Editor, Experienced Forum User, Published Author, Skilled player
(1411)
Joined: 3/31/2010
Posts: 2090
Well, I wasn't quite done yet...
Link to videoUser movie #638543431070982544
Turns out, you can use the breakable blocks glitch to create a crevice that is small enough for Simon to zip in, with the help of also scrolling some of the tiles offscreen. The damage boost you see to activate the zip is frame-precise.
It's useless here, but maybe there's some more potential in other places. Of course, the specifics will vary a lot depening on the surrounding geometry. Possible further things to look into would be doing the zip without a damage boost, doing the zip without doing any tile unloading, or doing the zip in such a way that you can get Simon below his starting position, so he zips along the floor in the ground, rather than through the elevated tiles. There's definitely more, I'm sure of it.
In doing so, I also managed to unravel a little more about how the breakable block glitch works in general. Get ready for an overly complex writeup:
There's actually 4 separate objects involved, a "spawner" object that manages the state of the entire wall, and which turns into the item when the wall is destroyed, and three separate block objects, one for each tile:
What now happens is that when the camera is more than 224 pixels away, the game begins unloading the objects. The game tries to unload one object on each frame. If an object cannot be unloaded because it is still in range, the game moves on to the next object on the next frame. Using this, it possible to exit the blocks' loading range on the exact frame the game tries unload a specific block. Using this, it is possible to partially unload pieces of the wall.
If the spawner object gets unloaded, not much of interest happens. The spawner gets reloaded once you enter its spawn range a second time and then creates a duplicate set of blocks over the still existing blocks. This creates the effect that the blocks still produce hit sparks even after they are already destroyed: the second block doesn't realize its corresponding tile was already destroyed and remains active.
Where things get interesting is when the spawner remains intact, but some of its blocks get unloaded. When the lowest block is missing, tile corruption can happen. What I still haven't figured out is why exactly the specific types of corruptions you see happen. There's a few different patterns, and which pattern you get is based on what other objects (excluding candles and some other "static" objects) are currently loaded when the block gets destroyed. This part still comes down to essentially trial and error, but at least the tile corruption itself is easier to set up now.
It's much easier to see what's going on with an object viewer. Object memory starts at 0x440, and each object is 64 bytes in memory. With this, it's easily possible to trigger tile corruption on demand, when previously it was random and incredibly finnicky and inconsistent to perform.
some massive idiot wrote:
Additionally, the glitch appears to have been fixed in the U version of the game.
This is false. So far as I can tell, the glitch works exactly the same on the U and E versions of the game.
By making the camera scroll upwards before jumping down, it is possible to get Simon very close to the lower edge of the screen. In theory, if Simon is low enough, this can be used to clip through any floor, although usually Simon will then just fall to his death.
The edges of these floors are actually sloped tiles that are less than 16 pixels tall. This seems to be the secret sauce that allows Simon to clip in.
The specifics are very precise, and it is hard to apply this trick to other areas, in part, because there's only few instances of free 4-way scrolling in the game, but also because it very much depends on the specific geometry of the area, Simon's position, and how exactly the camera behaves.
In this case, we got a new zip, although it's cut short by the spikes ahead. Some more understanding of how OOB movement works might allow Simon to go past the spikes.
I also suspect a damage boost may be used to clip Simon into any flat floor too, not just sloped floors as in this example, since they interrupt his fall, but haven't managed to get it to work yet.
some idiot wrote:
Turns out stair blocks aren't actually solid from below, but you need a lot of vertical speed to go through them. This can be done with a ring, as seen here:
A bit more poking reveals that you need at least -9 Y speed to go through, while Simon usually only gets -8 from rings. I suspect the ceiling is only 8 pixels thick at the staircase, and the ring's position allows Simon to partially clip into the ceiling, allowing him to make it through with just -8. Still no explanation why it's so ungodly precise with regards to subpixels though.
Update: A full zip is possible by using the aforementioned 'clip through the ground + damage boost' method using the lowest bone pillar in the stage.
Welp. See y'all in a few years again.
Editor, Experienced Forum User, Published Author, Skilled player
(1411)
Joined: 3/31/2010
Posts: 2090
Recently, I've been playing a bunch of custom Doom wads, and one of the things that stuck out most to me was the breadth and quality of custom soundtracks talented composers would make specifically for those wads. So here is just a small sampling of many of the fantastic MIDIs that dedicated composers made over the years. WAD or Map titles (where applicable) are put after each entry.
Mark Klem - Hidden Anger (Memento Mori)
Jeremy Doyle - Quarry (Icarus: Alien Vanguard)
James Paddock - Sanctified (Plutonia 2)
Petter Mårtensen - Temple of Judgment (Hell Revealed II)
B.P.R.D - The Mucus Flow (The Mucus Flow)
There's also the MIDI Pack for the Master Levels for Doom 2 by Peter Lawrence, which I cannot link a single track of, because so many of them are so good.
i could link more.
