I haven't posted here for a while, but recently digging through the Internet I found some very interesting information, and I think I managed to find the origin of the Select+Cancel glitch in Red/Green J and the Missingno. glitch, and I also have some speculation on how Trainer-Fly was originally found. Since I have nowhere better to post this, I'm posting here. Also, notice that my japanese is terrible, and I had to rely on google translate and OCR software to read printed japanese characters and understand what's going on. In any case, it's a giant wall of text, but there was a lot of detective work:
Select+Cancel:
27/02/1996 - Pokemon Red/Green was released in Japan.
05/04/1996 - Famimaga No. 7 is
published. It covers the Pokemon games, and in a section where readers can submit tricks, it shows a technique to clone Pokémon in a trade by shutting down one of the cartridges at a specific time.
15/04/1996 - CoroCoro comics for the first time acknowledges the existence of Mew, the 151st pokemon, and
announces a special distribution event where only 20 readers would be selected to receive it.
19/04/1996- Famimaga No. 8
publishes in its reader submitted tricks a Select glitch to allow to share experience between Pokemon in battle. More details later.
20/05/1996 - Kouichi Hiwasa
describes on usenet a Select glitch to create Mew. In the email he says he did not believe the Mew rumors initially, but this glitch convinced him it was real. Later he
credits his younger brother for the discovery, after experimenting with a Select glitch he learned from Famimaga.
Speculation about what happened:
The story of how Mew was put into the Pokemon games as told by the developers has changed several times. Sometimes it's said that it was a prank, and later they released it when a glitch allowed players to access it, and in other interviews it was said that they included it to create rumors to drive up sales.
According to this timeline, it looks like it was the latter.
The Select glitch published in Famimaga No.8 goes like this: say you want to give exp to the second pokemon. You place the Select cursor on the second item, and exit with B. Then you go to the pokemon menu, and the cursor shows up at the second. Then if you switch it with the first pokemon, the Pokemon in battle don't actually change, but after you win the battle the experience only goes to the one you switched over.
Now, Select glitches were pretty common in JRPG's at the time. For some reason developers couldn't implement swapping things in UI right. Anyway, it looks like shortly after the games release someone wrote to Famimaga about this select bug in Red and Green, and they published it. Now, when you know this, there's the natural question. Does this work outside battle? There are six pokemon in the party, but 20 items, what happens if you swap pokemon that do not exist? Nowadays we know the answer, this corrupts memory, and if you choose specific hex values, you could theoretically get other types of Pokemon, including Mew.
It's unclear to me why editors of Famimaga did not consider this. The most likely reason is that, since they had connections to Nintendo, they didn't want to publish things that would cause kids to lose their game, so they settled for publishing a safe variant. In any case, I think you can see how unreasonable is the story of Game Freak devs, that Mew was an accident discovered by a glitch. Although there really was a glitch that allowed players to see it, it was only published in a very mild form in Famimaga only AFTER they released it. Although Famimaga No. 7 does include a glitch, it's an improper trade cancellation glitch, not a select one. And even the select one published requires quite a bit of work to get Mew.
Even if we accept that somehow Famimaga knew about the glitch and told the Nintendo devs way before publishing, we are talking about someone finding Mew with a very difficult to exploit glitch, telling Famimaga everything, they contacting Nintendo, and then distributing it to only 20 people after finding out about it.
I think this is extremely unlikely, what most likely happened is that they put Mew in the game with the intention of distributing it to 20 people in CoroCoro almost two months later to create rumors. Then the glitch was found, and people heard about the rumors and could confirm that Mew existed in the game using the glitch. Then later the devs made up that story as a marketing strategy.
Missingno.
February 1999 - Nintendo Power No. 117
publishes in "Classified Information" section the Fight Safari Zone pokemon in the wild trick. This is a crucial source that most people miss. The Fight Safari Zone pokemon trick is described as running out of steps in the zone of the safari zone the pokemon you want appears, then surfing from the south of Fuschia to Seafoam Islands and surfing up and down in the east coast. It's incorrectly stated that if the player battles trainers along the way, the trick will not work.
