Atlas videos aren't created by hacking the game, they're created by taking the video footage and super-imposing it on a created map of the level it's going through, moving it around, etc. I hear they're a lot of hard work ;)
Let's see:
-You corrupt memory
-You wrong warp to real and glitched maps
-You end the final boss early by using a glitched move that instantly kills it
TECHNICALLY you fight the final boss, but not in a 'yep, he died fair and square' way.
I don't know if that makes it warp glitch or game end glitch.
http://pastebin.com/MGHHgfc4 Werster's Golden Sun FAQ explains it:
So you can reset, use psynergy, attack cancel, etc to manipulate RNG. There are also rooms with random scenery elements that advance the RNG every frame, but most rooms aren't like this (they make RNG manipulation impossible in RTA).
So you pretty much have godly RNG control all the time.
tl_plexa also has his own FAQ for Golden Sun with more information:
http://pastebin.com/mAJfK1FW
Here's werster's Golden Sun 1 All Djinn Speedrun in 3:57:42
Link to video
And here's tl_plexa's Golden Sun 1 'no intro any%' (basically you start from a savefile that skips the about half hour of boring intro cutscene/plot stuff) in 2:46:48
http://www.twitch.tv/tl_plexa/c/4415639
I think these are still the fastest times for their respective categories.
Also, the nico TASer uploaded another part:
Link to video
If a movie that produces the ending properly is submitted, even if it's slower, it should obsolete this. The main purpose of publishing this movie is thus to draw attention to 'hey, you can do a really cool game ending glitch in this game, check it out' basically.
There's also a third possibility, memory corrupting places in memory that the game uses as code.
Plus an even more diabolical fourth possibility - if a program deliberately uses self-modifying code, and you trick it into self-modifying in a way that produces ACE. There's no way to automatically detect such a situation as ACE that wouldn't flag the program's normal execution too!
The memory values per sage and the RNG occurring before the start of each level match my observations, but unless my newbishness is causing me to miss something here, the 7E1050 address doesn't look like the right location. I see it go to 15 every time on the first level (regardless of cutscene skip point and resulting sage type), go to 0 as the boss fight starts, and then remain unchanged as the second level starts. I haven't checked it later in the game yet, but I'm guessing it won't be helpful there either.
The impression I get from a description like this is that the memory value (or a block containing it) is dynamically allocated per stage, so you should look for places in memory that hold 7E1050 or any value closely lower than it, and see which of said places in memory change when you change the stage. If so, you've found a 'pointer' that will tell you where it's moved to.
These annotations are great. Would it be possible to have them as soft subtitles for the official encodes?
EDIT: I made some notes on your annotations while watching:
4:50: I thought it was actually that there's a pointer for 'last projectile that hit Alucard', in particular pointing to its X position relative to the center of the screen (center being 127, 0 being far left, 255 being far right), and the game uses that pointer to determine if Alucard is more to the left or right of the projectile when damage boosting him, and the toadstool erroneously uses the pointer without setting it again (because the toadstool is not actually a projectile, I guess it would have taken a lot more code to handle this case and no one could be assed). Usually the pointer points to a value 0, e.g. Alucard gets pushed right, but if you get hit by that specific bone in that specific room then go to the clock room, the pointer points to a value that's high enough that it's to the right of Alucard and he gets pushed left.
7:39: Slightly change wording to note that the fairy's speech prevents pausing in addition to the shop menu (it's implied, but you kind of stumble upon it after the whole paragraph is over)
8:04: The bottom line of the annotation is cut off. (In particular, depending on the resolution you view it at, more annotations besides this one have the bottom line cut off, because youtube doesn't conservatively resize the annotations. For example the one at 10:56 also has its bottom line cut off when using the small player. In general you have to make your annotations taller than you think they'd need to be to make sure this doesn't happen. Annoying but necessary.)
