in over 19 million posts!
You are currently viewing our boards as a visitor. Click here to sign up right now and start on your path in the Smash community!
I have stumbled across something VERY serious, everyone needs to read this! VID IS UP
- Thread starter LavisFiend
- Start date
Arc2012
Smash Cadet
So what you're saying is that the Kirby has become aware of its own existence, able to make a choice and freely determine the outcome of the match. My God... We are all going to die at the hand of Kirby.
Okay, I'm not reading the rest of the posts past the first page here, but I'm pretty sure that I know exactly what happened here.
It's the same thing that everyone else is saying.
When the game saves your replay, it doesn't save a *.avi file or anything like that. What it does is it saves the series of button presses that it detected from you.
This means that the file is a lot smallter. This also means that if the game glitches for even one command, the whole replay is thrown off the mark from then on.
They probably don't save the AI's button presses the same way. They probably just save the "random seed number" for the match and just assume that the exact same randomly generated numbers will be used at the exact same points in the match, and so the AI will just make the same randomly-generated choices. Which is a safe assumption, for the most part, provided that the replay NEVER messes up. This is why the AI still acts and moves like an AI while about half of your moves are just ignored. (I never saw the Toon Link 1st replay pull out bombs or use the boomerang in the first part of the match like he did in the 2nd)
To note: It seems like the player's specials, in particular, were ignored in that first replay. Of course, since the RNG (random number generator) never generated numbers for the randomized damage for the specials that never happened, then that would throw off the whole number sequence and cause the AI to behave very strangely. Also, the subtle (or not so subtle) changes in position caused by the Specials would never be taken into account since they never happened.
Now, without knowing more about how the replay system works, I can't give you a deeper answer, but relying upon what I do know about it, this is my best guess of what happened.
Edit: I just read the rest of the posts and pretty much everyone beat me to all of my individual points. Oh well, this is a Computer Science major's nerdy explaination of it. So there you have it!
Addendum: I still don't know why the replay swtiches between the "real" match and the "fake" match back and forth like that, but I can guess why the replays would be messed up like this.
When Real-Time games like SSB:B are played, the internal program uses a sequence of code that handles all of the graphics, sound, input, move processing, etc in the game. This sequence is played over and over again in a loop that only ends when the match ends.
Every time through the loop = 1 frame. One time the loop might take 0.017 seconds, and another time the loop might take 0.015 seconds, and if there is a sudden lag spike then the loop might even take 0.044 seconds. No matter how much "real" time has passed, as far as the game is concerned it's only been 3 frames. Now, it's a bit more complicated than this, and I'm leaving a few things out, but that's how I've seen it done before and that's what Brawl seems to be going with. I could be wrong on that last point, though.
Assume (because I don't know this for sure) that the button presses are saved via a "timestamp" and not a "framestamp". The reason that this is important is this: When the game processes the commands, it does them by frames (looping the sequence over and over) and not by time. If there is a sudden lag spike for any reason, then the timestamped data will be thrown off the mark. Maybe even for one or more frames at a time.
This MAY actually explain why it switches between the fake one and the real one. When the Real match plays, it still has to do a lot of processing, which means that a lot of temporary variables are being created and then, when those variables are not longer needed, set aside to be "garbage collected." But if it doesn't do the "garbage collection" in between the matches, and instead does it during the second match, then there will be a slight lag spike during the replay that manages to throw off the whole match.
But this is all a bunch of assumptions and "maybes". It's an educated guess, if you will.
It's the same thing that everyone else is saying.
When the game saves your replay, it doesn't save a *.avi file or anything like that. What it does is it saves the series of button presses that it detected from you.
This means that the file is a lot smallter. This also means that if the game glitches for even one command, the whole replay is thrown off the mark from then on.
