Will the AI be stronger at least!!!

[Takalth]

However, what's the point of AI when you can fight a real person at any time with WiFi?

4 reasons:

1) Go online, and likely as not, you'll end up against a noob or a pro. Plus you have waiting time for the game and other inconveniences. If I want to play Smash for 15 minutes while I'm waiting for something, I don't want to spend 3 of those minutes connecting and logging in, 2 finding a game, 5 playing people whose skill is ridiculously different than my own, and 5 in actual good gaming.

2) Online will not be able to hold up 60 frames per second. I usually play against CPU's when I want to practice tech skill under pressure (Once I can do it in training, I have to test whether I can do it while exerting at least a little effort in staying alive or getting through a defense). If they added good DI and more frequent teching to computers, you could actually use them for really good combo practice.
Anyway, if you try this online, it will be too hard to time because of the lag.

3) When I'm experimenting with something new that may or may not work, I want to get a feel for it before I go fight somebody and get 4-stocked because I keep suiciding.

4) Some people don't have high speed internet. Some areas don't even offer it. I'm not part of this very unfortunate group (though several members of my family are), and while it's not a large part of the smashers, it does include some.

 

[Thino]

2) Online will not be able to hold up 60 frames per second.

this is wrong

 

[Takalth]

this is wrong

The only times this might work is two people playing on connections with a bandwidth that generously exceeds its total usage at the time of the game with a very clear connection. Packets sent over the internet deal with routers, security, queues, and everything else, not to mention that if you're playing against somebody who is a significant distance away, the actual travel time of the signal puts lag into it. There is no way you're going to get the same game experience on a 60 FPS game online as when you're on the same machine, and some high precision techniques (powershielding and Smash DI most notably) will become significantly harder.

I'm not saying online play is a bad thing, but it can't replace same machine play.

 

[Eaode]

The things I really want fixed are just the dumb edge things, and adding teching/better DI.

I don't care about the smartness, CPUs are just for tech practice. Even if online is too laggy, I can still play with my SSBM friends at their houses....


EDIT: and as someone said, if you want laughs, play FFA's or 2v2

 

[lex.aut]

i'd be satisfied if there are levels from 1 to 9 and a special "custom" option in which you can configure how the cpu should behave, which attacks it uses the most, etc.
but of course, i'm also expecting an AI that's far better than the one in SSBM - especially if there's no possibility to play online.

 

[dynamic_entry]

i'd be satisfied if there are levels from 1 to 9 and a special "custom" option in which you can configure how the cpu should behave, which attacks it uses the most, etc.
but of course, i'm also expecting an AI that's far better than the one in SSBM - especially if there's no possibility to play online.

firstly, @Eaode: but not EVERYONE has such cool friends whereby they will give up half their day just to play SSB continuously...but it sure is a massively fun day when you do!

but then, unless you're a good programmer yourself (im not) then you're still gonna find the comp predictable, aren't you? i mean, you wrote his **** srcipt! i realise if you did it properly it prolly wouldnt matter, but honestly...

 

[lex.aut]

firstly, @Eaode: but not EVERYONE has such cool friends whereby they will give up half their day just to play SSB continuously...but it sure is a massively fun day when you do!

but then, unless you're a good programmer yourself (im not) then you're still gonna find the comp predictable, aren't you? i mean, you wrote his **** srcipt! i realise if you did it properly it prolly wouldnt matter, but honestly...

yes, the AI would be predictable, especially since it would only be possible to change a few aspects of the cpu to keep the menu simple and userfriendly.
but with it's inhuman reaction time the cpu would take almost no damage if you tell it to play very defensive and basically only do counterattacks.

of course, it's only an idea and probably not many people will appreciate it, but hey - discussion is what the forum is for, right?~

 

[dynamic_entry]

yes, the AI would be predictable, especially since it would only be possible to change a few aspects of the cpu to keep the menu simple and userfriendly.
but with it's inhuman reaction time the cpu would take almost no damage if you tell it to play very defensive and basically only do counterattacks.

of course, it's only an idea and probably not many people will appreciate it, but hey - discussion is what the forum is for, right?~

ahh, i didnt realise that you meant tey should design a UI for you. but i think you prolly pointed out its major flaw, that in order to keep the interface user frienldy, it would only merit the comps with an overly simplistic style of play. which could be great if you wanted to work on one aspect of your game, like how to deal with the ranged attack of a certain type of character and didnt have a friend handy to spam B for you, but in my eyes, this doesnt rreally seem like a viable alternative.

preference: sakurai makes lvl 10 and mebbe 11..........this one goes to eleven...