Editor, Experienced Forum User, Published Author, Skilled player
(1411)
Joined: 3/31/2010
Posts: 2090
I've dabbled in NESdev before and been interested in the limits of raw PCM on a NES, and all I can say is, I am completely stunned. This is so completely ridiculous I cannot even comprehend it. Interleaving APU and PPU writes like this is not something I would've thought would be possible, but the result looks and sounds stunning.
I am completely in awe. Hats off to you.
Thanks also for the really long and detailed write up explaining every step of the way. Breaking down all of the programming steps really helps to understand what is going on.
Editor, Experienced Forum User, Published Author, Skilled player
(1411)
Joined: 3/31/2010
Posts: 2090
Stunning work.
I have to admit it could get quite hard to follow everything that was going on with so many projectiles on screen and two players, but at the same time I couldn't believe how you get through all of this without getting hit.
Absolute yes vote from me as well. This was a treat.
Editor, Experienced Forum User, Published Author, Skilled player
(1411)
Joined: 3/31/2010
Posts: 2090
Before the inevitable deluge of April Fool's submissions, I'll just say that this run was quite enjoyable. The optimization looked pretty good to my layman's eyes, with lines being taken optimally, and you didn't ever seem to waste any time. The speed tricks were especially cool. I didn't think I'd ever see a boss falling to its death outside of a Dark Souls game.
Yes Vote!
Editor, Experienced Forum User, Published Author, Skilled player
(1411)
Joined: 3/31/2010
Posts: 2090
It's a bit unfortunate that you have to run past most enemies, letting them just hit you. Since all damage that you take slows you down, along with the creature having a pretty slow acceleration, it does hurt the flow of watching. Same goes for landing from a jump, and a lot of other animations you need to recover from.
Without knowing anything about this game, could you have picked up a stronger weapon in Stage 3 to finish the slug boss faster? Or is the stick all that you get?
Fwiw, the game was short enough and the TAS was interesting enough to hold my attention, so I'm giving it a yes.
Editor, Experienced Forum User, Published Author, Skilled player
(1411)
Joined: 3/31/2010
Posts: 2090
This was excellent. I'm so glad to finally see a submission for the memory corruption run. The gameplay was great. The optimization shows in every room, and I'll never get tired of seeing the zips. I still wish you could do the glitch earlier into the game, since it's so linear and you see so much from the full categories anyway, but either way, this run was fantastic.
Great work!
Editor, Experienced Forum User, Published Author, Skilled player
(1411)
Joined: 3/31/2010
Posts: 2090
Well, this TAS served to remind me why I love this game so much. Thank you for making it.
There's not much I can say about the gameplay. It seems optimal as far as I can tell, but then the actions you can do are very simple for the most part.
Either way, I certainly enjoyed this a lot. Yes vote.
Editor, Experienced Forum User, Published Author, Skilled player
(1411)
Joined: 3/31/2010
Posts: 2090
Hi Glitchman, and welcome to the site!
This was a fantastic run. I can always go for more Mario, and I'm glad to see the previous warpless run was improved. It's also cool to see a route change in the game.
Great work! I hope you make more TASes in the future.
Editor, Experienced Forum User, Published Author, Skilled player
(1411)
Joined: 3/31/2010
Posts: 2090
Fantastic! I was still holding out for a Haunted Castle TAS, and I'm so glad there finally is one.
I never got far into TASing this game, but I remember it being a veritable pain, since you have to slow down for everything. You seemed to handle it very well overall. Just one question, can you not take more damage in Stage 1 to get past some enemies faster? You seem to have enough health overall.
Anyhow, great work! Absolute yes vote from me. Also a huge thank you to CloakTheLurker for the encode!
Editor, Experienced Forum User, Published Author, Skilled player
(1411)
Joined: 3/31/2010
Posts: 2090
Hi and welcome to the site, Oopla, and congrats on your first TAS!
I really liked watching this, The movement was very fluid and every room was dispatched efficiently. I didn't even know about the shortcut before this. That was very cool to see. It might help that I'm a huge Morphcat fanboy, but this TAS really was a treat for me.
Great work! Yes vote, and keep it up!
Editor, Experienced Forum User, Published Author, Skilled player
(1411)
Joined: 3/31/2010
Posts: 2090
This was recorded on Windows XP Professional Service Pack 3, Version 2002.
And this is why I was hoping we could've ditched hourglass a decade ago.
Anyhow, this is a game I'm actually fairly partial to. Obviously the movement is quite simple, and even in six minutes, it kinda begins to drag on, but it's not too bad. I only wish there were more skips. Is it impossible to stand on the edge of the spikes at 4:52 in the encode?
