This is Emerald and not R/S, but I had a bunch of questions on how the "Access Pokemon beyond the sixth slot" glitch in the current submission worked.
From the submission text:
The "new" Pomeg Berry Glitch
-Make a team with 3 to 6 Pokemon. Your first Pokemon must be alive, with HP EVs and 1 HP. You'll give him a Pomeg Berry to KO it later in the process. Your last Pokemon must be alive, preferably with full health. All the other Pokemon (between first and last) must be KO or Eggs (considered as KO Pokemon because they can't fight).
-Go make a wild battle. In this battle, switch to your last Pokemon, then flee. The game will update the slot of the last Pokemon seen alive in Battle with your last Pokemon's slot.
-Deposit the last Pokemon. Now, there's no Pokemon on the slot previously memorized by the game, but your team still has an alive Pokemon remaining.
-Use a Pomeg Berry to KO your first Pokemon (and last alive Pokemon). Now your team is full of KO Pokemon.
-Go into a wild battle again. The game will first check your team to see if he finds an alive Pokemon. If he doesn't see one, he will launch the "Pokemon" in the slot of the last Pokemon seen alive, as in normal gameplay, it is automatically an alive Pokemon. But here, there is no Pokemon into this slot. The game will still follow its script and interprete the zeroes as a Pokemon. Thus, a ? will be launched, with no moves, 0 PPs, 0 HP, Lv 0, ID 0,... This Glitched Pokemon is called Decamark because of its summary name being ??????????.
-Go into the Pokemon selection. The game is puzzled : he found out that there are 4 (for example) "concrete" Pokemon into the party, but he also sees that there is 1 battling Pokemon, and 4 Pokemon in reserve. For now, the Pokemon selection pointer is set to select 4 Pokemon, so the last Pokemon of your team can't even be selected.
-Check the Summary of one of your Pokemon, and go back. Due to anti-cheat scripts, the game will now consider the Decamark as a "concrete" Pokemon. But under Decamark's data, there's a wide RAM area full of zeroes. Thus, the game will also interpret these zeroes as Pokemon, and the number of Pokemon the pointer can select becomes higher than 6.
-Push Up or Down to move the Pointer on the Cancel button. A Glitch happens here since the pointer is still selecting a Pokemon, but he's also selecting the Cancel Option. This makes the Pointer position glitch, and teleports it way down into RAM.
- Why do you need at least 3 Pokemon? What happens if you attempt to perform the glitch with only 2 Pokemon in the party, one Pomeg glitch'd to -1 HP and the one regular unfainted Pokemon?
- It looks like a major initiation point of the glitch is when the game figures out who to send out at the start of a battle. The submission text says that the game first checks the team, and if that fails, then checks the data of the Pokemon last sent out. But what's the point of the second check entirely? In normal gameplay the first check should always succeed.
- Similarly, I've performed the Pomeg glitch in the past, and what makes the 'regular' Pomeg glitch sometimes send out a Pokemon with 0 HP instead of a ??????????, and sometimes send out an Egg (for the competitive version of the glitch)?
- Two lines in the submission text mention the term "concrete" Pokemon; what does that mean? I have never heard this before.
- I'm guessing that pressing Up from the top of the party onto the Cancel button somehow forces the game to also select the "-1th"/"255th" Pokemon in the party, which is what teleports the pointer into the PC data and begins the corruption? Is this accurate? What does it mean for a pointer to be simultaneously selecting a Pokemon and the Cancel button (how does this work internally/in memory)?
- Below the part of the text I've quoted, it says that the party isn't healed after blacking out (allowing the Bad Egg to be sent out) because the Pokemon selection pointer has been glitched. How does the selection pointer have anything to do with the game being unable to find the party Pokemon? The top of the party should be in a fixed hardcoded location in memory.
I would greatly appreciate any help in understanding the mechanics of all this better. Thanks!
Not sure if this belongs in this subforum; feel free to move as needed.
I'd like to nominate 2504M (Pokemon Gold) for the Obsoletely Fabulous page under the "Obsolete Tricks" tab. It easily fits the "show neat tricks and glitches that are not present in their current equivalents" criterion, using a very visually entertaining map distortion glitch to skip past barriers to the end, without the arbitrary code execution glitch warp the current publication uses. There is also a sort of precedence as there is another Pokemon game on that page that completes the game with a glitch not used in its own current publication.
I'd also like to nominate 4563S (Super Mario Bros. Enhanced) and 3273S (Kaizo Mario World 2) for Gruefood Delight under the "Hacks/Homebrews" tab. For the former, it fits in about the same way as Full Moon Mario on that page, being a run that itself isn't bad at all in terms of optimal play but uses a not-very-noteworthy hack. The latter is almost self-explanatory; Kaizo 1's already there.
^The term "hell's portal" I believe actually refers to a very specific type of glitch where you do get flung very high into the air like in a heaven's portal but never fall back down, making you stuck ("hell") until you press A or B which results in death. The type where you slide kick too late for the endless stair skip, then fall through the floor and die is just a wall clip and not a "portal".