April 1999 - Nintendo Power No. 119
publishes in the PokéChat section, where they take questions for fans, two interesting answers. The first is an answer to a question regarding catching tough pokemon, that the player should try holding B the moment the ball explodes, which has no effect at all. Although this is not technically wrong advice (it's never explicitly stated that doing this will make catching easier), it's likely the source of the infamous myth that holding B will improve catching. The other interesting answer asks the player to try fishing on gym statues. This is indeed a trick that exists, since gym and pokemon league statues are considered water tiles, and you can fish and surf on them.
May 1999- Nintendo Power No. 120
acknowledges the existence of Missingno. and describes it as a glitch that could harm the game. It describes the way to encounter it as talking to the Old Man in Viridian City and using Fly to get to Seafoam Islands while avoiding contact with all land pokemon.
June 1999 - Nintendo Power No. 121
describes the Poké Doll trick in Pokémon Tower to skip the Silph Scope.
Sometime before Jan 2000 - Nintendo has an
FAQ about Red and Blue where they discuss missingno. It describes it as appearing when the player incorrectly performs the Fight Safari Zone trick. This trick, by the way, is also described incorrectly in another answer, saying that the trick will not work if the player faces enemies which are not the water pokemon on the way there.
Jan 2002 - Jolt135 creates a
strategy guide which has his discoveries about Missingno. He figures out that the encounters are controlled by the characters in the player's name.
Speculation about what happened:
For those who don't know, here's how this method for seeing Missingno. works. Tiles in gen 1 are actually composed of blocks of 2x2 subtiles. Due to a programming oversight, when the game calculates if you should have an encounter, it uses the bottom right subtile, but when it picks which pokemon you should actually see it uses the bottom left.
This doesn't matter in 99% of the cases. The only instance where this causes problems is when you have a shore tile, where the left subtiles are land, and the right subtiles are water. To compute the encounter density in these tiles, the game will pick the data for surf encounters, but when you eventually get the encounter it will pull that data from the land encounters. Although there are a lot of shore tiles with left land, and right water in the game, the only regions that have surf encounters are in the water routes between Pallet Town, going through Cinnabar and Seafoam, until Fuschia.
Because of this, this discrepancy only happens in three places. The left most shore tiles you can surf while leaving Pallet, the right coast of Cinnabar, and the right coast of Seafoam Islands. Pallet is completely useless because Sea Route 21 has land pokemon, so you will simply encounter them, which you could easily do by walking on the grass.
That leaves only Cinnabar and Seafoam, where you can potentially access uninitialized data, and today almost everyone does it in Cinnabar because it's easier to reach.
Now, one
guide in Gamefaqs states without source that Missingno. was first found in an in-game trade with the guy in the Cinnabar lab. This guide is cited in some places, and it does make sense, because you can think someone trading a pokemon there, surfing in the Cinnabar coast and finding a Missingno. However, in face of all of this, I think this is highly unlikely. Notice that nearly all early material on this glitch comes from Nintendo, and they only mention Seafoam Islands, both on their Safari Zone trick, the Old Man+Missingno. disclosure in Nintendo Power, and in their FAQ on pokemon.com, where they state that Missingno. is an effect of players doing the Safari Zone trick incorrectly.
Therefore, all evidence points to the glitch being disclosed by Nintendo, and it's not clear at all that they even knew the glitch could be performed in Cinnabar, because all their early references are in Seafoam. Even in their Missingno. disclosure in May 1999, where they correctly state that the player must avoid land pokemon, they specifically tell readers to use Fly and go to Seafoam, not Cinnabar, which would be easier. Because of this, I think it's much more probable that Nintendo knew that the old man could trigger the glitch before they knew it could be done in Cinnabar, so the story that Missingno. was found first without the old man and at Cinnabar is implausible.
Nowadays, it's generally acknowledged that Nintendo disclosed the bug on the May 1999 issue of the Nintendo Power magazine, but I think the one with most information is the Feb 1999 issue, with the Fight Safari Zone trick, and together with the others, we can have a pretty good idea of what was going on there.
