Firepower will become increasingly relevant as you reach situations where its more convenient to have a bomb land farther away from something, or hit two things at a time. don't forget the rolling bomb either, although i suspect it would only be of use for mass enemy destruction (ie. 100%). I think each stage needs to be heavily focused on as you're in it, and a different save file for after each, until you really get the hang of things.
There was a time I considered a 100% console run of this game but seeing as how it would have to be SS, it's difficult and very tedious in the long run, in addition to needing big boss skills. I think that without any real glitches or shortcuts that an any% tas of this will just wind up as not too interesting (though I have yet to see later stages where things must be collected), whereas a 100% TAS would be something to go for. This is one game I wouldn't mind TASing I suppose. Maybe I should try the 100%.
This guy also made a youtube video of getting behind the upstairs key door by BLJing on the highly sloped lobby ceiling directly above the door, if you want more evidence that he's using a code that allows standing on impossible slopes.
I think a camcorder video like he's been doing, in which the full N64, including the front of it and the 3 spare controller ports, the cartridge in it, and the cord that goes from it to the TV (need front input), and possibly the controller and controller cord, are all within the video frame at once, while such a trick is done, would be proof enough. If he stalls on that, then you know for sure...
Excellent. I have been looking forward to any new TAS of this game. All temples in sub 3 hours is a most interesting proposition and I am about to watch it :)
Oh, and about that key door thing, I remember that once the first key door did open the upper lobby, but it was graphically bugged. So it may be possible to do something about it.
[URL=http://www.abload.de/img/182stars17d.png]182 stars[/URL]
Every star has a number. Collecting a star with the number 7 while the current stage value is set to 17 (Bowser in the Fire Sea) gets you a normally unobtainable star.
There are still mistakes in this chart, as not every stage has a canon, but I'm too lazy to look it up and correct it. ;)
Keys are stored before stars in memory, so there's no way we could glitch keys by changing the stage value and then collecting a star. The key byte looks like this:
Bit 1 - ? Used
Bit 2 - ?
Bit 3 - ?
Bit 4 - ?
Bit 5 - Has Key 1
Bit 6 - Has Key 2
Bit 7 - Key 1 Used
Bit 8 - Key 2 Used
Using a key unsets bits 5 or 6, like previously mentioned by CBright.
@z0MG - That means I was editing the post, sorry for being unclear.
Random comments on image: I assume you meant to give the cannon bit to WMotR and not the aquarium. Also wondering, although it's totally irrelevant, whether bit 8 gets set for lobby/outside's cannon, or if the grate's occurance is a function of whether stars (>)= 120.
I doubt loading a game with current stage set to something other than 0 would do much. Loading from menu / game over does a few special things, including defaulting extra life count and spawning you at the beginning of the castle path, that no other activity can, and I'm guessing a routine covers this. However you now have me thinking of ways the intro cutscene could be avoided (fruitless so far, and that's just conceptually).
I wonder which bits correspond to which secret stars because I know during development they had more courses planned and that's probably what that extra room is there for.
So, it looks like things with coins is out.
What someone needs to try now (I'd do it but don't have the knowledge), is making the game perform an opening on a door that's already opened (has the bit set), to see if it overflows and sets an adjacent flag, thus unlocking a different door. Although I'd guess it flips the flag out of code efficiency, it's possible that it adds 1 to the flag since a situation where a 1 flagged door recieves the opening treatment again would never happen in normal gameplay.
180 stars? So have we determined if some of those extra flags are for other non-star conditions like bowser 1 defeated, etc? If so then having the first or second big key might not be a memory state in itself, but having defeated bowser 1 or 2 is...
It's good that information is shared and discussion occurs. I'm going out there with ideas because, well why not explore everything? I've messed around with mupen and toad's tool a bit in my own search for feasible things like lowering star count in console runs (to no avail so far, granted). Memory corruption doesn't look like it's going to happen; if it is, it'll probably have something to do with a stage with infinite coins, since that data is probably close to the stars and progress flags in memory, if not directly adjacent (recall the screen that shows coin counts). Then again I'm not familiar with 255/1000 coins and how the overflows for that work, or how they behave in different types of stages.
Not to change the subject or anything but I found the key thing fascinating. It's worth noting that in the past I used the 0 star run up till where bljs are used to launch downstairs to the basement door for entry directly out of BitDW. By modifying the path slightly I made mario travel through the doors into the black area behind, about face, fly back through, up, into the lobby, up the stairs, and stopped him in front of the 2nd key door.