 

[Micahc]

Making the AI unpredictable shouldn't be too difficult, different scripts for when an action happens, and randomly activate one of the scripts. Making the AI think on it's own however, would be difficult.

 

[TakDragon7]

Well, I don't think they'll do that. That'll be no fun if they thought on their own.

 

[Falco&Victory]

This argument is pointless. Alright everyone, do not try to argue about how smash is coded until you have downloaded ssbm onto your computer and hacked it. THEN i'll hear you out. AI is the hardest problem and largest setback of every game. In a game involving skill computers need to have inhuman precision to live. Computers have only limited options to attack and not suck. Sure, it's predictable, but they do not miss all the time.

Oh, about unreal 3, it's a coding engine, i didn't mean the unreal tournies. Oh, a mind bender for you. It is absolutely 100% impossible to make randomness. Even with the most advanced codes no super computer could ever surpass the simulation of randomness of the brain. It would add another 2 or 3 years of development to make the AI human. Better is easy. Harder to predict is hard. Human is impossible. Put your hand in front of your face and talk to it.

 

[Zauron]

Some of the ignorance in this thread is astonishing.

Like another poster said a bit ago, actual programming of AI is not nearly as time consuming as some have tried to claim, at least once you have the basic AI system set up. And I can say this, because its my job. I am an AI programmer for a games company. I've done fighting game AI before.

I'm not saying writing good AI is easy. Its not. However, the AI in SSBM could DEFINITELY be improved and it would not require anywhere near as much effort as some here have said. Code isn't done that way. Its modular. Code is built on top of other code. Well written code would require putting a small bit of code in one particular place and it would affect all situations.

As someone else pointed out, for example, you wouldn't need to change actual code to account for different stages. Instead, each stage has DATA that is read in and accessible to the AI. The AI reacts to this data. We call this table-driven or data-driven code, and it is the ideal way to code whenever it is feasible to do so.

If you wanted to have, say, a reaction to being near a ledge, all the code needs to know is where the ledge is in the world in comparison to the character. The ledge positions would be stored in memory easily accessible to the AI, and a function likely made called something like "GetNearestLedge" with other functions like "GetDistanceToNearestLedge" and so on. Lots and lots of functions like these are made and then used.

The data returned from these functions could then be used to determine actions. These actions would work on ANY stage because each stage's data would affect the results of those functions which would then determine the actions performed.

If there was any AI code written for a specific stage in Smash Bros. code I would be shocked. Its horrible coding practice commonly known as a "hack". Good AI code is set up to work no matter the environment, using data from that enviroment to make decisions. In a system set up correctly, adding new AI actions and reactions should be quick and simple - its just a matter of knowing which actions make sense to perform and that they are performed at times that make sense.

SSBM's AI is actually pretty decent at some stuff - pathing finding, for instance, is no easy task in a platforming-style game (you'll notice most platformers have AI that just stays in one area, it doesn't chase you around from platform to platform). Getting the character to chase another character down to attack them with multiple platforms from the stage data and not look like a complete idiot or get stuck in a loop running back and forth is not easy. There is plenty of info out there on pathfinding through a maze and such (like pathfinding in an RTS game) but I challenge you to find any comprehensive info on doing pathfinding in a platforming enviroment where gravity and jumping is involved instead of being able to move in any direction around a map. This is likely where most the AI work went into the game, even though most people in this thread probably don't realize it.

After pathfinding and a couple of impressive works of some characters attempting spikes or other advanced manuevers here and there, and some characters being pretty good at recovery, the AI is pretty sparse. Its clear they just didn't have time to make it better or decided it was not worth the effort when multiplayer was the focus of the game.