They're quite rare and unfortunately the only source that listed several of them (the SM64DS Glitch FAQ on Gamefaqs) is now deleted. I've pulled it off once before in V1 only, through the cannon on the opposite corner of the course in Bob-omb Battlefield. Aim as far down and to the left as possible (into the hill/wall), then fire for a heaven's portal. Aim very slightly away from that for a hell's portal.
This is the only video I could find, and unfortunately it doesn't show the touch screen (Mario's icon jumps between the corners of the screen in a strange way while "falling") or much of the full result. But you should be able to see that Mario is very high in the air, and he never comes back down.
https://www.youtube.com/watch?v=eefRFuQKf4k
ALAKTORN wanted his name to be first because he found the glitch which made the run possible, I respect that. Saying he improved 50 secs of the 52 secs? That doesn't sound right to me.
WOW. I stopped following the SM64DS thread for a time, and when this was submitted I decided to watch the encode first without reading the submission to see where the large time improvement came from. The BitFS skip totally came out of left field for me; I was like "You're getting 2 extra stars?...Okay, you're entering JRB, how's that useful? I suppose you could go OoB into BitDW but you already can skip that...wait, what?".
Love the "3 stars is faster than 1 star" incongruity and the fact that this run now uses 2 skips impossible in the original N64 version. Yes vote.
Encountering Missingno. is unarguably memory corruption since doing so writes garbage to the Hall of Fame records and writes/flips an unintended bit in the inventory to give you 128 more of your sixth item. Not sure if performing the Old Man glitch alone, or performing "vanilla" Trainer-Fly could be considered the same as the game is only reading garbage rather than writing it. No idea about the Brock Through Walls and Safari Zone walk through walls glitches; someone with a more detailed knowledge of what the game is actually doing behind the scenes could probably clarify better.
Edit: Come to think of it, watching the Old Man's tutorial at all will write the player's name to the "wild Pokemon on land" encounter data. The developers are simply assuming that it's impossible to encounter wild Pokemon on land without passing through an area that overwrites that data. Would this be considered writing garbage, and then would talking to the Old Man at all be automatically considered memory corruption?
A side note: This hack's title should be "Super Mario Bros. Enhanced". Peter_ac changed its name a few years ago to avoid confusion with the GBC game.
Sources:
http://rh.pac.am/hacks.php His site (links broken, but you can compare it to a web archive and see the name change)
https://www.youtube.com/watch?v=0Kxv5NpdcQ8 Youtube video on his channel
http://www.smwcentral.net/?p=showhack&id=5875 Updated download link submitted by him citing the name change
The only change I believe is to the title screen, so everything should still sync.
Looks like a "holding onto null sprite"-related glitch?
I smell SMW-style hyper-corruption (stunning, ACE) glitches in the future if this gets developed further.
Shouldn't there be something about the newly discovered glitch and maybe also the trick of ending input early in the description? It's a fairly strong glitch (the NES Tetris equivalent was also mentioned in its description) and deserves recognition rather than making it just look like a time optimization improvement.
^I believe the glitched Bob-omb clone's backwards momentum is needed to get off the pole, and there are no Bob-ombs that can be reached at the BitFS pole.
^I believe that emulator run made use of memory watch to make sure all the sprites were at the right pixel positions, which would make it technically tool-assisted.
Voting No for Vault since it doesn't play on hardest difficulty. If the AI still behaved this way on the hardest difficulty and a nearly identical movie were submitted I would vote Yes.
That's not the same question; I earlier asked if the specific TAS shown at AGDQ 2015 would be publishable, to which the answer was no. OmegaPaladin is asking if a (presumably more optimized) newly done TAS of SMW-SMB1 would be publishable.
No. By an extremely rough calculation, the N64 has 14 buttons, so on every frame there are 2^14 = 16,384 possible button press combinations. Super Mario 64 runs at 30 frames per second iirc, so in one second there are 16384^30 possible movie files, which is a number over 100 digits long. Even if we got a supercomputer that could process a quintillion movie files per second, it would still take several googol years to try them all. And that's just 1 second. Brute forcing just isn't possible.
Thanks for the explanations everyone; makes sense. Although I was vaguely looking forward to being able to read complicated explanations and submission texts of how everything got to work. For example, I recall Masterjun mentioning that they tried doing SMW-SMB1 in 2014, but there was some memory(?) limitation on how much code could be pumped in. How was that overcome?
Not sure if you've already used this, but:
http://www.gamefaqs.com/gamecube/516492-super-smash-bros-melee/faqs/17242
Here's an old SSBM HRC guide on GameFAQs that has non-Youtube video links to the world records at the time, as well as links to external sites (scroll down to Section 11). Some of them are surely broken by now, and I haven't thoroughly checked against the current records, but some of them might be useful for your compilation.
Are the SMW and Pokemon Red AGDQ TASes being submitted officially to the site this time around? I'm sure no one would object to them obsoleting a technically shorter run.
Apologies if this question is cutting into anyone's downtime as I know TASbot AGDQ was a monumental effort and a lot of people are taking breaks.
