OK done with "labyrinth" stage I managed a 1 frame improvement
http://rapidshare.de/files/29008828/Sonic_the_Hedgehog_1_ALL_EMERALDS.gmv.html
I gained one frame in three places but lost two on the jump to the penultimate button since Sonic kept hovering over for longer than he should and don't even mention the moving water platform jump, but still one frame gained is like one second gained.
TSA, I think you need to calm down since you're the one whose getting all worked up about it. I'm just happy that Guano finished it, you have to remember that the whole Kokiri Forest/ Deku Tree stuff started out as a testrun. I think you just seem annoyed that this run was "sneaked in" before you could finish yours. Anyway an improvment is on its way and hopefully its possible to break 2 hours. One final point, not publishing this run because of more up to date tricks being available would be of the upmost stupidity.
With optimization, a 100% run would be maybe an hour, to an hour and a half longer.
I figure the Skulltulas could be collected in their respected areas. And getting Epona would make this run quite a bit faster I believe. Definately faster than walking backward and such.
No way hosae could you ever do a 100% run in 4 hours, since a lot of the sequence breaks are negligable. I would be amazed if 6 hours was broken for a 100% run.
I don't think you could collect every skulltula in its relvent location, the duplicating trick probally saves about 20 mins in a run (even I personally disaprove of that trick) . There's not that much distance to travel with Epona, you could certainly get him very quickly without the need to race Ingo. It would simply be quicker to back run; whip out a bomb; shield drop and then do a superslide, which was something that Guano never took advantage of in his run.
AKA, now you know what I'm dealing with in OoT :) Buttons wont get pressed when you want them to. Directions get off by just a hair. Its annoying and very complicated.
I think its because even thought the game runs at 60fps it only takes input at 30fps, which either means delaying for a frame for button to register or the same button or movement being duplicated twice. I'm not sure if this applies to the Gc version. I actuallly think the Gc version is 1% faster than the NTST N64 version, however the pause lag does cause a loss in time on the Gc though.
Although there's no such emulator, there has already been a TAS of a GameCube game.
I've seen that movie and it was very cool, but it could certainly be improved, I think Captain Falcon would be a better choice overall. The movie didn't use the easy giant DK win strategy, but I guess the movie tries to go for a WTF factor than anything else. The furthest I've ever got with one life is the metal bros stage. I managed to get Mario on 350+ damage but he still wouldn't die. The furthest point I've manged to capture on video is me getting up to the Samus battle with one life, but its more about manipulating AI than being skillful at the game.
I have to admit when I was using frame advance in the the open sequence the timing seem to go like this when I pressed a button
A = *,A,A
A,B = *,A,A,B,B,A,A,B,B
It was also a bit anoying when I was try to roll since I held a direction and pressed A, but nothing happened. You must have to hold down a for longer than one frame for it to register properly.
100 skulltulla run would be very boring to watch as you would just use the endless skulltula collecting technique. Anyway it should be part of a more wider 100% run.
I highly doubt that. At least not a *generic* tasbot.
It may be possible to build a program which is highly optimized specifically for Mario 1 and uses algorithms and heuristics created for achieving the goals in that game. However, it's just not possible to create a *generic* bot which could complete a game (any game) in any feasible way. Heck, it would be really, really difficult to even make it search for the optimal play for 10 seconds of playing time. In fact, even if optimality is not the goal but just trying to search the goal by any path, it would still be very unlikely.
That's because the number of possible combinations of button presses increase exponentially in consecutive frames. To get an idea, let's just assume that it's enough for the program to test the 4 directional buttons to reach the desired goal:
*brackets mean example of how it appplies to a game
A Tasbot would know all the specific rules a certain game and like I said earlier it would be designed like a chess program, when a chess program is playing it would not think of every piece on the board and decide where it could be put and then say to itself "Oh darn I can't do that since its not legal". It would know all the legal moves it can do(game controller actions) and know where the oponents pieces are and how they move and where they can go in the next turn (walls and enemies and moving obstacles). So if a high value piece e.g. queen is being threatened is not going to go through every possible option of what it and the oponent could do since it obvious that the oponnent will take the queen next turn if it can do so. So the options will come tumbling down to either
1. Move queen to safety
2. Take piece thats threatening the queen
3. Threating the other queen
4. Put the oponent in check
5. Block path to queen with other piece
6. Take oponents queen
That would limit the situation down to about 15-20 moves and of those moves its likely that only 2 or 3 would be reasonable instead of searching through the
75068352983405739055035602592843 moves that you were sugesting. Of course the huge difference between a TASbot and chess program is a chess program tries to limit the number of reasonable moves that a oponent can do so to make the next 20-30 moves a little easier to compute. For example if would be completely unreasonable for a compute to test out possibilties such as left, right or move left at the start of a level unless there was some form of interaction that would dramatically speed up the level e.g. bouncer at the start of Spring Yard 2, but yeah a doubt a generic TASbot would be coming around anytime soon.