Oh, and as for the AI's "Level" - most AI of differing difficulty level is the same AI for all levels. The levels just are reference to a table of values that change the effectiveness of the AI. In the case of fightning games, the level generally just affects how often the AI does nothing useful versus attempting to attack the other player, and how often the AI properly reacts with the "correct" action. For example, you throw a projectile at an AI. At level 1, it has a low chance of avoiding the projectile. At level 9, it has a high chance of avoiding it by blocking, jumping, dodging, whatever. The code is the same either way, the difference is that in the data table for level 1 AI it has something like "AvoidChance = 10" and for level 9 AI it has "AvoidChance = 90". Again, data-driven code is the best way to code if it is feasible.

You can see this is what was done in Smash Bros. as well - notice how a level 1 AI, every once in a while, will do something you'd only expect a level 9 to do like pull off a spike. Its just that they do it FAR less often. The difference between 1 and 9 is just changing the chance of performing the actions (sometimes the chance is 0 to disable it as an option at all at certain levels) or changing the intensity of the correct action (i.e. how MUCH the AI uses DI as opposed to whether or not it does at all). The code, however, is the same regardless. Its all data-driven!

 

[Docablo]

Dope post ^^^^^

I think they should be a lot more inhumanly precise: powershielding every single attack, jumpcancelling to aerial instantly, perfect waveshine, perfect rising pound, perfect SHL. Add a new advanced AI, able to know which way to recover and which move to use. Maybe give them the ability to learn.

Make this the level 1 Cpu and improve it up to 9.

OMG lol!

maybe if you could customize the lvl 9 the way u want or something because that would be ungodly impossible if lvl 1's became lvl 9's. like this guy said only being able to change the values yourself would be great.

 

[LostAddict]

I thought this would be kind of a neat idea, but I have no idea if it would be incredibly hard to implement or what not. Something like a mode that watches you fight against the AI for like ... hrm, I don't know, 20 minutes. While you're fighting it analyzes patterns and things that you do, so if you wavedash a lot, it would record that as a gameplay tactic. This pattern recognition would become the new AI, so you could essentially face yourself. Once again, I'm no expert on programming, so I have no clue how difficult this would be.

 

[Phil Ashio]

OMG!!! This is truley wishful thinking, but I am totally 100% with you on this one, LostAddict. A computer that actually learns your attact patterns. But eventually you would be defeated, and your skill level will plateau...well until you start a new file or something. That would be cool if you can actually save your then newly created computer sparring partner. lol Wow, this would probably never happen, because it would be either A) too expensive or B) computer intelligence of that calibur I can never really picture would be for the SSB francise.

Another thought would be if you can save your computer sparring partner AI in the memory card, and then bring him over to another friend's house to spar against his AI sparring partner. Then play who is the better "master," "sparring partner" or all around better "player." I'm almost in tears because I know this will never happen...(sigh)

Yeah, keeping the dreaming going...wonder if you can even bring this new AI intelligence you have grown at home onto the online world. You and your AI creation against two other unfortunate souls. I should probably stop because the ideas are pretty much limitless.

I don't know much about computer programming because I'm the sorry sap that just enjoys the fruits of technology not knowing how it works and all. This thread really brought up some really good posts and ideas such as the one Addict brought up.

Zauron is this even remotely possible?

 

[Zauron]

This argument is pointless. Alright everyone, do not try to argue about how smash is coded until you have downloaded ssbm onto your computer and hacked it. THEN i'll hear you out. AI is the hardest problem and largest setback of every game. In a game involving skill computers need to have inhuman precision to live. Computers have only limited options to attack and not suck. Sure, it's predictable, but they do not miss all the time.

Oh, about unreal 3, it's a coding engine, i didn't mean the unreal tournies. Oh, a mind bender for you. It is absolutely 100% impossible to make randomness. Even with the most advanced codes no super computer could ever surpass the simulation of randomness of the brain. It would add another 2 or 3 years of development to make the AI human. Better is easy. Harder to predict is hard. Human is impossible. Put your hand in front of your face and talk to it.

Of course making the AI play like a human is impossible. That doesn't mean the AI can't be a LOT smarter than it is now. Its easy to see that, just look at other fightning games. Many fighting games feature AI that use a lot more of their moves than the ones in Smash do. The AI in Smash is just not very good, and could be improved to SEEM more human without as much effort as you claim.

I don't understand your statement on randomness either. Yes its true there is no such thing as true randomness in computers, but the random formula that's been used for years is close enough to true random as to be practically indistinguishable. Randomness in AI is crucial to unpredictable AI and is not at all difficult to do. All my AI uses the random function extensively!

