I want to make some more points regarding "fastest version whatever it is" that are similiar to one I made previously since nobody seems to think the FDS ending is valid.
SMB isn't a dead game. Nintendo is still rereleasing this game on every new console they bring out and they keep releasing almost identical games like "Super Luigi Bros." or "Speed Mario Bros.". The latter only differs by framerate and text on the titlescreen and it beats both NTSC NES and PAL NES.
Ok, you can say that doesn't count, because the name of the game is different, but what if the next VC release runs at 61fps because of a typo or because the real framerate of 60.098... gets rounded up? Then NTSC is faster again and Happylee's old NTSC video just has to be sped up 1.5% for the new publication that obsoletes PAL NES.
Fundamentally I think beating a TAS or speedrun by using a different version is like using a different definition of "circle" to proof that you can indeed square it.
Unless you meant the -3 Ending that's already obsoletes the well... negative loop and is a different topic entirely.
The -3 Ending runs the game ending function and unlocks hard mode and world selection. The author of the initial submission in that branch actually wanted it to obsolete the regular run.
I don't want to go off on a general rant too much, but I want to briefly say that I see in some responses here the same imho toxic mentality towards ports that has long plagued the real-time speedrunning community. This idea that "only the fastest version matters, no matter how obscure or how shitty of a port it is as long as it has the Nintendo™ Seal™ of™ Quality™ on it". That has nothing to do with trying to optimize the game, it's no different from setting a new 100m record on a 95m track.
Can those who support PAL obsoleting NTSC despite it being bad irrelevant port explain why they don't also support the FDS port obsoleting both runs? The issue seems to be the same for me: Both ports enable new glitches that allow you to finish the game quicker than the original version.
I haven't tried every pipes systematically, so now I only know 8-1, C-1 and 5-2 (probably not for RTA). I believe there might be more (for TAS only), but it takes time to find a hidden route.
Edit: OK, after looking through the map and some tests, I might be wrong. There could be nothing more that can be used in an FDS warpless run to save time.
Thanks for the answer.
dhxo wrote:
So, you have been talking about that pipe glitch with the pipes coming upside-down. I am not very much used to the physics in this game because the slightly differ from the SMB physics. Could someone explain how this pipe-glitch works? I have seen it in the warped run in world 5-2 if I remember correctly, and there the pipe takes you to the area where the vine leads to. I don't understand how it works in 8-1 though.
Physics are the same as in SMB1. The pipe glitch works, because the bottom part of an inverted pipe is still a top part of a regular pipe, just with a wall above it. You can enter it like a normal pipe, if you get into the wall above it. There is no destination set however, so it takes you to whatever values have been set by previous pipes you encountered. 2 values matter. One decides which screen you get put on and another decides where horizontally you are put on that screen. The first gets reset to 0 frequently, but the other only when a (non-inverted) new warp pipe is loaded.
https://www.speedrun.com/smbtll#Any_D-4
Luigi is faster than Mario in certain categories for RTA. Why is this not the case in TAS as Luigi seems to be strictly slower than Mario in all circumstances?
Luigi is faster in the SNES port than in the original, Luigi can go a faster route in 1-2 (irrelevant for TAS as it can go through the wall) and 8-3 and difficulty/luck matter in RTA and Mario has a harder time overcoming obstacles. That specific category hasn't been TAS yet.
lxx4xNx6xxl wrote:
Copying what I wrote on the comments:
The responsible memory adress is 0x0751, which determines where on a screen you are put after taking a pipe/vine (while 0x0750 determines which screen you end up on). The adress is set to 6 after passing a pipe in 7-2 and doesn't get cleared until 8-1. While in the warps route it is set to 0 in 5-2 and stays there. If you set other values via Cheats, it takes you to other parts of the level and higher values crash or go to the title screen.
The only useful places are 8-1 and C-1 in the warpless category. I could get a Crash without Cheats by game overing in 8-4 with the value at 17, continuing and doing the glitch in 8-1.
Can savestates be made to contain a screenshot? It's very irritating to load a savestate and have an old image shown, especially if you load onto a lag frame and the image lingers for a while.
Config>Rewind & States...>check "Save a screenshot in savestates"
Can savestates be made to contain a screenshot? It's very irritating to load a savestate and have an old image shown, especially if you load onto a lag frame and the image lingers for a while.