They probably don't save the AI's button presses the same way. They probably just save the "random seed number" for the match and just assume that the exact same randomly generated numbers will be used at the exact same points in the match, and so the AI will just make the same randomly-generated choices. Which is a safe assumption, for the most part, provided that the replay NEVER messes up. This is why the AI still acts and moves like an AI while about half of your moves are just ignored. (I never saw the Toon Link 1st replay pull out bombs or use the boomerang in the first part of the match like he did in the 2nd)
To note: It seems like the player's specials, in particular, were ignored in that first replay. Of course, since the RNG (random number generator) never generated numbers for the randomized damage for the specials that never happened, then that would throw off the whole number sequence and cause the AI to behave very strangely. Also, the subtle (or not so subtle) changes in position caused by the Specials would never be taken into account since they never happened.
Now, without knowing more about how the replay system works, I can't give you a deeper answer, but relying upon what I do know about it, this is my best guess of what happened.
Edit: I just read the rest of the posts and pretty much everyone beat me to all of my individual points. Oh well, this is a Computer Science major's nerdy explaination of it. So there you have it!
Addendum: I still don't know why the replay swtiches between the "real" match and the "fake" match back and forth like that, but I can guess why the replays would be messed up like this.
When Real-Time games like SSB:B are played, the internal program uses a sequence of code that handles all of the graphics, sound, input, move processing, etc in the game. This sequence is played over and over again in a loop that only ends when the match ends.
Every time through the loop = 1 frame. One time the loop might take 0.017 seconds, and another time the loop might take 0.015 seconds, and if there is a sudden lag spike then the loop might even take 0.044 seconds. No matter how much "real" time has passed, as far as the game is concerned it's only been 3 frames. Now, it's a bit more complicated than this, and I'm leaving a few things out, but that's how I've seen it done before and that's what Brawl seems to be going with. I could be wrong on that last point, though.
Assume (because I don't know this for sure) that the button presses are saved via a "timestamp" and not a "framestamp". The reason that this is important is this: When the game processes the commands, it does them by frames (looping the sequence over and over) and not by time. If there is a sudden lag spike for any reason, then the timestamped data will be thrown off the mark. Maybe even for one or more frames at a time.
This MAY actually explain why it switches between the fake one and the real one. When the Real match plays, it still has to do a lot of processing, which means that a lot of temporary variables are being created and then, when those variables are not longer needed, set aside to be "garbage collected." But if it doesn't do the "garbage collection" in between the matches, and instead does it during the second match, then there will be a slight lag spike during the replay that manages to throw off the whole match.
But this is all a bunch of assumptions and "maybes". It's an educated guess, if you will.
Dust319
Smash Apprentice
Everyone who said this earlier is correct. This is not a "recording," as we would know it. Rather it's simply a record of input commands. This was obvious when I realized that I could pause a Replay and take screenshots like I would a normal match.
Anyway, on to business. The only problem I've found thus far with the "generic comp of same lvl/ not recording the CPU's commands" is that this problem would be far more persistant if that were the case.
I just ran a few experiments with this, with Olimar as my sparing partner. I used a generic stage and a custom stage, and he was at level 9. On neither stage did the replay ever change dramatically as shown in the video. Olimar even used the same Pikmin against me each time. If it were simply substituting a CPU of appropriate level, I don't think it would go to that degree.
I did notice, however, that the replay did seem to change ever so slightly at certain times, but in my matches it was hardly enough to notice, and didn't change the overall outcome. It could be that one of these slight indiscrepincies, in combination with the stage being used (a stage builder stage that was not part of the original programing) may cause it to "freak out" and modify the game enough to change the outcome.
That being said, I'm not sure what would cause it to do it consistantly every other time.
Then again, the data could have been corrupted somehow as well.
Anyway, on to business. The only problem I've found thus far with the "generic comp of same lvl/ not recording the CPU's commands" is that this problem would be far more persistant if that were the case.
I just ran a few experiments with this, with Olimar as my sparing partner. I used a generic stage and a custom stage, and he was at level 9. On neither stage did the replay ever change dramatically as shown in the video. Olimar even used the same Pikmin against me each time. If it were simply substituting a CPU of appropriate level, I don't think it would go to that degree.