My hopes were that simply being near the basement door and possessing the first key would be enough to transport the game into a status where a generic "key_door" got permission to open for the first time. I could then manipulate mario to actually stop in front of the other door and hopefully trick the game into keeping its "why-yes-you-may-open-that-door" status. Unfortunately this did not work.
I'm still not entirely convinced that a door can't be tricked into unlocking. This is probably tested easier and faster by getting 1 star and then trying to use WF as a trigger to get the CCM door open. If you think about it, all the star doors have identical properties to the key doors, with the only exception being they don't lead to another map... there's an initial locked state and a particular set of requirements, probably stored in the door itself as class variables, which is required to make the door open for the first time... after which it behaves normally.
This tells me that a) all doors are seperate instances with unique criteria to initially unlock
b) once a door is open it permanently assumes an unlocked status
It appears we're going to have to try some sort of override or memory leak to either generate the key we want, or make it so BitDW gives us key 2. The other alternative is to find a way to trick star doors, perhaps using other, already-qualified star doors, into opening. we do that and we have a no keys run. if you can trick the CCM door, you can trick the upstairs one too.
Call me lame but I love watching simple TASes of games like this with finite possibilities. It may be a bit boring because of what it is, but the work behind the scenes is identical to a TAS like Panel Du Pon, which just happens to have quicker action...
Nice one for April 1st.
I've watched every published TAS ever made of this game, and that WIP kept me on the edge of my seat more than any so far have. For the first time in a long time every movement, route, and strategy appears to be the very fastest that could ever be possible. You've outdone yourself and N64 TASing in general... keep up the excellent work.
Its starting to seem like improvements are being found, and the run re-started at a rate greater then the progress in the run itself. It stands out because it is almost certainly the published run with the greatest amount of known improvements.
It will be a real thing of beauty when it finally does reach completion, but the pursuit of perfection can be a bitch sometimes.
Provided more than one person here is dedicated and intensely knowledgable about both the game and all these bleeding-edge discoveries, it would not be unrealistic to reach a point where you merely continue to completion rather than restart. Please don't quote the list of reasons because I don't want to hear it; the real world is a series of ever improving changes and although the possibilities of input inside a game are not endless and there is one theoretically fastest set of input, this is not an obfuscated C code wankfest and it's unlikely that at any given point our current best knowledge is truly the fastest way.
A TASer could knowingly shut himself off from the forum as a source of any new discoveries or information that would require a restarted run and create what is, in his personal microcosm of time, exposure, and known tricks, the fastest possible run, and this could be an acceptable improvement to the existing one. As it is now, the habits of excuses not to publish improvements, despite how the current posted run got where it is, is literally preventing this run from ever progressing. The authors involved in the previous record SMB movie were fully aware that at any time in the future a faster way could be discovered, but they produced the most perfect movie they could with their knowledge set. Because of the way this works you can extend that concept and cross barriers of time related to completing this particular run before an improvement is discovered. Until the run is completed we turn the TASer into schroedinger's cat and, logically anyway, satisfy everyone involved. While at the same time actually making progress of a meaningful sort rather than bending over and shitting out new tricks every week that drive the movie author insane with restarts.
Let the flaming commence; I just want to see this happen, and it's not going to as long as all you TAS geniuses (not sarcasm and not an insult) keep working on it and accidently discovering better ways to do things :)... take a break from it for a month or something, or do the opposite and deny our prospective movie author any new information, and create a movie that is perfect when rules of scope are obeyed!
the records for stunt courses are pretty high! I'd even say they're close to perfection.
You can do 4 or 5 stunts on a normal jump and you think they're close to perfection? The depreciation of points for repeated stunts is the only thing holding you back from doubling every one of those scores.
I do have the US manual somewhere. I can look next time I'm at my parents. Had a big box with MAD magazines, Nintendo Power magazines, all my manuals for SNES and N64, and all my players guides, from back in the day. I specifically remember getting the impression there were 4 Megas but I don't know if it was from the manual or from the score grid. Actually I think it was Nintendo Power that asked in one issue if anyone knew about the other 2 megas.
I look forward to seeing this TAS. If this game wasnt so overly repetitive I'd make a normal speedrun of it. It would be handy if the game had a code built in to unlock everything instead of having to use a hacked SRAM.