First, why on earth does Nintendo say in Feb 1999 that the Fight Safari Zone trick doesn't work if you fight trainers on the way to Seafoam? It's not only a mistake, but a very difficult one to make if you found the bug by testing. Say you found the trick, then if you're going to publish that it doesn't work if you fight a trainer, you do the obvious, you fight a trainer and go there, you figure out that it still works, and you don't publish!
The thing is: this mistake is very easy to make if you are a DEVELOPER. Judging from the other issues, Nintendo was disclosing tricks which are pretty difficult to find if you don't know the game's internals, but pretty easy if you know their logic. For example, the statue tiles being considered water so that you can fish, the pokedoll trick, and even one about holding the B button, which is wrong, but could potentially come from someone who understood the code wrong. Anyway, if you assume this wrong tip about the Safari Zone trick came from a programmer, the mistake is easy to justify. You know that the glitch happens because of uninitialized data in land pokemon, so you conceive of this trick, where you visit the Safari Zone and just go to Seafoam, while keeping it there. However, you don't know if other parts of the code can overwrite it, so you do the minimum amount of actions, you don't use Fly, you don't fight trainers, and try to get it working only with wild encounters. After you see that the trick works, congratulations, the bug in your code can now be put on a magazine as a neat trick.
What the programmer(s) who conceived of this trick did not know at the time is that it's extremely difficult to overwrite the active land encounter table. The only things that can do it are the Old Man, in-game trades and cable club battles and trades, and exploiting the last is not trivial either. Normal trainer battles don't override the encounter table, and since it persists when you save the game, resetting to access random data there also doesn't work. However, even if this tip was wrong, it might have led players to the glitch.
Suppose you are a pokemon enthusiast that regularly reads game magazines and you see the Fight Safari Zone trick as it was published. The trick seems to imply that the tiles at Seafoam are special, and with little effort you can conclude that you can encounter there pokemon from the last area you visited. But if they were programmed to do so, why does it include the condition to not fight trainers for the trick to work? If you do fight trainers, what's going to happen? From there, assuming that you are going to test this trick on a game where all trainers have already been fought, the easiest option is to simply go to the Old Man and go to these tiles and see what happens!
From this, I think the most probable timeline of events went like this:
1. While Nintendo is working on Yellow or G/S, they figure out the problem with the shore tiles, along with other things.
2. Someone at Nintendo gets creative and sees that these bugs can be refurbished as neat tricks to publish in Nintendo Power.
3. Through careless testing, the Fight Safari Zone pokemon trick is published, including an incorrect condition to avoid trainer battles.
4. Players, probably using a game without any trainers to fight, try the Old Man before going to Seafoam.
5. Shocked, they see all the oddities of the glitch, including Missingno.
6. At some point, Nintendo Power is contacted with questions about Missingno., if it's a new Pokemon, and why it appears after you talk to the Old Man when doing the Safari Zone trick.
7. The magazine contacts the programmers about this, who now analyze the code correctly.
8. Nintendo Power publishes the programmers' conclusion. Missingno. is simply a bug, it might have unintended side effects. They now describe the trick correctly. It's land pokemon that you should avoid, not trainers. They publish the procedure to get it and print its photo to show readers it's only a bug.
9. This disclosure has unintended consequences. When you tell people not to do something, they usually do it. People start encountering Missingno. and catching it.
10. Nintendo sees that the glitch almost never erases the game, and can have good side-effects, and could be exploited by cheaters, then tries to bury the issue by never giving details about Missingno. again, saying it's just a glitch when the Safari Zone trick is done incorrectly and questions should not be asked.
Trainer-Fly
Sometime around 1998 - Team PA starts
Pokemon analysis (everything except the BBS board is archived
here). They accumulate a lot of knowledge about the game, they reverse engineer how item, pokemon, moves, stats and encounter tables work, and include a page about how to
make Mew using Select+Cancel. Unlike Hiwasa, which I cited before, they know what they're doing. While Hiwasa got Mew's ID of 21 by using a water type without knowing about the ID, they get the value of 21 by using map coordinates.