8:29: You forget to explain how potion duping works (you wield a potion that the fairy tries to use at the same time, 0-1 = 255, so now you have 255 potions. same as gemstone sell duping)
10:20: When the animation of transformation is happening (the wavey thingy), whatever momentum you had is slowed down but the direction is preserved. But until the animation begins you have that full preserved momentum. So momentum conservation is a property of the transformation that you extend by postponing the animation (it can't process the wavey thing and 6 fire boomerangs at the same time, and apparently some other things too. Go PSX technology!)
16:52 It might be worth mentioning that you came up with a strategy to get the Bat Card when you fought this boss. What made it slower?
17:13 How come this RSL goes an extra room than other RSLs?
I'm still debating on whether or not I want to go back through the published movie and implement these boosts (and test a couple other things in the process). If I did, it wouldn't be for a while.
Finish the no health underflow glitch run, then if you're still hungry go back and improve the any% TAS with everything new you've learned in one big go.
Second, before I continue with the rest of the stage, I'd like some opinions: I'm considering making this run damageless to serve as a counterpart to the very... well, damage-full published run. The way I see it, a damageless run could have an entertainment advantage due to how insanely hard this game is and how it tries its best to hurt you at any given opportunity. The later levels especially would be fun to watch, with all the effortless dancing around enemies and bullets. The downside is that it would be quite a bit slower, likely minutes slower than a run that does use damage + death to save time. The hard hat deathwarp would be taken out, adding 20 seconds of slow swimming on its own, plus there's the obvious extra waiting around, and in a lot of cases there wouldn't be much I could do to make those waits entertaining.
I think showing that the entire game can be beaten no damage/no death would be incredibly entertaining/remarkable.
How do you continue working on a recorded TAS file on DS emu? Like I wanna shutdown my computer, but save my TAS to work on later. How would I do that? I obviously know how to record it, I just wanna know how to pick up where I left off.
I'd be interested in a run of Pac-Man 2: The New Adventures. I don't know how tedious 100% would be, but seeing just how fast the game could be pushed would be interesting enough to me.
This was great. I'm not too big on Portal OOB runs frankly, because it is so hard to follow what's going on in them, whereas this was a lot more watchable.
Voting Yes. Again.
All games on Source engine (CS:S; CS:GO; L4Dead 1,2; Portal 1,2; HL2 series exactly) can easily be speedrunned using in-game slow motion command host_timescale 1.0, where 1.0 is 100% speed; and recording and then playing in-game movie so I'm not surprised the WR is finally done! Nice run though!
In fact, this run was done entirely in realtime using no scripts, only segmenting input. It could be even faster with more TAS tools, but they don't need it! ;)
Link to video
Again, no TAS, but all fastest tech is on display here~
Despite being inbounds only, I think this is the most incomprehensible Portal segmented run yet XD
For those who are tl;dr:
23 frames wasted, avoids 2-3 real time minutes waiting for the boss to self-destruct after input ends.
But it's a slower TAS so it must be rejected.
If it's entertaining enough for moons, you can TAS whatever you like.
So do testruns with light TAS optimization and decide for yourself which has more potential :)
Well, whether a sports game is a good idea to TAS or not usually depends on how much playaround potential it has (See Superstar Soccer Deluxe), which isn't necessarily known if no one has tried TASing it yet.
Does anyone of you have a proof that this even works on console?
Not that it invalidates the run if it doesn't sync on console.
It's not invalidated if it doesn't sync on console, no.
But if it's not possible to do the glitch on console, then the run is no good.
It's not valid to do a TAS that exploits an emulator only glitch, because it's a flaw in the emulator, not in the game's programming.
At 0:35 is the door entered on the first possible frame?
At 0:41 can you start the ground pound any frames earlier/initiate it lower to the first box (by starting the jump earlier so you're lower at that same x position) and still break the boxes?
When you reach 0:47 it doesn't matter what your x position is, you have so much time to adjust it but you're waiting for your y position to be high enough to clear the cliff. So you should time your jumps so you make it onto the first wooden ledge (the one that's left of the four golden coins) on the earliest frame possible, etc.
0:56 do you finish the level faster if you approach the door from the side instead of fall onto it?