I like how the instruction booklet says there's 4 hidden stunts. Did they never make the last 2?
What good is Transform for this purpose on your own Pokemon? It's the opponent's special that you draw from in battle manipulation.
The point is to have Transform so that an enemy can use it with Mirror Move, to see if this is faster than getting to an encounter with a Ditto. Enemy goes first, we use transform, it fails, we switch to a pokemon we want, the enemy uses mirror move and does transform. The enemy using Transform means they borrow the special of a pokemon we switch to.
Also, AKA, keep in mind, you went across all of hyrule to get that 3 second gain. In swordlesslink's TAS I think there aren't going to be any segments where he needs to backwalk for that long, he would probably do a superslide. Sidehopping does have a much faster initial velocity that makes it better for short distances, since backwalking takes a few frames to set up and sidehopping is still superior on any slope.
Would need to be tested on a per-case basis, which he'd probably do at least once anyway. Just because of the nature of side hopping and the way it changes speed on different terrain, it should be somewhat obvious to test in every situation. Isn't it entirely possible, based on the up-down terrain sloping you encounter across most areas, that you might gain or lose several frames just by starting your sidehopping at a location a few pixels closer or further from the intended destination? Totally situation dependent, hybrid might be the fastest approach to try for.
Wouldn't it just be easier to upload at the end of each course?
I actually really like the single frame teaser screenshot. He could do that once for every level and still keep suspense at maximum. WIPs destroy suspense though...
If you mean him going through the door, thats because there is a small amount of programmable space behind the door for mario to walk into.
I know, but it looks like he modified the collision map of the door object so that he can walk further than normal through it. normally there's no way to get as deep inside the door as he did... even with hyperspeed there's a frame where you simply move through the whole thing rather than having the possibility to get frozen inside it. that just seems to work on the wood doors :\
mr_roberts_z wrote:
Somehow I doubt that's...possible.
Anyway, 120 Star update:
There will be a lack of updates. I've elected to not put up any WIPs, because I think the methods, routes, and most importantly, time saved would seem better in one full run. With WIPs, you've seen one, and then when the run is submitted, you fast forward right to the end to watch the final bit you haven't seen...I think that in the end, the run will be much better if people have no idea what's coming from the very beginning (and that will be the case; the JRB WIP from before is basically useless, and even getting to the castle is different).
Therefore, I'll just put up the amount of stars done (possibly in the form of screenshots) once in a while, so those interested can know where the progress is at (if lots of people object to this, I'll skip it).
The only reason I'd ever object to that would be that people who watch your WIP can give you ideas or suggestions of improvements before you've made another half hour of movie. You're one of the people skilled enough at optimizing this game that it doesn't really matter. Maybe a single early WIP after about 5 stars and then, when oyu're 2/3 in, a video of one or two impressive stars from anywhere in the video, on youtube, then nothing else till the full run is done, so we've only seen, for example, 7 stars but tehy weren't all in the first course.
I'm almost positive the door was modified to allow getting partially inside it somehow (like MIPS without the MIPS). Attempts at recreating the trick don't seem to work in the same way, and the game was so heavily modified I wouldn't doubt he messed with that. The doors appeared to be different in more ways than the texture used on them.
Those with a sizable amount of RAM could easily utilize some sort of auto save state functionality. You'd turn the feature on specifically for frame advance TASing. There's nothing wrong with the idea of having, say, 10-20 (user adjustable quantity) automatic save states created every n frames, where n is adjustable from 1 to 50, whatever the user wants. To 'rewind' while the game is paused, a hotkey is pressed, the current state is discarded, and replaced with either the immediately last savestate made, or perhaps a menu would allow selection of others (5 frames ago, 10 frames ago, etc). The user defined save states 1-9 would work as normal, and you'd be advised to save state before rewinding too far. What's wrong with that? Even storing the states entirely in computer RAM, your biggest system RAM to emulate is N64 with expansion pack, 8MB of RAM per savestate. Inefficient, yes, but we could still deal with it.
It is also worth noting that properly managing your savestates, and/or simply having more savestates, would make this unnecessary. The idea is that you have, at any given time, a slot saved within the last 30 frames, which is honestly not that hard even with 9 slots (depending heavily on experimentation level and the game itself, granted).
