I understand such brute-force calculations can become very computationally costly. There's 15 gems in the first level, but really only 13-ish such that have to be considered separately. But if the level had 20+ gems, would it still be within computationally reasonable range? Or is it not as bad as I'm thinking?
The total number of possible paths (on the order of 20!) is staggering, but the total number of likely to be good paths is much much smaller (since you never want to, for example, go from 1 to 3 then back to 2 that's about half way between them), and there are good heuristic approaches to the travelling salesman problem that aren't guaranteed to find the best, but find a great solution. (A trivial example is to create the path corresponding to always going to the nearest unused neighbour, then going around the path and trying every flip of nodes one by one to see if it improves it)
This information should help:
-Hourglass works by hooking (injecting a piece of code before every call to) many, many functions that do keyboard input, display graphics on screen, system calls, calls to rand() and so on. This way it can give keyboard input exactly as often as needed, freeze the process at consistent intervals for deterministic feedback, etc. However, if a function is not hooked by Hourglass it does not exist, and the full list of functions it hooks can be seen by examining its source code on google code.
--In particular, this means that the needed information is - what API/functions does Dustforce to get keyboard input on the lowest level?
-Hourglass does not currently support any networking or internet stuff 1) because it would be too difficult to make deterministic for playback, hard to make an input format for, etc 2) someone has to code it, anyway.
-It also doesn't support mouse input, but this isn't due to technical limitations, it's just because it hasn't been implemented, e.g. it would take a while but nothing prevents it otherwise.
Amazing TAS, but one question:
Is possible use the same system to find the address value to execute the end title screen in other games of same console or games from another consoles like Super Metroid, Sonic....
In theory, all it takes to jump to the credits of the game is writing the values into RAM that tell the game 'play the credits' or jumping to the instructions for the game directly that run the credits.
In practice, to do this you need to find a memory corruption AND/OR 'data as code' execution related glitch. - either a very strong, general one, or one that you get lucky with and has a way to jump to credits. Examples from other games are:
-Arbitrarily writing over RAM, and then jumping to the place in code you have written - the ideal case, arbitrary execution of any memory you want (Pokemon Yellow)
-Mapping arbitrary RAM as level data, allowing you to interact with sprites/blocks that shouldn't exist and trick the game into ending the level (Wario Land 2) or going straight to the end sequence (Super Mario Land 2)
-Scribbling a specific value over memory by going out of bounds, writing the fact that the final boss is dead to memory (Dawn of Sorrow) (also Ocarina of Time, but it can't be used to warp you, only to give you items and equipment)
-Running 'code' for a non-existent sprite/weapon/behaviour/player state, executing data as code at the arbitrary addressed pointed to (Kirby's Adventure), and then jumping to controller input to execute MORE code from there (Super Mario World) (if Super Metroid had a credits early glitch it would probably look like this, using one of the glitched beams to execute data as code)
-Trick the game into executing code for more objects than exist, such that it executes the code for the object 'end game' (Battletoads) or 'end level' (Mega Man)
-Read arbitrary memory and interpret it as text of a conversation, glitchily reach the debug menu, then use that to reach the end of the game (Earthbound)
I like snap0130.jpg's aesthetics, the only problem is that Link is barely in the shot.
snap0058.jpg is funny, and probably the way to go if you want a 'don't spoil the ending!' type screenshot :) That look Link is getting... haha
So I would vote for those two
Also, don't forget that tasvideos as designed is not meant to hold all TASes that could possibly be made for every single game, but only a collection of TASes that are both 1) very well optimized and 2) have some merit as judged by the community otherwise. So not every possible TAS will go on this site because it's not designed that way.
Is anyone uploading this to nicovideo? I really want to see what the comments will be like ^^
EDIT: Preferably it should have a controller input overlay, so you can see all four controllers being used at the end
I didn't expect this to be such a huge improvement in entertainment over the previous glitched movie! The seemingly random bizzare things you did had me laughing so hard knowing that it had to be part of the setup but having no clue how. Even knowing how it was done it is still crazy and entertaining. Yes vote :)
Link to video
This beats three gradius games and top gun on the same input.
(I looked around for a bit and couldn't find a general thread for one input many games beating TASes. If there is one feel free to merge this topic with it.)
The run also generates a miss for an attack which was not intended to be able to miss: Lt. Surge's Raichu using Thunderbolt in Vermillion Gym. Previously, I had listed this as luck manipulation, but after examining the disassembled code, I am also convinced that this is a programming error. Each attack has a miss ratio associated with it, which is simply stored as a value between 0 and 255. Attacks which are not intended to be able to miss have a miss ratio of zero. When determining if an attack has failed, a pseudo-random number is choosen, also between 0 and 255, and this value is subtracted from the miss ratio. If there was not a carry (if this value is less than or equal to the miss ratio), the attack will fail. Obviously, the programmer had intended to have these operands switched, and to check for carry instead; in the check for non-physical attacks, this is implemented correctly. This glitch was fixed in later generations of Pokémon.
The only exception is Swift which does not perform this check and, thus, cannot possibly miss.
Finally, I'm really certain X Accuracy boosted moves in RBY can never miss, because when werster streams RBY he's never worried about the 1/256 miss after using an X Accuracy in a fight, and werster knows his stuff :)
EDIT: Ok, now I have a question of my own.
I remember reading or hearing about a statment something like the following: In pokemon GSC, all status effect attacks used on your pokemon have an extra random chance of not working, except confuse ray which is still 100% accuracy. What am I thinking of?
EDIT: Now that the topics are merged I can see werster talking about it a few posts up. Hey hey!
Two random ideas
1) Roadkill spawns at a non-fixed point when you reach his panel, right? Does the game pick a random place and spawn him when you reach the panel, or has he been simulated starting from an earlier point in the level?
1a) In the case of being simulated, is there a certain point in his walking where, if you mess with collision detection earlier (like if you can shake the whole level, or if it's going through a panel at a certain frame, or something) he clips through the cage/through the pedestal?
1b) In the case of being spawned on the frame you enter the panel, is there a certain point he can be spawned at that makes him clip immediately?
2) Is it some kind of sprite overload/sprite messup glitch? Like, if you transition to Roadkill's panel on a certain frame relative to when the paper debris falling down the screen despawns does it mess up their two sprite properties and make something weird happen?
Possible time-saver for the TAS: Mecha_richter was playing Comix Zone and on one of his attempts, the rat on stage 1-1 spawned outside of the cage. This would save needing to destroy the cage. He'll put up a highlight of it here http://www.twitch.tv/mecha_richter/videos whenever he finishes streaming
EDIT: I also imagine he'd like to know exactly what conditions make the trapdoor skip on 1-1 work (where you do a jump kick move against the explosive crate and get recoiled off the trapdoor and fall through the floor), since he messes it up a lot
The principle to follow when working on projects on a computer is 'save early, save often'. I've been bitten by crashing programs and power outages too often.
If the only difference in PJ2.0 vs N64 is lag, then wouldn't it be able to be console verified still? In the same way the SM64 TAS was console verified, even though mupen had different amounts of lag in different places, because the game only polled on input frames and so the TAS could be a stream of input frames.
BisqBot had to be told that increasing x co-ordinate is. good, getting power ups is good, taking warp pipes is good, etc and dying is bad in SMB1. The novelty of this bot is that you don't tell it what to do - it watches you play and assigns itself its goals for the game. BisqBot is great but the idea of this bot is that it decides what it wants, in a way I haven't seen applied to game bots before.
(btw, I think this should be split to http://tasvideos.org/forum/viewtopic.php?t=13996 ) (Mod edit: Done --Mothrayas)