Couldn't (in a racing game) the game use the distance from the finish line and the velocity of the vehicle at the last whole frame to calculate centiseconds or milliseconds for the final time?
The questions is in the title...
So is one available since there is a lot of games that could be TASed. The one game I would like to do is the Super Monkey Ball, since it would be funny to see someone suicidally completing every stage, but you could also bet your life on someone wanting to beat Metroid Prime in under an hour.
[Title edited by Bisqwit; If one uses "search" for Game Cube, they wouldn't find this thread if it had only "Gc" in its title.]
I keep forgetting to bring this up but backwalking is not Links fastest form of movement, the fastest way to move is to switch between the sword and an item that can be equiped and unequipped during backwalking. These are
Deku stick
Singshot
Bomerang
Hookshot
Bow
Guano ommited this tactic since he couldn't figure out the timing and it would inevitabaly mean using frame advance for an entire backwalk. I would definitley want to see this as it would probally save about 2 mins over the entire course of the game and would be a little cooler and add to the WTF factor. It would be easy to do since all you do is find the correct line for the backwalk and then re-record using that tactic in frame advance. It would probally make backwalks a little more interesting since it can be done with seperate items. The reason its quicker is because Link gets given a mini boost every time he draws his sword.
TASbots could probally do a game like Mario 1, but I'm not so sure about something like Sonic, which has infinitely more possiblilties of movemenTs. It would probally be more conplex than a chess program and it would not look for glitches. I don't think it would be any better since the developers would snap this up, which effectivley means fewer no testers for games. TASbots would probally have to be programmed for each individiual game since it would have to know what the exact goal and route was, but yeah if it has to look for glitches then it would take till the year 3000. The way I've always thought of it is
Human = dum
Computer = dummer
Human + Computer = deadly combination
So basically I think humans using tools will always be better than a computer testing 100,000's of possibilties through a 5 minute game.
Actually Xebra was right and you were wrong and so am I. Although I've always questioned why the Sega translators spelt it as "Labyrinth" as opossed to "Labarinth".
EDIT:
xebra wrote:
At the beginning of Labyrinth 1, at the first place where you enter water, there is a switch you have to hit just before a gate that is blocking you. While the gate is rising, if you jump into it just right, the game will boost you ahead. I think this can save 10-15 frames tops, if even that. Good luck!
The boost seems non existent can you show me or is there a WIP you have of it.
The only question I have about the run is on the bonus stage you visit for Spring Yard 1. You seem to get hung up on several of the corners, is there a reason for this?
That special stage is highly questionable I know and will probally see a marked improvement if either me or someone else wants to do it but the trouble with the special stages that everything apart from the bouncers seems to be square based collision system which can often lead to huge inconsistencies of how sonic moves and re-acts, which is weird because I exploited a programming error in Marble 3 but cutting a corner of a circular based collision on the jump before going into the wall where the extra life is which saved about 10 frames. The special stages also jerk into the new postion every eight frames which can make seeing into the furure a little tricky. All of this leads to immense frustration, you also want to avoid jumping as much as possible since it loses about 90 frames for doing so. I'm not going to re-do that though unless someone else does it and can hex edit it in. Anyhow I think the other 5 stages are near enough optmized espeacially the first one. If I can just somehow avoid becoming another morgue TASer on the Labarinth stages.
Wasn't Bisquit try to encode it?
When I tried to encode a different AVI it ran at 10 fps when I was capturing it and then it cut out and wouldn't playback once 5 mins had been exceeded and only a few of the codecs work. I hope this gets uploaded to archive.org because I can't stand BitTorents or not being able to frame skip or rewind with Mupen.
How come no one appreaciates the view from behind, if you're watching carefully before the backwalk is lined up you can see exactly where Guano is going, you just have to use your imagination to work out where he is. If thats too hard for anyone just have a look behind Link to see if you can see obscure and funny things happening during the backwalk, but seriously backwalking will never be ommited for anyones sake since it probally saves and minimum of 5 minites. Its not like look down in Goldeneye which saves only 1 second for every minute.