07/08/1999 - Someone
reports encountering a Level 7 wild Mew in teamPA BBS board. The source for this is Kakeru, and he admits in his blog he can't give an objective source, because this BBS archive is private. Nevertheless, I think this is trustworthy, and is key material that I will discuss this and Kakeru's tweets later.
31/05/2002 - "Nintendo"
submits a code to JesseWorld detailing how to get Mew using the well known Gambler+Youngster method. It's dismissed as fake.
09/03/2003 - Daniel26
submits the same Gambler+Youngster technique to catch Mew on GameTalk. It doesn't pick up any attention and receives low ratings.
19/04/2003 - TheScythe
posts in GameFAQs boards the same Gambler+Youngster technique. He anticipates skepticism and says he saw it in two other websites, tried it because he was bored, and it does work.
20/04/2003 - Jolt135, a respected user who wrote the strategy guide posts "What I wonder: Where do they pull 7 and 21 from if this indeed exploits timing glitches? Oh, well. It doesn't matter. All that matters is... IT WORKS!!!!!" This validation proves to be critical, and everyone starts trying the method and seeing that it works.
20/04/2003 - Jolt135 posts a link to the GameFAQs thread to
Azure Heights, telling people about it.
29/04/2003 - White Cat posts an Eureka in Azure Heights, figuring out that the pokemon battled is controlled by the special of the last pokemon seen.
29/01/2004 - fifthヽ(´ー｀)ノ ◆Fi3PJTZKLQ posts in
2ch a link to his
webpage. This explains the glitch in some detail, like how the pokemon changes when you fight different trainers. This post greatly popularizes the trick, which becomes known as the "fifth method" in Japan.
Analysis:
The most interesting piece of info is the post on the teamPA BBS board in August 1999. The poster is playing the japanese Yellow version, and answers Mew stats for L64 and L68. When asked where he got it he never says the location explicitly, but it's strongly implied to be Rock Tunnel. That's because he mentions using Flash with Pikachu, being annihilated in a tunnel and going back in, and encountering Mew in the 2nd floor, in the place where girls show up one after the other. Rock Tunnel is the only dark cave in Gen I, the only place where you need Flash, and it does have in the final section three girl trainers. It's entirely possible that you die there in front of a trainer, get teleported out, and enter Rock Tunnel from Lavender to activate an encounter.
Kakeru comments in his tweet that he should have noticed the start menu not working after doing this, but actually in the Trainer-Fly variant where you die from an encounter in front of a trainer, the start menu still works afterwards. The poster mentions using a L20 Pikachu and paralyzing Mew and using 3 balls to catch it. These numbers also make sense. Now there's the question: why does the poster not mention dying in front of a trainer specifically? My guess is that this technique of dying to a Trainer in a cave was already known, and this is implied by the context.
To start, let's see how we could get Mew this way. If we die in Rock Tunnel for the first time, we need to reenter it from the north part of route 10, which is not the area with the girls. So, to do this, we need to already be past Rock Tunnel. So, this definitely did not happen by accident on the first Rock Tunnel visit. It must have happened in a revisit. Well, if you are revisiting Rock Tunnel with the intention of dying to a trainer there, the entrance closer to Lavender is much better, because the two girls there are optional, so you can skip them in a first visit, and they are pretty close to the entrance, so you can try dying to them.
Now, if we just do Trainer-Fly by dying in front of a trainer and going back immediately, the result is uninteresting, because we're simply going to fight the trainer. To have interesting results we need to battle a trainer or a wild pokemon first. It turns out that none of the trainers close to the south Rock Tunnel entrance give Mew. With respect to wild encounters in the yellow version, it's much more interesting. The grass behind the Cut trees in route 8 in Yellow have L20 Pidgey and L22 Pidgey as the most common Pokémon. Together they make about 40% of the encounters. L20 Pidgey has 3/16 chance of having the special to trigger Mew, and L22 Pidgey has 2/16 chance.