I just saw that Bizhawk is banned for the ACE category on https://www.speedrun.com/smb3 for emulation inaccuracy. I can't find anything else on that topic there though. Maybe that should be cleared up?
I looked into the disassembly once to see how the Bullet Bill Cannon works. I don't remember the exact details and memory values, but here's the general idea:
The cannon has a counter that starts at 14 and checks every frame if it should be reduced by 1 and shoots the bullet at 0 (or stays there if Mario is too close to the cannon for it to shoot). The check works by looking if there is an enemy in slot #3. If it's empty, a bitwise operation (I don't remember what exactly) is performed with the 3rd RNG value to see if the value will be reduced. If it reaches 0 at that time (and Mario is far enough away from the cannon), the bullet bill is spawned in the 3rd slot. If the 3rd enemy slot is already full, it goes to the 2nd slot and takes the 2nd RNG value and if that one is full too it goes to the 1st and takes the 1st RNG value.
So the outcome can be manipulated by trying to change what enemy slots enemies go into by varying when you despawn previous enemies. But it's probably too chaotic to just directly calculate what to do.
If the bullet shoots left just before you get there, you have it almost right. If Mario is within the range where it can't shoot when the value reaches 0, you will get a perfect shot right as soon as you pass it.
Does someone know why the pipe at the end of 1-2 of Super Mario Forever sometimes takes you to the end of the level and sometimes to an underground section? It seems you skipped the trigger for the proper location, if it takes you to the underground section. But I can't see any pattern or logic behind it.
I had tested it thoroughly in 4-2 and couldn't do it through the stairs. But I think it could be done in other levels such as 1-2 at the end but I don't know if it's faster than going through the wall.
I guess you mean this:
But why would this work in 1-2 but not in 4-2?
EDIT: Is it maybe that you need a moving platform to get the right jump height?
That is the PAL version and it's known to have glitchy floor collision detection although I only saw it after hitting a block before. A PAL TAS could maybe use this in 4-2 to speed up the warp zone section.
This may be unrelated, but do any of these tests correlate with how accurate lag emulation is or could something like this only be really tested by comparing movie files on nesbot and on emu?
I had a glitch happen once, where the game lagged in 6-4 just as I reached Bowser and his two halves weren't aligned. I didn't think to test, if his hitboxes were actually shifted or if it was just graphical though. If someone can recreate this and it is the first case, it would confirm that lag can cause more than just graphical glitches at least.
I can't quite spot the possible improvement you are talking about.
Can you help me out here?
If you can clip into the ground like that at the end of 1-1 in the NTSC version, you can save time with a flagpole glitch similiar to the bullet bill glitch in 8-2. If it's PAL-only it doesn't matter, since you can already do that flagpole glitch on PAL without going into the ground.
I just console verified this on the first try with (a slightly modified version of) micro500's bot. Maybe FCEUX improvements in the last few years have made it possible? I was using version 2.2.1 for Linux to strip the lag frames. What are the steps for getting this marked as console-verified?
Full disclosure: the second time I ran the movie it desynced on 2-1, and the third time it desynced 3-3. I think that was just from ESD or insufficient decoupling or something silly like that.
The input file remained the same, so fceux's development doesn't matter.
Lag frames in this game are easily visible since the text at the top of the screen gets distorted for several frames. For the the movie to synch correctly, there should be no lag frame at all, simply stripping the lag frames from the input file would not really mean synching it since it would then be longer on console than on the emulator or movie.
The lag frame in 3-1 under the new ppu can be eliminated by shooting the Koopa a frame earlier. This movie is the exact same except for that and it synchs under the new ppu:
http://dehacked.2y.net/microstorage.php/info/1091863116/newppu.fm2
Can someone try, if this one synchs on console?
I liked the run, because it showed off the glitch to get an overpowered Pokemon early without the battles getting stale as they would in a proposed "all badges" run.
I don't think this should get obsoleted by a run that only forgoes save corruption and still warps to the Hall of Fame without fighting the elite 4+rival.
Can you tell me how much leeway you have for the bullet to shoot in 5-2?
If I calculated it correctly the last frame the bullet can appear on (one pixel outside the cannon) to save the framerule is 48552. Is that correct?
But why would this work in 1-2 but not in 4-2?
EDIT: Is it maybe that you need a moving platform to get the right jump height?