Anyhow, I enjoyed this enough to vote yes. Keep it up, and huge props for actually having the patience to deal with hourglass.
Editor, Experienced Forum User, Published Author, Skilled player
(1411)
Joined: 3/31/2010
Posts: 2090
So I went and did the thing I probably should've done 7 years ago and actually mapped out all of the ringboosts you get for a ring. Since the positions don't normally change between rings, you can just create a heatmap and show it in-game, which is what I've done.
This allows to quickly get a visual reference for where you need to be to activate good ringboosts. It also finally clears up where exactly good ringboosts are and aren't.
I've written a Lua script that lets you overlay the heatmap over any ring of your choice. You can check it out here (GitHub link).
This is the current heatmap. Black shows high-value negative ring boosts, white shows high-value positive ringboosts. Red shows ringboosts that are above the speed cap of 48 / -48. You can clearly see the radial pattern. You could still extend it further to the sides if you wanted.
Here is the raw data. (CSV format, About 5 MiB, currently)
Editor, Experienced Forum User, Published Author, Skilled player
(1411)
Joined: 3/31/2010
Posts: 2090
BigBoct wrote:
1. The Metroids are immune to Power Bombs in this hack, so you need the Ice Beam, which is at Mama Turtle, which is gated by a metal door linked to the Red Space Pirates in the room below Mt. Everest, which can only be destroyed by the Screw Attack.
2. No, there's not. Ridley's death unlocks that metal door.
Editor, Experienced Forum User, Published Author, Skilled player
(1411)
Joined: 3/31/2010
Posts: 2090
Well, this is Super Metroid Impossible. I have to say, being this early, the hack honestly doesn't hold up very well to me. Most of the rooms are almost completey unchanged, save for the extra spikes, so the TAS can go through many areas just the same way as in the base game. That means that there is not really that much novelty left. Sure, you have to hellrun, but other routes through the main game have that too.
The gameplay and optimization looked mostly fine to my layman's eyes. I was especially surprised to see how smooth most of suitless Maridia went. There were some parts that were questionable, but there's probably a good reason I'm not aware of.
For example:
- Why position yourself so far to the right when picking up the Charge Beam in Upper Red Brinstar?
- Why did you position yourself so far to the right when the Golden Torizo fight ends?
- Is it really necessary to pick up the screw attack?
- Is there really no way to pick up the Wave Beam earlier?
Overall, I liked the movement and gameplay enough to vote yes for Entertainment. Staying on 0 HP for so long an especially cool and unexpected part. Keep it up!
Editor, Experienced Forum User, Published Author, Skilled player
(1411)
Joined: 3/31/2010
Posts: 2090
This was fantastic. There is something that never gets old to me about the game stumbling and going "You Win" after something seemingly innocuous happens.
Yes vote for sure.
What now happens is that when the camera is more than 224 pixels away, the game begins unloading the objects. The game tries to unload one object on each frame. If an object cannot be unloaded because it is still in range, the game moves on to the next object on the next frame. Using this, it possible to exit the blocks' loading range on the exact frame the game tries unload a specific block. Using this, it is possible to partially unload pieces of the wall.
If the spawner object gets unloaded, not much of interest happens. The spawner gets reloaded once you enter its spawn range a second time and then creates a duplicate set of blocks over the still existing blocks. This creates the effect that the blocks still produce hit sparks even after they are already destroyed: the second block doesn't realize its corresponding tile was already destroyed and remains active.
Where things get interesting is when the spawner remains intact, but some of its blocks get unloaded. When the lowest block is missing, tile corruption can happen. What I still haven't figured out is why exactly the specific types of corruptions you see happen. There's a few different patterns, and which pattern you get is based on what other objects (excluding candles and some other "static" objects) are currently loaded when the block gets destroyed. This part still comes down to essentially trial and error, but at least the tile corruption itself is easier to set up now.
It's much easier to see what's going on with an object viewer. Object memory starts at 0x440, and each object is 64 bytes in memory. With this, it's easily possible to trigger tile corruption on demand, when previously it was random and incredibly finnicky and inconsistent to perform.
This allows to quickly get a visual reference for where you need to be to activate good ringboosts. It also finally clears up where exactly good ringboosts are and aren't.
I've written a Lua script that lets you overlay the heatmap over any ring of your choice. You can check it out here (GitHub link).
This is the current heatmap. Black shows high-value negative ring boosts, white shows high-value positive ringboosts. Red shows ringboosts that are above the speed cap of 48 / -48. You can clearly see the radial pattern. You could still extend it further to the sides if you wanted.
Here is the raw data. (CSV format, About 5 MiB, currently)