I did notice, however, that the replay did seem to change ever so slightly at certain times, but in my matches it was hardly enough to notice, and didn't change the overall outcome. It could be that one of these slight indiscrepincies, in combination with the stage being used (a stage builder stage that was not part of the original programing) may cause it to "freak out" and modify the game enough to change the outcome.
That being said, I'm not sure what would cause it to do it consistantly every other time.
Then again, the data could have been corrupted somehow as well.
ChromeSonic00
Smash Apprentice
That is so weird. I wonder when Nintendo may fix this.
regorris
Smash Apprentice
It's good to see some Comp-Sci people looking into this thread.
I think that Paingel may have hit the nail on the head with the connect-delay-frames notion. The match time progresses at 1second/second no matter what happens. I agree that the issue probably arises when a character's personal time time slows down, but the match time--and therefore the player's recorded move-marco--progresses as normal.
This hypothesis relies on the assumption that the delay in question is calculated at the time of the hit, and not all of the variables used come from clear data (like damage %) or off of the match's random seed.
On the Judgment/GreenMissile issue. The only way to preserve the type of hit would be to have some random seed calculated before any move is performed. The question that we need to ask is whether the random seed is for the match, or for the player. If the seed is calculated by the match, it could be possible for the replay of a all CPU G&W or Luigi game to be totally different from the original match.
I think that Paingel may have hit the nail on the head with the connect-delay-frames notion. The match time progresses at 1second/second no matter what happens. I agree that the issue probably arises when a character's personal time time slows down, but the match time--and therefore the player's recorded move-marco--progresses as normal.
This hypothesis relies on the assumption that the delay in question is calculated at the time of the hit, and not all of the variables used come from clear data (like damage %) or off of the match's random seed.
On the Judgment/GreenMissile issue. The only way to preserve the type of hit would be to have some random seed calculated before any move is performed. The question that we need to ask is whether the random seed is for the match, or for the player. If the seed is calculated by the match, it could be possible for the replay of a all CPU G&W or Luigi game to be totally different from the original match.
The more I think about it, the more it makes sense. Timing the replay of the moves via a time stamp would cause this problem. Consider the following scenario:
(Key)
B: Button: Timestamp of Button (in seconds)
F: Frame: Move done in Frame: Timestamp of Frame
Button and Frame Timings from Actual Match
(Starting with no buttons pressed)
F: Standing : 15.000
B: A Pushed : 15.010
B: Stick Tilted Right : 15.011
F: F-tilt to the Right : 15.015
B: A Released : 15.017
B: Stick Neutral : 15.022
B: Stick Tilted Left : 15.023
F: Move Left : 15.030
F: Move Left : 15.047
F: Move Left : 15.063
F: Move Left : 15.082
F: Move Left : 15.100
B: Stick Neutral : 15.102
F: Standing : 15.117
What happens in Match:
F: Standing : 15.000
F: F-tilt to the Right : 15.015
F: Dash Left : 15.030
F: Dash Left : 15.047
F: Dash Left : 15.063
F: Dash Left : 15.082
F: Dash Left : 15.100
F: Standing : 15.117
Button Sequence Saved for Player A
B: A Pushed : 15.010
B: Stick Tilted Right : 15.011
B: A Released : 15.017
B: Stick Neutral : 15.022
B: Stick Tapped Left : 15.023
B: Stick Neutral : 15.102
Button and Frame Timings from Replay Match
(Starting with no buttons pressed)
F: Standing : 15.005
B: A Pushed : 15.010
B: Stick Tilted Right : 15.011
B: A Released : 15.017
F: Move Right : 15.020
B: Stick Neutral : 15.022
B: Stick Tapped Left : 15.023
F: Dash Left : 15.035
F: Dash Left : 15.052
F: Dash Left : 15.068
F: Dash Left : 15.087
B: Stick Neutral : 15.102
F: Standing : 15.105
F: Standing : 15.122
What Happens in Replay
F: Standing : 15.005
F: Move Right : 15.020
F: Dash Left : 15.035
F: Dash Left : 15.052
F: Dash Left : 15.068
F: Dash Left : 15.087
F: Standing : 15.105
F: Standing : 15.122
Now, in both cases, the buttons were pressed at exactly the same time in both cases, but the timing of the frames were slightly off. In the Actual match, the player went from standing, to doing an F-tilt, to moving to the Right. In the Replay match, the F-tilt never happens, and the player moves Right a little bit before Dashing Left. Also, the Dash 1 frame shorter in frame time.