For example. Your AI checks the position of the other fighters. It finds one is above it whose velocity indicates it will collide with the AI soon. Your AI character is currently in an idle or moving state and thus free to perform an action.

Now it can make a decision, based on a table of random values affected by the difficulty level the AI is set to, as well as other factors that may add weight to the randomness. These could include things such as:

- Move out of the way of the character, in case they will land with an attack.
- Jump and perform a jumping attack.
- Charge an up-smash and prepare release it when they are near.
- Stand there and do nothing.
- Ignore the situation and check for other possible situations to react to from other fighters.
- Perform a completely stupid move such as attempting a grab even though there's nothing to grab.

Each of these actions has a random chance of being selected, with the chance based on difficulty and perhaps on a log of previous actions (so, for example, a difficult AI could "learn" by logging what it did the last time this situation came up and if its action was succesful or not, and change the random percent such that the action that failed last time is less likely to occur this time).

The random function returns a value that is compared to the actions table generated in memory, and the result determines which action to take. Then more code is run to perform that action succesfully, such as comparing the position and velocity of the character to know how much DI to use so the jump will put the character in the correct position to perform an aerial attack. The ranges and movement speed is all data in tables that the code acts on, so the character knows which air attack would work best based on the position. Different characters can have AI data about preferred moves for damage and which direction they'd like to send the character they are attacking if the hit connects, and again select which attack based on weighted random values.

Good AI does not need to be like a human. But a lot can be done to make it seem more human-like and keep it fresh and somewhat unpredictable. The more actions added with different chances of performing them the less predictable the AI becomes. It takes time and planning, but not anywhere near as much code effort as you claim.

 

[Zauron]

There's been games that have done AI based on recording player actions before, but its never been done very well. Is it possible? Sure. But it would be a huge undertaking and probably not work as well as you'd hope.

The problem isn't learning the sequences but recognizing when the AI should use them. For example, it might figure out how to wavedash, but it wouldn't know WHY you should wave dash. An AI system that could do that would be pretty amazing, and require some hefty memory and CPU resources to do at all well.

Generally AI's of this type just record statistics on how often the player employs tactics it recognizes. So, for example, if you tend to spam a lot of projectiles, it will record that. If you tend to shield grab a lot, it will record that you tend to shield grab. However, if you instead dodge everything and hardly ever shield, it will note that instead.

Then it will try to play like you by performing the same actions with the same frequency it sees you do it, by modifying the weighting on its random selection of actions. So like in my above example, if there is a situation response set up for a character coming toward the AI with an air attack, and the choices for response include shield grabbing or rolling out of the way, the AI would be more likely to pick the response that corresponded to the action you generally used more.

However, each thing it records and tries to duplicate in this system must be set up by the coder beforehand. Thus if you do something the coder never considered as a strategy, the AI trying to imitate you would just ignore that input. So, for example, if the AI didn't know beforehand about Wavedashing, it wouldn't realize that it could use a back Wavedash followed by a forward Smash to counter the incoming air attack. It would not be one of the options it was set up to track.

A truly awesome AI of this type could combine learning moves by recording your actual input and applying those moves by recording what situations you use them in and adding them to its list of responses. That, however is extremely complicated and most game developers would not consider the payoff worth the coding time and system resources to do it.

 

[BackItch]

I'm just gonna sit back and watch Zauron rip apart everyone. There is no need to add to his awesome posts.

The man knows what he's talking about and you would be wise to listen to him. Heven forbid, you might actualy learn something!

 

[Falco&Victory]

Um, Zauron, I agree with you. It's not like they cant be smarter. However, making them act randomly, making them more human(not smarter, but being able to react), is a time consuming process. There would be too many variables to consider. Sorry everyone if I was giving the wrong impression, I apologize. But the level of AI everyone is asking for would take so long, and the code needed might even lag the game occasionally(I dont know if the Wii is packed with as much RAM as a GCube, I should check), so stop asking for challenging computers. If you want a challenge, it will be an inhumanly precise robot.

Zauron, I'll talk to my own hand. Cant back talk to an AI programmer when I'm a 15-year old self-learned programmer.

 

[Buttcrust]

Then why did you just try and say he's wrong?