OK finally done it
Here's the WIP for Spring Yard 3, I added in some of my own stylistic choices and collected one less ring which to me qualifies as a better level, but no improvement.
Here's how it went
I lost one frame on the skid before the invincibilty and speed boots.
I then gained two frames on the jump to the button, but because of the setup of the moving stairs before robotnik I had to use a delayed jump into the bouncer before the stair section
ENJOY !
http://rapidshare.de/files/28523231/Sonic_the_Hedgehog_1_ALL_EMERALDS.gmv.html
Now to shit my pants on the (Fatal) Labarinth stages.
I also put this up as well just to shown that nothing is ever random
http://www.youtube.com/watch?v=CtoEgl1akU0
Hah I checked
Guano from Power on till last text box disapearing took
12991 frames
I used frame advance and got
12967 frames
Difference is 12991 - 12967 = 24 frames
And time wise 0.40 or 0.3999999984 seconds. So yep whenever any text box appears use frame advance which is actually
B,*blank,A,A,B,B,A,A,B,B and not A,B,A,B,
Since buttons seem to be active for two frames and not one. I realised as soon as I tried to lleave Links house that I would never be able to TAS this game on a keyboard.
Question: How do you toggle between frame advance and slowdown?
AND when is Mupen getting updated, a resume recording feature would be good for a start.
Guano, I've done you a favour and have optimized the opening sequence for up to the point where Link jumps out of bed, it'll save you a few minutes of heartace and will allow you to jump straight into action I'll update the ruppee route wiki now.
http://rapidshare.de/files/28316973/THE_LEGEND_OF_ZELDA__USA_.m64.html
Remember to rename it to prevent other people thinking that it was me who done the whole run (lol).
EDIT: There it is updated of course slight variations showed be tested around the reccomendations
http://mwlin.freehostia.com/wiki/index.php?title=TAS_Corrections#Kokiri_Forest
Might as well give an update since I haven't posted in a while (this topic). I've have spent well over 1200+ re-records on this stage that I thought would be relativley straight forward . I'm currently at the boss on Spring Yard 3 but I've encountered some annoyances during it. Firstly the obstacles are timed in a way that forces you to wait so there is a limit to how fast you can really be. I lose a frame on the on the skid just before you're meant to get the invincibilty and speed boots. this is because when me and JXQ skid my character always floats for one frame longer despite the fact I synced with his run for 5 seconds so I could overcome this. However, I gained two frames on the jump to the button imeadiatley after that, I could very nearly land on the button without jumping if sonic stayed in the air for one frame longer, but during the penultimate jump to the robotnik section, the platforms are a one frame further away from me since I'm a frame quicker, this forces me to alter my jump and because of that I lose two frames to JXQ and the lag before Robotnik appears two frames earlier for him. So my point is, is it OK if I accept a one frame loss and move onto the (fatal) labyrinth since its destroying me.
p.s. I'm actually on Robotnik, but its unlikely that I can do it one frame quicker.
Pretty nice commentry JXQ, but its best to say something rather nothing as I ended up skipping some of it and had to spend a while working out what point you were commentating on. Getting the bombchu's early in Spirit saves at least 15 mins. It would take me roughly 1:35 to become an Adult without heading to Spirit first, as opposed to GuanoBowl becoming an Adult in 1:15,the seam walk you saw at Hyrule castle can be easily replicated on a console with practice and the bombchu trick across the moat was a bombchu jumpslash (pressing Z+R and then pressing B) is also easily replicated on a console with practice. I look forward to hearing what you have to say for the Adult stuff. Guano does have a save state somewhere which starts at the beggining of the Adult adventure if you spend ages looking for it.
Bottle Adventure? I couldn't find anything about it in
mwl's link.
BA is very complicated and version specific 1.0. it requires the rod to be stolen, it simply writes the light arrows to B. I think Acryte has mentioned something complicated where the light arrows written to the inventory and so would would enable Link to go straight to Ganon's castle
Is there a way to edit that Zelda trick page there is a much better way to do the "fish holding down a switch trick". I tried to emphasis that it could be done without the bottle dumping animation.
Simply
1. Empty fish and capture
2. Back flip and press bottle fish button in mid air, then press the B button
3 . Link will land and hold the sword in his mouth and it'll make it a fully playable ocarina
4. Press B again to cancel playing ocarina animation
5. Door or event will stay successfully open