See? Slight lag+ Timestamping moves = Inaccurate Replays.
It would be nice if someone could get some real data for this so I won't have to make up numbers.
(Key)
B: Button: Timestamp of Button (in seconds)
F: Frame: Move done in Frame: Timestamp of Frame
Button and Frame Timings from Actual Match
(Starting with no buttons pressed)
F: Standing : 15.000
B: A Pushed : 15.010
B: Stick Tilted Right : 15.011
F: F-tilt to the Right : 15.015
B: A Released : 15.017
B: Stick Neutral : 15.022
B: Stick Tilted Left : 15.023
F: Move Left : 15.030
F: Move Left : 15.047
F: Move Left : 15.063
F: Move Left : 15.082
F: Move Left : 15.100
B: Stick Neutral : 15.102
F: Standing : 15.117
What happens in Match:
F: Standing : 15.000
F: F-tilt to the Right : 15.015
F: Dash Left : 15.030
F: Dash Left : 15.047
F: Dash Left : 15.063
F: Dash Left : 15.082
F: Dash Left : 15.100
F: Standing : 15.117
Button Sequence Saved for Player A
B: A Pushed : 15.010
B: Stick Tilted Right : 15.011
B: A Released : 15.017
B: Stick Neutral : 15.022
B: Stick Tapped Left : 15.023
B: Stick Neutral : 15.102
Button and Frame Timings from Replay Match
(Starting with no buttons pressed)
F: Standing : 15.005
B: A Pushed : 15.010
B: Stick Tilted Right : 15.011
B: A Released : 15.017
F: Move Right : 15.020
B: Stick Neutral : 15.022
B: Stick Tapped Left : 15.023
F: Dash Left : 15.035
F: Dash Left : 15.052
F: Dash Left : 15.068
F: Dash Left : 15.087
B: Stick Neutral : 15.102
F: Standing : 15.105
F: Standing : 15.122
What Happens in Replay
F: Standing : 15.005
F: Move Right : 15.020
F: Dash Left : 15.035
F: Dash Left : 15.052
F: Dash Left : 15.068
F: Dash Left : 15.087
F: Standing : 15.105
F: Standing : 15.122
Now, in both cases, the buttons were pressed at exactly the same time in both cases, but the timing of the frames were slightly off. In the Actual match, the player went from standing, to doing an F-tilt, to moving to the Right. In the Replay match, the F-tilt never happens, and the player moves Right a little bit before Dashing Left. Also, the Dash 1 frame shorter in frame time.
See? Slight lag+ Timestamping moves = Inaccurate Replays.
It would be nice if someone could get some real data for this so I won't have to make up numbers.
Steck
Smash Journeyman
Do the character's playing look like CPUs? Maybe the game got mixed up and made up a CPU Toon Link vs Kirby with 2 stock match?
Xtreme Starfox
Smash Ace
- Joined
- Dec 12, 2007
- Messages
- 622
Toon Link in no way looked like he became a CPU. -_-
My personal experience with this
http://www.youtube.com/watch?v=HDWGi3PmrsA
I recreated Lavis' stage exactly (Known as PwnagePanda on NSider2)
This was the result
http://www.youtube.com/watch?v=HDWGi3PmrsA
I recreated Lavis' stage exactly (Known as PwnagePanda on NSider2)
This was the result
Xtreme Starfox
Smash Ace
- Joined
- Dec 12, 2007
- Messages
- 622
It could but the game can also record that like buttons eg.