Big Z is absolutely right. The stuff is pretty common sense as far as what he's saying (however doing the actual programming itself I can't even imagine being able to do) and I don't know the first thing about programming. Anyone who doesn't see what he is saying as right is downright dumb. I'm just glad that someone who actually knows what he's talking about came in here and straightened you all up. Big Z gets two thumbs up from the Butt.

 

[Mamush]

But the level of AI everyone is asking for would take so long, and the code needed might even lag the game occasionally.

Okay these are professionals doing the coding so lets get the whole difficulty of it all out of the way. They know what they're doing and the game has been in the works for over a year now so time is not of the essence as you say it is. The fact is everyone wants the AI to have a little more common sense and be less dumb. Which chances are pretty good for the AI to be better considering it wasn't perfect in the first place.

 

[dynamic_entry]

i sincerely doubt lag will be an issue in this game.

 

[Dantarion]

Oh, about unreal 3, it's a coding engine, i didn't mean the unreal tournies.

Unreal 3 is a game engine. I don't even know what a coding engine is.


Okay, now heres my thing about smash AI.

If you don't think it is possible to have human like AI in smash, you have to be kidding me.
Ever play Tekken, or Madden, or anything else?

Ever try playing UT2004 bots on Godlike?

Smash has HORRIBLE AI. In a complex 2d fighter like smash, there should never be one thing the computer always does. However, in smash, the computer ALWAYS does almost the exact same thing in any situation.

There are also a lot of moves that are completely unused by the computer, such as Falcon/Gannondorf using down+b to regain the second jump, Mario's Cape, Marth's over+b, and other moves which the testing team probably found out could be used for recovery.

i sincerely doubt lag will be an issue in this game.

60 Frames per second = 1/60 for each frame.

1/60 of a second = 16.66667 milliseconds.

Okay, so, disregaurding any network overhead, smash could be lag free if you have under 17ms ping between the people you are playing :D

 

[Luk^Skywalker]

I would just like to say that the comps in SSBM are not smart, however, their timing in attacks are very accurate,or inhuman. For example, i can't jump and use vitamin with doc and then waveland towards the comp to attack. Anyway, i would like the AI to be increased at least a bit more, but even if the AI remained the same, I wouldn't mind it at all.

 

[Takalth]

If you don't think it is possible to have human like AI in smash, you have to be kidding me.
Ever play Tekken, or Madden, or anything else?

Ever try playing UT2004 bots on Godlike?

Smash has HORRIBLE AI. In a complex 2d fighter like smash, there should never be one thing the computer always does. However, in smash, the computer ALWAYS does almost the exact same thing in any situation.

There are also a lot of moves that are completely unused by the computer, such as Falcon/Gannondorf using down+b to regain the second jump, Mario's Cape, Marth's over+b, and other moves which the testing team probably found out could be used for recovery.

You're partially right. It is true that random behavior should be in, but for the reasons Zauron pointed out, Smash can't reasonably be compared to Tekken or other similar games because there are so many other things you have to deal with such as platforms, elaborate air combat, recoveries, etc.

Essentially, I think Zauron's posts (if you take the time to read them) cover what is wrong, why it's wrong, and how it can be improved.

 

[Dantarion]

Smash can't reasonably be compared to Tekken or other similar games because there are so many other things you have to deal with such as platforms, elaborate air combat, recoveries, etc.

That is true, however, there are so many untouched things in the AI that it is pretty obvious that the people working on smash didn't even try to make the AI do some things.

For example, dash.

Elaborate air combat isn't really that hard for a computer. Since the computer knows things like how long you are stunned in midair, what direction you are moving, how long it takes to jump, and the range of your attacks, the AI could be able to do things like link moves together in midair if they really wanted it to.

 

[Fartemup]

Well said Zauron, looks like you pretty much cleaned up this post.

Just a question, I've used programming before, with randomness, but I had no idea that it wasn't EXACTLY randomness, I have realized that computers couldn't exactly pick a random number because they wouldn't be able to decide, because they're not human, but I've just ignored that and continued without questioning. Regardless, this brings me to my question, how exactly does the computer pick a random number then?

EDIT: Oh, and also, I think AI that thinks for itself would be really cool, but I don't think that the computers would be able to learn things like the Ganondorf's Down-B giving you another jump, because it would only think that you're doing this randomly, because the creators of Smash wouldn't have made these moves give you another jump on purpose, and as a result, the computer would probably randomly perform Down-B on its way back to the platform, becoming a disadvantage.