Just to confirm all of this, I loaded the save file from FractalFusion's cancelled
Yellow TAS and setup all I needed to make this work, leveling Pikachu up a bit, catching 10 pokemon, getting Flash, etc. I could get a wild encounter in front of the girl in Rock Tunnel in my fifth attempt. It's pretty easy, you can simply save in the diagonal of the trainer, walk to the spot and if you don't get an encounter, reset immediately and try again. Then I just had Pikachu use Thunder Wave on the Zubat until it got killed. After that I got the cut slave and went to route 8 to find encounters. I only went back to Rock Tunnel when I encountered a Pidgey, and in my tenth attempt I got Mew! It's certainly reasonable to obtain Mew this way if you put some effort.
One important thing also is that the post in 1999 has in its title that it's a wild LEVEL 7 Mew. And unlike other trainer-fly reports, teamPA does take it seriously and asks questions. It looks like the most probable thing is that it was already known to teamPA members that you could encounter level 7 wild pokemon by dying in front of a trainer, facing a battle and going back, but they did not know what controlled it. Since it was already known how they could make Mew using select glitches, the novelty of the report is that Mew could be encountered with the wild level 7 method. It was plausible to them that this could occur, because they didn't understand it, so they immediately asked what the person did, so that they could replicate it. The poster did not know either, so he wrote about Pikachu's paralysis status, etc.
Speculation about what happened:
1. It was already known to enthusiasts in Japan in 1999 or before that you could die in front of a trainer to cancel a battle, and resuming it after facing a Pokemon would spawn level 7 encounters, but why this happened wasn't known.
2. By trial and error, they started testing around Rock Tunnel, where it's pretty easy to die in front of a trainer.
3. The trick was considered minor at the time, so it never became popular.
4. Eventually by trying to figure out what controlled the encounters, someone found Mew without knowing how it works, which drew attention to the glitch.
5. In the time between 1999 and 2002, people figured out optimizations. Like Flying in front of the gambler, which is much more reliable to cancel the battle instead of dying, and realizing that fighting a trainer always gave the same Pokemon.
6. It was not before the Youngster was found, which gave Mew 100% of the time, that the trick had any hope of getting mainstream. At the time there were lots of hoaxes claiming to give Mew, which only made people waste their time.
7. In some way, someone might have seen the trick in Japanese forums, or it was spread without the Internet until someone heard about it and posted in its final Gambler+Youngster form, which was not annoying to execute, and you could try it out if you were bored.
8. Eventually someone who was familiar with the game internals looked at it and thought it was reasonable, verified it worked, and the word spread.
I had a look at why fifth is credited for the method in Japan. It looks like the source of this is fifth himself. In his webpage, he claims that he bought the game the day it came out, researched it, came up with that trick after playing for a few years, and was only posting at the time because he did not know how the web worked. To be honest, I highly doubt any of this is true. While he does seem to know that the pokemon that shows up depends on the trainer you use, he doesn't seem to know that there's nothing special about the Gambler, any long-range trainer, of which there are many in the game, work. And a person who found the glitch by researching would know this. Also, his post is nine months after the glitch was common knowledge in the west. To me, it looks like he simply found it on the web, made a convincing tutorial in Japanese and claimed it as his own.
From this, I believe the best place to keep looking for the origins of the Trainer-Fly Mew glitch is to search japanese glitch forums and websites for a level 7 wild pokemon trick or something like that, but as I said before, my Japanese is terrible, and I'm not qualified to do it.
Now, assume, just in theory, of course I'm not trying to imply anything here, that when they show you the daily value of this fund, instead of calculating it from the assets there, they just use the following Python code, with a Gaussian of mean 7500 and standard deviation 3000, which is summed to the value each day:
The question is: what is the probability that this code will increase the value of the fund each day it runs? Do you think this pattern is reasonable for a fund consisting of an asset whose price fluctuates like this?
Please, don't try to connect this problem to current events. This is a purely academic question :D
Running it through VNC there's some horrible lag, though. But anyway, from this setup we have
(1) the xvfb process with the state of the virtual X server
(2) the container wine processes with the state of the virtualization layer
(3) setting up xscope between the container and xvfb we know exactly when input events are sent, and can pause it anytime
After some googling, I found the criu utility which can checkpoint processes and containers, maybe it's already implemented in docker itself, I don't know. It definitely looks like we can control input, video, audio and savestates. It looks like the best way to see if there will be any desyncs is to actually implement this...