A, B+up, Run to right, Trip, roll, A........
You get the idea?
MildlyTemperate
Smash Cadet
This is very strange.
KosukeKGA
Smash Champion
Expect glitches to occur in the first version of this game...
_X_
Smash Lord
That's pretty funny. And a little freaky without any good explanation.
This is how the whole "Jesus" thing happened. Something happened with no explanation behind it and people assumed divine involvement.
As a few people were saying if the game does in fact record the presses of the buttons it's possible it's a kinda flawed system epically if you're doing a whole bunch of weird advanced ****.
Non the less, you were right to be a little freaked.
This is how the whole "Jesus" thing happened. Something happened with no explanation behind it and people assumed divine involvement.
As a few people were saying if the game does in fact record the presses of the buttons it's possible it's a kinda flawed system epically if you're doing a whole bunch of weird advanced ****.
Non the less, you were right to be a little freaked.
redgreenblue
Smash Ace
This happens to me in Excitetruck. Man that was a weird replay.
missedwithtruestrike
Smash Journeyman
the same thing happened to my friends except they stopped moving and the replay goes longer than 3 mins. the problem is we dont know how to post it. (its on an sd card if that halps)
WilldaBeast
Smash Apprentice
Strange, but I doubt it's anything to get concerned about. Just a first version glitch.
Kirby Gotenks absorbed
Smash Apprentice
This isn't something to be scared about, it's just a malfunction. The same thing happened to me when I played Sonic R, I watched the replay... And then my character just kept on bumping into a wall.
SirroMinus1
SiNiStEr MiNiStEr
im pretty freaked out myself. i was thinking what if someone sent that to you but thats clearly not the case.
Darkslash
Smash Master
Thats goddam creepy, Whats more creepy is that it switches on a regular basis 
UltiMario
Out of Obscurity
I have a small question though, what If you ran out of time for your 3 mins? What if you and your foe had the same amount of stocks left too?
Would it go into sudden death and continue?
Would it stop with no winner?
would the game freeze?
Would it go into sudden death and continue?
Would it stop with no winner?
would the game freeze?
I don't feel like reading past the first page, yet all 4 pages so I am going to post what happened
it's simple
SSBB keeps the file size of replays down by recording the moves of each character in the fight. What you encountered was a glitch in the software. Might have been something to do with the custom stage. Anyway, thats why you are rolling everwhere and swiping at the air.
whoady4shoady
Smash Apprentice
Dont freak out, replays just work the same as Halo 3.
Bacon Man11
Smash Journeyman
This is real weird. But I'm pretty sure that it is just a glitch. Never happend to me before though.
Sir Brane Damuj
Smash Rookie
That happened a lot in my NASCAR games. Instead of actually racing, I just turned around and crashed. However the game, which doesn't start recording until AFTER the green flag, and I turned around on the pace lap, took all of my actions and translated them into me going forward. So I was turning into walls, backing into the grass, etc.
Linkster47
Smash Apprentice
Like others have said, the game records characters' moves, not video footage. The game probably lagged and caused the character's actions to be timed incorrectly, similar to how a movie's audio will sometimes not match up with its video footage, but that still does not explain why you were doing completely differant moves than what you originally performed...
Anyway, you shouldn't get to worried about this. It's just a minor glitch that occured in one recording, but if it happens again, you should get your game checked out.
Anyway, you shouldn't get to worried about this. It's just a minor glitch that occured in one recording, but if it happens again, you should get your game checked out.
gnosis
Smash Lord
So say we have a replay between two human opponents. And early on in this replay, some glitch occurs that does this 'desynch' we're seeing, and the game carries on with the players' (now ineffective) button presses.