 

[Red Exodus]

It doesn't matter how smart you make them, you'll just adapt to their weaknesses and do maximum damage. Right now they don't recovery twice if the first one was canceled right? Now if they fix that and give them a new weakness we'll just capitalize on it. Not to mention better AI would add months if not years to Brawls release date.

I don't exactly remember but computers follow algorithms to get random values. Algorithms are steps taken to complete a task so at the end of the randomization steps it just repeats them. I could be wrong though.

 

[Tera253]

nah, Melee didn't have that much production time, and the Melee level 3 CPUs were better than the SSB64 CPUs.

~Tera253~

 

[Opfer]

Pseudorandom number generators for computers always create number sequences. They start with a number, let's say 3, then the perform a more or less complex operation on it, like (3 + 19)² = 484, then they take let's say the last digit, in this case 4. Then they repeat the process with the new number: (4 + 19)² = 529, take the last digit, and so on. You get a sequence of numbers, in this case 3, 4, 9, ... which seems to be random, but you'll note that the sequence repeats itsself at some point.

This example is strongly simplified over the real thing, it repeats itsself after just 3 numbers. Modern algorithms can generate sequences that don't repeat even if you let them create thousands of numbers every second for several billion years.

So technically, it wouldn't be a problem to make them unpredictable.

 

[TheChaos]

yeah, more advanced tech learning skills would be nice. you could turn that on and off, like an option:
AI: ON/OFF
a level 9 fox waveshined me and l-cancelled once. 0_o

 

[BrTarolg]

Multiply that by about 10 lines for a complete action, 30-50 times for every stage, 3-5 times per ledge on each stage, 40-50 times per character, 2-8 times per direction of recovery, and 9 times for varying comp lvls. Congrats, 1 action slightly changed. Do you program unreal 3, at all?

saying this is just proof that youre not a very good programmer yet.

thats like saying "oh making a factorial function is stupid because youd have to write down in a database every single factorial going to infinity"

a good programmer will make very simply code to do things which may at first seem very hard.

a crap one would do exactly what you stated, and have 100's of levels of cyclical complexity and nobody would be able to understand it.

 

[StarshipGroove]

Fixing the basic errors would be easy in my opinion.

For example, making C.Falcon use his kick to recover his second jump would only take a few lines of coding, I think: just make him use it after he has used his second jump and only when he's still distant enough from the bottom of the screen, then make him redo his second jump and then his third. They've already shown that the characters can do basic strings of moves.

Probably the hardest thing would be finding the collateral effects to these AI touches. For example every character dodging in Fourside without any reason.

 

[Thino]

Fixing the basic errors would be easy in my opinion.

For example, making C.Falcon use his kick to recover his second jump would only take a few lines of coding, I think: just make him use it after he has used his second jump and only when he's still distant enough from the bottom of the screen, then make him redo his second jump and then his third. They've already shown that the characters can do basic strings of moves.

this is not basic the 2nd jump recovery is a glitch

 

[Buttcrust]

Why the balls is everyone still talking about coding? Big Z already told you what's up. Quit debating it.

Red Exodus, your post is just so silly. "We'll just adapt to their new weaknesses!" Really? Wow. That sounds exactly like playing anyone, human or not, in any game ever created. The point is to not make the weaknesses so glaring and to give the ability to change somewhat to attempt to cover them regardless of success of the change.

 

[Zauron]

this is not basic the 2nd jump recovery is a glitch

Do you have evidence to support that Thino? Like a dev quote or something?

I'm pretty sure its intentional, based on what would have to happen for a glitch like that to occur. The flag "hasDoubleJumped" (or whatever they called it) has to have been cleared. Its very simple to see that this flag is only cleared in one specific case in the game - when the character has hit the ground. As far as I can recall, there is no other way to get your double jump back but to hit the ground - EXCEPT for C. Falcon and Gannondorf's down+B.

This is very, very straight forward code. The flag clearing only ever needs to be in one place - in the landing handler. How would one specific character have a move that accesses the landing handler without doing a whole bunch of other weird stuff? They wouldn't, it would cause a ton of odd issues beyond this. That means it can't be accessing the code that clears the flag through the landing handler. Which means, the flag is getting cleared by some other code. But there's no reason to clear double jump anywhere else in code - you don't get your double jump back by getting hit or any other action besides landing.

With that in mind, I'd have to say the only way I could see the flag being cleared by this move is if this move intentionally has a special, seperate line of code that clears it.

The only way this could be a bug, in my opinion, is if the extra line of code that causes it to happen was accidentally left in the handler for that move from a copy-paste human error or was put in on purpose at some point and then they forgot to remove it. I don't see how it could be the same as most bugs which are a result of a combination of factors the coders didn't account for causing odd behaviour. Getting the double jump back after the down+B would require new code to be put in for it to happen.

Now, what could have happened is they cleared the flag intentionally in this move as a fix to another bug. Perhaps they had problems with the special landing result of this particular B-move somehow bypassed the normal landing handler (not many B moves have a special reaction when you land while doing it, but this one does as it causes a long recovery once you hit the ground). Thus you'd do the move, hit the ground, then jump, and for some reason you wouldn't be able to double jump unless you landed and jumped again. That would be a pretty bad bug. Maybe someone made a mistake when trying to fix the bug and instead of clearing the flag only when they landed with the down+B, they just had it clear whenever they did the down+B at all - assuming that you'd either hit the ground or die when using it, and didn't think about the fact that it could be used to assist recovery by allowing a second double-jump if you didn't actually hit the ground or die after the move.

I still think it was intentional though, as one way or the other the code had to be explicitely added to do this. So if it was not intentional, it was a pretty blantant programming error compared to the other very obscure and hard-to-find bugs discovered in the game.

 

[Red Exodus]

They're computers, what do you expect them to do?

I've been through topics like this already and the all end the same way, people just stick to humans. I have almost no one to play yet I don't ***** about lack of skill in AI. If you ask me the weaknesses aren't that big of a deal at all.

 

[Thino]

dont talk like you know how ssbm was coded , everythnig you said is an ocean of hypothesis. there was no reason for the devs to have character recover their 2nd jump. as many glitches in this game , they most likely acknowledged it but left it in the final release , its not such a blatant programming error that would make the game crash or comparing it to black holes , soul stunner and other freeze glitches

 

[Zauron]

I am making no more assumptions than you are my friend. You claim its a glitch. You have no evidence beyond your own beliefs of the devs intents. I claim its intentional, with at least as much evidence to support my claim. Unless you can provide a dev quote to back up your claim that it is an unintentional glitch, you have no more ground to stand on than I do. You'll note that I stated in my reply that my statements were my opinions based on educated guesses about the coding of the game. You, on the other hand, made your statement as if it was accepted fact without any evidence beyond your own opinion.

 

[Thino]

im basing my statement on the fact that every character except kirby and jiggz have one 2nd jump . if devs intended to give falcon and ganon 2 2nd jumps then they should have been able to do it without the down+b. and i quote you here :

The only way this could be a bug, in my opinion, is if the extra line of code that causes it to happen was accidentally left in the handler for that move from a copy-paste human error or was put in on purpose at some point and then they forgot to remove it. I don't see how it could be the same as most bugs which are a result of a combination of factors the coders didn't account for causing odd behaviour. Getting the double jump back after the down+B would require new code to be put in for it to happen.

Now, what could have happened is they cleared the flag intentionally in this move as a fix to another bug. Perhaps they had problems with the special landing result of this particular B-move somehow bypassed the normal landing handler (not many B moves have a special reaction when you land while doing it, but this one does as it causes a long recovery once you hit the ground). Thus you'd do the move, hit the ground, then jump, and for some reason you wouldn't be able to double jump unless you landed and jumped again. That would be a pretty bad bug. Maybe someone made a mistake when trying to fix the bug and instead of clearing the flag only when they landed with the down+B, they just had it clear whenever they did the down+B at all - assuming that you'd either hit the ground or die when using it, and didn't think about the fact that it could be used to assist recovery by allowing a second double-jump if you didn't actually hit the ground or die after the move.

I still think it was intentional though, as one way or the other the code had to be explicitely added to do this. So if it was not intentional, it was a pretty blantant programming error compared to the other very obscure and hard-to-find bugs discovered in the game.

while claiming its intentional , you still accept the fact this bug could have happened , being a really bad one.