What if this desynched set of presses doesn't result in either player winning by the time they've run their course? Do the characters just stand there until the replay ends?
edit: I am curious as to why his replays perfectly alternate. Perhaps it does have something to do with a rounding of some variable the game does, and the variable it's rounding ends with .5, and half the time the engine rounds a .5 down and the other half it'll round a .5 up... but probability doesn't work like that so that still doesn't explain it.
What if this desynched set of presses doesn't result in either player winning by the time they've run their course? Do the characters just stand there until the replay ends?
edit: I am curious as to why his replays perfectly alternate. Perhaps it does have something to do with a rounding of some variable the game does, and the variable it's rounding ends with .5, and half the time the engine rounds a .5 down and the other half it'll round a .5 up... but probability doesn't work like that so that still doesn't explain it.
IGN Forums
Smash Journeyman
scary!! stuff....
DVD Smith
Smash Journeyman
This has happened to me too. I saved an incredible replay of my friend and I playing as Link and Sonic repsectively on Green Hill, and it was an incredibly close match (The score was +1, -1 in my favour in the end), and it worked fine for ages. It's only been today when I watched the replay again and everyone kept suiciding, Link would run to the edge of the screen, start using A-attacks and kill himself. Same with Sonic on the opposite side. Then there was a period where they stood about doing nothing. The final score shown by the replay was -5, 0. It's really freaky, and so crappy because that was the best Sonic I have ever played.
EDIT: and it NEVER shows the original match, unlike the Toon Link/Kirby match. I've watched it over 20 times now and all I get is the corrupted one. But note that this was between two HUMAN players, not a CPU.
EDIT: and it NEVER shows the original match, unlike the Toon Link/Kirby match. I've watched it over 20 times now and all I get is the corrupted one. But note that this was between two HUMAN players, not a CPU.
Milos
Smash Lord
everyone chill out it's just a replay glitch it happens in Sonic Riders as well, and most games where replays are recorded by button and not onto a video.
Impact009
Smash Journeyman
It's funny how doujin game programmers can get this right (see Melty Blood or any Touhou game), but the well-known, big companies can never seem to get this down.
Crizthakidd
Smash Champion
celda and staffy lmao
zelda? BRAWL SUCKS!!!!
if it says i love melee.........
zelda? BRAWL SUCKS!!!!
if it says i love melee.........
Taymond
Smash Journeyman
Yeah, I recall a similar issue in Starcraft a few years back. A random match against a couple computers and a couple humans. The match went on for a few hours, and one of the humans won. The replay, though, showed a zerg computer winning in about half the time, and then just lounging around with nobody to fight for the remainder of the video's length, mining and making units.
Was pretty funny to pretend the AI was unhappy with its loss and decided to change the replay to destroy all evidence.
Was pretty funny to pretend the AI was unhappy with its loss and decided to change the replay to destroy all evidence.
Eternal Yoshi
I've covered ban wars, you know
- Joined
- Mar 3, 2007
- Messages
- 5,450
- Location
- Playing different games
- NNID
- EternalYoshi
- 3DS FC
- 3394-4459-7089
I had that happen before. When a Zero Suit Samus online and I went to Sudden Death. I managed to win after I got her with a grenade after a few dodges and just barely avoided her attack. In the replay, my character got caught by her attack and lost. You are not alone.
It never showed the real match so I deleted it.
It never showed the real match so I deleted it.
brawlmaniac
Smash Lord
I saw the video and I believe you. That's really freaky to see. The first vid had you dying from a suicide dair. The second vid had Kirby dying from the falling block, and you recovering by walljumping. Aside from all this... I have no possible explanation in my mind on what could have happened
Judge Judy
Smash Lord
- Joined
- May 18, 2008
- Messages
- 1,638
Lol, the computer switched videos to make itself look better.
Mario FTW!
Mario FTW!
The Halloween Captain
Smash Master
I might just be adding to clutter, but it looks like the error invloved an early wall-jump in the false video, If you want to pinpoint the problem spot.